THE ROLE OF ANTI-MÜLLERIAN HORMONE IN ASSISTED REPRODUCTION IN WOMEN

Transcription

1 THE ROLE OF ANTI-MÜLLERIAN HORMONE IN ASSISTED REPRODUCTION IN WOMEN A thesis submitted to the University of Manchester for the degree of MD in the Faculty of Medical and Human Sciences 2014 OYBEK RUSTAMOV SCHOOL OF MEDICINE

2 TABLE OF CONTENTS Abstract...3 Publications arising from the thesis..5 Chapter 1. General Introduction & Literature review...8 Chapter 2. Evaluation of the Gen II AMH Assay: between-sample variability and assay- method comparability Anti-Müllerian hormone: serum levels and reproducibility in a large cohort of subjects suggest sample instability AMH Gen II assay: A validation study of observed variability between repeated AMH measurements...65 Chapter 3. The measurement of anti-müllerian hormone: a critical appraisal...78 Chapter 4. Extraction, preparation and collation of datasets for the assessment of the role of the markers of ovarian reserve in female reproduction and IVF treatment Chapter 5. Assessment of determinants of anti-müllerian hormone in infertile women The effect of ethnicity, BMI, endometriosis and the causes of infertility on ovarian reserve The effect of salpingectomy, ovarian cystectomy and unilateral salpingoopherectomy on ovarian reserve Chapter 6. Assessment of determinants of oocyte number using large retrospective data on IVF cycles and explorative study of the potential for optimization of AMH-tailored stratification of controlled ovarian hyperstimulation Chapter 7. General Summary Authors and affiliations Acknowledgments

3 ABSTRACT The University of Manchester Dr Oybek Rustamov Degre: MD Title: The role of anti-müllerian hormone in assisted reproduction in women Date: 30 March 2014 Anti-Müllerian hormone appears to play central role in regulation of oocyte recruitment and folliculogenesis. Serum AMH concentration was found to be one of the best predictors of ovarian performance in IVF treatment. Consequently many fertility centres have introduced AMH for the assessment of ovarian reserve and as a tool for formulation of ovarian stimulation strategies in IVF. However published evidence on reliability of AMH assay methods and the role of AMH-tailored individualisation of ovarian stimulation in IVF appear to be weak. Consequently, I decided to conduct a series of studies that directed towards an improvement of the scientific evidence in these areas of research. The studies on performance of Gen II AMH assay revealed the assay suffers from significant instability and provides erroneous results. Consequently, the manufacturer introduced a modification on assay method. In view of the observed issues with Gen II assay, I conducted a critical appraisal of all published research on the previous and current assay methods that reported AMH variability, assay method comparison and sample stability. The literature indicated clinically important variability between AMH measurements in repeated samples, which was reported to be more significant with Gen II assay. The studies on between-assay conversion factors derived conflicting conclusions. Correspondingly, the review of studies on sample stability revealed conflicting reports on the stability of AMH under normal storage and processing conditions, which was reported to be more significant issue in Gen II assay. In view of above findings, we concluded that AMH in serum may exhibit pre-analytical instability, which may vary with assay method. Therefore robust, international standards for the development and validation of AMH assays are required. In the analysis of determinants of ovarian reserve, I evaluated the effect of ethnicity, BMI, endometriosis, causes of infertility and reproductive surgery on AMH, AFC and FSH measurements using data on a large cohort of infertile patients. Using robust multivariable regression analysis in a large cohort of IVF cycles, I established the effect of age, AMH, AFC, diagnosis, attempt, COS protocol changes, gonadotrophin type, USOR operator, regime and initial dose of gonadotrophins on oocyte yield. Then, I examined effect of gonadotrophin dose and regime on total and mature oocyte numbers. The study found that, after adjustment for all above variables, there was no increase in oocyte yield with increasing gonadotrophin dose categories beyond the very lowest doses. This suggests that there may not be significant direct dose-response effect and consequently strict protocols for tailoring the initial dose of gonadotrophins may not necessarily optimize ovarian performance in IVF treatment. In summary, studies described in this thesis have revealed instability of Gen II assay samples and raised awareness of the pitfalls of AMH measurements. These studies have also demonstrated the effect of clinically measurable factors on ovarian reserve and provided data on the effect of AMH, other patient characteristics and treatment interventions on oocyte yield in cycles of IVF. Furthermore, a robust database and statistical models have been developed, which can be used in future studies on ovarian reserve and IVF treatment interventions. 3

4 DECLARATION No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning; COPYRIGHT STATEMENT i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the Copyright ) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the Intellectual Property ) and any reproductions of copyright works in the thesis, for example graphs and tables ( Reproductions ), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see in any relevant Thesis restriction declarations deposited in the University Library, The University Library s regulations (see and in The University s policy on Presentation of Theses 4

5 PUBLICATIONS ARISING FROM THE THESIS Journal Articles 1. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, Philip W. Pemberton The measurement of Anti-Müllerian hormone: a critical appraisal. The Journal of Clinical Endocrinology & Metabolism, 2014 Mar;99(3): A. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. Anti-Müllerian hormone: poor assay reproducibility in a large cohort of subjects suggests sample instability. Human Reproduction Oct; 27(10): B. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. Human Reproduction. Dec2012, Vol. 27 Issue 12, p3641 5

6 Conference presentations 1. O. Rustamov, S. Roberts, C. Fitzgerald Ovarian endometrioma is associated with increased AMH levels Annual Meeting of European Society of Human Reproduction and Embryology (ESHRE), July 2014, Munich Poster Presentation. 2. O. Rustamov, M. Krishnan, R. Mathur, S. Roberts, C. Fitzgerald. The effect of BMI to the ovarian reserve Annual Meeting of British Fertility Society, January 2014, Sheffield Oral presentation: Dr O. Rustamov. 3. M. Krishnan, O. Rustamov, R. Mathur, S. Roberts, C. Fitzgerald. The effect of the ethnicity to the ovarian reserve Annual Meeting of British Fertility Society, January 2014, Sheffield, Oral Presentation: Dr M. Krishnan. 4. O. Rustamov, M. Krishnan, S. Roberts, C. Fitzgerald. Reproductive surgery and ovarian reserve Annual Meeting of European Society of Human Reproduction and Embryology (ESHRE), July 2013, London Oral presentation: Dr O. Rustamov. 5. C. Fitzgerald, O. Rustamov, P. Pemberton, A. Smith, A. Yates, M. Krishnan, R. Russell, L. Nardo, S.Roberts. AMH assays: A review of the literature on assay method comparability Annual Meeting of European Society of Human Reproduction and Embryology (ESHRE), July 2013, London Oral presentation: Dr C. Fitzgerald. 6. M. Krishnan, O. Rustamov, R. Russell, C. Fitzgerald, S. Roberts The role of the ethnicity and the body weight in determination of AMH levels in infertile women. Annual Meeting of European Society of Human Reproduction and Embryology (ESHRE), July 2013, London 6

7 Poster presentation. 7. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. AMH Gen II assay - can we believe the measurements? 8th Biennial Conference of UK Fertility Societies, January 2013, Liverpool Poster presentation. 8. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. Old and new AMH assays: Can we rely on current conversion factor? 8th Biennial Conference of UK Fertility Societies, January 2013, Liverpool Poster presentation. 9. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. Random AMH measurement is not reproducible 8th Biennial Conference of UK Fertility Societies, January 2013, Liverpool Poster presentation. 10. Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. The reproducibility of serum Anti-Müllerian hormone: AMH Gen II assay Annual Meeting of European Society of Human Reproduction and Embryology (ESHRE), July 2012, Istanbul Oral Presentation: Dr O. Rustamov. 7

8 GENERAL INTRODUCTION AND LITERATURE REVIEW 1 8

9 CONTENTS I. LITERATURE REVIEW...10 GENERAL BACKGROUND OVARIAN RESERVE Primordial Follicle Assembly Oocyte recruitment Theory of neo-oogenesis MARKERS OF OVARIAN RESERVE Ovarian reserve markers with limited clinical value Inhibin B Basal oestradiol Dynamic tests Ovarian volume Ovarian reserve markers in routine clinical use Chronological age Basal FSH Antral follicle count ANTI-MÜLLERIAN HORMONE Biology of anti-müllerian hormone The role of AMH in the ovary AMH in women with polycystic ovary syndrome AMH Assay Variability of AMH measurements Role of AMH in assessment of ovarian reserve Prediction of poor and excessive ovarian response in IVF Prediction of live birth in cycles of IVF Role of AMH in ovarian stimulation for cycles of IVF MULTIVARIATE TESTS SUMMARY 28 II. GENERAL INTRODUCTION..29 REFERENCES 31 9

10 I. LITERATURE REVIEW GENERAL BACKGROUND Infertility is a disease of the reproductive system defined by the failure to achieve a pregnancy after 12 months of regular unprotected sexual intercourse, although the criteria for the duration vary between different countries (NICE 2013). Worldwide prevalence of infertility estimated to be around 72.4 million couples and around 40 million of those seek medical care (Hull et al 1985). In the UK, 15% couples present with infertility with an annual incidence of 1.2 couples per 1000 general population (Scott et al 2009). The main causes of infertility are tubal disease, ovulatory disorders, male factor and poor ovarian reserve. In a third of couples the cause of failure to achieve pregnancy is not established which is known as unexplained infertility (NICE 2013). Effective treatment options include improving lifestyle factors, medical and/or surgical treatment of underlying pathology, induction of ovulation and Assisted Reproductive Technology (ART). Assisted Reproduction consist of intrauterine insemination (IUI) and in vitro fertilisation (IVF) cycles with or without introcytoplasmic sperm injection (ICSI) as well as treatment involving donated gametes. It is estimated that, 75% of infertile couples presenting at primary care centres in the UK are referred to fertility specialists based at secondary or tertiary care centres and nearly 50% of those are subsequently offered IVF ICSI treatment (Scott et al 2009). This is supported by figures of Human Fertility and Embryology Authority (HFEA) which indicates more than 50,000 IVF treatment cycles are performed in the UK annually (HFEA 2008). An IVF treatment cycle involves a) pituitary down regulation, b) controlled ovarian stimulation, c) oocyte recovery, c) in vitro fertilisation of eggs with sperm, d) transfer of resulting embryo(s) back to uterus and c) luteal phase support (NICE 2013). Prevention of premature surge of luteinising hormone during controlled ovarian stimulation (COS) is achieved by pituitary down regulation using either preparations of gonadotrophin releasing hormone agonist, which is widely known as Agonist cycle or gonadotrophin releasing hormone antagonist which is known Antagonist cycle (Figure 1 and 2). Controlled ovarian stimulation involves administration of gonadotrophins to encourage the development of supernumerary preovulatory follicles followed by administration of exogenous human chorionic gonadotropin (hcg) or 10

11 recombinant luteinising hormone (rlh) to assist in maturation of oocytes hours prior to egg collection which is usually conducted with guidance of transvaginal ultrasound scanning. Subject to sperm parameters, the fertilisation of oocytes is conducted by in vitro insemination or intracytoplasmic sperm injection. The resulting embryo(s) are cultured under strict laboratory conditions and undergo regular qualitative and quantitative assessments before transferring the best quality embryo(s) back into uterus during its cleavage (Day 2 or Day 3) or blastocyst (Day 5 or Day 6) stage of development. In natural menstrual cycles, under the influence of HCG, progesterone secreted by the ovarian corpus luteum ensures proliferative changes in the endometrium providing the optimal environment for implantation of embryo(s) (van der Linden et al 2011). However in IVF treatment cycles, owing to pituitary down regulation and lack of HCG, progesterone levels are not in sufficiently high concentration to ensure an adequate endometrial receptivity and therefore exogenous analogues of this hormone is administered following transfer of embryo(s). This is called luteal phase support and, in patients with viable pregnancy, usually lasts till 12 th week of gestation when placenta starts producing progesterone in sufficient quantities (van der Linden et al 2011). In IVF programmes the success of the treatment often, defined as achieving a live birth following IVF cycle and expressed using Live Birth Rate (LBR). In general, success in IVF predominantly determined by woman s age, cause(s) of infertility, ovarian reserve, previous reproductive history and lifestyle factors (NICE 2013; Taylor 2003; Lintsen et al 2005). However, effectiveness of medical interventions as well as the quality of care play important role in determining the outcome of IVF treatment. This is evident from significant variation in live birth rates among fertility clinics, given for instance in the UK LBR for women younger than 35 years of age after IVF cycles varies from 15% to 61% (HFEA 2008; HFEA 2007). The provision of effective interventions, in both clinical and laboratory aspects of the care, appears to be the key in achieving high success rates. Identification of patients with sufficient ovarian reserve who benefit from IVF cycles, followed by providing optimal ovarian stimulation regimens may be useful in improving the outcomes of IVF programmes. According to HFEA data, around 12% of IVF cycles are cancelled due to poor or excessive ovarian response (Kurinczuk et al 2010). Availability of reliable markers for assessment of ovarian reserve and tailoring ovarian stimulation regimens to the need of each individual patient 11

12 may improve selection of patients with sufficient ovarian reserve and reduce the rate of cycle cancellation, consequently improving the success of IVF cycles (Yates et al 2011). Assessment of ovarian reserve can be achieved using various biomarkers, and four of those are currently used by most clinics: woman s chronological age (Age), serum follicle stimulating hormone (FSH), antral follicle count (AFC) and serum anti-müllerian hormone (AMH). More recently, AMH has been a focus of interest, given it is the only available endocrine marker that is suitable for direct assessment of the activity of ovarian follicles in their noncyclical stage development, providing a window to FSH independent phase of follicular recruitment. Furthermore, it appears to be reliable biomarker for a) both the assessment of ovarian reserve and the optimisation of ovarian stimulation regimens (Yates et al 2011; La Marca et al 2009), b) screening and diagnosis of polycystic ovarian syndrome (PCOS) (Cook et al. 2002); c) monitoring of disease activity in women with a history of granulosa cell tumours (Lane et al 1999); d) prediction of the age of diminished fertility and the menopause (van Disseldorp et al 2008; Broer et al 2011) and finally (e) assessment of the long term effect of chemotherapy on ovarian reserve (Anderson 2011). In this review, I first discuss current knowledge on factors that determine ovarian reserve, including the formation and loss of oocyte pool. Then, characteristics of the markers of ovarian reserve are reviewed. Finally I examine current understanding of biology of anti-müllerian hormone and its role in management of infertility. 1. OVARIAN RESERVE It is important to recognize that there is no universal definition for the term ovarian reserve and the term can have various meanings depending on the context in which it is used. For instance, the scientific literature describing the biology of ovarian reserve usually refers to the total number of remaining oocytes in the ovaries, which consists of the number of resting primordial follicles and growing primary, pre-antral and antral follicles (Gleicher et al 2011). In contrast, the use of the term in the context of clinical studies may refer to clinically measurable ovarian reserve established using available biomarkers of ovarian reserve. For the purpose of clarity in this thesis the 12

13 term ovarian reserve refers to clinically measurable ovarian reserve whilst true biological ovarian reserve will be termed biological ovarian reserve. Recent studies have demonstrated that ovarian reserve is highly variable between women due to the variation in the size of initial ovarian reserve at birth as well as the rate of loss of ovarian reserve thereafter (Wallace et al 2010). Interestingly the rate of oocyte loss appears to be mainly determined by the initial ovarian reserve, which is believed to be facilitated by most potent ovarian growth factor, anti-müllerian hormone. Similarly, the size of the initial ovarian reserve is mainly underpinned by the rate of primordial follicle assembly in the embryo, which is also regulated by AMH. Both primordial follicle assembly and the rate of oocyte loss appear to be primarily under the influence of genetic factors, although developmental and environmental factors are also believed to play a role (Nilsson et al 2010, Shuh-Huerta et al 2012). 1.1 Primordial follicle assembly The process of assembly of primordial follicles in the female embryo spans from the early embryonic to the early postnatal period and formation of primordial follicles consists of following stages: 1) primordial germ cell (PGC), 2) oogonia, 3) primary oocyte and 4) primordial follicle. In the human female fetus around a hundred cells that differentiated from extra-embryonic ectoderm form early PGCs on the yolk sac and migrate via hindgut to gonadal ridges during 4 th - 6 th weeks of gestation (MC et al 1953; Donovan 1998). Once arrived to the gonadal ridges these cells are called primary oogonia, which consequently undergo several rounds of mitotic division during 6 th - 28 th weeks of gestation. Interestingly, the numbers of oogonia reach as high as six million during its highest rate of mitotic division at around 20 weeks of gestation. Following the last round of mitotic division oogonia enter meiosis, which marks their new stage of development-primary oocyte. Formation of primordial follicles starts as early as at 8 th week of gestation and is characterised by meiosis of primary oocyte, that arrest in diplotyne stage, and surrounding of the oocyte by somatic granulosa cell (Baker et al 1963; Maheshwari and Fowler 2010). Indeed, the primordial follicle is the cardinal unit of the biological ovarian reserve and therefore the rate of formation of primordial follicles is the main determinant of initial biological ovarian reserve at birth. Interestingly, the process of loss of oogonia and oocytes, which is also one of the main determinants of the initial ovarian reserve, takes place 13

14 throughout the period of follicle assembly. The formation of the granulosa cell layer around the oocyte prevents the oocyte from subsequent atresia. The oocyte enveloped in a single layer of granulosa cells, which is also known as primordial follicle, remains quiescent until recruitment of the follicle for growth which may not take place for a number of decades after the formation of a particular primordial follicle (Skinner 2005; Maheshwari and Fowler 2010). 1.2 Oocyte recruitment Follicle growth in women consists of two stages a) the initial non-cyclical recruitment of primordial follicles and the formation of a primary and a preantral follicles and b) cyclical development of antral follicles with subsequent selection of, usually, a single dominant follicle. The initial recruitment of primordial follicles is continuous non-cyclical process that starts as early as from weeks of gestation and lasts till the depletion of follicle pool, which later results in the menopause (McGee and Hsueh 2000). Transformation of flat granulosa cells into cuboidal cells increases the diameter of the oocyte and the formation of zona pellicuda completes the stage of formation of a primary follicle. During pre-antral stage, oocytes increase in diameter and mitotic division of granulose cells create a new layer of cells-theca cells. The mechanism of initial recruitment of oocytes is not well understood, but it is clear that the process is independent of influence of pituitary gonadotrophins and appears to be governed by the genetically pre-programmed interaction of the oocyte with local growth factors, the most important of which appears to be anti-müllerian hormone, and cytokines (McGee and Hsueh 2000). The cyclical phase of development of oocytes is characterised by the transformation of secondary follicle into antral follicle and subsequent growth of antral follicles into pre-ovulatory stages. In general, the process of cyclic recruitment starts from puberty under the influence of rising levels of pituitary follicular stimulating hormone (FSH). During the antral stage oocyte increases in size even further and the formation of a fluid filled space in follicle is observed. Under the influence of FSH, luteinising hormone (LH) and local growth factorsselection of a single dominant follicle occurs which followsby an ovulation (McGee and Hsueh 2000). Oocyte loss is a continuous process and occurs due to atresia of oocytes during primary, secondary and antral stages of development. The rate of oocyte loss appears to increase until the age of around 14 and declines 14

15 thereafter until the age of the menopause, when around 1000 primordial follicles remain (Hansen et al 2008; Oktem and Oktayl 2008). Furthermore, by the age of 30 years, the average age at which women of western societies plan to start a family, around 90% of initial primordial follicles are lost, which illustrates that formation and maintenance of ovarian reserve is wasteful process in humans (ONS 2012; Wallace and Kelsey 2010) As mentioned above, there is a wide individual variation in both sizes of initial primordial follicular pool and the rate of oocyte loss, which explains variation in the reproductive lifespan in women. Evidently, the number of primordial follicles at birth ranges between around 35,000 to 2.5 million per ovary and similarly, the rate of oocyte loss during its peak at 14 years of age may range between 100 to 7,500 primordial follicles per month which is believed to be inversely proportional to initial size of primordial follicle pool (Wallace and Kelsey 2010). 1.3 Theory of neo-oogenesis The traditional view of oogenesis states that the process of the creation and the mitotic division of oogonia with subsequent formation of primordial follicles takes place only during embryonic and foetal life (Zuckerman 1951). According to this central theory of mammalian reproductive biology, females are born with a certain number of germ cells that is gradually lost, but not renewed, during postnatal period. However, Johnson et. al. have recently challenged this view and reported that adult mammalian ovary may possesses mitotically active germ cells that continuously replenish the primordial follicle pool (Johnson et al 2004). The group reported that, ovaries of juvenile and young adult mice contained large ovoid cells, which resemble germ cells of foetal mouse ovaries. Interestingly, immunohistochemical staining for a gene, which is expressed exclusively in germ cells, have been reported to have confirmed that these large ovoid cells were of germline lineage. Furthermore, application of a mitotic germ cell toxicant, busulphan, appeared to have eliminated primordial follicle reserve by early adulthood but did not induce atresia, suggesting the presence of proliferative germ cells in postnatal mouse ovary (Johnson et al 2004; Bazer 2004). The study has generated enormous amount of interest as well as debate among reproductive biologists (Notarianni 2011). Some other groups have also reported an evidence of postnatal oogenesis (Pacchiarott et al 2010; Zou et al 2009; Bukovsky et al 2004)) while 15

16 others do not support the theory (Bristol-Gould et al 2006; Byskov et al 2005; Begum et al 2008). Furthermore, some authors argued that adult mouse germline stem cells exist and remain quiescent in physiologic conditions and neo-oogenesis occurs only in response to ovotoxic damage (Tilly et al 2007; De Felici 2010). Although consensus has yet to emerge, to date there is no conclusive evidence on validity of theory of neo-oogenesis. 2. MARKERS FOR ASSESMENT OF OVARIAN RESERVE Biological ovarian reserve is defined as the number of primordial and growing follicles left in the ovary at any given time and therefore, only counting the number of primordial follicles by histological assessment can accurately determine ovarian reserve, which is clearly not feasible in clinical setting. However, ovarian reserve can be estimated using various biomarkers, dynamic clinical tests and implied from the outcomes of ART cycles. Although a wide range of clinical (age, ovarian response in previous IVF cycles), biochemical (basal FSH, Inhibin B, basal oestradiol, AMH), ultrasound (ovarian volume, antral follicle count (AFC)) and dynamic (clomiphene challenge test, exogenous FSH ovarian reserve test, GnRH analogue stimulating test) tests of ovarian reserve exist, only a few of the markers are reliable and practical enough to be of use in routine clinical practice. In this chapter, first I discuss the research evidence on the assessment of the markers and/or tests of ovarian reserve that have limited clinical value. Then I evaluated more reliable markers that are in routine clinical use, Age, FSH, AFC, and combination of these markers in multivariable tests. Finally, I conducted detailed review of biology of AMH and the role AMH measurement in the management of infertility. 2.1 Ovarian reserve markers with limited clinical value Inhibin B Inhibins are members of TGFβ family and expressed in granulosa cells of growing follicles. Principal role of inhibins is thought to be the negative feedback regulation of pituitary FSH secretion and therefore the serum level of circulating hormone is believed to reflect the state of folliculogenesis. 16

17 Consequently several groups have studied the role of serum Inhibin β in the assessment of ovarian reserve. Although initial reports were encouraging (Seifer et al 1997), more robust studies demonstrated that serum Inhibin β was less reliable than chronological age or basal FSH (Creus et al 2000; Urbancsek 2005). The systematic review of nine studies demonstrated that accuracy of the Inhibin β test for predicting poor ovarian response and non-pregnancy in IVF cycles was modest, even at a very low threshold level (Broekmans et al 2006). Therefore, it is recommended that inhibin β, at best, can be used as only screening test in the fertility centers where other more reliable markers are not available (Broekmans et al 2006) Basal oestradiol Some studies suggested that elevated basal oestradiol levels indicate low ovarian reserve and are associated with poor fertility prognosis (Johannes et al 1998; Licciardi and Rosenwaks 1995). Johannes et. al. demonstrated basal oestradiol in conjunction with serum FSH is more reliable than serum FSH alone in prediction of cycle cancellation due to the poor response in IVF cycles (Johannes et al 1998). However, there are no published data on the comparison of basal oestradiol to more reliable markers such as AMH or antral follicle count (AFC). Moreover, a recent systematic review has demonstrated that basal oestradiol has very low predictive value for poor response and has no discriminatory power for accuracy of non-pregnancy prediction (Broekmans et al 2006) Dynamic tests of ovarian reserve The dynamic tests of ovarian reserve are based on assessment of ovarian response by measuring serum FSH and oestradiol levels following administration of exogenous stimulation. The following tests are reported in literature: Clomiphene Citrate Challenge Test (CCCT), Exogenous FSH Ovarian Reserve Test (EFORT) and GnRH agonist stimulation test. A recent systematic review and meta-analysis on the accuracy of these tests showed that none of them can adequately predict poor response or non-pregnancy in IVF cycles and therefore are not recommended for use in routine clinical practice (Maheshwari et al 2009). 17

18 2.1.4 Ovarian volume There is some evidence that, increased age is associated with decreased ovarian volume and women with smaller ovaries are more likely to have cancellation of their IVF cycles due to poor ovarian response (Syrop et al 1995; Syrop et al 1999; Templeton 1995). However, a meta-analysis of the published studies, on the accuracy of ovarian volume as a predictor of poor response and non-pregnancy in IVF cycles, failed to demonstrate clinical usefulness of the test and suggested the test is not reliable enough for use in a routine clinical practice (Broekmans et al 2006). 2.2 Ovarian reserve markers in routine clinical use Chronological age Owing to the biological age-related decline of the quantity, and arguably the quality, of oocytes the chronological age can be used as a marker of ovarian reserve. Studies have demonstrated that ovarian reserve (Wallace and Kelsey 2010; Kelsey 2011), natural fecundity (Islam et al 1989; and outcomes of ART (Templeton et al 1996; van Kooij et al 1996) decline significantly from age of 35 when it is believed the ovarian reserve undergoes accelerated decline. Although there is a strong association between chronological age and reduction in fertility, evidently there is a significant variation in age-related ovarian reserve indicating chronological age alone may not be sufficient to estimate the individual woman s ovarian reserve reliably (Broekmans et al 2006) Basal FSH Basal FSH was one of the first endocrine markers introduced in ART programs and is still utilized in many fertility clinics, albeit in conjunction with other markers which are considered more reliable (Creus et al 2000). Secretion of FSH is largely governed by the negative feedback effect of steroid hormones, primarily oestradiol, and inhibins which are expressed in granulosa cells of growing ovarian follicles. Consequently, decreased or diminished recruitment of ovarian follicles is associated increased serum FSH measurements and high, particularly very high basal FSH reading is considered as a good marker of very low or diminished ovarian reserve (Abdalla et al 2006). However, unlike some other markers, FSH measurements do not appear to have discriminatory power for categorisation of patients to various 18

19 bands of ovarian reserve. Given between-patient variability FSH measurement (CV 30%) is similar to its within-patient variability (27%), stratification of patients to various ranges of ovarian reserve does not appear to be feasible (Rustamov et al 2011). Indeed, a recent systematic review of 37 studies on the prediction of poor response and non-pregnancy in IVF cycle has concluded that, basal FSH is an adequate test at very high threshold levels and therefore has limited value in modern ART programs (Broekmans et al 2006) Antral follicle count Antral follicle count estimation involves ultrasound assessment of ovaries between 2 nd and 4 th day of menstrual period and counting follicles, which corresponds to antral stage of folliculogenesis (Broekmans et. al. 2010). The test provides direct quantitative assessment of growing follicles and is known as one of the most reliable markers of ovarian reserve (Broekmans et al 2006). AFC measurement has been reported as having a similar sensitivity and specificity to AMH in prediction of poor and excessive ovarian response in IVF cycles (Broekmans et al 2006; Broer et al 2010; Jayaprakasan et al 2010). Given AFC measurement is available instantly and allows patients to be counseled immediately, the test eliminates the need for an additional patient visit prior to IVF cycle. However, AFC is normally performed only in the early follicular phase of the menstrual cycle, given most published data on measurement of AFC are based on studies that assessed antral follicles during this stage of the cycle (Broekmans et al 2010a). Interestingly, more recent studies suggest that variability of AFC during menstrual cycle is small, particularly when follicles between 2-6mm are counted, and therefore assessment of AFC without account for the day of menstrual cycle may be feasible (Deb et al 2013). One of the main drawbacks of AFC is that the cut off levels for size of counted follicles remains to be standardised (Broekmans 2010b). Initially, follicles of 2-10mm were introduced as the range for AFC and many studies were based on this cut off. Later, counting follicles of 2-6mm was reported to provide most accurate assessment of ovarian reserve (Jayaprakasan et al 2010b; Haadsma et al 2007) and therefore some newer studies are based on AFC measurements that used this criterion. Consequently, direct comparison of the outcomes of various studies on assessment of AFC requires careful analysis. 19

20 3. ANTI-MÜLLERIAN HORMONE 3.1 Biology of Anti-Müllerian hormone AMH is a member of transforming growth factor β superfamily which was discovered by Jost et al., in 1947 and was initially known for its is role in regression of Müllerian ducts in sex differentiation of the male embryo. In women, AMH is believed to be solely produced by ovaries and expressed in granulosa cells of growing follicles of 2-6 mm in size which corresponds to primary, pre-antral and early antral stage of follicular development. Although there has been a report of expression of AMH in endometrial cells, to date there is no other published evidence that supports this finding (Wang et al 2009). Indeed studies that evaluated half-life of AMH in serum have demonstrated that in women who had bilateral salpingo-oopherectomy AMH becomes undetectable within 3-5 days of following surgery, suggesting ovaries are the only source of secretion of AMH in appreciable quantity (La Marca et al 2005b). Anti-Müllerian hormone is a dimeric glycoprotein, which is composed of a long N-terminus and short C-terminus and was believed to be secreted in serum only in this dimeric form (AMH-N, C). Like other members of TGF-β family, which includes inhibins, activins, bone morphogenic proteins (BMPs) and growth and differentiation factors (Massague et al 1990), AMH binds to two type of serine/threonine kinase receptors referred to as type I and type II. In order to activate AMH signaling pathway, both receptors have to form a heteromeric complex. When AMH binds to the type II (AMHR-II) receptor (Massague et al 2000), this will phosphorylate and activate a type I receptor (ALK2, -3 and/or -6) which subsequently activates the SMAD pathway, through phosphorylation of SMAD 1, 5 and/or 8. These activated SMADs interact with SMAD4 and translocate to the nucleus, regulating the expression of different genes, inhibiting the recruitment of primordial follicles and reducing FSH sensitivity in growing follicles. In addition, AMH receptors, as well as the other members of TGF-β family, can activate MAPK and PI3K/AKT pathways. Studies on AMHR II-deficient male mice demonstrated lack of regression of Müllerian ducts, suggesting that type II receptor is essential in AMH signaling (Mishina et al 1996). Similarly, Type I receptors, which includes three members of activin receptor-like kinase (ALK2, ALK3 and ALK6) also appear to play an important role in the regression of Müllerian ducts, although 20

21 the role of ALK 6 in AMH signaling appears not to be crucial (Visser 2003; Clarke et al 2001). The signal transduction pathway of AMH in the ovary is largely not understood. In postnatal mice ovary AMHR-II receptor was expressed in both granulosa and theca cells of pre-antral and antral follicles (Visser 2003). AMH type I receptors ALK 2 and ALK 3 is expressed in foetal as well as adult mouse ovary, while ALK 6 is expressed in only adult ovary (Visser 2003) The role of AMH in the ovary In the mammalian ovary, the role of AMH appears to be one of a regulation of size of the primordial follicle pool by its inhibitory effect on the formation, as well as the growth of, primordial follicles (Nilsson et al 2011). In the embryonic mouse ovary, AMH inhibits the initiation of the assembly of follicles, when the process of apoptosis of the majority of oocytes is observed (Nilsson et al 2011). Consequently, AMH reduces the rate of oocyte loss, which plays an important role in the determination of the size of initial follicle pool. Similarly, in the adult mouse ovary, AMH plays a central role in maintaining the follicle pool. AMH inhibits both the processes of the initial (non-cyclical) recruitment of primordial follicles and subsequent FSHdependent cyclical growth of antral follicles (Figure 3). Inhibition of the initial recruitment of a new cohort of follicles is believed to be achieved by a paracrine negative feedback effect of the rising levels of AMH secreted from already recruited growing follicles (Durlinger et al 1999). Durlinger et al compared the complete follicle population of AMHnull mice and wild type mice of different ages of 25 days, 4 months old and 13 months old and found that the ovaries of 25 day and 4 months old AMHnull females contained significantly higher number of growing, pre-antral and antral follicles but significantly fewer primordial follicles compared to wild-type females (Durlinger et al 1999). Interestingly, almost no primordial follicles were detected in 13 months old AMHnull mice ovaries, suggesting AMH is a potent inhibitor of the recruitment of primordial follicles and in the absence of AMH ovaries undergo premature depletion of primordial follicles due to an accelerated recruitment. Subsequent study conducted by the group demonstrated that, in addition to its inhibitory effect to the resting follicles, AMH also suppresses the development of the growing follicles (Durlinger et al 2001; Durlinger et al 2002; Themmen 2005). It appears that AMH inhibits 21

22 FSH-induced follicle growth by reducing the sensitivity of growing follicles to FSH, which has been confirmed by, in vivo as well as in vitro studies (Durlinger et al 1999; Durlinger et al 2001). In the initial study the group observed that, despite lower levels of serum FSH concentration, ovaries of AMHnull mice contained more growing follicles than that of their wild-type littermates, which has been supported by the findings of subsequent in vitro study (Durlinger et al 1999). Addition of AMH to the culture inhibited FSH-induced follicle growth of pre-antral mouse follicles due to reduction in granulosa cell proliferation (Durlinger et al 2001). In the human embryo, the expression of AMH commences in the late foetal life and can be detected only from 36 weeks of gestation (Rajpert-De et al 1999; Lee et al 1996). Following a small decline in first two years of life, AMH levels gradually increase to peak at (mean 5 ng/ml) around age of 24 years. In line with the pattern of oocyte loss serum hormone levels gradually decline with increasing age and become undetectable around 5 years prior to menopause (Kelsey et al 2011; Nelson et al 2011). It has been suggested that anti-müllerian hormone plays a central role in determining the pace of recruitment of primordial follicles, hence maintaining the primordial follicle pool of postnatal mammalian ovary. Consequently, a reduction in the concentration of circulating AMH signals the exhaustion of the primordial follicle pool and the decline of ovarian function AMH in women with polycystic ovary syndrome Polycystic ovary syndrome (PCOS) endocrine abnormality characterised by increased ovarian androgen secretion, infrequent ovulation and the appearance of polycystic ovaries on ultrasound scan (Dunaif 1997; Homburg et al 1993). It is the commonest endocrine abnormality in women of reproductive age and affects around 15-20% of women. PCOS is also one of the main causes of anovulation and subsequent sub-fertility (Webber et al 2003). Although the role of anti-müllerian hormone in the development of PCOS is not fully understood, it is becoming increasingly evident that the hormone plays an important role in its pathogenesis (Pehlivanov et al 2011). There is a strong association between serum AMH levels and PCOS and it appears that women diagnosed with PCOS have two to three fold higher serum AMH concentration compared to normo-ovulatory women (Cook et al 2002; Pigny et al 2003). Similarly, women with PCOS are found to have 22

23 significantly higher number antral follicles. Interestingly, the expression of AMH in granulosa cells of follicles were found to be 75 times higher in women with PCOS compared to those without a the disease suggesting increased serum AMH in PCOS may be due to increased secretion of hormone per follicle rather than due to an increased number of antral follicles (Pellat et al 2007). High AMH concentrations may act as the main facilitator of abnormal folliculogenesis in PCOS, given the follicles appear to arrest when they reach an antral stage (2-6mm) of development (Rajpert-De et al 1999). Indeed, the studies of Durlinger et al., have demonstrated that AMH inhibits selection of dominant follicle when follicles reach antral stage of development (Durlinger et al 2001). Serum AMH levels appear to decrease with treatment of PCOS, which may play important role in restoration of ovulatory cycles. Studies have reported a significant reduction in serum concentration of AMH following treatment of PCOS with metformin and laparoscopic ovarian diathermy (Falbo et al 2010; Amer et al 2009; Elmashad 2011). Similarly, reduction of BMI following intensified endurance exercise training for treatment of PCOS may also lead to a significant reduction in serum AMH levels (Moran et al 2011). This suggests that there is strong association between serum concentration of AMH and abnormal folliculogenesis in PCOS and therefore understanding the molecular mechanisms of this interaction should be one of the priorities of future research. 3.2 AMH Assays Enzyme-linked immunosorbent assay specific for measurement of anti- Müllerian hormone was first developed in 1990 and was recognised as a significant step in the assessment of ovarian reserve (Hudson et al 1990). Subsequently, a number of non-commercial immunoassays were developed which were mainly used in research settings (Lee et al 1996). Later, Diagnostic Systems Ltd (DSL) and Immunotech, Beckman Coulter Ltd (IOT) introduced two commercial immunoassays for the routine clinical assessment of ovarian reserve, which are known as first generation AMH assays (Nelson and La Marca 2011). These assays employed two different antibodies against AMH and used different standards for calibration providing non-comparable measurements (Nelson and La Marca 2011). Consequently, several studies attempted to develop a reliable between-assay conversion factor, which interestingly revealed from five-fold higher with the IOT assay to assay 23

24 equivalence causing significant impact to reliability of AMH measurements and interpretation of research findings (Hehenkamp et al 2006; Freour et al 2007; Bersinger et al 2007; Taieb et al 2008; Lee et al 2011). Later, the manufacturer of IOT assay (Beckmann Coulter Ltd.) consolidated the manufacturer of the DSL assay (Diagnostic Systems Laboratories Inc.) and introduced a new assay Gen II AMH assay which is only available commercial immunoassay in most countries, including the UK. AMH Gen II assay was developed using the antibodies derived from first generation DSL assay and calibrated using the standards used for IOT assay and was believed to be considerably more stable compared to the first generation immunoassays providing more reliable measurements (Kumar et al 2010; Nelson and La Marca 2011). The manufacturer as well as initial external validation study recommended, when compared to old DSL assay, AMH Gen II assay provides around 40% higher measurements and therefore previously reported DSL-based clinical cut-off levels for estimation of ovarian reserve should be increased by 40% in order to use Gen II-based AMH results (Kumar et al 2010; Wallace et al 2011; Nelson and La Marca 2011). 3.3 Variability of AMH measurements It is generally believed that AMH values do not change throughout the menstrual cycle and early studies reported that variation in AMH measurements between repeated measurements of same patient was negligible (van Disseldorp et al 2010; La Marca 2010). On the basis of these studies sampling at a random time in the menstrual cycle was introduced as a method for measurement of AMH in routine clinical practice. However, the methodologies of some of these studies do not appear to be robust enough to reliably estimate sample-to-sample variability of AMH, which is mainly due to small sample sizes (Rustamov et al 2011). Consequently, in a recent study we assessed sample-to-sample variability of AMH using DSL assay and found that within-subject coefficient of variation (CV) of AMH between samples were as high as 28%, which cannot be attributed to any patient or cycle characteristics (Rustamov et al 2011). Although there is no consensus in the causes of this observed variability in AMH measurements, we believe it is largely attributable to instability of AMH samples given initial recruitment of primordial follicles and growth of AMH producing pre-antral and antral follicles are continuous process and therefore the true biological variation between samples is unlikely 24

25 to be high. However, given the importance of establishing true variability of AMH, in both understanding of the biology of hormone and clinical application of the test, future studies should be conducted to establish the source of variability in the clinical samples The role of AMH in the assessment of ovarian reserve Prediction of poor and excessive ovarian response in cycles of IVF A number of studies have assessed the role of AMH in the prediction of poor ovarian response in IVF cycles using first generation AMH assays and found that AMH and AFC were the best predictors of poor ovarian response compared to other markers of ovarian reserve. Nardo et al showed that the predictive value of AMH in receiver operating characteristic curve (ROC) analysis was similar to (AUC 0.88) that of AFC (AUC 0.81) and found that AMH cut offs of >3.75 ng/ml and <1.0 ng/ml would have modest sensitivity and specificity in predicting the extremes of response (Nardo et al 2009). These findings were largely supported by subsequent prospective studies and a systematic review (Nelson et al 2007; Jayaprakasan et al 2010; Broer et al 2011). Similarly, comparison of chronological age, basal FSH, ovarian volume, AFC and AMH found that only AMH (AUC 0.90) and AFC (AUC 0.93) were reliable predictors of poor ovarian response in cycles of IVF. Subsequent combination of the effect of AMH and AFC using multivariable regression analysis did not improve the level of prediction of poor ovarian response significantly (AUC 0.94); suggesting both AMH and AFC can be used as independent markers (Jayaprakasan et al 2010). Similarly, most studies agree that AMH and AFC are the best predictors of excessive ovarian response and ovarian hyperstimulation syndrome (OHSS) compared to other clinical, endocrine and ultrasound markers (Nardo et al 2009; Nelson et al 2007). Broer et al. compared these two tests in systematic review of 14 studies and reported that the summary estimates of the sensitivity and the specificity for AMH were 82 and 76% respectively and for AFC 82% and 80% respectively (Broer et al 2011). Consequently, the study concluded that AMH and AFC were equally predictive and the difference in the predictive value between the tests was not statistically significant. 25

26 Prediction of live birth rate (LBR) in cycles of IVF Lee at al. reported that AMH and chronological age were more accurate than basal FSH, AFC, BMI and causes of infertility in the prediction of live birth rate (Lee et al 2009). Similarly La Marca et al. suggested that odds of live birth could be reliably predicted using AMH (La Marca et al 2010b), although, subsequent review of the study questioned strength of the evidence (Loh and Maheshwari 2011). A study conducted by Nelson et al., found that higher AMH levels had stronger association with increased live birth rate compared to age and FSH (Nelson et al 2007). However, the study also suggested that this association was mainly confined in the women with low AMH levels and there was no additional increase in live birth in women with AMH levels of higher than 7.10 pmol/l. This may suggest that achieving a live birth may be under the influence of number of other factors and that markers of ovarian reserve alone may not be able predict this outcome reliably. 3.5 The role of AMH in individualisation of ovarian stimulation in IVF cycles Prediction of ovarian response to the stimulation of ovaries in cycles of IVF plays an important role in the counseling of couples undergoing treatment programmes, and hence many clinical studies on AMH have focused on the prognostic value of AMH measurements. However, data on using AMH as a tool for improving the clinical outcomes in IVF cycles appear to be lacking, considering AMH may be useful tool in tailoring treatment strategies to an individual patient s ovarian reserve. Unlike most other markers, AMH has discriminatory power in determining various degrees of ovarian reserve due to significantly higher between patient (CV 94%) variability compared to its within-patient (CV 28%) variation (Rustamov et al 2011), which allows stratification of patients into various degrees of (e.g. low, normal, high) ovarian reserve. Subsequently, most optimal ovarian stimulation protocol may be established for each band of ovarian reserve. Consequently, reference ranges on the basis of distribution of AMH in infertile women were developed which were subsequently adopted by fertility clinics for a tailoring the mode of 26

27 ovarian stimulation and daily dose of gonadotrophins in IVF (The Doctors Laboratory However, currently available clinical reference ranges are based on the first generation DSL assay and may not be reliably convertible to currently available Gen II assay measurements (Wallace et al 2011). Indeed, the findings of the studies on comparability of the first generation AMH assays suggest that establishing a reliable between assay conversion factor between AMH assays may not be straightforward. Furthermore the reference ranges appear to reflect the distribution of AMH measurements within a specific population and may therefore not be directly applicable for the prediction of response to ovarian stimulation in IVF patients (The Doctors Laboratory 2008). More importantly, despite lack of good quality evidence on the effectiveness of AMH-tailored ovarian stimulation protocols, a number of fertility clinics appear to have introduced various AMH-based COH protocols in their IVF programs. At present, research evidence on AMH-tailored ovarian stimulation in IVF is largely based on two retrospective studies (Nelson et al 2009; Yates et al 2012). Both of these studies display considerable methodological limitations, including small sample size and centre-related or period-related selection of their cohorts. In this context AMH is used as a tool for therapeutic intervention and therefore the research evidence should ideally be derived from randomised controlled trials. However, recruitment of large enough patients in IVF setting may take considerable time and resources. In the meantime, given AMH-tailored ovarian stimulation has already been introduced in clinical practice and there is urgent need for more reliable data, the studies with a larger cohorts and robust methodology should assess the role of AMH in individualisation of ovarian stimulation in IVF treatment cycles. 4 Multivariate models of assessment of ovarian reserve. In view of the fact there is not a single marker of ovarian reserve that can accurately predict ovarian response, various models for combination of multiple ovarian markers have been developed (Verhagen et al 2008). A number of studies reported that multivariate models are better predictors of poor ovarian response in IVF compared to a single marker (Bancsi et al 2002; Balasch et al 1996; Creus et al 2000; Durmusoglu et al 2004). However, a metaanalysis showed that when compared to a single marker (AFC) multivariate 27

28 model has a similar accuracy in terms of prediction of poor ovarian response (Verhagen et al 2008). In contrast, a more recent study demonstrated that multivariate score was superior to chronological age, basal FSH or AFC alone in predicting likelihood of poor ovarian response and clinical pregnancy (Younis et al 2010). However, the study did not include one of the most reliable markers, AMH, in either arm, necessitating further assessment of the role of combined tests, which include all reliable biomarkers. 4. SUMMARY During the last two decades a significant leap has been taken towards understanding the biology of anti-müllerian hormone and its role in female reproduction (Durlinger et al 2002; Themmen et al 2005). Availability of commercial AMH assays has resulted in significant increase in interest in the role of the measurement of serum AMH in the assessment of ovarian reserve, which has been followed by the introduction of the test into routine clinical practice (Nelson et al 2011). However, more recent studies suggest that current methodologies for the measurement of AMH may provide significant sampling variability (Rustamov et al 2011). Furthermore, the studies that compared first generation commercial assay methods appear to provide non-reproducible results, suggesting there may be underlying issues with assay methodologies (Lee et al 2011). Similarly, despite lack of sufficient evidence in the role of AMH in individualisation of ovarian stimulation protocols in IVF, AMHtailored IVF protocols have been introduced in routine clinical practice of many fertility clinics around the world. Consequently, it appears that clinical application of AMH test has surpassed the research evidence in some aspects of fertility treatment and therefore future projects should be directed toward areas where gaps in research evidence exist. On the basis of the review of literature we believe that, evaluation of the performance of assay methods, understanding the role of AMH in assessment ovarian reserve and establishing its role in individualisation of ovarian stimulation protocols should be research priority. 28

29 II. GENERAL INTRODUCTION On the basis of the review of published literature I have identified that the following areas of research on the clinical application of AMH in the management of infertility requires further investigation: 1) Within-patient variability of measurement of AMH using Gen II assay method, 2) Establishment of clinically measurable determinants of AMH levels and 3) The role of AMH in individualisation of ovarian stimulation in IVF treatment cycles. In our previous study we estimated that, there was significant sample-tosample variation (CV 28%) in AMH measurements when the first generation DSL assay was used (Rustamov et al 2011). The source of variability is likely to be related to the assay method given that biological within-cycle variation of AMH is believed to be small (La Marca et al 2006). Therefore assessment of sample-to-sample variability of AMH using the newly introduced Gen II assay, which is believed to be significantly more stable and sensitive compared to that of DSL assay, should enable us to establish the measurement related variability of AMH. Furthermore, given I am planning to use data from both DSL and Gen II assays, I need to establish between-assay conversion factor for these assays using data on clinical samples. There appears to be a lack of good quality data on the effect of ethnicity, BMI, causes of infertility, reproductive history and reproductive surgery on ovarian reserve. Therefore, I am planning to ascertain the role of above factors on determination of ovarian reserve by analysing AMH measurements of a large cohort of patients. There is a strong correlation between AMH and ovarian performance in IVF treatment when conventional ovarian stimulation using GnRH agonist regimens with a standard daily dose of gonadotrophins are used (Nelson et al 2007; Nardo et al 2007). Furthermore, studies suggest tailoring the ovarian stimulation protocols to AMH measurement may improve ovarian performance and subsequently the success of IVF treatment (Nelson et al 2011; Yates et al 2012). However, given methodologies of the published studies the effectiveness of currently proposed AMH-tailored ovarian stimulation protocols remains unknown. Therefore, I am planning to develop individualised ovarian stimulation protocols by establishing the most optimal mode of pituitary down regulation and starting dose of gonadotrophins for 29

30 each AMH cut-off bands using a robust research methodology. However, development of individualised ovarian stimulation protocols on the basis of retrospective data requires a reliable and validated database containing a large number of observations. In the IVF Department of St Mary s Hospital we have data on a large number of patients who underwent ovarian stimulation following the introduction of AMH. However, the data on various aspects of investigation and treatment of patients is stored in different clinical data management systems and may not be easily linkable. In addition it appears that data on certain important variables (e.g. causes of infertility, AFC) are available only in the hospital records, necessitating searching for data from the hospital records of each patient. Consequently, I designed a project for building a research database, which will have comprehensive and validated datasets that are necessary for investigation of the research questions of the MD programme. In conclusion, I am planning to conduct a series of studies to improve the understanding of the role of AMH in the management of women with infertility. Specifically I am intending to evaluate 1) sample-to-sample variability of Gen II AMH measurements; 2) conversion factor between DSL and Gen II assays in clinical samples; 3) the effect of ethnicity, BMI, causes of infertility, endometriosis, reproductive history and reproductive surgery to ovarian reserve and explore AMH-tailored individualisation of ovarian stimulation in IVF cycles. 30

31 References Abbeel E., The Istanbul consensus workshop on embryo assessment: proceedings of an expert meeting. Human reproduction, : p Abdalla, H.T., M Y, Repeated testing of basal FSH levels has no predictive value for IVF outcome in women with elevated basal FSH. Human reproduction, (1): p Amer SA, L.T., Ledger WL, The value of measuring anti-mullerian hormone in women with polycystic ovary syndrome undergoing laparoscopic ovarian diathermy Human reproduction, : p Anderson RA. Pretreatment serum anti-mullerian hormone predicts long-term ovarian function and bone mass after chemotherapy for early breast cancer. J Clin Endocrinol Metab 2011;96: Balaban B, B.D., Calderón G, Catt J, Conaghan J, Cowan L, Ebner T, Gardner D, Hardarson T, Lundin K, Cristina Magli M, Mortimer D, Mortimer S, Munné S, Royere D, Scott L, Smitz J, Thornhill A, van Blerkom J, Van den Baker, A quantitative and cytological study of germ cells in human ovaries. Proc R Soc Lond B Biol Sci, : p Balasch J, C.M., Fabregues F, Carmona F, Casamitjana R, Ascaso and V.J. C, Inhibin, follicle-stimulating hormone, and age as predictors of ovarian response in in vitro fertilization cycles stimulated with gonadotropin-releasing hormone agonist-gonadotropin treatment. Am J Obstet Gynecol, : p Bancsi LF, B.F., Eiijekemans MJ at al.,, Predictors of poor ovarian response in in vitro fertilisation: a prospective study comparing basal markers of ovarian reserve. Fertility and Sterility, : p Bazer, F.W., Strong science challenges conventional wisdom: new perspectives on ovarian biology. Reprod Biol Endocrinol, : p. 28. Begum, S., V.E. Papaioannou, and R.G. Gosden, The oocyte population is not renewed in transplanted or irradiated adult ovaries. Hum Reprod, (10): p Bersinger NA, Wunder D, Birkhäuser MH, Guibourdenche J. Measurement of anti-mullerian hormone by Beckman Coulter ELISA and DSL ELISA in assisted reproduction: differences between serum and follicular fluid. Clin Chim Acta 2007;384: Bristol-Gould, S.K., et al., Fate of the initial follicle pool: empirical and mathematical evidence supporting its sufficiency for adult fertility. Dev Biol, (1): p Broekmans, F.J., et al., A systematic review of tests predicting ovarian reserve and IVF outcome. Hum Reprod Update, (6): p

32 Broekmans Frank J. M., de Ziegler Dominique, Howles Colin M., Gougeon Alain, Trew Geoffrey, and Olivennes Francois. The antral follicle count: practical recommendations for better standardization. Fertility and Sterility, : p Broer SL, Eijkemans MJ, Scheffer GJ, van Rooij IA, de Vet A, Themmen AP, Laven JS, de Jong FH, Te Velde ER, Fauser BC, Broekmans FJ. Anti-mullerian hormone predicts menopause: a long-term follow-up study in normoovulatory women. J Clin Endocrinol Metab Aug;96(8): Broer, S.L., et al., AMH and AFC as predictors of excessive response in controlled ovarian hyperstimulation: a meta-analysis. Hum Reprod Update (1): p Bukovsky, A., et al., Origin of germ cells and formation of new primary follicles in adult human ovaries. Reprod Biol Endocrinol, : p. 20. Byskov, A.G., et al., Eggs forever? Differentiation, (9-10): p Clarke, T.R., et al., Mullerian inhibiting substance signaling uses a bone morphogenetic protein (BMP)-like pathway mediated by ALK2 and induces SMAD6 expression. Mol Endocrinol, (6): p Cook CL, Siow Y, Brenner AG, Fallat ME. Relationship between serum Mullerian inhibiting substance and other reproductive hormones in untreated women in PCOS and normal women. Fertil Steril 2002;77: Creus M, P.J., Fa bregues F, Vidal E, Carmona F, Casamitjana R, and B.J. Vanrell JA, Day 3 serum inhibin B and FSH and age as predictors of assisted reproduction treatment outcome. Human reproduction, : p Creus M, P.a.J., Fabregues F, Vidal E, Carmona F, Casamitjana R, et al., Day 3 serum inhibin B and FSH and age as predictors of assisted reproduction treatment outcome. Human reproduction, : p Cook CL, S.Y., Brenner AG, et al., Relationship between serum Mullerian inhibiting substance and other reproductive hormones in untreated women in PCOS and normal women. Fertility and Sterility, : p Deb S, Campbell B. K., Clewis J.S., Pincott-Allen C and Raine-Fenning N.J. Intracycle variation in number of antral follicles stratified by size and in endocrine markers of ovarian reserve in women with normal ovulatory menstrual cycles Ultrasound Obstet Gynecol 2013; 41: De Felici, M., Germ stem cells in the mammalian adult ovary: considerations by a fan of the primordial germ cells Mol Hum Reprod. 16(9): p Donovan, P.J. (1998). The germ cell the mother of all stem cells. Int. J. Dev. Biol. 42, Dunaif A. Insulin resistance and the polycystic ovary syndrome: mechanism adn implications for pathogenesis. Endocr Rev, : p

33 Durlinger, A.L., et al., Control of primordial follicle recruitment by anti- Mullerian hormone in the mouse ovary. Endocrinology, (12): p Durlinger, A.L., et al., Anti-Mullerian hormone attenuates the effects of FSH on follicle development in the mouse ovary. Endocrinology, (11): p Durlinger, A.L., J.A. Visser, and A.P. Themmen, Regulation of ovarian function: the role of anti-mullerian hormone. Reproduction, (5): p Durmusoglu F, E.K., Yoruk P, Erenus M., Combining day 7 follicle count with the basal antral follicle count improves the prediction of ovarian response. Fertility and Sterility, : p Ebner, T., et al., Basal level of anti-mullerian hormone is associated with oocyte quality in stimulated cycles. Hum Reprod, (8): p Elmashad A.I. Impact of laparoscopic ovarian drilling on anti-müllerian hormone levels and ovarian stromal blood flow using three-dimensional power Doppler in women with anovulatory polycystic ovary syndrome. Fertility and Sterility, : p Falbo A, R.M., Russo T, D'Ettore A, Tolino A, Zullo F, Orio F, Palomba S., Serum and follicular anti-mullerian hormone levels in women with polycystic ovary syndrome (PCOS) under metformin. J Ovarian Resere, Jul: p. 16. Fanchin, R., et al., Anti-Mullerian hormone concentrations in the follicular fluid of the preovulatory follicle are predictive of the implantation potential of the ensuing embryo obtained by in vitro fertilization. J Clin Endocrinol Metab, (5): p Fasouliotis, S.J., A. Simon, and N. Laufer, Evaluation and treatment of low responders in assisted reproductive technology: a challenge to meet. J Assist Reprod Genet, (7): p Freour T, Mirallie S, Bach-Ngohou K, Denis M, Barriere P, Masson D. Measurement of serum anti-mullerian hormone by Beckman Coulter ELISA and DSL ELISA: comparison and relevance in assisted reproduction technology (ART). Clin Chim Acta 2007;375: Gleicher, N., A. Weghofer, and D.H. Barad, Defining ovarian reserve to better understand ovarian aging. Reprod Biol Endocrinol. 9: p. 23. Haadsma ML, B.A., Groen H, Roeloffzen EM, Groenewoud ER, Heineman MJ, et al., The number of small antral follicles (2 6 mm) determines the outcome of endocrine ovarian reserve tests in a subfertile population. Human reproduction, : p Hansen KR, Knowlton NS, Thyer AC, Charleston JS, Soules MR, Klein NA. A new model of reproductive aging: the decline in ovarian non-growing follicle number from birth to menopause. Hum Reprod. 2008;23:

34 Hansen KR, Hodnett GM, Knowlton N, Craig LB. Correlation of ovarian reserve tests with histologically determined primordial follicle number. Fertil Steril. 2011;95: Hazout, A., et al., Serum antimullerian hormone/mullerian-inhibiting substance appears to be a more discriminatory marker of assisted reproductive technology outcome than follicle-stimulating hormone, inhibin B, or estradiol. Fertil Steril, (5): p Hehenkamp WJ, Looman CW, Themmen AP, de Jong FH, te Velde ER, Broekmans FJ. Anti-Mullerian hormone levels in the spontaneous menstrual cycle do not show substantial fluctuation. J Clin Endocrinol Metab 2006;91: HFEA, Fertility Figures HFEA, HFEA Fertility Facts and Figures HFEA Homburg R, B.D., Levy T, Feldberg D, Ashkenazi J, Ben-Rafael Z, In vitro fertilisation and embryo transfer for the treatment of infertility associated with polycystic ovary syndrome. Fertility and Sterility, : p Hudson, P.L., et al., An immunoassay to detect human mullerian inhibiting substance in males and females during normal development. J Clin Endocrinol Metab, (1): p Hull MG, G.C., Kelly NJ, et al., Population study of causes, treatment and outcome of infertility. Br Med J Clin Res Ed, : p Islam, M.N. and M.M. Islam, Biological and behavioural determinants of fertility in Bangladesh: Asia Pac Popul J, (1): p Jayaprakasan, K., et al., A prospective, comparative analysis of anti-mullerian hormone, inhibin-b, and three-dimensional ultrasound determinants of ovarian reserve in the prediction of poor response to controlled ovarian stimulation.(2010a). Fertil Steril (3): p Jayaprakasan et al. (2010b). The cohort of antral follicles measuring 2 6 mmreflects the quantitative status of ovarian reserve as assessed by serum levels of anti-mullerian hormone and response to controlled ovarian stimulation. Fertil Steril_ 2010;94: Johannes L. H. Evers, M.D., Peronneke Slaats, M.S., Jolande A. Land, M.D., John C. M. Dumoulin, Ph.D., and Gerard A. J. Dunselman, M.D., Elevated Levels of Basal Estradiol-17β Predict Poor Response in Patients with Normal Basal Levels of Follicle-Stimulating Hormone Undergoing In Vitro Fertilization. Fertility and Sterility, 1998(69): p Johnson, J., et al., Germline stem cells and follicular renewal in the postnatal mammalian ovary. Nature, (6979): p Kelsey, T.W., et al., A validated model of serum anti-mullerian hormone from conception to menopause. PLoS One (7): p. e

35 Kumar, A., et al., Development of a second generation anti-mullerian hormone (AMH) ELISA. J Immunol Methods. 362(1-2): p Kurinczuk JJ et. al. Fertility Treatment in A Statistical Analysis. HFEA, La Marca A, De Leo V, Giulini S, Orvieto R, Malmusi S, Giannella L, Volpe A. Anti-Mullerian hormone in premenopausal women and after spontaneous or surgically induced menopause. J Soc Gynecol Investig 2005b;12: La Marca, A., et al., Normal serum concentrations of anti-mullerian hormone in women with regular menstrual cycles (2010a). Reprod Biomed Online (4): p La Marca, A., et al., Anti-Mullerian hormone-based prediction model for a live birth in assisted reproduction (2010b). Reprod Biomed Online (4): p La Marca, A., et al., Anti-Mullerian hormone (AMH) as a predictive marker in assisted reproductive technology (ART). Hum Reprod Update. 16(2): p La Marca, A., et al., Anti-Mullerian hormone (AMH): what do we still need to know? Hum Reprod, (9): p Lane AH, Lee MM, Fuller AF Jr, Kehas DJ, Donahoe PK, MacLaughlin DT. Diagnostic utility of Mullerian inhibiting substance determination in patients with primary and recurrent granulosa cell tumors. Gynecol Oncol 1999;73:51. Lee, M.M., et al., Mullerian inhibiting substance in humans: normal levels from infancy to adulthood. J Clin Endocrinol Metab, (2): p Lee, T.H., et al., Impact of female age and male infertility on ovarian reserve markers to predict outcome of assisted reproduction technology cycles. Reprod Biol Endocrinol, : p Lee JR, Kim SH, Jee BC, Suh CS, Kim KC, Moon SY. Antimullerian hormone as a predictor of controlled ovarian hyperstimulation outcome: comparison of two commercial immunoassay kits. Fertil Steril 2011;95: Licciardi FL, L.H., Rosenwaks Z., Day 3 estradiol serum concentrations as prognosticators of ovarian stimulation response and pregnancy outcome in patients undergoing in vitro fertilization.. Fertility and Sterility, : p Lie Fong, S., et al., Anti-Mullerian hormone: a marker for oocyte quantity, oocyte quality and embryo quality? Reprod Biomed Online, (5): p Lintsen, A.M., et al., Effects of subfertility cause, smoking and body weight on the success rate of IVF. Hum Reprod, (7): p Maheshwari, A. and P.A. Fowler, Primordial follicular assembly in humans-- 35

36 revisited. Zygote, (4): p Maheshwari, A., et al., Dynamic tests of ovarian reserve: a systematic review of diagnostic accuracy. Reprod Biomed Online, (5): p Massague, J., et al., TGF-beta receptors and TGF-beta binding proteoglycans: recent progress in identifying their functional properties. Ann N Y Acad Sci, : p Massague, J. and Y.G. Chen, Controlling TGF-beta signaling. Genes Dev, (6): p Mc KD, H.A., Adams EC, Danziger S, Histochemical observations on the germ cells of human embryos. Anat Rec, : p McGee, E.A. and A.J. Hsueh, Initial and cyclic recruitment of ovarian follicles. Endocr Rev, (2): p Mishina, Y., et al., Genetic analysis of the Mullerian-inhibiting substance signal transduction pathway in mammalian sexual differentiation. Genes Dev, (20): p Moran LJ, H.C., Hutchinson SK, Stepto NK, Strauss BJ, Teede HJ, Exercise decreases anti-mullerian horomone in anovulatory overweight women with polycystic ovary syndrome-a pilot study. Horm Metab Res, October. Nardo, L.G., et al., Circulating basal anti-mullerian hormone levels as predictor of ovarian response in women undergoing ovarian stimulation for in vitro fertilization. Fertil Steril, (5): p Nelson, S.M., R.W. Yates, and R. Fleming, Serum anti-mullerian hormone and FSH: prediction of live birth and extremes of response in stimulated cycles-- implications for individualization of therapy. Hum Reprod, (9): p Nelson, S.M., et al., Anti-Mullerian hormone-based approach to controlled ovarian stimulation for assisted conception. Hum Reprod, (4): p Nelson, S.M. and A. La Marca, The journey from the old to the new AMH assay: how to avoid getting lost in the values Reprod Biomed Online. Nelson, S.M., et al., External validation of nomogram for the decline in serum anti-mullerian hormone in women: a population study of 15,834 infertility patients. Reprod Biomed Online (2): p NICE, Assessment and treatment for people with fertility problems. NICE Guidelines, Nilsson EE, Savenkova MI, Schindler R, Zhang B, Schadt EE, et al. (2010) Gene Bionetwork Analysis of Ovarian Primordial Follicle Development. PLoS ONE 5(7): e Nilsson EE, Savenkova MI, Schindler R, Zhang B, Schadt EE, et al. (2010) Gene Bionetwork Analysis of Ovarian Primordial Follicle Development. PLoS 36

37 ONE (7): Nilsson, E.E., et al., Inhibitory actions of Anti-Mullerian Hormone (AMH) on ovarian primordial follicle assembly. PLoS One (5): p. e Notarianni, E., Reinterpretation of evidence advanced for neo-oogenesis in mammals, in terms of a finite oocyte reserve J Ovarian Res. 4(1): p. 1. Office of National Statistics, , 2011, Live Births in England and Wales by Characteristics of Mother Oktem, O. and B. Urman, Understanding follicle growth in vivo. Hum Reprod. 25(12): p Oktem, O. and K. Oktay, The ovary: anatomy and function throughout human life. Ann N Y Acad Sci, : p Ottosen, L.D., et al., Pregnancy prediction models and eset criteria for IVF patients--do we need more information? J Assist Reprod Genet, (1): p Pacchiarotti, J., et al., Differentiation potential of germ line stem cells derived from the postnatal mouse ovary. Differentiation (3): p Paternot G, W.A., Thonon F, Vansteenbrugge A, Willemen D, Devroe J, Debrock S, D'Hooghe TM, Spiessens C., Intra- and interobserver analysis in the morphological assessment of early stage embryos during an IVF procedure: a multicentre study. Reprod Biol Endocrinol, : p Pehlivanov B, O.M., Anti-Müllerian hormone in women with polycystic ovary syndrome. Folia Medica, : p Pellat L, H.L., Brincat M, et al., Granulosa cell production of anti-müllerian hormone is increased in polycystic ovaries. J Clin Endocrinol Metab, : p Pigny P, M.E., Robert Y, et al., Elevated serum level of anti-mullerian hormone in patients with polycystic ovary syndrome: relationship to the ovarian follicle excess and the follicular arrest. J Clin Endocrinol Metab, : p Porter, R.N., et al., Induction of ovulation for in-vitro fertilisation using buserelin and gonadotropins. Lancet, (8414): p Rajpert-De Meyts, E., et al., Expression of anti-mullerian hormone during normal and pathological gonadal development: association with differentiation of Sertoli and granulosa cells. J Clin Endocrinol Metab, (10): p Rustamov O, Pemberton PW, Roberts SA, Smith A, Yates AP, Patchava SD, Nardo LG. The reproducibility of serum anti-müllerian hormone in subfertile women: within and between patient variability. Fertil Steril 2011;95:

38 Scott Wilkes, Murdoch Alison D.C and Greg Rubin. Epidimiology and management of infertility: a poppulation-based study in UK primary care. Family Practice, : p Seifer DB, L.-M.G., Hogan JW, Gardiner AC, Blaza AS, Berk CA., Day 3 serum inhibin-b is predictive of assisted reproductive technologies outcome. Fertility and Sterility, : p Skinner, M.K. (2005). Regulation of primordial follicle assembly and development. Hum. Reprod. Update 11, Syrop, C.H., et al., Ovarian volume may predict assisted reproductive outcomes better than follicle stimulating hormone concentration on day 3. Hum Reprod, (7): p Syrop, C.H., A. Willhoite, and B.J. Van Voorhis, Ovarian volume: a novel outcome predictor for assisted reproduction. Fertil Steril, (6): p Taieb J, Belville C, Coussieu C, Guibourdenche J, Picard JY, di Clemente N. Two immunoassays for antimullerian hormone measurement: analytical and clinical performances. Ann Biol Clin (Paris) 2008;66: Taylor A. ABC of subfertility. Making a diagnosis. Br Med J Clin Res Ed, : p Templeton, A., J.K. Morris, and W. Parslow, Factors that affect outcome of invitro fertilisation treatment. Lancet, (9039): p Templeton, A., Infertility-epidemiology, aetiology and effective management. Health Bull (Edinb), (5): p TDL test update. AMH: Stability, Hormones and OCPs. The Doctors Laboratory Guide 2008 page 29. Themmen, A.P., Anti-Mullerian hormone: its role in follicular growth initiation and survival and as an ovarian reserve marker. J Natl Cancer Inst Monogr, 2005(34): p Tilly, J.L. and J. Johnson, Recent arguments against germ cell renewal in the adult human ovary: is an absence of marker gene expression really acceptable evidence of an absence of oogenesis? Cell Cycle, (8): p Urbancsek, J., Use of serum inhibin B levels at the start of ovarian stimulation and at oocyte pickup in the prediction of assisted reproduction treatment outcome. Fertility and Sterility, (2): p van der Linden M, B.K., Farquhar C, Kremer JAM, Metwally M, Luteal phase support for assisted reproduction cycles (Review). Cochrane Library, October. 38

39 van Disseldorp, J., et al., Comparison of inter- and intra-cycle variability of anti-mullerian hormone and antral follicle counts. Hum Reprod (1): p van Kooij, R.J., et al., Age-dependent decrease in embryo implantation rate after in vitro fertilization. Fertil Steril, (5): p van Rooij, I.A., et al., Serum anti-mullerian hormone levels: a novel measure of ovarian reserve. Hum Reprod, (12): p Velde ER, Fauser BC, Broekmans FJ. Anti-mullerian hormone predicts menopause: a long-term follow-up study in normoovulatory women. J Clin Endocrinol Metab 2011;96: van Disseldorp J, Kwee C.B.L., J, Looman C.W.N, Eijkemans M.J.C and F.J. Broekmans. Comparison of inter- and intra-cycle variability of anti-mu llerian hormone and antral follicle counts. Human reproduction, : p Verberg, M.F., et al., Predictors of low response to mild ovarian stimulation initiated on cycle day 5 for IVF. Hum Reprod, (7): p Verhagen, T.E., et al., The accuracy of multivariate models predicting ovarian reserve and pregnancy after in vitro fertilization: a meta-analysis. Hum Reprod Update, (2): p Visser, J.A., AMH signaling: from receptor to target gene. Mol Cell Endocrinol, (1-2): p Wallace, W.H. and T.W. Kelsey, Human ovarian reserve from conception to the menopause. PLoS One. 5(1): p. e8772. Wallace, A.M., et al., A multicentre evaluation of the new Beckman Coulter anti-mullerian hormone immunoassay (AMH Gen II). Ann Clin Biochem (Pt 4): p Webber L J, S.S., Stark J, Trew G H, Margara R, Hardy K, Franks S, Formation and early development of follicles in the polycystic ovary. Lancet, (September): p Yates AP, Rustamov O, Roberts SA, Lim HY, Pemberton PW, Smith A, Nardo LG. Anti-Mullerian hormone-tailored stimulation protocols improve outcomes whilst reducing adverse effects and costs of IVF. Hum Reprod 2011;26: Younis, J.S., et al., A simple multivariate score could predict ovarian reserve, as well as pregnancy rate, in infertile women. Fertil Steril (2): p Zou, K., et al., Production of offspring from a germline stem cell line derived from neonatal ovaries. Nat Cell Biol, (5): p Zuckerman, The number of oocytes in the mature ovary. Recent Prog. Horm. Res., (63-108). 39

40 Figure 1. Schematic representation of a long GnRH agonist cycle In a long agonist cycle, pituitary down regulation is achieved by administration of a daily dose of GnRH agonist preparations starting from mid-luteal phase of the preceding menstrual cycle till the day of administration of HCG. Cycle Started Menstrual Period Daily GnRH agonist From midluteal phase Daily GnRH agonist Menstrual Period Daily GnRH agonist & Daily hmg Day 2-10 HCG USOR & ET

41 Figure 2. Schematic representation of GnRH antagonist cycle In an antagonist cycle, pituitary down regulation is achieved by administration of a daily dose of GnRH antagonist preparations starting from the 5 th day of IVF cycle till the day of administration of HCG. Therefore, an Antagonist cycle is significantly shorter than an Agonist cycle. Cycle Started Menstrual Period Daily GnRH antagonist (Day 5-10) & Daily hmg (Day 2-10) HCG USOR & ET 41

42 Figure 3. The role of AMH in regulation of oocyte recruitment and folliculogenesis. It appears that AMH plays an important role in a) the recruitment of primordial follicles and b) the selection of a dominant follicle from a cohort of antral follicles. AMH is believed to be the main regulator of ovarian reserve, which is achieved by its paracrine negative feedback effect to resting primordial follicles (Durlinger et al., 1999). AMH was found to play an important role in the regulation of the selection of a dominant follicle by inhibition of the FSH-induced follicle growth (Durlinger et al., 2001). 42

43 EVALUATION OF THE GEN II AMH ASSAY: BETWEEN-SAMPLE VARIABILITY AND ASSAY-METHOD COMPARABILITY 2

44 ANTI-MÜLLERIAN HORMONE: SERUM LEVELS AND REPRODUCIBILITY IN A LARGE COHORT OF SUBJECTS SUGGEST SAMPLE INSTABILITY Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, Philip W. Pemberton Human Reproduction 2012a; 27:

45 Title Anti-Müllerian hormone: serum levels and reproducibility in a large cohort of subjects suggest sample instability Authors Oybek Rustamov a, Alexander Smith b, Stephen A. Roberts c, Allen P. Yates b, Cheryl Fitzgerald a, Monica Krishnan d, Luciano G. Nardo e, Philip W. Pemberton b Institutions a Department of Reproductive Medicine, St Mary s Hospital, Central Manchester Foundation Trust, Manchester M13 0JH, UK; b Department of Clinical Biochemistry, Central Manchester Foundation Trust, Manchester M13 9WL, UK; c Health Sciences - Methodology, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 9PL, UK; d School of Medicine, University of Manchester, Manchester, M13 9WL, UK; e Reproductive Medicine and Gynaecology Unit, GyneHealth, Manchester M3 4DN, UK Corresponding author Oybek Rustamov, MRCOG Research Fellow in Reproductive Medicine Department of Reproductive Medicine, St Mary s Hospital, Central Manchester Foundation Trust, Manchester M13 0JH, UK oybek.rustamov@cmft.nhs.uk, oybek_rustamov@yahoo.co.uk Word count: 3909 Conflicts of Interest: There are no potential conflicts of interest. Acknowledgement of financial support Dr Steve Roberts is supported by the NIHR Manchester Biomedical Research Centre. 45

46 Declaration of authors roles: OR led on clinical aspects of this study with responsibility for collation of the clinical database and the analysis of the clinical data. OR prepared the first draft of the clinical work and was involved in preparation of the whole paper and submission of the final manuscript. CF and LGN contributed to clinical data analysis, draft preparation and approval of the final manuscript. MK was involved in clinical data collation and approval of the final draft. PWP was the laboratory lead responsible for all of the laboratory based experiments and for the routine analysis of clinical samples. PWP prepared the first draft of the laboratory work and was involved in the preparation of the whole paper and submission of the final manuscript. AS suggested the sample stability studies and was involved in discussion, draft preparation and approval of the final manuscript. APY was involved in some of the routine clinical analyses, and progression of drafts to approval of the final manuscript. SAR was involved in clinical study design, oversaw the statistical analysis and progression of drafts through to approval of the final manuscript. OR and PWP should be considered as joint first authors. 46

47 ABSTRACT Title Anti-Müllerian hormone: serum levels and reproducibility in a large cohort of subjects suggest sample instability Study question What is the variability of anti-müllerian hormone (AMH) concentration in repeat samples from the same individual when using the Gen II assay and how do values compare to Gen I (DSL) assay results? Summary answer Both AMH assays displayed appreciable variability, which can be explained by sample instability. What is known already AMH is the primary predictor of ovarian performance and is used to tailor gonadatrophin dosage in cycles of IVF/ICSI and in other routine clinical settings. A robust, reproducible and sensitive method for AMH analysis is of paramount importance. The Beckman Coulter Gen II ELISA for AMH was introduced to replace earlier DSL and Immunotech assays. The performance of the Gen II assay has not previously been studied in a clinical setting. Study design, size, and duration For AMH concentration study, we studied an unselected group of 5007 women referred for fertility problems between 1 st September 2008 to 25 th October 2011; AMH was measured initially using the DSL AMH ELISA and subsequently using the Gen II assay. AMH values in the two populations were compared using a regression model in log(amh) with a quadratic adjustment for age. Additionally, women (n=330) in whom AMH had been determined in different samples using both the DSL and Gen II assays (paired samples) identified and the difference in AMH levels between the DSL and Gen II assays was estimated using the age adjusted regression analysis. In AMH variability study, 313 women had repeated AMH determinations (n=646 samples) using the DSL assay and 87 women had repeated AMH determinations using the Gen II assay (n=177 samples) were identified. A mixed effects model in log (AMH) was utilised to estimate the sample-to- 47

48 sample (within-subject) coefficients of variation of AMH, adjusting for age. Laboratory experiments including sample stability at room temperature, linearity of dilution and storage conditions used anonymised samples. Main results and the role of chance In clinical practice, Gen II AMH values were ~20% lower than those generated using the DSL assay instead of the 40% increase predicted by the kit manufacturer. Both assays displayed high within-subject variability (Gen II assay CV=59%, DSL assay CV=32%). In the laboratory, AMH levels in serum from 48 subjects incubated at RT for up to 7 days, increased progressively in the majority of samples (58% increase overall). Pre dilution of serum prior to assay, gave AMH levels up to twice that found in the corresponding neat sample. Pre-mixing of serum with assay buffer prior to addition to the microtitre plate gave higher readings (72% overall) compared to sequential addition. Storage at -20ºC for 5 days increased AMH levels by 23% compared to fresh samples. The statistical significance of results was assessed where appropriate. Limitations, reasons for caution The analysis of AMH levels is a retrospective study and therefore we cannot entirely rule out the existence of differences in referral practices or changes in the two populations. Wider implications of the findings Our data suggests that AMH may not be stable under some storage or assay conditions and that this may be more pronounced with the Gen II assay. The published conversion factors between the Gen II and DSL assays appear to be inappropriate for routine clinical practice. Further studies are urgently required to confirm our observations and to determine the cause of the apparent instability. In the meantime, caution should be exercised in the interpretation of AMH levels in the clinical setting. Key Words Anti-Müllerian hormone; Müllerian inhibitory substance; AMH; AMH Gen II ELISA; DSL Active MIS; AMH ELISA; sample stability. 48

49 INTRODUCTION AMH in women is secreted by the granulosa cells of pre-antral and small antral follicles (Vigier et al 1984; Themmen 2005) and circulating levels reflect the ovarian pool from which follicles can be recruited (Loh & Maheshwari 2011). Measurement of AMH has become of paramount significance in clinical practice in IVF units to assign candidates to the most suitable controlled ovarian hyperstimulation protocol and its level is used to predict poor or excessive ovarian response (Nelson et al 2007; Nardo et al 2009; Yates et al 2011). It is also of increasing importance in (a) prediction of live birth rate in IVF cycles (La Marca et al 2011); (b) screening/diagnosis of polycystic ovarian syndrome (Cook et al 2002); (c) follow up of women with a history of granulosa cell tumours (Lane et al 1999); (d) prediction of the age of onset of infertility due to the menopause (van Disseldorp et al. 2008, Broer et al 2011) and finally (e) assessment of the long term effect of chemotherapy on fertility (Anderson 2011). Following development of the first laboratory AMH assay in 1990 (Hudson et al 1990; Lee et al 1996) first generation, commercially available immunoassays were introduced by Diagnostic Systems Ltd (DSL) and Immunotech Ltd (IOT). These assays used different antibodies and standards (Nelson & La Marca 2011) and the resulting AMH concentrations obtained using the IOT assay were found to be higher than those produced using the DSL assay by most, but not all, authors (Freour et al 2007;Taieb et al 2008; Lee et al 2011). The AMH Gen II Assay (Beckman-Coulter Ltd) replaced both of these assays, using the DSL Gen I antibody with the IOT standards. AMH values obtained using this kit were predicted to correlate with, but be higher than, those using the old DSL kit (Kumar et al 2010; Nelson & La Marca 2011). This was confirmed (Wallace et al 2011) with the AMH Gen II assay giving values approximately 40% higher than the DSL assay. The recommended conversion factor of 1.4 (AMH Gen II = DSL x 1.4) was also applied to the DSL reference ranges but this recommendation does not appear to have been independently validated. It is generally accepted that serum AMH concentrations are highly reproducible within and across several menstrual cycles and therefore a single blood sampling for AMH measurement has been accepted as routine practice 49

50 (Hehenkamp et al 2006; La Marca et al 2006; Tsepelidis et al 2007). However, we recently challenged this view and reported significant sample-to-sample variation in AMH levels using the DSL assay in women who had repeated measurements; 28% difference between samples taken from the same patient with a median time between sampling of 2.6 months and taking no account of menstrual cycle (Rustamov et al 2011). Although we could not explain the cause of this variability, we speculated that it might be due to true biological variation in secretion of AMH or due to post-sampling, pre-analytical instability of the specimen. Given the widespread adoption of AMH in Clinical Units, it is critical that the sources of variability in any AMH assay are understood and quantified. This paper presents the results of clinical and laboratory studies on routine clinical samples using the new AMH Gen II assay, specifically comparing assay values with the older DSL assay, assessing between sample variability and investigating analytical and pre-analytical factors affecting AMH measurement. METHODS Study population Samples were obtained from women of years of age attending for investigation of infertility requiring AMH assessment at the secondary (Gynecology Department) and tertiary (Reproductive Medicine Department) care divisions of St Mary s Hospital, Manchester from 1 st September 2008 to 25 th October Samples which were lipaemic or haemolysed and samples not frozen within 2 hours of venepuncture were excluded from the study. Anonymised samples from this pool of patients were used for stability studies after routine AMH measurements had been completed. The full dataset comprised AMH results on 5868 samples from 5007 women meeting the inclusion criteria. Additionally, we identified women in whom AMH had been determined in different samples using both the DSL and Gen II assays (paired samples from 330 women). 50

51 Sample processing Collection and handling of all AMH samples was conducted according to the standards set out by the manufacturers and did not vary between the different assays. Serum samples were transported immediately to the Department of Clinical Biochemistry, based in the same hospital, and separated within 2 hours of venepuncture using the Modular Pre-Analytics Evo (Roche Diagnostics, Burgess Hill, West Sussex, UK). Samples were frozen in aliquots at -20 C until analysis, normally within one week of receipt. The laboratory participates in the pilot National external quality assessment scheme (UKNEQAS) for AMH in Edinburgh and performance has been satisfactory. AMH analysis All AMH assays were carried out strictly according to the protocols provided by the manufacturer and sample collection and storage also conformed to these recommendations. All AMH samples were analysed in duplicate and the mean of the two replicates was reported as the final result. 1) The DSL AMH assay. The enzymatically amplified two-site immunoassay (DSL, Active MIS/AMH ELISA; Diagnostic Systems Laboratories, Webster, Texas) was used for measurement of AMH prior to 17 th November The working range of the assay was up to 100pmol/L with a minimum detection limit of 0.63pmol/L. The intra-assay coefficient of variation (CV) (n=16) was 3.9% (at 10pmol/l) and 2.9% (at 56pmol/l). The inter-assay CV (n=60) was 4.7% (at 10pmol/l) and 4.9% (at 56pmol/l). 2) The Beckman Coulter Gen II assay. After 17th November 2010, AMH was measured using the enzymatically amplified two-site immunoassay (AMH Gen II ELISA, Beckman Coulter, Inc., Brea, CA, USA). The working range of the assay is up to 150pmol/L with a minimum detection limit of 0.57pmol/L. The intra-assay CV (n=16) is 2.92% (at 18pmol/l) and 2.03% (at 60pmol/l). The inter-assay coefficient of variation (n=28) is 3.57% (at 18pmol/l) and 3.64% (at 60pmol/l). Sample Stability Studies (1) Stability of AMH in serum at room temperature (RT): serum samples (n = 48) were allowed to thaw and then left at RT for one week. At 0, 1, 2, 4 and 7 days, 100µl aliquots were removed and immediately stored at -80 ºC in 51

52 2ml screw-capped polypropylene tubes (Alpha Laboratories, Eastleigh, UK). Two freeze/thaw cycles had no effect on AMH concentration (results not shown). Samples from individual subjects were analysed for AMH on the same GenII microtitre plate to eliminate inter-assay variability. Results were expressed as a percentage of the day 0 value. (2) Linearity of Dilution: 100µl fresh serum (n = 9) was added to 100µl AMH Gen II sample diluent, incubated for 30min at RT and the mixture analysed using the standard GenII assay procedure. (3) Comparison between the Standard Assay method and an equivalent procedure: in the standard GenII ELISA assay method, the first steps involve the addition of calibrators, controls or serum samples to microtitration wells coated with anti-amh antibody; Assay buffer is then added to each well. As a comparison, serum and assay buffer were mixed in a separate tube, incubated for 10min at RT and then added in exactly the same volume and proportions to the microtitre plate. Thereafter, the assay was performed using the standard protocol. (4) Stability of AMH during storage: fresh serum samples (n = 8) analysed on the day of reception were compared with aliquots from the same samples that had been frozen for 5 days either in polystyrene tubes at -20 C or polypropylene tubes at -80 C Statistical Analysis Data analysis was performed using the Stata 12 analytical package (StataCorp, Texas, USA). Data management and analysis of clinical data was conducted by one of the researchers (OR) and verified independently by another member of the research team (SR) using different statistical software (R statistical environment). Approval for the use of the data was obtained from the Local Research Ethics Committee (UK-NHS 10/H1015/22). The agerelated relationship of the DSL and Gen II assays to AMH was visualised using scatter plots and quadratic fit on a logarithmic scale (Nelson et al 2011). The age adjusted regression analysis of paired samples was used to estimate the difference in AMH levels between the DSL and Gen II assays. A mixed effects model in log (AMH) was utilised to estimate the sample-to-sample (withinsubject) coefficients of variation of AMH levels in women who had repeated 52

53 measurements, within a 1 year period from the patient s first AMH sample, adjusting for age as above. In the sample stability studies, percentage changes are expressed as mean ± SEM. In the stability of AMH in serum at RT study, a paired t-test determined the level of significance between baseline and subsequent days. RESULTS Population studies and variability: AMH concentration Table 1 summarizes the results of AMH determinations in our population of women attending the IVF Clinic prior to the 17 th November, 2010 (using the DSL assay) and after that date (using the Gen II assay). A second analysis compares AMH levels in women who had AMH measured using both assays at different times. Results were consistent with lower serum levels of AMH observed when samples were analysed using the Gen II assay compared to the DSL assay. Figure 1 shows the correlation of AMH with age for the unselected groups. After adjustment for age, the total cohorts showed Gen II giving AMH values 34% lower than those for DSL. Analysis restricted to patients with AMH determinations using both assays gave an age-adjusted difference of 21%. AMH variability During the study period, 313 women had repeated AMH determinations (n=646 samples) using the DSL assay with 295 patients having two samples, 17 three samples and one five samples. The median time between samples was 5.1 months. Eighty seven women had repeated AMH determinations using the Gen II assay (n=177 samples) with 84 women having two samples and 3 having three samples. The median interval between repeat samples was 3.2 months. Both assays exhibit high sample-to-sample variability (CV); this was 32% in the DSL assay group (our previous finding (Rustamov et al 2011) in a smaller group was 28%); variability in the Gen II assay group was much higher (59%). 53

54 AMH [pmol/l] Table 1. Median and inter-quartile range for the two assays in the different datasets, along with the mean difference from an ageadjusted regression model expressed as a percentage. all data paired sample s DSL n age AMH (pmol/l ) (29, 36) (7.8,25.0 ) (29, 36) 14.9 (7.4, 24.7) Gen II n Age AMH (pmol/l ) (29, 36) (4.5, 21.6) (30, 37) 11.0 (5.6, 20.9) difference (%) (-39.5 to ) (-36.2 to -6.4) Figure 1. Unselected AMH values from DSL (circles) and Gen II (triangles) assays as a function of age. Lines show the regression fits of log(amh) against a quadratic function of age; solid lines Gen II, broken lined DSL. DSL Gen II Age 54

55 Sample stability studies (1) Stability of AMH in serum at room temperature: AMH levels in 11 of the 48 individuals remained relatively unchanged giving values within ±10% of the original activity over the period of a week and one patient had an undetectable AMH at all time points. The remaining 36 serum samples had AMH values that increased progressively with time. In the 47 samples with detectable AMH, levels increased significantly (p<0.001) for each time interval compared to baseline, the increase at day 7 being ± 7.6 % (Figure 2). Figure 2. Stability of AMH in serum at RT Results at each time interval are expressed as a percentage of the patient s AMH concentration at day 0. Means ± SEM are indicated. 55

56 (2) Linearity of Dilution: In a group of nine anonymised samples, proportionality with two-fold sample dilution does not hold and, on average, there is a 57.4 ± 12.3% increase in the apparent AMH concentration on dilution, compared to neat sample (see table 2a). Two samples which gave the highest increases were diluted further. It was apparent that, after the anomalous doubling of AMH concentration on initial two-fold dilution, subsequent dilutions gave a much more proportional result (see Table 2b). Linearity of dilution was maintained only in samples that showed no initial increase on two-fold dilution. Table 2a. Proportionality with two-fold dilution of serum AMH (pmol/l) sample no. neat serum x2 dilution recovery % Table 2b. Linearity with multiple dilution of serum AMH (pmol/l) sample no dilution Measured expected recovery (%) 1 x x x x x x x x x x

57 (3) Comparison between the Standard Assay method and an equivalent procedure: Serum samples that had been pre-mixed with buffer prior to addition gave, on average, 71.8 ± 4.8% higher readings than those added sequentially using the standard procedure (see table 3). Table 3. Comparison between equivalent ELISA procedures AMH (pmol/l) sample no. A B B/A (%) A = 20µl serum added directly to the plate followed by 100µl assay buffer B = 60µl serum + 300µl assay buffer mixed & incubated at RT for 10min. 120µl mixture added to the plate (4) Stability of AMH during storage: AMH levels in samples stored at -20 C showed an average increase of 22.5 ± 11.1% over 5 days compared with fresh values, while those samples stored at -80 C showed no change (1.8 ± 3.1%) (see Table 4). Table 4. Stability of AMH in serum on storage AMH (pmol/l) sample fresh -20ºC, PS -80ºC, PP no PS = polystyrene LP4 tube; PP = polypropylene 2ml tube 57

58 DISCUSSION This publication arose from two, initially separate, pieces of work in the Clinical IVF Unit at St Mary s Hospital and in the Specialist Assay Laboratory at Central Manchester Foundation Trust. The IVF Unit had become concerned with their observed increase in variation in AMH values and consequently with the reliability of their AMH-tailored treatment guidance. The Laboratory wished to establish whether the practice of sending samples in the post (which has been adopted by many laboratories, rather than frozen as specified by Beckman) was viable. It soon became clear that these anomalies observed in clinical practice might be explained by a marked degree of sample instability seen in the Laboratory which had not previously been reported and which may, or may not, have been an issue with previous AMH assays. The data contained in this paper represents the largest retrospective study on the variability of the DSL assay and the first study on the variability of the Gen II assay. Early studies reported insignificant variation between repeated AMH measurements, suggesting that a single AMH measurement may be sufficient in assessment of ovarian reserve (La Marca et al 2006; Tsepelidis et al 2007). However, these recommendations have been challenged by a number of groups (Lahlou et al 2006; Wunder et al 2008; Rustamov et al 2011). The current study in a large cohort of patients has demonstrated substantial sample-to-sample variation in AMH levels using the DSL assay and an even larger variability using the Gen II assay. We suggest that this variability may be due to sample instability related to specimen processing given that a) AMH is produced non-cyclically and true biological variation is believed to be small (Fanchin et al 2005; van Disseldorp et al 2009) and b) the intra-and inter assay variation in our laboratory for both the DSL and Gen II assays is small (<5.0%) suggesting that the observed variation is not due to poor analytical technique. The population data presented in this paper also suggests that, in routine clinical use, the Gen II assay provides AMH results which are 20-40% lower compared to those measured using the DSL assay. This is in contrast to validation studies for the Gen II assay which showed that this assay gave AMH values ~40% higher than those found with the DSL assay (Kumar et al 2010; Preissner et al 2010, Wallace et al 2011). 58

59 All samples in this retrospective study were subject to the same handling procedures, and analyzed by the same laboratory; the two populations were comparable, with the same local referral criteria for investigation of infertility and we are unaware of any other alterations in practice which might produce such a large effect on AMH. we cannot rule out the possibility of other changes in the population being assayed that were coincident in time with the assay change. However any such change would have to be coincident and produce a 50% decrease in observed AMH levels to explain our findings. We did note a weak trend towards decreasing AMH over calendar time; assuming a linear trend in the analysis implies that AMH values might be 12% (2-22%) lower when the Gen II assay was being used compared to the Gen I assay. This suggests that the age adjusted analysis of repeat samples on individuals, showing a 21% decrease in AMH with the Gen II assay, is currently the best estimate of the assay difference. This is the first study to compare AMH assays in a routine clinical setting in a large group of subjects, and as such is likely to reflect the true nature of the relationship between AMH measured by two different ELISA kits and avoids some of the issues in other published studies. Previous laboratory studies have compared AMH assays in aliquots from the same sample which only provides data on the within-sample relationship between the two assays (Kumar et al 2010, Preissner et al 2010, Wallace et al 2011). Although it is difficult to give a definitive explanation for the discrepancy between the previously published studies (on within-sample relationships) and this study (on between-sample relationships), we suggest that it may be due to degradation of the specimen in one (or both) of the assays. If AMH in serum is unstable under certain storage and handling conditions, this might result in differing values being generated because of differential sensitivity of the two assays to degradation products. Unfortunately we cannot suggest which step of sample handling might have caused this discrepancy, since the published studies did not provide detailed information. The present study used samples which were frozen very soon after phlebotomy and analysed shortly thereafter, hopefully minimising storage effects. The most striking change followed incubation over a period of 7 days at RT; this showed a substantial increase in AMH levels, rather than the expected decline. Previously, Kumar et al (2010) had shown that the average variation between fresh serum samples and those stored for seven days to be 59

60 approximately 4% at 2-8ºC and <1% at -20ºC but presented no data on RT stability. Zhao et al (2007) reported that AMH values were likely to differ by <20% in samples incubated at RT for 2 days compared to those frozen immediately. Several supplementary experiments were performed in order to investigate this observed increase in AMH when samples were incubated at RT. These included (1) addition of the detergent Tween-20 to assay buffer to disclose potential antibody-binding sites on the AMH molecule; (2) the removal of heterophilic antibodies from serum using PEG precipitation or heterophilic blocking tubes. None of these approaches affected AMH levels significantly (results not shown) Examination of the data presented here shows that, in some samples, AMH levels tend towards twice those expected, while results greater than that only occur in two outliers found in Figure 2. The AMH molecule is made up of two identical 72kDA monomers which are covalently bound (Wilson et al 1993; di Clemente et al 2010). During cytoplasmic transit, each monomer is cleaved to generate 110-kDa N-terminal and 25-kDa C-terminal homodimers, which remain associated in a noncovalent complex. The C-terminal homodimer binds to the receptor but, in contrast to other TGF-β superfamily members, AMH is thought to require the N-terminal domain to potentiate this binding to achieve full bioactivity of the C-terminal domain. After activation of the receptor, the N-terminal homodimer is released (Wilson et al 1993). One possible explanation for our findings is that the N-and C-terminal homodimers dissociate gradually under certain storage conditions and that either the two resulting N- and C-terminal components bind to the ELISA plate or a second binding site on the antigen is exposed by the dissociation, effectively doubling the concentration of AMH. It has been shown (di Clemente et al 2010) that no dissociation occurs once the complex is bound to immobilised AMH antibodies. The observation that, in some of our samples, there was no change after one week at RT might be explained by the supposition that in those samples AMH is already fully dissociated. A mixture of dissociated and complex forms in the same sample would, therefore, account for the observed recoveries between 100% and 200% in the experiments presented in this paper. Rapid sample processing and storage of the resulting serum in a different tube type at -80ºC might slow down this breakdown process. 60

61 The change in ionic strength or ph that occurs on dilution also seems to have the same effect in increasing apparent AMH levels and again may be due to dissociation or exposure of a second binding site. Our results contradict those reported by Kumar et al (2010) who showed that serum samples in the range of 36-93pmol/l of AMH when diluted in Gen II sample diluent showed linear results across the dynamic range of the assay with average recoveries on dilution close to 100%. This might be explained if Kumar s samples were already dissociated before dilution. Linearity is one of the cornerstones of assay validation and it is essential that a proportional response is obtained on dilution of sample, but our results do not seem to support this. These findings have significant clinical relevance, given the widespread use of AMH as the primary tool for assessment of ovarian reserve and as a marker for tailoring the dose of gonadotrophins in cycles of IVF/ICSI. As no guideline studies have been published using the new Gen II assay, some ART centres have adopted modified treatment cut off levels for ovarian stimulation programs based on the old DSL assay based cut off levels multiplied by a conversion factor of 1.4 (Nelson et al 2007; Nelson et al 2009; Wallace et al 2011). The data presented in this paper suggest that this approach could result in patients being allocated to the wrong ovarian reserve group. Poor performance of the Gen II assay in terms of sample-to-sample variability (up to 59%), could also lead to unreliable allocation to treatment protocols. It is a matter of some urgency, therefore, that any possible anomalies in the estimation of AMH using the Gen II assay be thoroughly investigated and that this work should be repeated in other centres. 61

62 References Anderson RA. Pretreatment serum anti-mullerian hormone predicts long-term ovarian function and bone mass after chemotherapy for early breast cancer. J Clin Endocrinol Metab 2011;96: Broer SL, Eijkemans MJ, Scheffer GJ, van Rooij IA, de Vet A, Themmen AP, Laven JS, de Jong FH, Te Velde ER, Fauser BC, Broekmans FJ. Anti-mullerian hormone predicts menopause: a long-term follow-up study in normoovulatory women. J Clin Endocrinol Metab 2011;96: Cook CL, Siow Y, Brenner AG, Fallat ME. Relationship between serum Mullerian inhibiting substance and other reproductive hormones in untreated women in PCOS and normal women. Fertil Steril 2002;77: di Clemente N, Jamin SP, Lugovskoy A, Carmillo P, Ehrenfels C, Picard JY, Whitty A, Josso N, Pepinsky RB, Cate RL. Processing of anti-mullerian hormone regulates receptor activation by a mechanism distinct from TGFbeta. Mol Endocrinol 2010;24: Freour T, Mirallie S, Bach-Ngohou K, Denis M, Barriere P, and Masson D. Measurement of serum anti-mullerian hormone by Beckman Coulter ELISA and DSL ELISA: comparison and relevance in assisted reproduction technology (ART). Clin Chim Acta 2007;375: Fanchin R, Taieb J, Mendez Lozano DH, Ducot B, Frydman R, Bouyer J. High reproducibility of serum anti-mullerian hormone measurements suggests a multi-staged follicular secretion and strengthens its role in the assessment of ovarian follicular status. Hum Reprod 2005; 20: Hehenkamp WJ, Looman CW, Themmen AP, de Jong FH, Te Velde ER, Broekmans FJ. Anti-Mullerian hormone levels in the spontaneous menstrual cycle do not show substantial fluctuation. J Clin Endocrinol Metab 2006;91: Hudson PL, Dougas I, Donahoe PK, Cate RL, Epstein J, Pepinsky RB, MacLaughlin DT. An immunoassay to detect human mullerian inhibiting substance in males and females during normal development. J Clin Endocrinol Metab 1990;70: Kumar A, Kalra B, Patel A, McDavid L, Roudebush WE. Development of a second generation anti-müllerian hormone (AMH) ELISA. J. Immunol Methods 2010;362: La Marca A, Stabile G, Artenisio AC, Volpe A. Serum anti-mullerian hormone throughout the human menstrual cycle. Hum Reprod 2006;21: La Marca A, Nelson SM, Sighinolfi G, Manno M, Baraldi E, Roli L, Xella S, Marsella T, Tagliasacchi D, D'Amico R, Volpe A. Anti-Mullerian hormonebased prediction model for a live birth in assisted reproduction. Reprod Biomed Online 2011; 22: Lahlou N, Chabbert-Buffet E, Gainer E, Roger M, Diphasic pattern of anti- Mullerian hormone (AMH) in ovulatory cycles as evidenced by means of ultra-sensitive assay: new insights into ovarian function. Fertil Steril 2006;86:S11. Lane AH, Lee MM, Fuller AF Jr, Kehas DJ, Donahoe PK, MacLaughlin DT. Diagnostic utility of Mullerian inhibiting substance determination in patients with primary and recurrent granulosa cell tumors. Gynecol Oncol 1999;73:

63 Lee JR, Kim SH, Jee BC, Suh CS, Kim KC, Moon SY. Antimullerian hormone as a predictor of controlled ovarian hyperstimulation outcome: comparison of two commercial immunoassay kits. Fertil Steril 2011;95: Lee MM, Donahoe PK, Hasegawa T, Silverman B, Crist GB, Best S, Hasegawa Y, Noto RA, Schoenfeld D, MacLaughlin DT. Mullerian inhibiting substance in humans: normal levels from infancy to adulthood. J Clin Endocrinol Metab 1996;81: Loh JS, Maheshwari A. Anti-Mullerian hormone--is it a crystal ball for predicting ovarian ageing? Hum Reprod 2011;26: Nardo LG, Gelbaya TA, Wilkinson H, Roberts SA, Yates A, Pemberton P, Laing I. Circulating basal anti-müllerian hormone levels as predictor of ovarian response in women undergoing ovarian stimulation for in vitro fertilization. Fertil Steril 2009; 92: Nelson SM, Yates RW, Fleming R. Serum anti-müllerian hormone and FSH: prediction of live birth and extremes of response in stimulated cycles-- implications for individualization of therapy. Hum Reprod 2007;22: Nelson SM, Yates RW, Lyall H, Jamieson M, Traynor I, Gaudoin M, Mitchell P, Ambrose P, Fleming R. Anti-Mullerian hormone-based approach to controlled ovarian stimulation for assisted conception. Hum Reprod 2009;24: Nelson SM, La Marca A. The journey from the old to the new AMH assay: how to avoid getting lost in the values. Reprod Biomed Online 2011;23: Nelson SM, Messow MC, Wallace AM, Fleming R, McConnachie A. Nomogram for the decline in serum antimullerian hormone: a population study of 9,601 infertility patients. Fertil Steril 2011;95: Preissner CM MD, Gada RP, Grebe SK. Validation of a second generation assay for anti-mullerian hormone. Clin Chem 2010;56:A54. Rustamov O, Pemberton PW, Roberts SA, Smith A, Yates AP, Patchava SD, Nardo LG. The reproducibility of serum anti-müllerian hormone in subfertile women: within and between patient variability. Fertil Steril 2011;95: Taieb J, Coussieu C, Guibourdenche J, Picard JY, and di Clemente N. Two immunoassays for antimullerian hormone measurement: analytical and clinical performances. Ann Biol Clin (Paris) 2008;66: Themmen AP. Anti-Mullerian hormone: its role in follicular growth initiation and survival and as an ovarian reserve marker. J Natl Cancer Inst Monogr 2005;(34): Tsepelidis S, Devreker F, Demeestere I, Flahaut A, Gervy C, Englert Y. Stable serum levels of anti-mullerian hormone during the menstrual cycle: a prospective study in normo-ovulatory women. Hum Reprod 2007;22: van Disseldorp J, Faddy MJ, Themmen AP, de Jong FH, Peeters PH, van der Schouw YT, Broekmans FJ. Relationship of serum antimullerian hormoneconcentration to age at menopause. J Clin Endocrinol Metab 2008;93: van Disseldorp J, Lambalk CB, Kwee J, Looman CWN, Eijkemans MJC, Fauser BC, Broekmans FJ. Comparison of inter- and intra-cycle variability of anti-mullerian hormone and antral follicle counts. Hum Reprod 2010; 25:

64 Vigier B, Picard JY, Tran D, Legeai L, Josso N. Production of anti-mullerian hormone: another homology between Sertoli and granulosa cells. Endocrinology 1984;114: Wallace AM, Faye SA, Fleming R, Nelson SM. A multicentre evaluation of the new Beckman Coulter anti-musllerian hormone immunoassay (AMH Gen II). Ann Clin Biochem 2011;48: Wilson CA, Di Clemente N, Ehrenfels C, Pepinsky RB, Josso N,Vigier B, Cate RL. Müllerian inhibiting substance requires its N-terminal domain for maintenance of biological activity, a novel finding within the transforming growth-factor-beta superfamily. Mol Endocrinol 1993;7: Wunder DM, Yared M, Kretschmer R, Birkhauser MH. Statistically significant changes of anti-mullerian hormone and inhibin levels during the physiologic menstrualcycle in reproductive age women. Fertil Steril 2008;89: Yates AP, Rustamov O, Roberts SA, Lim HY, Pemberton PW, Smith A, Nardo LG. Anti-Mullerian hormone-tailored stimulation protocols improve outcomes whilst reducing adverse effects and costs of IVF. Hum Reprod 2011;26: Zhao J, Ng SY, Rivnay B, Leader BS. Serum Anti-Mullerian hormone (AMH) measurement inter-assay agreement and temperature stability. Fertil Steril 2007; 88:S17. 64

65 AMH GEN II ASSAY: A VALIDATION STUDY OF OBSERVED VARIABILITY BETWEEN REPEATED AMH MEASUREMENTS Oybek Rustamov, Richard Russell, Cheryl Fitzgerald, Stephen Troup, Stephen A. Roberts

66 Title AMH Gen II assay: A validation study of observed variability between repeated AMH measurements. Authors Oybek Rustamov 1, Richard Russell 2, Cheryl Fitzgerald 1, Stephen Troup 2, Stephen A. Roberts 3 Institutions 1 Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospitals NHS Foundation Trust, Manchester, M13 9WL, UK; 2 Hewitt Fertility Centre, Liverpool Women s NHS Foundation Trust Hospital, Crown Street, Liverpool, L8 7SS; 3 Centre for Biostatistics, Institute of Population Health, University of Manchester, Manchester, M13 9PL, UK. Word count: 1782 Conflict of interest Authors have nothing to disclose. Acknowledgment The authors would like to thank the Biomedical Andrology Laboratory team at the Hewitt Fertility Centre for their assistance. 66

67 Declaration of authors roles: OR coordinated the study, conducted the statistical analysis and prepared first draft of the manuscript. RR extracted data, prepared the dataset, assisted in preparation of first draft of manuscript. CF, ST and SR involved in study design, oversaw statistical analysis, contributed to the discussion and preparation of the final version of the manuscript. 67

68 ABSTRACT Objective To study the within patient sample-to-sample variability of AMH levels using the Gen II assay reproduced in an independent population and laboratory? Design: Retrospective cohort analysis. SettingTertiary referral IVF Unit in the United Kingdom Patients: Women being investigated for sub-fertility. Interventions Retrospective measurements were obtained from women who had AMH measurements using Gen II assay during routine investigation for infertility at a tertiary referral unit during a 1-year period. The patients who had repeated AMH measurements were identified and within-patient coefficient of variation (CV) calculated using a mixed effects model with quadratic adjustment for age. Main Outcome Measures The within-patient coefficient of variation (CV) calculated using a random effects model with quadratic adjustment for age. Results There was in total of 76 samples from 38 women with repeated AMH measurements during the study period. The within-patient sample-to-sample variation (CV) was found to be 62%. Conclusions The study has confirmed that even when samples are processed promptly and strictly in accordance with the manufacturer's instructions, substantial variability exists between repeated samples. Thus caution is recommended in the use of these newer assays to guide treatment decisions. Further work is required to understand the underlying cause of this variability. Key Words Anti-Müllerian hormone; Müllerian inhibitory substance; AMH; AMH Gen II ELISA; AMH ELISA; sample variability. 68

69 INTRODUCTION Anti-Müllerian hormone is a dimeric glycoprotein that is produced by the granulosa cells of pre-antral and early antral follicles, and has been found to be the primary regulator of oocyte recruitment and folliculogenesis (Durlinger et al 1999, Durlinger et al 2001). Strong correlation between AMH levels and primordial follicle count (Hansen et al 2011), and hence a reflection of ovarian response has promised a valuable tool in the reproductive specialists armory. The development of commercially available AMH immunoassay assay kits has heralded the widespread introduction and routine usage of AMH assessment in the clinical setting. Several studies have demonstrated that AMH serves as a good predictor of ovarian response to gonadotrophin stimulation during IVF treatment (van Rooij et al 2002, Nelson et al 2007, Nardo et al 2009). AMH testing has also been shown to identify patients at risk of excessive ovarian response and ovarian hyperstimulation syndrome (Yates et al 2011) with consequent reduction in per cycle treatment costs by adopting an antagonist approach during controlled ovarian stimulation. Sensitivity and specificity of AMH in detecting extremes of response has been shown to be comparable to antral follicle count without the apparent technical limitations of the latter (Broer et al 2009, Broer et al 2011). It is stated that the sample-to-sample variation of AMH concentration in individual women is small and therefore, a single AMH measurement has been recommended as standard practice (La Marca et al 2006, Hehenkamp et al 2006). However recent studies, based on data from a single centre, recently published in Human Reproduction, found that larger variability between repeated samples exists, which is particularly profound when currently available second generation AMH assay (AMH Gen II ELISA, Beckman Coulter, Inc., Brea, CA, USA) is used (Rustamov et al 2012a, Rustamov et al 2012b, Rustamov et al 2011). The trial team had 2 objectives; firstly to assess whether the controversial findings from the above study (Rustamov et al 2012a) were reproducible when performed in the data based on the samples from a different laboratory with differing populations. If our study reached similar conclusions, concerns regarding the AMH Gen II assay and / or manufacturers recommendations on handling and sampling processes would be validated. Alternatively, if non- 69

70 similar findings were reported, the laboratory performance in the initial study ought to be questioned. Secondly and more importantly if the repeat samples are found to be within acceptable parameters then the current clinical standard of a single random AMH measurement in patients is appropriate. If the results of repeated samples are significantly different, following adjustment for age, it would suggest that AMH measurement is not a true estimation of the patient s ovarian reserve. In view of clinical and research implications of these findings, we undertook to replicate the variability study in a second fertility centre. The authors wish to note that Beckman Coulter recently issued a worldwide STOP SHIP order on all AMH Gen II Elisa assay kits until further notice due to manufacturing and quality issues. MATERIALS AND METHODS Population Women had serum AMH measurements using Gen II AMH assay from 15 April 2011 to 25 May 2012 for investigation of infertility at the "Hewitt Fertility Centre" in the Liverpool Women's NHS Foundation Trust Hospital, tertiary referral unit, were identified using the Biochemistry Laboratory AMH samples database and all women within age range of years were included in the study. The main reasons for repeating the samples were a) obtaining upto-date assessment of ovarian reserve, b) patient request and c) for formulation of a treatment strategy prior to repeat IVF cycles. Institutional Review Board approval was granted by the Audit Department; Liverpool Women s NHS Foundation Trust Hospital. Assay procedure Samples were transported immediately to the in-house laboratory of Liverpool Women s Hospital for the processing and analysis. The serum was separated within 8 hours from venipuncture and frozen at -50C until analyzed 70

71 in batches. The sample preparation and assay methodology strictly followed the manufacturer's guidelines. The AMH analysis of laboratory is regularly monitored by external quality assessment scheme (UKNEQAS) and performance has been satisfactory. The samples were analyzed using enzymatically amplified two-site immunoassay (AMH Gen II ELISA, Beckman Coulter, Inc., Brea, CA, USA). The intra-assay CV was 5.21% and inter-assay CV (n=9) was 2.76% (low controls) and 6.57% (high controls). The working range of the assay was 150pmol/L and the minimum detection limit was 0.57pmol/L. The main difference in the assay preparation in this study is that the samples were processed within 8 hours, whilst the samples in the previous study were processed within 2 hours (Rustamov 2012a). Importantly, the kit insert of Gen II AMH assay does not state any maximum duration of storage of unprocessed samples or any constraints on the transportation of unprocessed samples. Therefore, there appears to be considerable variation in practice of sample processing between clinics, which ranges from processing samples immediately to shipping unfrozen whole samples to long distances. Statistical analysis The dataset was obtained from the Biomedical Andrology Laboratory of the hospital and anonymised by one of the researchers (RR). Data management and analysis of the anonymised data followed the same procedures as the previous study (13) and were performed using Stata 12 Statistical Package (StataCorp, Texas, USA). Approval for data management, analysis and publication was obtained from the Research and Development Department of Liverpool Women s Hospital. Between and within-subject sample-to-sample coefficient of variability (CV) as well as the intra correlation coefficient (ICC) was estimated using a mixed effects model in log (AMH) with quadratic adjustment for age. AMH levels of the samples that fell below minimum detection limit of the assay (<0.57 pmol/l) were arbitrarily assigned a value of 0.31 pmol/l in line with the previous analysis (Rustamov et al 2012a). 71

72 RESULTS During the study period in total of 1719 women had AMH measurements using Gen II assay. Thirty-eight women had repeated AMH measurements; with a total number of 76 repeat samples (Figure 1). The median age of the women was 31.8 (IQR: ). The median AMH level was 5.2pmol/L (IQR: ). The median interval between samples was 93 days (IQR: ) with range of days. Age-adjusted regression analysis of samples of these women showed that within-patient sample-to-sample coefficient of variation (CV) of AMH measurements was 62%, while betweenpatient CV was 125%. An age adjusted intra-correlation coefficient was Figure 1. The repeated AMH measurements by date, lines join the repeats from the same patients (AMH in pmol/l). 72

73 DISCUSSION A number of studies have recently been published that have expressed concerns regarding the stability and reproducibility of AMH results. Whilst technical issues regarding reproducibility between assays were known, more recently the reproducibility of results regarding the current Gen II assay has raised significant concern (Rustamov et al 2012a, Rustamov et al 2012b, Rustamov et al 2011). Proponents of the assay have proposed that poor sample handling and preparation are responsible for these observed concerns (Nelson et al 2013). Several studies have observed the stability of samples at room temperature. Kumar et al. (Kumar et al 2010) observed a 4% variation in results after 7 days storage compared with those samples analysed immediately. These results were consistent with studies by Fleming and Nelson who also reported no change in AMH concentration over a period of several days (Fleming et al 2012). However Rustamov et. al., reported a measured AMH increase of 58% in samples stored at room temperature over a seven day period (Rustamov et al 2012a). Similar concerns were raised regarding the appropriate freezing process, whilst samples frozen at -20C demonstrated variation in results of between 6 and 22% (Durlinger et al 1999, Rustamov et al 2012a), freezing at -80C obviated a significant variation in assay results (Al- Qahtani et al 2005, Rustamov et al 2012a). Several studies initially reported good linearity of dilution (Kumar et al 2012, Preissner et al 2010, Fleming et al 2012) which was contradicted by reports that demonstrated poor linearity in dilution when fresh samples were utilized (Rustamov et al 2012a). This study suggested a tendency of AMH results to double with dilution. More recently Beckman Coulter issued a warning on their Gen II AMH ELISA kits that the dilution of sample may give an erroneous result, confirming non linearity of dilution (King Dave 2012). A number of studies have looked at the variability of AMH in repeated samples, without account to the menstrual cycle, utilizing different assays. Dorgan et. al. in analyzing DSL samples frozen for prolonged periods demonstrated a variability of 31% (ICC 0.78; 95% CI ) between two samples, with a median-sample interval of one year (Dorgan et al 2012). Rustamov et. al. presented a larger series of 186 infertile patients with a median between-sample interval of 2.6 months and a CV of 28% in DSL samples 73

74 (ICC 0.91: 95% CI (Rustamov et al 2011). In a follow-up study utilizing the Gen II assay in a group of 84 infertile patients, the coefficient variation of repeated results was 59% (ICC of 0.84, 95% CI ), a substantial increase in the observed variability of the studies reporting for the DSL assay (Rustamov et al 2012a). The most recent study to cast doubt on current practice suggested that repeated measurement of AMH, using Gen II assay, resulted in a within-subject variability of 80% (CV) (Hadlow et al 2012). As a result 7 out of 12 women were subsequently reclassified according to their originally predicted ovarian response. Our study outlined above involving 76 samples from 38 infertile patients demonstrated a within-patient sample-tosample coefficient of variation (CV) of AMH measurements was 62%. Overall these results suggest that there is significant within patient variability that may be more pronounced in the Gen II assay. Whilst biological variation has been demonstrated to play a part within this, the appreciative effects of sample handling, storage and freezing play a significant part in the results, and it may be that the Gen II assays may be more susceptible to these changes. This study has confirmed that there is significant within-patient sample-to-sample variability in AMH measurements when the Gen II AMH assay is used, which is not confined to a single population or laboratory. It is important to note that the samples reported by both Rustamov et. al., 2012, and this study were processed and analyzed strictly according to manufacturer s recommendations in their respective local laboratories without external transportation (Rustamov et al 2012a). Therefore it seems reasonable to suggest that AMH results from other centers and laboratories are likely to display similar significant sampling variability. Reproducibility of AMH measurements is of paramount importance given that a single random AMH measurement is used for triaging patients unsuitable for proceeding with IVF/ICSI and determining the dose of gonadotrophins for ovarian stimulation for those patients who proceed with treatment. Similarly other clinical applications of AMH, such as an assessment of the effect of chemotherapy to fertility and follow up of women with history of granulosa cell tumors, also rely on accurate measurement of circulating hormone levels. The present work confirms the high between-sample withinpatient variability. The recent warning from Beckman Coulter utilizing their Gen II ELISA assay kits may give an erroneous result with dilution of samples, further questions the stability of the assay (King David 2012). Subsequently, 74

75 the manufacturer recalled the assay kits due to issues with the instability of samples and introduced modified protocol for preparation of Gen II assay samples. Given there can be a substantial difference between two samples from the same patient, the use of such measurements for clinical decision-making should be questioned and caution is advised. 75

76 References Al-Qahtani A, Muttukrishna S, Appasamy M, Johns J, Cranfield M, Visser JA, Themmen AP and Groome NP. Development of a sensitive enzyme immunoassay for anti-mullerian hormone and the evaluation of potential clinical applications in males and females. Clin Endoc 2005; 63: Broer SL, Dolleman M, Opmeer BC, Fauser BC, Mol BW, Broekmans FJM. AMH and AFC as predictors of excessive response in controlled ovarian hyperstimulation: a meta-analysis. Hum Reprod Update 2011;17:46-54 Broer SL, Mol BWJ, Hendricks D, Broekmans FJM. The role of antimullerian hormone in prediction of outcome after IVF: comparison with the antral follicle count. Fertil Steril 2009;91: Dorgan JF, Spittle CS, Egleston BL, Shaw CM, Kahle LL and Brinton LA. Assay reproducibility and within-person variation of Mullerian inhibiting substance. Fertil Steril 2010;94: Durlinger AL, Gruijters MJ, Kramer P, Karels B, Kumar TR, Matzuk MM, Rose UM, de Jong FH, Uilenbroek JT, Grootegoed JA and Themmen AP. Anti-Mullerian hormone attenuates the effects of FSH on follicle development in the mouse ovary. Endocrinology 2001;142: Durlinger AL, Kramer P, Karels B, de Jong FH, Uilenbroek JT, Grootegoed JA and Themmen AP. Control of primordial follicle recruitment by anti- Mullerian hormone in the mouse ovary. Endocrinology 1999;140: Fleming R and Nelson SM. Reproducibility of AMH. Hum Reprod 2012;27: Hadlow Narelle, Longhurst Katherine, McClements Allison, Natalwala Jay, Brown Suzanne J. and Matson Phillip L. Variation in antimüllerian hormone concentration during the menstrual cycle may change the clinical classification of the ovarian response. (Article in press) Fertil Steril 2012 Hansen KL, Hodnett GM, Knowlton N, Craig LB. Correlation of ovarian reserve tests with histologically determined primordial follicle number. Fertil Steril 2011;95:170-5 Hehenkamp WJ, Looman CW, Themmen AP, de Jong FH, Te Velde ER, Broekmans FJ. Anti-Mullerian hormone levels in the spontaneous menstrual cycle do not show substantial fluctuation. J Clin Endocrinol Metab 2006;91: King Dave, URGENT FIELD SAFETY NOTICE- FSN 20434, AMH Gen II ELISA (REF A79765), Beckman Coulter United Kingdom Limited, November 27, 2012 Kumar A, Kalra B, Patel A, McDavid L and Roudebush WE. Development of a second generation anti-müllerian hormone (AMH) ELISA. J Immunol Methods 2010;362: La Marca A, Stabile G, Artenisio AC, Volpe A. Serum anti-mullerian hormone throughout the human menstrual cycle. Hum Reprod 2006;21: Nardo LG, Gelbaya TA, Wilkinson H, Roberts SA, Yates A, Pemberton P and Laing I. Circulating basal anti-müllerian hormone levels as predictor of ovarian response in women undergoing ovarian stimulation for in vitro fertilization. Fertil Steril 2009; 92:

77 Nelson S. Biomarkers of ovarian response: current and future applications. Fertil and Steril 2013;99: Nelson SM, Yates RW and Fleming R. Serum anti-müllerian hormone and FSH: prediction of live birth and extremes of response in stimulated cycles-- implications for individualization of therapy. Hum Reprod 2007;22: Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, and Philip W. Pemberton. Anti-Müllerian hormone: poor assay reproducibility in a large cohort of subjects suggests sample instability. Hum. Reprod. 2012a;27: Oybek Rustamov; Alexander Smith; Stephen A. Roberts; Allen P. Yates; Cheryl Fitzgerald; Monica Krishnan; Luciano G. Nardo; Philip W. Pemberton Reply: Reproducibility of AMH. Hum. Reprod. 2012b:27: Rustamov O, Pemberton PW, Roberts SA, Smith A, Yates AP, Patchava SD, Nardo LG. The reproducibility of serum anti-müllerian hormone in subfertile women: within and between patient variability. Fertil Steril 2011;95: Preissner CM, Morbeck DE, Gada RP and Grebe SK. Validation of a second generation assay for anti-mullerian hormone. Clin Chem 2010;56:A54. Sunkara SK, Rittenberg V, Raine-Fenning N, Bhattacharya S, Zamora J, Coomarasamy A. Association between the number of eggs and live birth in IVF treatment: an analysis of treatment cycles. Hum Reprod 2011;26: van Rooij IA, Broekmans FJ, te Velde ER, Fauser BC, Bancsi LF, de Jong FH and Themmen AP. Serum anti-mullerian hormone levels: a novel measure of ovarian reserve. Hum Reprod 2002;17: Yates AP, Rustamov O, Roberts SA, Lim HY, Pemberton PW, Smith A, Nardo LG. Anti-Mullerian hormone-tailored stimulation protocols improve outcomes whilst reducing adverse patient variability. Fertil Steril 2011;95:

78 THE MEASUREMENT OF ANTI-MÜLLERIAN HORMONE: A CRITICAL APPRAISAL Oybek Rustamov, Alexander Smith, Stephen A. Roberts, Allen P. Yates, Cheryl Fitzgerald, Monica Krishnan, Luciano G. Nardo, Philip W. Pemberton The Journal of Clinical Endocrinology & Metabolism 2014 Mar; 99(3):

79 Title The measurement of Anti-Müllerian hormone: a critical appraisal. Authors Oybek Rustamov a, Alexander Smith b, Stephen A. Roberts c, Allen P. Yates b, Cheryl Fitzgerald a, Monica Krishnan d, Luciano G. Nardo e, Philip W. Pemberton b Institutions a Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH, UK; b Department of Clinical Biochemistry, Central Manchester University Hospitals NHS Foundation Trust, Manchester M13 9WL, UK; c Centre for Biostatistics, Institute of Population Health, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 9PL, UK; d Manchester Royal Infirmary, Central Manchester University Hospitals NHS Foundation Trust, Manchester M13 9WL, UK;, e Reproductive Medicine and Gynaecology Unit, GyneHealth, Manchester M3 4DN, UK Key terms Anti-Müllerian hormone; AMH; Active MIS/AMH ELISA, Diagnostic Systems Laboratories; AMH/MIS ELISA, Immunotech; AMH Gen II assay, Beckman Coulter. Word Count: 3947 (intro general summary, text only (no headings) Corresponding author & reprint requests Dr. Oybek Rustamov, Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH Grants or fellowships: No funding was sought for this study. Disclosure summary: There were no potential conflicts of interest. 79

80 Declaration of authors roles The idea was developed during discussion between O.R., C.F and S.A.R. O.R. conducted the initial appraisal of the studies, prepared and revised the manuscript. S.A.R. and C.F. contributed to the discussion and interpretation of the studies and oversaw the revision of the manuscript. P.W.P., A.Y. M.K. and A.S. reviewed the data extraction and interpretation, contributed to the discussion of the studies and revision of the manuscript. L.G.N. contributed to the discussion of the studies and revision of the manuscript. 80

81 ABSTRACT Context Measurement of AMH is perceived as reliable, but the literature reveals discrepancies in reported within-subject variability and between-assay conversion factors. Recent studies suggest that AMH may be prone to preanalytical instability. We therefore examined the published evidence on the performance of current and historic AMH assays in terms of the assessment of sample stability, within-patient variability and comparability of the assay methods. Evidence Acquisition Studies (manuscripts or abstracts) measuring AMH, published between and in peer-reviewed journals, using appropriate PubMed/Medline searches. Evidence Synthesis AMH levels in specimens left at room temperature for varying periods, increased by 20% in one study and almost 60% in another, depending on duration and the AMH assay used. Even at -20 C, increased AMH concentrations were observed. An increase over expected values, of 20-30% or 57% respectively was observed following two-fold dilution in two linearity-ofdilution studies, but not in others. Several studies investigating within-cycle variability of AMH reported conflicting results, although most studies suggest variability of AMH within the menstrual cycle appears to be small. However between-sample variability without regard to menstrual cycle as well as withinsample variation appears to be higher using the Gen II AMH assay than with previous assays, a fact now conceded by the kit manufacturer. Studies comparing first generation AMH assays with each other and with the Gen II assay reported widely varying differences. Conclusions: AMH may exhibit assay-specific pre-analytical instability. Robust protocols for the development and validation of commercial AMH assays are required. 81

82 INTORDUCTION In the female, AMH, produced by granulosa cells of pre-antral and early antral ovarian follicles, regulates oocyte recruitment and folliculogenesis (1, 2). It can assess ovarian reserve (3-5) and guide gonadotrophin stimulation in assisted reproduction technology (ART) (6). AMH is also used as a granulosa cell tumour marker, a marker of ovarian reserve post-chemotherapy (7, 8) and to predict age at menopause (9,10). AMH immunoassays, first developed by Hudson et al in 1990 (11), were introduced commercially by Diagnostic Systems Laboratories (DSL) and Immunotech (IOT). These assays were integrated into a second-generation AMH assay, GenII (12) by Beckman-Coulter but recent work suggests that this new assay exhibits clinically important, within-patient, sample variability (13-15). Beckman Coulter have recently confirmed this with a field safety notice (FSN ), they cite, without showing evidence for, complement interference as the problem. True AMH variability comprises both biological and analytical components (Figure 1), and given the varying antibody specificity and sensitivity of different AMH assays, then logically different kits will respond to these components to varying degrees. This review considers the published literature on AMH measurement using previous and currently available assays. Potential sources of variation and their contribution to observed AMH variability were identified. Review structure This review has been divided into logical subgroups. We first address the stability of AMH at different storage temperatures, then the effects of freeze/thaw cycles and finally AMH variability in dilution studies. Secondly, the within-person variability of AMH measurement is considered, encompassing intra- and inter-menstrual cycle variability and repeat sample variability in general. The final section covers AMH method comparisons, comparing older methods to each other and to the newer, now prevalent GenII method, finishing with data on published guidance ranges concerning the use of AMH in ART. A general summary concludes the paper. 82

83 Systematic review The terms anti-müllerian hormone, "AMH", "Müllerian Inhibiting Substance" and "MIS" were used to search the PubMed/Medline MeSH database, between 1 st January 1990 and 1 st August 2013, for publications in English commenting on AMH sample stability, biological and sample-tosample variability or assay method comparison in human clinical or healthy volunteer samples. Titles and/or abstracts of 1653 articles were screened to yield the following eligible publications; ten stability studies, 17 intra/intercycle variability studies and 14 assay method comparability studies. Sample stability Recent work has established that the GenII-measured AMH is susceptible to significant preanalytical variability (13, 14), not previously acknowledged which may have influenced results in previous studies with this assay. Stability of unfrozen samples Five studies examined AMH stability in samples stored either at room or fridge temperature (Table 1) (13, 16-19). Al-Qahtani et al (16), assessing the precursor of the DSL ELISA, reported that immunoreactivity survived the storage of samples unfrozen for 4 days but did not record storage temperature or sample numbers. Evaluating the GenII assay, Kumar et al (18) stored 10 samples at 2-8 C, for up to a week and found an average 4% variation compared to samples analysed immediately. However their specimens, originally reported as fresh, appear to have been kept cool and transported overnight. Fleming & Nelson (19) reported no significant change in the GenII-assayed AMH from 51 samples stored at 4 C. Methodological information was limited but interrogation of their data by Rustamov et al (14) suggested that AMH levels rose by an average of 27% after 7 days storage. Zhao et al (17) reported a difference of less than 20% between DSL-assayed AMH in 7 serum samples kept at 22 C for 48 hours when compared to aliquots from the same samples frozen immediately at -20 C. Rustamov et al (13) measured AMH (GenII) daily in 48 serum samples at room temperature for 7 days and observed an average 58% increase (from 0 to >200%), whilst others (20) reported a 31% mean rise in GenII-assayed AMH in whole blood 83

84 after 90hrs at 20 o C whereas serum AMH was virtually unchanged after prolonged storage at 20 o C. Sample stability at -20 o or -80 o C and the effects of freeze/thaw Rey et al (21) reported a significant increase in AMH (in-house assay) in samples stored at -20 C for a few weeks, attributing this to proteolysis which could be stabilised with protease inhibitor (see discussion below). Kumar et al (18) saw 6% variation between GenII-assayed AMH levels from 10 fresh and 10 frozen samples whilst Rustamov et al (13) observed a 22% increase in AMH (GenII) on re-analysis of 8 serum samples after 5 days storage at -20 C. These authors saw no AMH increase in serum stored at -80 C for the same period. Linearity of dilution Six studies examined linearity of dilution on observed AMH concentrations. Long et al (22) recovered between 84 and 105% of the expected AMH concentration (IOT, n=3). AMH dilution curves, parallel to the standard curve, were reported by others (16),,Kumar et al (18) (n=4) and Preissner et al (23) ) (n=7) reported GenII-assayed AMH recoveries from 95% to 104% and 96% respectively. Sample handling information was limited in some of these studies (16, 23). Fleming & Nelson (19) (GenII, n=10) reported variances of 8% using assay diluent and 5% using AMH-free serum following 2-fold dilution; however, interrogation of their data reveals an apparent dilutional AMH increase of 20-30% in samples stored prior to dilution and analysis. Rustamov et al (13) (GenII, n=9) in freshly collected serum, observed an average 57% increase in apparent AMH concentration following two-fold dilution, but with considerable variation. Discussion: Sample stability Sample stability can be a major analytical problem and detailed examination suggests that previous evidence stating that commercially measured AMH is stable in storage and exhibits linearity of dilution (12, 16, 18, 19), is weak or conflicting. No study looking at room temperature storage on IOT-assayed AMH was found, and only one using DSL-assayed AMH, which showed an increase 84

85 of less than 20% during storage (17). Studies using the GenII assay to investigate the effect of storage on AMH variability at room temperature, in the fridge and at C reach differing conclusions, ranging from stable to an average 58% increase in measured levels. It is important to note here that sample preparation and storage prior to these experiments was different and could account for the observed discrepancies. The most stable storage temperature for AMH in serum appears to be -80 C (13, 16). Linearity of dilution studies were also conflicting (13, 18, 19, 23); those reporting good linearity used samples transported or stored prior to baseline analysis, whereas dilution of fresh samples showed poor linearity. In late 2012, Beckman Coulter accepted that the GenII assay did not exhibit linear dilution and issued a warning on kits that samples should not be diluted. They now suggest that with the newly introduced pre-mixing protocol, dilution should not be a problem. This review highlights the fact that assumptions about AMH stability in serum were based on a limited number of small studies, often providing limited methodological detail (impairing detailed assessment and comparison with other studies) using samples stored or transported under unreported conditions. Furthermore, conclusions derived using one particular AMH assay have been applied to other commercial assays without independent validation. The available data suggests that dilution of samples and/or storage or transport in sub-optimal conditions can lead to an increase in apparent AMH concentration. The conditions under which this occurs in each particular AMH assay are not yet clear and more work is required to understand the underlying mechanisms. Two alternative hypotheses have been proposed: firstly, that AMH may undergo proteolytic change as postulated by Rey et al (21) or conformational change as proposed by Rustamov et al (13,14) during storage, resulting in stabilisation of the molecule in a more immunoreactive form; secondly, Beckman have postulated the presence of an interferent (complement), which degrades on storage (Beckman Coulter field safety notice FSN ). A recent case report found that a falsely high AMH level was corrected by the use of heterophylic antibody blocking tubes (24), but this does not explain elevation of AMH on storage (13). Whatever the mechanism responsible, two solutions are available; either inhibit the process completely or force it to completion prior to analysis. 85

86 Rustamov et al (13) and Han et al (15) both suggest pre-dilution of samples to force the process, a protocol now adopted by Beckman Coulter in their revised GenII assay protocol. Any solution must be robustly and independently validated both experimentally and clinically prior to introduction in clinical practice. Fresh optimal ranges for interpretation of AMH levels in ART will be needed and the validity of studies carried out using unreported storage conditions may have to be re-evaluated. Within-person variability The biological components of AMH variability, such as circadian and inter/intra-cycle variability have been extensively studied (Table 2 & Supplementary table 1). Circadian variation Bungum et al (25) evaluated circadian variability, measuring AMH (IOT) two hourly over 24hrs, within day 2 6 of the menstrual cycle in younger (20-30 years) and older (35-45 years) women. Within-individual CVs of 23% (range %) in the younger group and 68% (range %) in the older group were observed. Variability within the menstrual cycle Cook et al (26) observed significant (12%) variation in mean AMH (inhouse) levels in 20 healthy women throughout different phases of the menstrual cycle. Intra-cycle variability of IOT-assayed AMH was reported in three publications (27-29). In two, sequential samples were stored at -20 C until analysis (27, 28). Streuli et al (29) did not report on storage. La Marca et al (27) saw no difference in mean follicular phase AMH levels (days 2, 4 and 6) in untreated, spontaneous menstrual cycles from 24 women. This group went on to report a small, insignificant change (14%) in within-group AMH variability throughout the whole menstrual cycle in 12 healthy women. However, this analysis does not appear to allow for correlations within samepatient samples. Streuli et al (29) studied intra-cycle variation of AMH throughout two menstrual cycles in 10 healthy women and also reported no significant changes (<5%). 86

87 The DSL assay was used in eight studies assessing intra-cycle variability (30-37). Four studied sample storage at -20 C (30,32,34,37) and two studied samples storage at -80 C (33,35). No sample storage data was given in two publications (31, 36). Hehenkamp et al (30) assessed within-subject variation of AMH in 44 healthy women throughout two consecutive menstrual cycles and reported an intra-cycle variation of 17.4%. Lahlou et al (31) reported a diphasic pattern of AMH, with a significant decrease in levels during the LH surge from 10 women at various cycle phases. Tsepelidis et al (32) reported a mean intra-cycle coefficient of variation of 14%, comparing group mean AMH levels in 20 women during various stages of the menstrual cycle. Wunder et al (33) reported an intra-cycle variability of around 30% in 36 healthy women, sampling on alternate days. They saw a marked fall around ovulation, which might have been missed with less frequent sampling intervals, as in other studies. Sowers et al (35) studied within-cycle variability in 20 healthy women but did not compute an overall estimate, instead they selected subgroups of low and high AMH and reported significant within-cycle variability for women with high AMH but not those with low AMH - an analysis that has been questioned (38, 39). Robertson et al (36) subgrouped mean AMH levels in 61 women, observing that AMH levels were stable in women of reproductive age and ovulatory women in late reproductive age, whilst AMH in other women in late reproductive age was much more variable. Using the data from Hehenkamp et al (30), van Disseldorp et al (34) calculated intra-class correlation (ICC) and reported a within-cycle variability of 13%, although this was not clearly defined. Using the same data, Overbeek et al (37) analyzed the absolute intra-individual difference in younger ( 38 years) and older (>38 years) women. This study concluded that the AMH concentration was more variable in younger women ( g/l) compared to older women ( g/l) during the menstrual cycle (P=0.001), thus a single AMH measurement may be unreliable. A recent study using the GenII assay reported 20% intra-cycle variability in AMH measurements in women (n=12) with regular ovulatory cycles (40). All the reports considered have findings consistent with a modest true systematic variability of 10-20% in the level of AMH in circulation during the menstrual cycle. Whilst there have been suggestions that this variability may differ between subgroups of women, these 87

88 have been based on post-hoc subgroup analyses and there is no convincing evidence for such subgroups (38). Variability between menstrual cycles Three studies (Supplementary table 1) evaluated AMH variability in samples taken during the early follicular phase of consecutive menstrual cycles (10,29,41) and three studies have reported on the variability of AMH in repeat samples from the same patient taken with no regard to the menstrual cycle (13,42,43). One study employed an in-house assay (41), one study used the IOT assay (29), three studies used the DSL assay (10, 42, 43) and one study (13) used the GenII assay. In four infertile women, Fanchin et al (41) assessed the early follicular phase AMH (in-house) variability across three consecutive menstrual cycles; they concluded that inter-sample AMH variability was characterised by an ICC of 0.89 (95% CI: ). Streuli et al (29) calculated a between-sample coefficient of variation of 28.5% in AMH (IOT) in 10 healthy women. In 77 infertile women, van Disseldorp et al (10) found an inter-cycle AMH (DSL) variability of 11%. In summary, these studies suggest that the overall inter-cycle variability of AMH ranges from 11% (DSL) to 28% (IOT): this figure will include both biological and measurement-related variability. Variability between repeat samples Variability between repeat samples without regard to menstrual cycle phase was examined in three studies (Supplementary table 1). In a group of 20 women, using samples frozen for prolonged periods, Dorgan et al (42) demonstrated a variability of 31% (ICC 0.78; 95% CI, ) between two samples, with a median between-sample interval of one year. In a larger series of 186 infertile women, Rustamov et al (43) (DSL) found a CV of 28% between repeated samples, with a median between-sample interval of 2.6 months (ICC 0.91; 95% CI, ). Rustamov et al (13) found that the coefficient of variation of repeated GenII-assayed AMH in a group of 84 infertile women was 59% (ICC of 0.84; 95% CI, ), substantially higher than that reported using the DSL assay. Similarly, a recent study by Hadlow et al (40) found a within-subject GenII-assayed AMH variability of 80%. As a 88

89 result, 5 of the 12 women studied crossed clinical cut-off levels following repeated measurements. Discussion: Within-patient variability Evidence suggests that repeated measurement of AMH can result in clinically important variability, particularly when using the GenII assay. This questions the assumption that a single AMH measurement is acceptable in guiding individual treatment strategies in ART. The observed concentration of any analyte measured in a blood (serum) sample is a function of its true concentration and the influence of a number of other factors (Figure 1). Studies examining the variability of AMH by repeated measurement of the hormone will therefore reflect both true biological variation and measurement-related variability introduced by sample handling and/or processing. Thus within-sample inter-assay variability, used as an indicator of assay performance may not reflect true measurement-related variability between samples since it does not take into account the contribution from pre-analytical variability. Measurement-related between-sample variability can be established in part using blood samples taken simultaneously (to avoid biological variability) from a group of subjects, although even this does not reflect the full variability in sample processing and storage inherent in real clinical measurement. Since AMH is only produced by steadily growing ovarian follicles, it is plausible to predict a small true biological variability in serum, reflected in the modest 1-20% variability found within the menstrual cycle. In contrast, it appears that the magnitude of measurement-related variability of AMH is more significant: a) within-sample inter-assay variation can be as high as 13%; b) different assays display substantially different variability and c) AMH appears to be unstable under certain conditions of sample handling and storage (Table 1). Consequently, any modest variation in true biological AMH concentration may be overshadowed by a larger, measurement-related variability and careful experimental designs are required to characterise such differences. In general the reported variability in published studies should be regarded as a measure of total sample-to-sample variability i.e. the sum of biological and measurementrelated variability (Figure 1). 89

90 In repeat samples, the available evidence confirms that there is a significant level of within-patient variability between measurements which is assay-dependent, greater than the estimates of within cycle variability and therefore likely to be predominantly measurement-related. Evidence from several sources suggests that the effects of sample handling, storage and freezing differ between commercial assays and that the newer GenII assay may be more susceptible to these changes under clinical conditions. When it has been established that the modified protocol for the GenII assay can produce reproducible results independent of storage conditions, then it will be necessary to re-examine intra and inter cycle variability of AMH. Assay method comparability AMH assay comparisons have either used same sample aliquots or used population-based data with repeat samples. Study population characteristics, sample handling, inter-method conversion formulae and results from these comparisons are summarised in Table 3. AMH levels were almost universally compared using a laboratory based within-sample design. The Rustamov et al study (13) was population-based, comparing AMH results in two different samples from the same patient at different time points using 2 different assays. IOT vs. DSL Table 3 summarises 8 large studies (17, 29, 30, 44-48) that compared the DSL and IOT AMH assays. They demonstrate strikingly different conversion factors; from five-fold higher with the IOT assay to assay equivalence. Most studies carried out both analyses at the same time to avoid analytical variation (Figure 1). However this does mean that samples were batched and frozen at - 18 C to -80 C prior to analysis which, as already outlined may influence preanalytical variability and contribute to the observed discrepancies in conversion factors. IOT vs. GenII Three studies have compared the IOT and Gen II assays (Table 3). Kumar (18) reported that both assays gave identical AMH concentrations. However, Li et al (48) found that the IOT assay produced AMH values 38% 90

91 lower than the Gen II assay whilst Pigny et al (49) found levels that were 2-fold lower. DSL vs. GenII Four studies analysed same-sample aliquots using the DSL and GenII assays, either simultaneously or sequentially (33, 48, 50, 51). Only Li et al (48) gave details of sample handling (Table 3). All four studies found that AMH values that were 35 50% lower using the DSL compared to the GenII assay. Rustamov et al (13) carried out a between-sample comparison of the assays, measuring AMH in fresh or briefly stored clinical samples from the same women at different times, with values adjusted for patient age (Table 3). In contrast to within-sample comparisons, this study found that the DSL assay gave results, on average, 21% higher than with the GenII assay. Whilst this comparison is open to other bias, it does reflect the full range of variability present in clinical samples and avoids issues associated with longer term sample storage. Discussion: Assay method comparability It is critical for across-method comparison of clinical studies that reliable conversion factors for AMH are established. In-house assays aside, three commercially available AMH ELISAs have been widely available (IOT, DSL and GenII) and the literature demonstrates considerable diversity in reported conversion factors between first-generation assays (DSL vs. IOT), and between first and second-generation immunoassays (DSL/IOT vs. GenII). Although most studies appear to follow manufacturers protocols, detailed methodological information is sometimes lacking. The assessment of within-sample difference between the two assays involved thawing of a single sample and simultaneous analysis of two aliquots with each assay. Both aliquots experience the same pre-analytical sample-handling and processing conditions, therefore the results should be reproducible, provided the AMH samples are stable during the post-thaw analytical stage and the study populations are comparable. However, this review has identified significant discrepancies between studies, perhaps due to either significant instability of the sample or significant variation in assay performance. Studies comparing AMH levels measured using different assays in populations during routine 91

92 clinical use have also come to differing conclusions (13, 51). Given the study designs that workers have used to try to ensure that samples are comparable, the finding of significant discrepancies in the observed conversion factors between assays is consistent with the proposal that AMH is subject to instability during the pre-analytical stage of sample handling. This, coupled with any differential sensitivity and specificity between these commercial assays, could give rise to the observed results i.e. some assays are more sensitive than others to pre analytical effects. AMH guidance in ART AMH guidance ranges to assess ovarian reserve (52) or subsequent response to treatment (53, 54) have been published. The Doctors Laboratory, using the DSL assay advised the following ranges for ovarian reserve (< 0.57pmol/L-undetectable; pmol/l-very low; pmol/l-low; pmol/l-satisfactory; pmol/L-optimal; >48.5pmol/L-very high), ranges that supposedly increased by 40% on changing to the GenII assay (51). More recently other authors have attempted to correlate AMH levels with subsequent birth rates. Brodin et al (53), using the DSL assay, observed that higher birth rates were seen in women with an AMH level > 21 pmol/l and low birth rates were seen in women who had AMH levels < 1.43 pmol/l. In the UK, the National Institute for Health and Care Excellence (NICE) have recently issued guidance on AMH levels in the assessment of ovarian reserve in the new clinical guideline on Fertility (54). They advise that an AMH level of 5.4 pmol/l would indicate a low response to subsequent treatment and an AMH 25.0 pmol/l indicates a possible high response. Although not specifically stated, interrogation of the guideline suggests that these levels have been obtained using the DSL assay, which is no longer available in the UK. As discussed above, the initial study of comparability between the DSL and GenII assays reported that GenII generated values 40% higher compared to the DSL assay; clinics were therefore recommended to increase their treatment guidance ranges accordingly (51). However, a more recent study, using fresh samples, found that the original GenII assay may actually give values which are 20-30% lower; suggesting that following the above recommendation may lead to allocation of patients to inappropriate treatment groups (13). The apparent disparity in assay comparison studies implies that 92

93 AMH reference ranges and guidance ranges for IVF treatment which have been established using one assay cannot be reliably used with another assay method without full, independent, validation. Similarly, caution is required when comparing the outcomes of research studies using different AMH assay methods. General Summary Recent publications have suggested that GenII-assayed AMH is susceptible to pre-analytical change leading to significant variability in determined AMH concentration, an observation now accepted by the kit manufacturer. However, this review suggests that all AMH assays may display a differential response to pre-analytical proteolysis, conformational changes of the AMH dimer or presence of interfering substances. The existence of appreciable sample-to-sample variability and substantial discrepancies in between-assay conversion factors, suggests that sample instability may have been an issue with previous AMH assays but appears to be more pronounced with the currently available GenII immunoassay. The observed discrepancies may be explicable in terms of changes in AMH or assay performance that are dependent on sample handling, transport and storage conditions, factors under-reported in the literature. We strongly recommend that future studies on AMH should explicitly report on how samples are collected, processed and stored. If it can be clearly demonstrated that the new GenII protocol drives this process to completion in all samples ensuring stability then a reexamination of reference and guidance ranges for AMH interpretation will be necessary. There is a clear need for an international reference standard for AMH and for robust independent evaluation of commercial assays in routine clinical samples with well-defined sample handling and processing protocols. These issues of sample instability and lack of reliable inter-assay comparability data should be taken into account in the interpretation of available research evidence and the application of AMH measurement in clinical practice. 93

94 References 1. Durlinger AL, Kramer P, Karels B, de Jong FH, Uilenbroek JT, Grootegoed JA, Themmen AP. Control of primordial follicle recruitment by anti-mullerian hormone in the mouse ovary. Endocrinology 1999;140: Durlinger AL, Gruijters MJ, Kramer P, Karels B, Kumar TR, Matzuk MM, Rose UM, de Jong FH, Uilenbroek JT, Grootegoed JA, Themmen AP. Anti- Mullerian hormone attenuates the effects of FSH on follicle development in the mouse ovary. Endocrinology 2001;142: van Rooij IA, Broekmans FJ, te Velde ER, Fauser BC, Bancsi LF, de Jong FH, Themmen AP. Serum anti-mullerian hormone levels: a novel measure of ovarian reserve. Hum Reprod 2002;17: Nelson SM, Yates RW, Fleming R. Serum anti-müllerian hormone and FSH: prediction of live birth and extremes of response in stimulated cycles-- implications for individualization of therapy. Hum Reprod 2007;22: Nardo LG, Gelbaya TA, Wilkinson H, Roberts SA, Yates A, Pemberton P, Laing I. Circulating basal anti-müllerian hormone levels as predictor of ovarian response in women undergoing ovarian stimulation for in vitro fertilization. Fertil Steril 2009;92: Yates AP, Rustamov O, Roberts SA, Lim HYN, Pemberton PW, Smith A, Nardo LG. Anti-Mullerian hormone-tailored stimulation protocols improve outcomes whilst reducing adverse effects and costs of IVF. Hum Reprod 2011;26: Lane AH, Lee MM, Fuller AF Jr, Kehas DJ, Donahoe PK, MacLaughlin DT. Diagnostic utility of Mullerian inhibiting substance determination in patients with primary and recurrent granulosa cell tumors. Gynecol Oncol 1999;73: Anderson RA. Pretreatment serum anti-mullerian hormone predicts longterm ovarian function and bone mass after chemotherapy for early breast cancer. J Clin Endocrinol Metab 2011;96: Broer SL, Eijkemans MJ, Scheffer GJ, van Rooij IA, de Vet A, Themmen AP, Laven JS, de Jong FH, te Velde ER, Fauser BC, Broekmans FJ. Antimullerian hormone predicts menopause: a long-term follow-up study in normoovulatory women. J Clin Endocrinol Metab 2011;96: van Disseldorp J, Lambalk CB, Kwee J, Looman CW, Eijkemans MJ, Fauser BC, Broekmans FJ. Comparison of inter- and intra-cycle variability of anti-müllerian hormone and antral follicle counts. Hum Reprod 2010;25: Hudson PL, Dougas I, Donahoe PK, Cate RL, Epstein J, Pepinsky RB, MacLaughlin DT. An immunoassay to detect human mullerian inhibiting substance in males and females during normal development. J Clin Endocrinol Metab 1990;70:

95 12. Nelson SM, La Marca A. The journey from the old to the new AMH assay: how to avoid getting lost in the values. Reprod Biomed Online 2011;23: Rustamov O, Smith A, Roberts SA, Yates AP, Fitzgerald C, Krishnan M, Nardo LG, Pemberton PW. Anti-Mullerian hormone: poor assay reproducibility in a large cohort of subjects suggests sample instability. Hum Reprod 2012; 27: Rustamov O, Smith A, Roberts SA, Yates AP, Fitzgerald C, Krishnan M, Nardo LG, Pemberton PW. Reply: reproducibility of AMH. Hum Reprod 2012;27: Han X, McShane M, Sahertian R, White C, Ledger W. Pre-mixing samples with assay buffer is an essential pre-requisite for reproducible Antimullerian Hormone (AMH) measurement using the Beckman Coulter Gen II assay (Gen II). Hum Reprod 2013;28 (suppl 1):i76-i78 (abstract). 16. Al-Qahtani A, Muttukrishna S, Appasamy M, Johns J, Cranfield M, Visser JA, Themmen AP, Groome NP. Development of a sensitive enzyme immunoassay for anti-mullerian hormone and the evaluation of potential clinical applications in males and females. Clin Endoc 2005;63: Zhao J, Ng SY, Rivnay B, Leader BS. Serum Anti-Mullerian hormone (AMH) measurement inter-assay agreement and temperature stability. Fertil Steril 2007;88:S17 (abstract). 18. Kumar A, Kalra B, Patel A, McDavid L, Roudebush WE. Development of a second generation anti-müllerian hormone (AMH) ELISA. J Immunol Methods 2010;362: Fleming R, Nelson SM. Reproducibility of AMH. Hum Reprod 2012;27: Fleming R, Fairbairn C, Blaney C, Lucas D, Gaudoin M. Stability of AMH measurement in blood and avoidance of proteolytic changes. Reprod Biomed Online 2013;26: Rey R, Lordereau-Richard I, Carel JC, Barbet P, Cate RL, Roger M, Chaussain JL, Josso N. Anti-Mullerian hormone and testosterone serum levels are inversely related during normal and precocious pubertal development. J Clin Endocrinol Metab 1993;77: Long WQ, Ranchin V, Pautier P, Belville C, Denizot P, Cailla H, Lhomme C, Picard JY, Bidart JM, Rey R. Detection of minimal levels of serum anti- Mullerian hormone during follow-up of patients with ovarian granulosa cell tumor by means of a highly sensitive enzyme-linked immunosorbent assay. J Clin Endocrinol Metab 2000;85: Preissner CM, Morbeck DE, Gada RP, Grebe SK. Validation of a second generation assay for anti-mullerian hormone. Clin Chem 2010;56:A54 (abstract). 24. Cappy H, Pigny P, Leroy-Billiard M, Dewailly D, Catteau Jonard S. Falsely elevated serum antimüllerian hormone level in a context of heterophilic 95

96 interference. Fertil Steril 2013;99: Bungum L, Jacobsson AK, Rosén F, Becker C, Yding Andersen C, Güner N, Giwercman A. Circadian variation in concentration of anti-mullerian hormone in regularly menstruating females: relation to age, gonadotrophin and sex steroid levels. Hum Reprod 2011;26: Cook CL, Siow Y, Taylor S, Fallat ME. Serum müllerian-inhibiting substance levels during normal menstrual cycles. Fertil Steril 2000;73: La Marca A, Malmusi S, Giulini S, Tamaro LF, Orvieto R, Levratti P, Volpe A. Anti-Müllerian hormone plasma levels in spontaneous menstrual cycle and during treatment with FSH to induce ovulation. Hum Reprod 2004;19: La Marca A, Stabile G, Carduccio Artenisio A, Volpe A. Serum anti- Mullerian hormone throughout the human menstrual cycle. Hum Reprod 2006;21: Streuli I, Fraisse T, Chapron C, Bijaoui G, Bischof P, de Ziegler D. Clinical uses of anti-mullerian hormone assays: pitfalls and promises. Fertil Steril 2009;91: Hehenkamp WJ, Looman CW, Themmen AP, de Jong FH, te Velde ER, Broekmans FJ. Anti-Mullerian hormone levels in the spontaneous menstrual cycle do not show substantial fluctuation. J Clin Endocrinol Metab 2006;91: Lahlou N, Chabbert-Buffet N, Gainer E, Roger M, Bouchard P. Diphasic pattern of anti-mullerian hormone (AMH) in ovulatory cycles as evidenced by means of ultra-sensitive assay: new insights into ovarian function. Fertil Steril 2006;86:S11 (abstract). 32. Tsepelidis S, Devreker F, Demeestere I, Flahaut A, Gervy C, Englert Y. Stable serum levels of anti-mullerian hormone during the menstrual cycle: a prospective study in normo-ovulatory women. Hum Reprod 2007;22: Wunder DM, Bersinger NA, Yared M, Kretschmer R, Birkhauser MH. Statistically significant changes of anti-mullerian hormone and inhibin levels during the physiologic menstrual cycle in reproductive age women. Fertil Steril 2008;89: van Disseldorp J, Faddy MJ, Themmen AP, de Jong FH, Peeters PH, van der Schouw YT, Broekmans FJ. Relationship of serum antimullerian hormone concentration to age at menopause. J Clin Endocrinol Metab 2008;93: Sowers M, McConnell D, Gast K, Zheng H, Nan B, McCarthy JD, Randolph JF. Anti-Müllerian hormone and inhibin B variability during normal menstrual cycles. Fertil Steril 2010; 94: Robertson DM, Hale GE, Fraser IS, Hughes CL, Burger HG. Changes in serum antimüllerian hormone levels across the ovulatory menstrual cycle in late reproductive age. Menopause 2011;18: Overbeek A, Broekmans FJ, Hehenkamp WJ, Wijdeveld ME, van 96

97 Disseldorp J, van Dulmen-den Broeder E, Lambalk CB. Intra-cycle fluctuations of anti-mullerian hormone in normal women with a regular cycle: a re-analysis. Reprod Biomed Online 2012;24: Roberts SA. Variability in anti-mullerian hormone levels: a comment on Sowers et al, Anti-Mullerian hormone and inhibin B variability during normal menstrual cycles. Fertil Steril 2010;94:e Sowers M, McConnell D, Gast K, Zheng H, Nan B, McCarthy JD, Randolph JF. Reply of the authors: Variability in anti-müllerian hormone levels: a comment on Sowers et al. Anti-Müllerian hormone and inhibin B variability during normal menstrual cycles. Fertil Steril 2010;94:e Hadlow N, Longhurst K, McClements A, Natalwala J, Brown SJ, Matson PL. Variation in antimüllerian hormone concentration during the menstrual cycle may change the clinical classification of the ovarian response. Fertil Steril 2013;99: Fanchin R, Taieb J, Lozano DH, Ducot B, Frydman R, Bouyer J. High reproducibility of serum anti-müllerian hormone measurements suggests a multi-staged follicular secretion and strengthens its role in the assessment of ovarian follicular status. Hum Reprod 2005;20: Dorgan JF, Spittle CS, Egleston BL, Shaw CM, Kahle LL, Brinton LA. Assay reproducibility and within-person variation of Mullerian inhibiting substance. Fertil Steril 2010;94: Rustamov O, Pemberton PW, Roberts SA, Smith A, Yates AP, Patchava SD, Nardo LG. The reproducibility of serum anti-müllerian hormone in subfertile women: within and between patient variability. Fertil Steril 2011;95: Freour T, Mirallie S, Bach-Ngohou K, Denis M, Barriere P, Masson D. Measurement of serum anti-mullerian hormone by Beckman Coulter ELISA and DSL ELISA: comparison and relevance in assisted reproduction technology (ART). Clin Chim Acta 2007;375: Bersinger NA, Wunder D, Birkhäuser MH, Guibourdenche J. Measurement of anti-mullerian hormone by Beckman Coulter ELISA and DSL ELISA in assisted reproduction: differences between serum and follicular fluid. Clin Chim Acta 2007;384: Taieb J, Belville C, Coussieu C, Guibourdenche J, Picard JY, di Clemente N. Two immunoassays for antimullerian hormone measurement: analytical and clinical performances. Ann Biol Clin (Paris) 2008;66: Lee JR, Kim SH, Jee BC, Suh CS, Kim KC, Moon SY. Antimullerian hormone as a predictor of controlled ovarian hyperstimulation outcome: comparison of two commercial immunoassay kits. Fertil Steril 2011;95: Li HW, Ng EH, Wong BP, Anderson RA, Ho PC, Yeung WS. Correlation between three assay systems for anti-mullerian hormone (AMH) 97

98 determination. J Assist Reprod Genet 2012;29: Pigny P, Dassonneville A, Catteau-Jonard S, Decanter C, Dewailly D. Comparative analysis of two-widely used immunoassays for the measurement of serum AMH in women. Hum Reprod 2013; 28:i (abstract). 50. Gada R, Hughes P, Amols M, Amols M, Preissner C, Morbeck D, Coddington C. Validation and comparison of AMH serum levels using the original active MIS/AMH ELISA to the new active AMH Gen II ELISA. Fertil Steril 2011;95:S23 (abstract). 51. Wallace AM, Faye SA, Fleming R, Nelson SM. A multicentre evaluation of the new Beckman Coulter anti-mullerian hormone immunoassay (AMH Gen II). Ann Clin Biochem 2011;48: The Doctors Laboratory. Lab Report newsletter Winter 2007/2008 AMH. 53. Brodin T, Hadziosmanovic N, Berglund L, Olovsson M, Holte J. Antimullerian hormone levels are strongly associated with live birth rates after assisted reproduction. J Clin Endocrinol Metab 2013;98(3): National Institute for Care and Health Excellence. NICE clinical guideline CG156 Fertility. 98

99 Figure 1. Biological and analytical variability of AMH 99

100 Table 1. AMH assay validation: effect of sample storage conditions, fresh/thaw cycles and linearity of dilution Study Assay Method Result Rey et al (21) in-house effect of. Long-term storage at -20 C (n=4) AMH levels in archival samples were 230% higher than original value Long et al (22) IOT linearity up to 16-fold dilution (n=3) observed AMH was % of expected AMH Al-Qahtani et al (16) Zhao et al (17) in-house DSL a. freeze/thaw stability; storage unfrozen for 4 days b. linearity up to 32-fold dilution (n=6) serum frozen immediately at -20 C compared to aliquots stored at 4 C or 22 C for up to 2 days (n=7) a. immuno-reactivity survived both multiple freeze-thaw cycles and storage unfrozen for 4 days b. dilution curves were parallel to the standard curve AMH levels increased by 1% at 4 C and 9% at 22 C after 2 days compared to sample frozen immediately Kumar et al (18) Gen II a. serum or plasma stored at 2-8 C or -20 C for up to 7 days (n = 20) b. serum or plasma underwent up to three freeze/thaw cycles (n=20) c.. linearity of dilution (n=4) a. AMH levels were stable for up to 7 days at 2-8 C or -20 C b. AMH increased by 15% in serum and by 5% in plasma after 3 cycles c. linear results obtained across the dynamic range of the assay Preissner et al (23) Gen II linearity of dilution (n=7) average agreement with expected result was 97% Rustamov et al (13) Gen II a. stability at RT for up to 7 days (n=48). b. storage for 5 days at -20 C or -80 C compared to fresh sample (n=8) c. linearity on 2-fold dilution (n=9) a. AMH levels increased by an average of 58% over 7 days b. AMH levels increased by 23% at -20 C but were unchanged at -80 C. c. AMH levels were on average 157% higher than expected Fleming & Nelson (19) Gen II a. serum stored at 4 C for 7 days (n=48) b. linearity of dilution (n=10) a. AMH levels increased by an average of 27% b. AMH was 28% & 33% higher on 2-fold & 4-fold dilution resp. Fleming et al (20) Gen II a. whole blood stored for up to 90 hours at 4 C (n=32) or 20 C (n=21) b. serum stored for 5 days at 20 C and 2 days at 4 C (n=13) a. AMH increased by 11% at 4 C and by 31% at 20 C b. only 1% increase in AMH compared to original value Han et al (15) Gen II serum from non-pregnant (n=13) or early pregnant (n=7) women stored at RT, -20 C or -80 C for up to 7 days In non-pregnant women, AMH increased by 26% after 7 days at RT, but was unchanged at -20 C or -80 C. In pregnant women, AMH increased by 50% at RT and 27% at -80 C after 48 hours. 100

101 Study Cook et al (26) La Marca et al (27) La Marca et al (28) Lahlou et al (31) Hehenkam p et al (30) van Disseldorp et al (10) Overbeek et al (37) Subjects healthy age regular cycle (n=20) healthy age regular cycle (n=24) healthy age18-24 regular cycle (n=12) a. cycles b. day sampled Assay a. 1 cycle inhouse b. day 2/3, LH surge, LH surge +7 d a. follicular IOT phase b. alternate days placebotreated (n=12) healthy fertile regular cycle (n=44) data from Hehenkamp et al (30) data from Hehenkamp et al (30) a. 1 cycle b. alternate days day 0 = day of LH surge a. 1 cycle b. every 3 days a. 2 cycles b. AMH measured at each of 7 cycle phases IOT Table 2. Intra-cycle variability of AMH a. storage b. freeze/thaw c. measurement Result Authors Conclusion a. -80 C b. once c. inter-assay variation a. -20 C b. once a. -20 C b. once day 3: AMH = ng/ml mid cycle: AMH = ng/mL mid luteal: AMH = ng/mL AMH did not change from days 2 to 6 in spontaneous cycles but decreased progressively in FSH-treated cycles low mean AMH = ng/mL (day 14) high mean AMH = ng/mL (day 12) DSL NR 7 days pre LH surge: AMH = pmol/L; peak : AMH = pmol/L; 10 days post LH surge: AMH = pmol/L DSL a. -20 C a. sine pattern fitted to AMH data was not significant (p=0.40) b.72% repeat AMH values fell within the same quintile, 28% in adjacent quintile. Intra-cycle within-subject variation of AMH was only 13% compared to 31-34% for AFC (dependent on follicle size) Fluctuations were larger than 0.5µg/L in one cycle in significantly (p = 0.001) more women in the younger group than the older one Fluctuations significant (p<0.008). AMH may have a regulatory role in folliculogenesis AMH levels did not change significantly during follicular phase of the menstrual cycle. AMH levels did not change significantly throughout menstrual cycle AMH levels exhibited a diphasic pattern with levels declining significantly (p<0.05) during the LH surge. AMH shows no consistent fluctuation through the cycle compared to FSH, LH & E2 AMH displays less intra-cycle variability than AFC. AMH can fluctuate substantially in younger women during menstrual cycle so a single measurement could be unreliable. 101

102 Tsepelidis et al (32) Wunder et al (33) Streuli et al (29) healthy age regular cycles (n=20) healthy age regular cycles (n=36) healthy mean age=24.1 regular cycles (n=10) a. 1 cycle b. days 3, 7, 10-16, 18, 21 & 25 a. 1 cycle b. alternate days a. 1 cycle b. before (LH -10,-5,-2,-1) and after LH surge (LH +1,+2,+10) DSL a. -20 C b. once Within-cycle differences not significant (p=0.408) DSL a. -80 C AMH levels were statistically higher in the late follicular phase than at the time of ovulation (p= 0.019) or in the early luteal phases (p<0.0001) IOT a. -18 C AMH levels were statistically lower during the early luteal phase compared to early follicular phase (p=0.016) and late luteal phase levels (p=0.02). AMH levels do not vary during the menstrual cycle. AMH levels vary significantly during the menstrual cycle In clinical practice, AMH can be measured at any time during the menstrual cycle Sowers et al (35) Robertson et al (36) Hadlow et al (40) healthy age regular cycles (n=20) a. age regular cycles (n=43) b. age: variable cycles (n=18) age: regular cycles non-pcos (n=12) a. 1 cycle b. daily a. 1 cycle + initial stages of succeeding cycle b. three times weekly a. 1 cycle b. 5-9 samples per subject DSL a. -80 C b. once c. simultaneous Higher AMH levels with significant variation between days 2-7 in the younger ovary. Low AMH levels with little variation in the aging ovary. DSL NR No intracycle variation in AMH level was found in women in mid reproductive life or in 33% women with regular cycles in late reproductive age. In the remaining cycles, there was a significant (p<0.01) two-fold decrease in AMH in 11 cycles and a significant (p<0.01) 4.2- fold increase between the follicular & luteal phases Gen II a. -20 C within 4 hours of sampling b. once c. simultaneous 7/12 women could be reclassified depending on when AMH was measured during the cycle; 2/12 crossed cut-offs predicting hyperstimulation. AMH varies across the menstrual cycle in the younger ovary. When AMH levels are substantially reduced, they become less reliable markers of ovarian reserve AMH cycles varied during menstrual cycle and clinical classification of the ovarian response was altered 102

103 Table 3. Variability in AMH levels between menstrual cycles Study Fanchin et al (41) Subjects infertile age regular cycles (n=47) a. cycles b. day sampled Assay Storage Result Authors Conclusion a. 3 cycles -80 C b. day 3 in-house (Long et al 2000) AMH showed significantly higher reproducibility than inhibin B (p<0.03), E2 (p<0.0001), FSH (p<0.01) and early AFC (p<0.0001) AMH showed improved cycle-tocycle consistency compared to other markers of ovarian follicular status Streuli et al (29) van Disseldorp et al (10) Dorgan et al (42) Rustamov et al (36) healthy mean age = 24.1 regular cycles (n=10) infertile median age =33 PCOS excluded (n=77) blood donors age collected (n=20) infertile women age (n=186) a. 2 cycles b. before (LH -10,-5,-2,-1) and after LH surge (LH +1,+2,+10) a. average 3.73 cycles b. day 3 two samples collected during the same menstrual cycle phase at least 1yr apart random sampling median interval = 2.6 months IOT -18 C Inter-cycle variability of 28.5% DSL -80 C AMH showed a withinsubject variability of 11% compared to 27% for AFC DSL -70 C between-subject variance in AMH of 2.19 was large compared to the withinsubject variance of 0.31 DSL -70 C within-subject CV for AMH was 28% compared to 27% for FSH AMH fluctuations during the cycle were smaller than or equal to the variability between two cycles AMH demonstrated less individual inter-cycle variability than AFC AMH was relatively stable, over 1 year in pre-menopausal women AMH showed significant sampleto-sample variation Rustamov et al (13) infertile women age (n=87) random sampling median interval = 5.1 months Gen II -20 C within-subject CV for AMH was 59% AMH demonstrated a large sample-to-sample variation 103

104 Study Assays Subjects Table 4. Within-subject comparison between AMH methods Simultaneous Analysis Regression Summary Freour et al (44) DSL vs. IOT 69 infertile women age Yes IOT = 4.01 x DSL (µg/l) (Deming regression) DSL = 22% IOT (p<0.0001) Hehenkamp et al (30) DSL vs. IOT 82 healthy women NR DSL= x IOT DSL = 49.5% IOT Bersinger et al (45) a. DSL vs. IOT b. DSL vs. IOT a. 11 infertile women b. 55 infertile women a. yes b. no a. DSL= x IOT b. DSL= x IOT a. DSL = 18% IOT b. DSL= 33% IOT Zhao et al (17) DSL vs. IOT 38 donors NR IOT = 1.5 x DSL (ng/ml) DSL = 66% IOT Taieb et al (46) DSL vs. IOT 104 samples NR DSL = 1.04 x IOT DSL = 96% IOT Streuli et al (29) DSL vs. IOT 153 normal and infertile No IOT = 1.07 x DSL DSL = IOT Kumar et al (18) IOT vs. Gen II 60 female, 60 male volunteers NR IOT =1.0 Gen II IOT=Gen II Gada et al (50) DSL vs. Gen II 42 women NR NR DSL = 63% Gen II Preissner et al (23) DSL vs. Gen II 206 samples NR Gen II = 1.53 x DSL DSL = 66% Gen II Lee et al (47) DSL vs. IOT 172 infertile women Yes IOT = x DSL DSL = IOT Wallace et al (51) DSL vs. Gen II 271 women NR Gen II = 1.40 x DSL DSL = 71% Gen II Li et al (48) a. DSL vs. IOT b. DSL vs. Gen II c. IOT vs. Gen II 56 women with PCOS or sub-fertility Yes a. IOT = 0.97 x DSL b. Gen II = 1.33 x DSL c. Gen II = 1.38 x IOT a. DSL = IOT b. DSL = 67% Gen II c. IOT = 62% Gen II Rustamov et al (13) DSL vs. Gen II female IVF patients (n=330) median of 2yr between samples No NR DSL = 127% Gen II (age-adjusted) Pigny et al (49) IOT vs. Gen II 59 women: 32 controls, 27 with PCOS Yes NR IOT = 200% Gen II 104

105 Appendix I Flow-chart of the search for publications. Database search for sample stability, measurement variability and assay-method comparability was conducted simultaneously using the MeSH database of PubMed/Medline using the search terms of anti-müllerian hormone, "AMH", "Müllerian Inhibiting Substance" and "MIS" which identified n=1653 studies on AMH. The initial step of identification involved screening of articles by reading titles and/or abstracts. Further search involved identification of studies from the reference sections of the initially identified studies. Database Search n=1653 Sample Stability Measurment Variability Method comparability Screening Titles n=6 Screening Titles n=14 Screening Titles n=10 Further Search n=4 Further Search n=3 Further Search n=4 Total n=10 Total n=17 Total n=14 105

106 EXTRACTION, PREPARATION AND COLLATION OF DATASETS FOR THE ASSESSMENT OF THE ROLE OF THE MARKERS OF OVARIAN RESERVE IN FEMALE REPRODUCTION AND IVF TREATMENT Oybek Rustamov, Monica Krishnan, Cheryl Fitzgerald, Stephen A. Roberts Research Database 4 106

107 Title Extraction, preparation and collation of datasets for the assessment of the role of the markers of ovarian reserve in female reproduction and IVF treatment Authors Oybek Rustamov a, Monica Krishnan b, Cheryl Fitzgerald a, Stephen A. Roberts c Institutions a Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH, UK; b Manchester Royal Infirmary, Central Manchester University Hospitals NHS Foundation Trust, Manchester M13 9WL, UK; c Centre for Biostatistics, Institute of Population Health, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 9PL, UK; NHS Research Ethics Approval North West Research Ethics Committee, (10/H1015/22). Word count: 5088 Grants or fellowships No funding was sought for this study Acknowledgements The authors would like to thank colleagues Dr Greg Horne (Senior Clinical Embryologist), Ann Hinchliffe (Clinical Biochemistry Department) and Helen Shackleton (Information Operations Manager) for their help in obtaining datasets for the study. 107

108 Declaration of authors roles OR prepared the protocol, extracted data from electronic sources and hospital notes, prepared datasets and prepared all versions of the chapter. MK assisted in collection of data from hospital notes. SR and CF oversaw and supervised preparation the protocol, extraction of data preparation of datasets and reviewed the chapter. 108

109 CONTENTS I. PROTOCOL Introduction 110 Methods 111 Objectives.111 Inclusion Criteria 111 Datasets Demography dataset..112 Biochemistry dataset Surgery dataset 113 AMH dataset 113 IVF dataset 114 FET dataset Embryology dataset. 115 RH dataset. 115 AFC dataset Folliculogram dataset..117 Data management Data cleaning and coding Merging datasets Data security and storage..119 II. RESULTS Introduction. 120 Data extraction and management 121 Demography dataset Biochemistry dataset Surgery dataset 121 AMH dataset 122 IVF dataset 122 FET dataset Embryology dataset. 123 RH, AFC and Folliculogram datasets Merging Datasets 124 Conclusion

110 I. PROTOCOL INTRODUCTION The aim of the project is to create a series of reliable and validated datasets, which contain all relevant data on the ovarian reserve markers (AMH, AFC, FSH), ethnicity, BMI, reproductive history, causes of infertility, IVF treatment parameters for patients that meet inclusion criteria as described below. The datasets will be used for the subsequent research projects of the MD programme and future research studies on ovarian reserve. Most data can be obtained from following existing clinical electronic records; a) Patient Administration System (PAS), b) Biochemistry Department data management system, c) the hospital database for surgical procedures and d) AMH dataset and e) ACUBase IVF data management system. Following obtaining original datasets from the administrators of the data management systems in their original Excel format, the datasets will be converted into Stata format and prepared by a) checking and recoding spurious data, transforming the dates from string to numeric format which will be consistent across all datasets (Day Month Year) and stored in Stata format under following names: Demography, Biochemistry, AMH, Surgery, IVF FET, Embryology. Copies of original datasets will be kept in the password-protected and encrypted computer located in the Clinical Records Room of Reproductive Medicine Department, Central Manchester University Hospitals NHS Foundation Trust, which is maintained by IT department of the Trust (Figure 1). Data not available in electronic format will be collected from the hospital records of each patient by researchers, Dr Oybek Rustamov and Dr Monica Krishnan, and entered into following datasets: Reproductive history (RH), antral follicle count (AFC) and Folliculogram. The hospital notes of all included patients will be hand-searched. The datasets will be transferred to Stata and each step of data preparation will be recorded using Stata Do files and the files will be stored under the filenames of History, AFC, Folliculogram in Stata format. In order to ensure the robustness of the data and for the purpose of validation of the datasets, electronic scanned copies of all available reports of pelvic ultrasound assessments for AFC and folliculograms will be obtained and stored in the password-protected and 110

111 encrypted computer located in the Clinical Records Room of Reproductive Medicine Department. Ethics approval for collection of data has already been obtained (UK-NHS 10/H1015/22). The datasets will be merged and datasets for each research project with all available data nested with IVF cycles nested within patients will be created. METHODS Objectives The aim of the project is to build a robust database, which can reliably used for the following purposes: 1. To estimate the effect of ethnicity, BMI, endometriosis and the causes of infertility on ovarian reserve using cross sectional data (Chapter 5.1). 2. To estimate the effect of salpingectomy, ovarian cystectomy and unilateral salpingo-oopherectomy on ovarian reserve using cross sectional data (Chapter 5.2). 3. To determine the effect of age, AMH, AFC, causes of infertility and treatment interventions on oocyte yield (Chapter 6). 4. To explore the potential for optimization of AMH-tailored individualisation of ovarian stimulation using retrospective data (Chapter 6). Inclusion criteria In order to capture the populations for all three studies the database will have broad inclusion criteria. All women from 20 to 50 years of age referred to Reproductive Medicine Department of Central Manchester University Hospitals NHS Foundation Trust will be included if a) they were referred for management of infertility or fertility preservation and b) had AMH measurement during the period from 1 September 2008 till 16 November

112 Datasets PAS dataset The dataset contains information on the hospital number, surname, first name, date of birth and the ethnicity of all patients referred to Reproductive Medicine Department, CMFT (Table 1). The data are originally entered during registration of the patient for clinical care by administrative staff of Gynaecology and Reproductive Medicine Departments. The dataset will be obtained from the administrators of the Information Unit. The dataset will be obtained in Excel format and transferred into Stata 12, Data Management and Statistical Software. The date values (referral date and date of birth) will be converted into numeric variable using Date Month Year format (DMY). Ethnicity will be coded using numeric variables in alphabetical order as pre-specified in the Table 2a. Biochemistry dataset The dataset contains all blood test results, specimen numbers, the names of the tests and the date of sampling of women who had assays for follicle stimulating hormone (FSH), oestradiol (E2), luteinizing hormone (LH) and AMH during the study period (Table 1). Data entries were conducted by the clinical scientists, the technicians and the members of administrative team of the Biochemistry Department. The dataset will be obtained from an administrator of the database. The date of sampling and analyses will be converted to the numeric DMY format. The specimen number will be kept unaltered in the string variable format and used to link the tests that were taken in the same sample tube. The name of the test will be kept as described in the original format AMH FSH, LH, and Oest. In the original dataset, the samples sent from Reproductive Medicine Department are coded as IVF which will be kept unaltered and the remaining observations will be divided into Gynaecology Department, Non-IVF/Gynaecology and Unknown categories using the code of referred ward and the names of the consultants. The test results will be converted into numeric format and the results with minimum detection limit will be coded as 50% of the minimum detection limit as follows: AMH <0.61 = 0.31 pmol/l, FSH <0.5 = 0.25 mlu/ml, LH 112

113 <0.5 =0.25 mlu/ml, Oest <50 =0.25 pg/ml. The test results that are higher than the assay ranges will be set to 150% of the maximum range. Interpretation of serum FSH results in conjunction with serum oestradiol levels is important in establishing true early follicular phase hormone levels. The test results are believed to be inaccurate if serum oestradiol levels higher than 250pmol/L at the time of sampling and therefore a new variable for FSH results with only serum FSH observations that meet above criteria will be created and used subsequently. All ambiguous data will be checked using electronic pathology data management system, Clinical Work Station (CWS). Surgery dataset The electronic dataset will be obtained from Information Department in Excel format. The dataset created using clinical coding software and data entry conducted during patient treatment episodes by theatre nursing and medical staff. In order to evaluate effect of past reproductive surgery to ovarian reserve, all patients had ovarian cystectomy, drainage of ovarian cyst, salpingectomy, salpingo-oopherectomy during 1 January November 2011 at Central Manchester University Hospitals NHS Foundation Trust will be included in the dataset. The dataset contains following variables: hospital number, surname, first name, date of birth, date of operation, name of operation, laterality of operation and name of surgeon. The final dataset will be stored in Stata.dta format (Figure 1). The dataset will be used to validate data on reproductive surgery that was collected from hospital records in the RH dataset. AMH dataset The dataset contains the AMH results, the dates of sampling, the dates of analyses and the assay generation (DSL or Gen II) for all patients included in the study (Table 1). The dataset will be obtained from the senior clinical scientist Dr Philip Pemberton, Specialist Assay Laboratory, who is responsible for the data entry and updating of the dataset. There are two separate primary Excel based AMH data files: 1) DSL dataset and 2) Gen II dataset. The datasets will be transferred to Stata 12 software separately and following preparation of the datasets, which logged using Stata Do file, Stata versions of the data files will be stored under DSL 113

114 and Gen2 names. Then, the files will be combined by appending DSL to Gen2 in order to create a new combined AMH dataset. The date variables, the sample date, the assay date and the date of birth, will be converted into numeric DMY format. The samples sent from other NHS trusts and private clinics will be excluded from the dataset, alongside the records from male patients and the patients outside of the age range of years of age. The manufacturers of the assays suggest that haemolysed and partly haemolysed samples may provide inaccurate test readings. Therefore a new variable without these samples will be created and used in the analyses for all studies. All the ambiguous data will be checked and verified using duplicate datasets obtained from Biochemistry dataset and the hospital records of the patients. IVF dataset The IVF dataset will be downloaded from ACUBase Data management system in original Excel format and contains detailed information on causes of infertility, sperm parameters, treatment interventions, assessment of oocyte quantity and quality, assessment of embryo quantity and quality and the outcomes of treatment cycles, (Table 1).Data entry to ACUBase was performed by members of administrative, nursing, embryology and medical staff of the Reproductive Medicine Department at the point of care. This is only electronic data management system for ART cycles and used for monitoring of the clinical performance of the department by internal and external quality assessment agencies and regulators (e.g. HFEA, CQC). Therefore, the quality of data entry for the main indicators of the performance of IVF ICSI programs (the treatment procedures, the outcomes of the cycles and assessment of embryos) should be fairly accurate. Table 2b describes the coding of the treatment outcomes and the practitioners of ICSI, the ultrasound-guided oocyte retrieval (USOR) and the embryo transfer (ET) procedures. In addition to the main patient identifier (Hospital Number) this dataset contains in-built cycle identifier (IVF Reference Number) which will be used to link the original IVF cycles to corresponding Frozen Embryo Transfer (FET) cycles and the embryos originating from the index cycle using FET and Embryo datasets respectively. 114

115 FET dataset The dataset provides information on the quality and the quantity of transferred embryos, the date of embryo transfer and the outcome of the cycle in frozen embryo transfer cycles (Table 1). Primary data entry was performed by the members of the clinical embryology team during the treatment of patients and will be downloaded from ACUBase by Dr O. Rustamov. Together with IVF dataset it can be used to study cumulative live birth rate (LBR) of index cycles. The treatment outcomes as well as ICSI, USOR and ET practitioners will be converted to numeric variables using the codes which are shown in Table 2b. The dataset can be linked to the index fresh IVF cycles as well as to embryos of FET cycles using the IVF Reference number. Embryology dataset The dataset has comprehensive information on the quality and the quantity of embryos on each day of their culturing; including embryos that were cryopreserved and those that were discarded (Table 1). The dataset also includes patient identifiers (name, date of birth, IVF reference number) and the dates of embryo transfer. The primary data entry into this dataset was conducted by the members of clinical embryology team during the clinical episodes and will be downloaded from ACUBase by Dr O. Rustamov. The dataset can be linked to index fresh IVF cycle and FET cycles using IVF Reference numbers of corresponding datasets. RH dataset This dataset will be created and data entry will be conducted during the search of the hospital notes. Following identification of included patients using AMH dataset, Excel electronic data collection file will be created. The hospital notes of each patient will be searched for by systematically checking all filed hospital records in Clinical Records Room of Reproductive Medicine Department by the order of their hospital number. Further search for missing notes will be conducted by checking all hospital notes located in the offices of nurses, doctors and secretaries. Electronic hospital notes filed in Medisec Digital Dictation Database will be used for data extraction for the patients whose hospital notes were not located. 115

116 All available diagnosis will be recorded under the following columns; 1) female referral diagnosis, 2) male referral diagnosis, 3) female initial clinic diagnosis, 4) female final clinic diagnosis, 5) diagnosis prior 2 nd IVF cycle, 6) diagnosis prior 3 rd IVF cycle. Furthermore, other relevant information on pathology of reproductive system will be documented. For instance all possible iatrogenic causes of poor ovarian reserve (e.g. oophorectomy, ovarian cystectomy, salpingectomy, chemotherapy and radiotherapy) will be recorded. In order to establish the existence of polycystic ovary syndrome (PCOS), the history of oligomenorrhea, amenorrhea and diagnosis of polycystic ovaries (PCO) on pelvic ultrasound scan will be collected and used in conjunction with serum LH levels of Biochemistry dataset (Table 1). Male infertility will be defined as severe male factor if the sperm parameters were low enough to meet criteria (<0.5 mln/ml or retrograde ejaculation) for Multiple Ejaculation Resuspension and Centrifugation test (MERC) as part of investigation for infertility. A variable for patients diagnosed with azoospermia will be created and the diagnosis will be recorded. The patients diagnosed with male factor infertility but with the sperm parameters that did not reach criteria for MERC will be diagnosed with mild male factor infertility. Patients diagnosed with severe and/or stage IV and/or stage III endometriosis will be categorized as severe endometriosis, while patients diagnosed with mild or moderate endometriosis will be coded as mild endometriosis group. In diagnosing the tubal factor infertility, only patients with history of bilateral salpingectomy and the patients with evidence of bilateral tubal blockage on a laparoscopy and dye test will be diagnosed as severe tubal factor. The patients with history of unilateral salpingectomy, unilateral tubal block in laparoscopy and dye test or unilateral/bilateral tubal block on hysterosalpingogram will be categorized as mild tubal factor infertility. Diagnosis of polycystic ovarian syndrome (PCOS) will be based in Rotterdam criteria: existence of two of the following features; 1) oligo- or anovulation, 2) clinical and/or biochemical signs of hyperandrgoenism, 3) polycystic ovaries. Referral for fertility preservation will be defined as referral for consideration of obtaining oocytes or/and embryos and/or sperm prior to chemotherapy, radiotherapy or surgical management of a malignant disease. The length of infertility will be recorded as per proforma of initial consultation for the patients attended initial clinic appointment following introduction of serum AMH test, 1 September For patients 116

117 attended initial consultation prior to introduction of AMH test, the length of infertility will be documented as per the initial clinic proforma plus years till the patient s first AMH test. The patient s body mass index (BMI) documented at initial assessment will used for patients who had assessment after introduction of AMH test, 1 September 2008, whereas the most up to date BMI result is collected for the patients seen prior to this date. AFC dataset Data will be extracted from the hospital notes. The data on the assessment of AFC will be obtained from the pelvic ultrasound scan reports. The date of assessment, the AFC in each ovary, the name of sonographer will be recorded (Table 1). Furthermore, other relevant ultrasound findings such as, ovarian cyst, hydrosalpynx and submucous uterine fibroids will also be entered in the dataset. To permit data validation, scanned copies of ultrasound scan report of each AFC investigation will be stored in PDF format in the computer that located in the Clinical Notes Room. The department uses a stringent methodology for the assessment of AFC, which consist of counting of all antral follicles measuring 2-6mm in longitudinal and transverse cross sections of both ovaries using transvaginal ultrasound scanning at early follicular phase (Day 0-5) of the menstrual cycle. The ultrasound assessments are conducted by qualified sonographers, who use the same methodology for the measurement of AFC. However, it is well known that the counting of antral follicles may be prone to significant interoperator variability. Therefore, the name of sonographers will be recorded during primary data collection and coded (Table 2a) so that the estimates of within- and between-operator variability can be obtained if necessary. Folliculogram dataset Although most data on IVF/ICSI cycles are available in IVF dataset, certain important data on IVF treatment are recorded only in the hard copy IVF folliculograms. Consequently data on ultrasound follicle tracking, the reasons for changing the doses of stimulation drugs are only available in the folliculograms. Furthermore, the length of the coasting and the causes for cycle cancellation are usually recorded in both folliculograms and IVF dataset, which can be used to validate accuracy of IVF dataset. Therefore 117

118 these data will be collected using the folliculograms that filed in the hospital notes and the scanned copies of each folliculograms will be stored in the computer located Clinical Records Room for data validation purposes (Table 1). The number of follicles on Day 8 and Day 10 ultrasound scans will be recorded according to the size of the follicles: 10-16mm and 17mm. Numeric variables for the follicle numbers will be created and used for assessment of ovarian response in IVF cycles. Data management Data cleaning and coding All datasets will be obtained in Excel format and transferred in the original unaltered condition into Stata 12 data management and statistical package (Stata 12, StataCorp, Texas, USA) and all steps of the data cleaning and the coding will be recorded using Stata Do files to create audit trails of the data management process. Both original Excel and cleaned Stata versions of data files will be stored in computer that is located in Clinical Records Room at Reproductive Medicine Department. Uniformity of hospital numbers in all datasets will be achieved by converting a) leading lower case prefixes s to upper case S, b) dropping suffixes z and Z and c) dropping all leading zeros in the second part of the hospital number (e.g. s10/00235z = S10/235 ). The coding of the datasets is shown in the Table 2a and the Table 2b. All ambiguous data will be checked using electronic data management systems (e.g. CWS, Medisec) and hospital notes. Merging the datasets The datasets will be structured as such that the data files can be used independently or merged at a) patient or b) IVF cycle levels using the patient identifier, cycle identifier and date variables (Figure 1). This allows analysis of outcomes of both Fresh IVF cycles and study the cumulative outcomes of Fresh IVF and Frozen Embryo Transfer cycles originating form index IVF cycles. Each dataset will contain two main patient identifiers and patient number (Patient ID), which will be used for linking the datasets in a patient 118

119 level. At the initial stages of the data management the hospital numbers will be used as the main patient identifier. The accuracy of the hospital numbers in each dataset will be validated using PAS dataset by checking patient surname, first name and date of birth. Following methodology will be used to add study numbers into each dataset. First, all dataset will be merged in a wide format using the hospital numbers which creates Master Datasets for each of the research projects. Then an accuracy of the merger will be checked using DOB, surname and first name. Once the dataset is validated, several copies of the Patient ID variable will be created and distributed to each dataset. Finally the datasets will be separated and stored as independent datasets alongside Master Datasets for each research projects. IVF, FET and Embryology datasets contain cycle specific IVF reference numbers, which were allocated during the clinical episodes on ACUBase. Using IVF reference number, new ID variable (Cycle ID) will be created and allocated to all datasets using closest observation prior to the IVF cycle in Master Research Dataset. Consequently, by using cycle reference number, all patient and cycle related data can be linked in the IVF, FET cycle and embryo level. Data security and storage The encrypted and password protected hospital computer will be used to process the data until all the patient identifiers have been removed and the datasets have been anonymised. Once the Master Research Datasets are validated and research team is satisfied with the quality of the data, the dataset will be anonymised by dropping variables for following patient identifiers; hospital number, surname, first name, date of birth and IVF reference number. The study number and the cycle reference numbers will be used as a patient and a cycle identifiers and only this anonymised dataset will be used for statistical analysis. A copy of non-anonymised dataset will be stored in the computer located in Clinical Records Room for data verification and a reference purposes. The datasets will be stored within IVF unit for the duration of the research projects of the MD programme. The necessity of storage of the datasets and measures of data security will be reviewed every three years thereafter. 119

120 II. RESULTS INTRODUCTION According to the protocol, all women from 20 to 50 years of age referred to Reproductive Medicine Department of Central Manchester University Hospitals NHS Foundation Trust for management of infertility or fertility preservation and had AMH measurement during the period from 1 September 2008 till 16 November 2011 have been included in the database. In total of 4,506 patients met the inclusion criteria, with 3,381 patients in DSL AMH assay group and 1,125 patients Gen II assay group. The following datasets have been extracted from the clinical electronic data management systems: PAS, Biochemistry, Surgery, IVF, FET and Embryology. Data extraction from the paper-based hospital records of 3,681 patients (n=3,130 DSL and n=551 Gen II) were performed by two researchers: Dr O.Rustamov (n=2801) and Dr M. Krishnan (n=880). In addition, data collection using Medisec Digital Dictation Software for the notes that were not located in DSL group (n=251 patients) was completed by Dr O. Rustamov. In view of the issues with validity of Gen II assay measurements, which were observed in the earlier study of the MD Programme (Chapter 2: AMH variability and assay method comparison), I decided to base subsequent work for the last three projects (Chapter 5-7) of the MD programme only on DSL assay measurements and not to include samples based on Gen II AMH Assay. Therefore I decided not to collect data from the hospital notes for the patients that had AMH measurements using exclusively Gen II Assay where the notes were not found during the first round of data collection (n=575). As a result in DSL group, all datasets for 3130 patients were completed and all but AFC and Folliculogram datasets were completed for 251 (Figure 2). In Gen II group, all datasets were completed for 551 patients and all but RH, AFC and Folliculogram datasets were obtained for 575 patients (Figure 2). As described above the studies of the last three projects (Chapter 5-7) are based on DSL assay, which is no longer in clinical use. The review of literature presented in Chapter 3 suggests that DSL assay appears to have provided the most reproducible measurements of AMH compared to that of other assays. Therefore, AMH measured using DSL assay is perhaps most 120

121 reliable in terms addressing the research questions. In all three chapters, estimates of the effect sizes are provided in percentage terms and therefore the results are convertible to any AMH assay. Datasets Demography dataset The dataset was obtained from Mr. Peter Hoyle, Senior Data Analyst of Information Unit, CMFT on 16 October The dataset includes all patients referred to Reproductive Medicine Department between 1 January 2006 and 31 August 2012 and contains 5573 patients. I created a dataset, Demography, in Stata format using the steps of data cleaning, coding and management as per protocol. The audit trial of the data management was created using Stata Do file (Figure 1). Biochemistry dataset The biochemistry data file was obtained from Dr Alexander Smith, Senior Clinical Scientist, Biochemistry Department on 24 January The dataset contains the results of all serum AMH, FSH, LH and E2 samples conducted from 01 September 2008 to 31 December The dataset was in Excel format that consisted of two datasheets: 1) Biochemistry and 2) Biochemistry The datasheets transferred to Stata 12 in original unaltered condition and a single Stata Biochemistry dataset was created by combining datasheets by appending them to each other. The dataset contains in total of blood results of patients with 6643 AMH, FSH, LH and E2 results. A wide format of the dataset was prepared by transferring all blood results of each patient to a single row. A variable, which indicates valid FSH results was created by coding FSH results as missing if corresponding E2 levels were higher than 250 pmol/l. The audit trial of the data management was created using a Stata Do file. Surgery dataset Data management was conducted according to the protocol. In total, dataset contained 2096 operations in 1787 patients. Data on all operations on 121

122 Fallopian tubes (e.g. salpingectomy, salpingostomy) and ovaries (e.g. cystectomy, drainage of cyst) at Central Manchester NHS Foundation Trust from 1 January 2000 to 16 January 2011 are available in the dataset. The dataset will be used to validate the data on history of reproductive surgery of Reproductive History dataset AMH dataset Both AMH datasets were received from Dr Philip Pemberton, Senior Clinical Scientist of the Specialist Assay Laboratory on 13 January 2012 and transferred to Stata 12 software in the original format. All steps of the data cleaning and the management were recorded using Stata Do file. There were 3381 patients in DSL dataset and 1,125 patients in Gen II dataset. Cleaning and coding of the datasets were achieved using the methodology described in above protocol and new, AMH, dataset has been created. IVF dataset The dataset was downloaded from ACUBase by Dr Oybek Rustamov on 08 October 2012 and following importing the dataset into Stata 12 in original format dataset was prepared according to the protocol. The dataset contains all IVF ICSI cycles that took place between 01 January 2004 and 01 October 2012; including the cycles of women who acted as egg donors and egg recipients. There were in total of 4323 patients who had 5737 IVF ICSI cycles with 4123 IVF ICSI cycles using own eggs, 10 embryo storage, 40 oocyte donation, 7 oocyte storage, 55 oocyte recipient cycles. The dataset has anonymised unique patient (Patient ID) and cycle identifiers (Cycle ID) and therefore can be linked to all other datasets; including all FET cycles and embryos originated from the index IVF cycle. FET dataset The dataset was downloaded from ACUBase by Dr Oybek Rustamov in Excel format on 20 October 2012 and transferred to Stata 12 Software in the original condition. The data managed as per above protocol and each step of the process of preparation of the dataset was recorded in Stata Do file. The dataset comprised of all FET cycles (n= 3709) of all women (n=1991) 122

123 conducted between 01 January 2004 and 01 October 2010 and the Stata version of FET dataset contains complete data on number of thawed, cleaved, discarded and research embryos for all patients. The dataset contains unique patient identifier (Patient N) and unique cycle identifiers (Cycle N) and therefore can be linked to all datasets in patient and cycle levels including index IVF cycle and embryos. Embryology dataset The Excel dataset was downloaded from ACUBase by Dr Oybek Rustamov on 20 October 2012 and transferred into Stata 12 Software in unaltered condition. The data was managed according to the above protocol. The dataset has details of all 65,535 (n=4305 women) embryos that were created between 01 January 2004 and 01 October The dataset contains complete data on quantity and the assessment of embryo quality, which includes grading of number, evenness and defragmentation of the cells for each day of culturing of the embryos. Furthermore the destination of each embryo (e.g. transferred, cryopreserved, discarded and donated) and the outcomes of cycles for transferred embryos are available in the dataset. Given that the Embryology dataset has the unique patient as well as the cycle identifiers, this dataset is nested within patients and IVF cycles. Consequently, each embryo can be linked to patient, index Fresh IVF cycle and subsequent FET cycles. Reproductive History, AFC and Folliculogram datasets The hospital notes of all patients (n=4,506) were searched during the period of 1 April 2012 to 15 October 2012 for collection of data for Reproductive history, AFC and Folliculogram datasets as per protocol. All case noted filed in the Clinical Records Room, the Nurses Room, the Doctors Room and the Secretaries Room of Reproductive Medicine Department were searched and relevant notes were pulled and searched for data. All ultrasound scan reports containing data on AFC and all IVF ICSI folliculograms of patients were scanned and electronic copy of scanned documents were stored in the password protected NHS computer located in the Clinical Records Room. 123

124 The first round of data gathering achieved following result. In DSL dataset there were in total of 3381 patients, with 3130 patients had complete data extraction from their hospital notes and hospital records of 251 patients were not found. There were in total of 1126 patients in Gen II dataset, 551 of whom had complete data extraction from their hospital records and the case notes of 575 patients were not located (Figure 2). The main reason for missing case notes was found to be the use of hospital records by clinical, laboratory and administrative members of staff at the time of data collection in patients undergoing investigation and treatment. In the meantime the results of our previous research study indicated that Gen II samples provide erroneous results (Chapter II) and therefore we decided to use only data from the patients in DSL group. Data on reproductive history for the remaining patients in the DSL group (n=251) with missing hospital records were collected using digital clinic letters stored in Medisec Digital Dictation Software (Medisec Software, UK). The data file, that contained combined datasets of reproductive history and AFC, was transferred to Stata 12 in original condition and data management was conducted according to the protocol. All steps of data management was recorded using Stata do file for audit trail and to ensure reproducibility of the management of the data. Similarly, the management of Folliculogram dataset was achieved using the procedures described in the protocol and all steps of data management was logged using Stata Do file. As result of above data collection and management I created three Stata datasets: RH (reproductive history), AFC and Folliculogram. Merging Datasets First the datasets were merged using a unique patient identifier (hospital number) as per protocol. Validation of the merger using additional patient identifiers (NHS number, name, date of birth) revealed existence of duplicate hospital numbers in patients transferred from secondary care infertility services to IVF Department of Central Manchester University Hospitals NHS Foundation Trust. I established that, in the datasets, the combination of the patient s first name, surname and date of birth in a single string variable could be used as a unique identifier. Hence, I used this identifier to merge all datasets, achieving a robust merger of all independent datasets into combined 124

125 final Master Datasets for each of the research projects. Following the creation of an anonymised unique patient identifier (Patient ID) for each patient and anonymised unique cycle identifier (Cycle ID) for each IVF cycle, all patient identifiers (e.g. surname, forename, hospital number, IVF ref number) were dropped (Figure 1). The anonymised independent datasets (e.g. AMH, AFC, IVF etc.) and anonymised Master Datasets were stored as per protocol. Subsequently, these anonymised datasets were used for the statistical analyses of the research projects. The original unanonymised data files were stored in two password protected NHS hospital computers in the Clinical Records Room and Doctors Room of Reproductive Medicine Department and archived according to the Trust policies thereafter. Only members of clinical staff have access to the computers and only nominated clinical members of the research group who have specific approval can have access to unanomysed. Fully anonymised datasets have been made available to other members of the research team, with the stipulation that the datasets are stored on secure password protected servers or fully encrypted computers. Fully anonymised datasets may in the future be shared with other researchers following consideration of the request by the person responsible for the datasets (Dr Cheryl Fitzgerald) and appropriate ethical and data protection approval. CONCLUSION Following extraction and management of the data I have built comprehensive validated datasets which will enable to study ovarian reserve in a wide context; including a) assessment of ovarian reserve, b) evaluation of the performance of ovarian biomarkers, c) study individualization of ovarian stimulation in IVF, d) association of the biomarkers of ovarian reserve with outcomes of IVF (e.g. oocytes, embryo, live birth). The database will be used to address the research questions posed in the subsequent chapters of this thesis and beyond that for future studies on the assessment of ovarian reserve and IVF treatment. 125

126 Figure 1. Data and program files. Datasets and programme files created in preparation of the research datasets. File names and types are provided in the brackets. 126

127 Table 1a. Available vriables available identifiers, variables and the source of data for following datasets: Ethnicity, RH, AMH, AFC, Biochemistry, OHSS, Folliculogram The Datasets Clinical ID Study ID Variables Source Demography Hospital N, Surname First name, DOB Patient ID Ethnicity Information Department (PAS) RH (Reproductive History) Hospital N, Surname First name DOB Surgery Hospital N, Surname First name DOB AMH Hospital N, Surname First name, DOB AFC Hospital N, Surname First name, DOB Biochemistry Hospital N, Surname First name, DOB Folliculogram Hospital N, Surname First name, DOB Patient ID 1. Diagnosis: Referral Female, Referral Male Clinic Female, Clinic Male Post Cycle 1, Post cycle 2, Post cycle 3 2. Iatrogenic causes of loss of ovarian reserve Ovarian surgery, tubal surgery, chemotherapy, radiotherapy 3. BMI, 4. PCOS (PCO, oligomenorrhea, amenorrhea, hirsutism) Patient ID Date Patient ID Date Patient ID Date Patient ID Date Patient ID Date Procedure Date Operator Date of sample, Date of assay, AMH level, Assay generation AFC (up to six AFC scans) Left ovary, Right ovary, Date of Scan, Sonographer Comments (Ovarian cyst, hydrosalpynx, fibroid, poorly visualized etc.) Oestradiol (Date of sample, Date of assay serum level) FSH (Date of sample, Date of assay serum level) LH (Date of sample, Date of assay serum level) Folliculogram (up to 3 cycles) Date (1 st day of ovarian stimulation) Day 8( 10-16mm), Day 8 (>17mm) Day 10 (10-16mm), Day 8 (>17mm) Comments (Day of HCG, OHSS, Cancellation, Ovarian cyst, Hydrosalpynx, Coasting, etc.) Hospital Records Information Department AMH dataset of Specialist Assay Lab Hospital Records Biochemistry Electronic Database Hospital Records 127

128 Table 1b. Available variables The available identifiers, variables and the source of data for IVF dataset. Datasets Clinical ID Study Variables Source IVF Hospital N, Surname First name DOB PCT code Patient ID Cycle ID Date GENERAL Attempt Type Protocol DaysStim InitDose Outcome OutcomeDt Age PartnerAge EggCollect TreatDate ETransfer Add_Drug1 Add_Drug2 Add_Drug3 Add_Drug4 Add_Drug5 Add_Drug6 Add_Drug7 EGG RECOVERY SNumber Follicles TotEgg EggNumber FERTILISATIO N IVFEgg IVFCleaved ICSICleaved Cleaved PN2 IVFPN2 ICSI2PN ICSICl ICSIEgg ICSIFPN IVFFPN IVFTransfer ICSITransfer IVFLysed ICSILysed IVFMetII IVFMetI IVFAtretic IVFAbnormal IVFEmptyZona IVFG_Vesicle ICSIMetII ICSIMetI ICSIAtretic ICSIAbnormal ICSIEmptyZona ICSIG_Vesicle OUTCOME sacs Hearts Preg ICSIPract STORAGE Frozen IVFFroz ICSIFroz SpermSource SortKeySTA R HISTORY cat_tubal cat_ovfail cat_utprob cat_unex cat_ MF cat_meno cat_genetic cat_endo cat_anov cat_nomale Inf_Since MaleInf CoupleInf Preg24Wk MiscTOP Ectopic LiveBirth FSH AMH Emb_Recip Surrogate Sperm_Recip StoreEggs EggThaw Treat_Reason IgnoreKPI EMBRYOLO GY D1LteClCells1 D1LteClCells2 D2Cells2 D2Cells3 D2Cells4 D2Even2 D2Even3 D2Even4 D2Frag2 D2Frag3 D2Frag SPERM Conc_Init MotA MotB Conc_ Prep MotAP MotBP SemenSource SemenAnalysis STIMULATIO N BMI TotDose GonadUsed Incubator ICSIRigg AMHBand DHEA EGG Egg_Recip Own_Eggs Altruistic_D ACUBASE Electronic Database 128

129 Table 1c. Available variables The available identifiers, variables and the source of the data for FET and Embryo datasets. Datasets Clinical ID Study ID Variables Source FER Hospital N Surname First name Patient ID Cycle ID Date GENERAL treatdate transfer ETDate OUTCOME preg IUP Outcome OutcomeDt EMBRYOLO GY Thawed Survived Cleaved Discarded Research STORAGE NumStored DtCreated CLINICIAN ETClinician ETEmbryologi st OrigCycle ACUBASE Electronic Database Embryo Hospital N, Surname First name DOB Patient ID Cycle ID Date GENERAL TreatDate Injected Destination CELLS CellsD1 CellsD2_AM CellsD2_PM CellsD3_AM CellsD3_PM EVENNES EvenD2_AM EvenD2_PM EvenD3_AM EvenD3_PM FRAGMENT FragD1 FragD2_AM FragD2_PM FragD3_AM FragD3_PM OUTCOMES ICSIPract Maturity PosPreg Hearts SpermSource Age ACUBASE Electronic Database 129

130 Table 2a. Coding The codes used to convert ethnicity and diagnosis variables from string to numeric format in PAS and RH datasets. 130

131 Table 2b. Coding The codes used to convert treatment outcomes from string to numeric format in IVF and FET datasets. Datasets Codes for outcomes: IVF FET Biochemical Pregnancy =1 Cancel (other) =2 Cancel Hyperstimulation =3 Cancel Poor response =4 Cancelled no sperm on day of EC =5 CONVERTED IVF TO IUI =6 Delayed Miscarriage =7 Donated =8, Ectopic =9 Egg donation =10 Embryos for storage =11 Empty Sac =12 Failed Fertilisation =13 For donation =14 Freeze All =15 Freeze All (OHSS) =16 Freeze All (Other) =17 Late Miscarriage =18 lost to contact =19 lost to follow up =19 No Eggs =20 No Sperm =21 No Normal Embryos =22 Not Pregnant =23 Ongoing Singleton =24 Ongoing Twin =25 Positive hcg =26 Singleton Birth=27 Twin Birth =28 Triplet Birth =29 Still Birth =30The 131

132 Figure 2. Data collection from hospital records Completeness of data collection from hospital records for RH, AFC and Folliculogram datasets All patients DSL Gen II (n=3381) (n=1126) All Datasets Complete AFC and Folliculogram not complete All Datasets Complete RH, AFC, Follicologram not complete n=3130 n=251 n=551 n=

133 Table 3. Results: Datasets and observation Summary of the number of patients, observations, IVF/FET cycles and data entry period for all datasets. Datasets Patients Observations Cycles Period AMH DSL: 3381Gen II: 1126 DSL-3,913 DSL: 01 Sep Nov 2010 Gen II: 16 Nov Nov 2011 Demography Jan Aug 2012 Biochemistry Total 78, Sep Dec AMH, 19,175-FSH 28,677-LH, 23,920-E2 RH DSL-3381 DSL Sep Oct 2012 Surgery Jan Nov 2011 AFC DSL: 2411 DSL Total: Sep Oct 2012 Single measurement:2411, Repeats: ; 3-370; 4-105; 5-25; 6-7; 7-1 Folliculogram Sep Oct 2012 IVF/ICSI Total own eggs-4123, oocyte 01 Jan Oct 2012 recipients-55, oocyte donors-40 Embryo storage-10, oocyte storage-7 FET Jan Oct 2012 Embryology embryos - 01 Jan Oct

134 Figure 3. Merging datasets The process of merging datasets in patient and cycle levels using patient, date and cycle IDs. 134

135 ASSESSMENT OF DETERMINANTS OF ANTI-MÜLLERIAN HORMONE IN INFERTILE WOMEN 5 135

136 THE EFFECT OF ETHNICITY, BMI, ENDOMETRIOSIS AND THE CAUSES OF INFERTILITY ON OVARIAN RESERVE Oybek Rustamov, Monica Krishnan, Cheryl Fitzgerald, Stephen A. Roberts To be submitted to: Fertility and Sterility

137 Title: The effect of ethnicity, BMI, endometriosis and the causes of infertility on ovarian reserve. Authors Oybek Rustamov a, Monica Krishnan b, Cheryl Fitzgerald a, Stephen A. Roberts c Institutions a Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH, UK; b Manchester Royal Infirmary, Central Manchester University Hospitals NHS Foundation Trust, Manchester M13 9WL, UK; c Centre for Biostatistics, Institute of Population Health, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 9PL, UK; Corresponding author & reprint requests Dr. Oybek Rustamov, Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH Word count: 4715 Grants or fellowships: No funding was sought for this study Disclosure summary: There were no potential conflicts of interest. Clinical Trial registration number: Not applicable. Acknowledgements: The authors would like to thank colleagues Dr Greg Horne (Senior Clinical Embryologist), Ann Hinchliffe (Clinical Biochemistry Department) and Helen Shackleton (Information Operations Manager) for their help in obtaining datasets for the study. 137

138 Declaration of authors roles OR prepared the dataset, conducted statistical analysis and prepared all version of the manuscript. MK assisted in data extraction, contributed in discussion and the review of the manuscript. SR and CF oversaw and supervised preparation of dataset, statistical analysis, contributed in discussion and reviewed all versions of the manuscript. 138

139 ABSTRACT Objective To estimate the effect of ethnicity, BMI, endometriosis and the causes of infertility on ovarian reserve. Design: Single centre retrospective cross-sectional study. Setting Women referred to secondary and tertiary level referral centre for management of infertility. Participants A total of 2946 patients were included in the study of which 65 did not have data on ethnicity leaving 2881 women in the sample Interventions: Serum AMH, AFC and basal FSH measurements. Main outcome measure Serum AMH, serum basal FSH and basal AFC measurements. Results Multivariable regression excluding BMI showed that woman of Black ethnicity and the group defined as Other ethnicity had significantly lower AMH measurements when compared to that of White (-25%; p=0.013 and -19%; p=0.047) and overall ethnicity was a significant predictor of AMH (p=0.007) However, inclusion of BMI in the model reduced these effects and the overall effect of ethnicity did not reach statistical significance (p=0.08). AFC was significantly reduced in Pakistani and women of Other ethnicities ; although the effect sizes were small (10-14%) and the overall effect of ethnicity was significant in both models (p=0.04 and p=0.03). None of the groups showed a statistically significant difference in FSH, although women of Other Asian ethnicity appear to have lower FSH measurements (12%) which was close to statistical significance (-.12%; p=0.07). 139

140 Obese women had higher AMH measurements (16%; p=0.035) compared to that with normal BMI and the overall effect of the BMI was significant (p=0.03). In the analysis of the effect of BMI to AFC measurements, we did not observe differences that were statistically significant. However FSH results showed that there is a modest association between BMI and FSH with both overweight and obese women having significantly lower FSH measurements compared to lean women (-5%; p=0.003 and -10%; p=0.003). In the absence of endometrioma, endometriosis was associated with lower AMH measurements, although this did not reach statistical significance. Neither AFC nor FSH was significantly different in the endometriosis group compared to those without endometriosis. In contrast we observed around 31% higher AMH levels in the patients with at least one endometrioma (p=0.034), although this did not reach statistical significance (21%; p=0.1) in the smaller subset after adjustment for BMI. AFC and FSH did not show any statistically significant association with endometrioma. There were no differences in the AMH measurements between patients diagnosed with unexplained infertility compared to the ones who did not have unexplained infertility; except the analysis that did not include BMI as a covariate, which found a weakly positive correlation (10%; p=0.03). Similarly, the estimation of the effect of the diagnosis of unexplained infertility to AFC as well as FSH showed that there were weak association between the markers and diagnosis of unexplained infertility. There was no significant difference in AMH, AFC and FSH measurements of women with mild and severe tubal infertility in the models, which included all covariates; except the analysis of FSH and mild tubal factor where we found weakly negative correlation between these variables. Women diagnosed with male factor infertility had significantly higher AMH and lower FSH measurements; the effect sizes of which were directly proportional to the severity of the diagnosis. In the analysis of AFC we did not found significant difference in the measurements between patients with male factor infertility and to that of non-male factor. 140

141 Conclusions Ethnicity does not appear to play a major role in determination of ovarian reserve as measured by AMH, AFC and FSH, whereas there is a significant positive association with BMI and these markers of ovarian reserve. Women with endometriosis appear to have lower AMH, whilst patients with endometrioma have significantly higher AMH and lower FSH measurements. The study showed that the association between markers of ovarian reserve and unexplained infertility, as well as tubal disease, is weak. In contrast women diagnosed with male factor infertility have higher ovarian reserve. Key Words Ovarian reserve, AMH, AFC, FSH, ethnicity, BMI, infertility, endometriosis, endometrioma. 141

142 INTRODUCTION The ovarian reserve consists of a total number of resting primordial and growing oocytes, which appears to be determined by the initial oocyte pool at birth and the age-related decline in the oocyte number (Hansen et al 2008; Wallace and Kelsey 2010). Both of these factors appear to be largely predetermined genetically, although certain environmental, socioeconomic and medical factors likely to play a role in the rate of the decline (Schuh-Huerta et al 2012b; Kim et al 2013; Dolleman et al 2013). The understanding of the formation and the loss of ovarian reserve have been improved greatly due to recently published data on the histological assessment of ovarian reserve (Hansen et al 2008). Furthermore the use of the biomarkers has enabled the evaluation of ovarian reserve in larger population-based samples. Biomarkers such as AMH and AFC can only assess the measurement of growing pre-antral and early antral follicle activity. However, some studies suggest that there is a close correlation between the measurements of these markers and the number of resting primordial follicles (Hansen et al 2011). Studies on age related decline of AMH and AFC have played important roles in understanding the decline of ovarian reserve, although most of the data have been derived from heterogeneous population without full account for characteristics of individual patients (Nelson et al 2011; Seifer et al 2011; Shebl et al 2011). These studies have demonstrated that there is a significant between-subject variation in ovarian reserve, beyond that due to chronological age (Kelsey et al 2011). More recent studies reported interesting findings on the role of demographic, anthropometric and clinical factors in the determination of ovarian reserve. Although these studies have employed better-described samples, some have small sample sizes and lack power for the estimation of the effect of these factors. Consequently, studies on large and well-characterised populations are necessary for evaluation of the determinants of ovarian aging as well as to provide normative data for the individualisation of the assessment of ovarian reserve. There have been reports of measurable disparities in the reproductive aging and reproductive endocrinology between various ethnicities. For instance, according to a large prospective study White, Black and Hispanic women reported higher rates of premature ovarian failure compared to 142

143 Chinese and Japanese (Luborsky et al 2002). In contrast the prevalence of PCOS, which is associated with higher ovarian reserve, has been reported to be significantly lower in Chinese (2.2%) compared to British (8%) women (Michelmore et al 1999; Chen et al 2002). Although these disparities may partially be due to the differences in the local diagnostic criteria, it is plausible to believe that the ethnicity may play a role in the determination of the reproductive aging. With regard to the effect of ethnicity to the markers of ovarian reserve, Seifer et al found that African American and Hispanic women have lower AMH levels compared to White (Seifer et al 2009). In contrast, Randolph et al reported that African American women had significantly higher ovarian reserve compared to that of White when determined by FSH measurements (Randolph et al 2003). These studies indicate that ethnicity may play a role in the determination of ovarian reserve and therefore warrants further investigation, which should include other major ethnic groups. Body weight appears to be closely associated with human female reproduction, which is evident by its effect on the natural fecundity as well as the success of the assisted conception treatment cycles (Maheshwari et al 2007). Indeed the relationship of increased body mass index (BMI) and PCOS is well described, although the mechanism of this is not yet fully understood. Consequently a number of recent studies have assessed the effect of BMI to the various aspects of reproductive endocrinology, including ovarian reserve. Studies on the influence of BMI on the markers of ovarian reserve have provided conflicting results, probably due to the limited statistical power in most of these studies and the difficulties encountered in properly accounting for confounding factors such as age, ethnicity and medical diagnosis (Buyuk et al 2011; Freeman et al 2007; Su et al 2008; Seifer et al 2008; Sahmay et al 2012; Skalba et al 2011). Therefore there is a need for studies with large datasets and good adjustment for confounding factors. We therefore designed and undertook a study to estimate the effect of ethnicity, BMI, endometriosis and causes of infertility on ovarian reserve, as measured by AMH, AFC and FSH, using a robust dataset from a large cohort of patients referred for infertility investigation and treatment in a single centre. 143

144 METHODS Objectives The objectives of the study were to assess the role of the ethnicity, BMI and endometriosis and the causes of infertility on ovarian reserve as assessed by the biomarkers AMH, AFC and FSH using a large clinical data obtained retrospectively. Sample All women between 20 to 45 years of age referred to the Women s Outpatient Department (WOP) and the Reproductive Medicine Department (RMD) of Central Manchester University Hospitals NHS Foundation Trust for management of infertility from 1 September 2008 to 16 November 2010 and who had the measurement of AMH using DSL assay (DSL, Active MIS/AMH ELISA; Diagnostic Systems Laboratories, Webster, Texas) were included in this study. Patients referred for fertility preservation (e.g. prior to or after the treatment of a malignant disorder) and patients with a history of tubal or ovarian surgery (salpingectomy, ovarian cystectomy, salpingo-oopherectomy) and patients diagnosed with polycystic ovaries on ultrasound were excluded. The sample size was determined on pragmatic grounds and represents all available patients meeting the inclusion criteria. Measurement of AMH All patients referred to RMD had a measurement of AMH prior to management of their infertility whereas the patients seen at WOP had AMH measurements if they had a clinical indication for an assessment of ovarian reserve. Blood samples for the measurement of AMH were taken at an initial patient visit, without regard to the day of the menstrual cycle and transported to the in-house Biochemistry Laboratory. All samples were processed and analysed strictly according to the assay kit insert provided by the manufacturer. Serum samples were separated within two hours from venipuncture and frozen at -20C until analysed in batches using the enzymatically amplified two-site immunoassay (DSL, Active MIS/AMH ELISA; Diagnostic Systems Laboratories, Webster, Texas). The working range of the assay was up to 144

145 100pmol/L with a minimum detection limit of 0.63pmol/L. The intra-assay coefficient of variation (CV) (n=16) was 3.9% (at 10pmol/l) and 2.9% (at 56pmol/l). The inter-assay CV (n=60) was 4.7% (at 10pmol/l) and 4.9% (at 56pmol/l). In patients with repeated AMH measurements the first measurement was selected for this study. Measurement of FSH Patients had measurement of basal FSH, LH and oestradiol levels (E2) during the early follicular phase (Day 2-5) of their menstrual cycle as a part of their initial work up. Blood samples were transported to the in-house Biochemistry Laboratory within two hours of venipuncture for sample processing and analysis. Serum FSH levels were measured using specific immunoassay kits (Cobas, Roche Diagnostics, Mannheim, Germany) for use on an autoanalyser platform (Roche Modular Analytics E170, Roche, USA). The intra-assay and inter-assay CVs were 6.0% and 6.8%, respectively. FSH measurements in samples with high E2 levels (>250) were defined as nonrepresentative of early follicular phase and were not included in this study. Where patients had repeated FSH measurements the measurement with the closest date to that of AMH measurement was used. Measurement of AFC Measurement of AFC was conducted in all patients undergoing assisted conception. The department uses a stringent protocol for the assessment of AFC, which consists of counting all antral follicles measuring 2-6mm in longitudinal and transverse cross sections of both ovaries using transvaginal ultrasound scanning at early follicular phase (Day 0-5) of the menstrual cycle. Fully qualified sonographers conducted the ultrasound assessments. Where patients had repeated AFC measurements, the AFC closest to the date of the AMH measurement was used. Data collection Data was extracted from hospital electronic clinical data management systems and from written hospital notes of each patient. AMH and FSH measurements were obtained from the Biochemistry Department of the hospital and validated by checking results of randomly selected 50 patients 145

146 against the results available in electronic clinical data management system (Clinical Workstation). Data on AFC, BMI, the causes of infertility, the duration of infertility, the history of reproductive pathology and reproductive surgery were gathered from the hospital case notes. Data on the ethnicity was obtained from the hospital s administrative database (PAS). The datasets were merged using a unique patient identifier (hospital number) and the validity of the linkage was validated using other patient identifiers (NHS number, patient s name and date of birth). Definitions and groups In our hospital, the ethnicity of the patient is established using a patient questionnaire based on the UK census classification. The body mass index (BMI) of patients was categorised using NHS, UK cut-off reference ranges: Underweight (<18.5), Normal ( ), Overweight ( ) and Obese (30-40). Causes of infertility were established by searching hospital records, including referral letters, clinical entries and the letters generated following initial and follow up clinic consultations. Patients with a history of bilateral tubal block, which was confirmed by laparoscopy and dye test, and patients with a history of bilateral salpingectomy were categorised as having severe tubal factor infertility. Patients with unilateral tubal patency or unilateral salpingectomy were categorised as having mild tubal factor infertility. Patient s with laparoscopic diagnosis of stage III and Stage IV endometriosis (AFS) were categorised as diagnosed with severe endometriosis whilst patients with Stage I and Stage II endometriosis were allocated to group of mild endometriosis. Severe male factor infertility was defined as azoospermia or severe oligospermia, which necessitated Multiple Ejaculation Resuspension and Centrifugation test (MERC) for assisted conception. The criteria for MERC were a) sperm count of <0.5 mln/ml or b) retrograde ejaculation. Patients with abnormal sperm count but who did not meet above criteria were classified as mild male factor infertility. Statistical analysis Firstly univariate analyses of the effect of age, ethnicity, BMI, endometriosis with and without endometrioma, causes of infertility and duration of infertility were conducted using two sample t test. Then a 146

147 multivariate linear regression model that included age, ethnicity, BMI, endometriosis, presence of endometrioma and the causes of infertility was specified for the analyses of the effect of these factors to AMH, AFC and FSH. Logarithmically transformed values were used for the statistical analysis of AMH, AFC and FSH The precise age on the day measurement of each of the marker of ovarian reserve (AMH, AFC and FSH) was used and age adjustment utilised a quadratic function following centring to 30 years of age. Differences between the groups were considered significant at p Interactions between all explanatory variables were tested at a significance level of p<0.01. In order to estimate the effect of BMI we fitted two different models with a) BMI not included and b) BMI included in the model. Duration of infertility did not show any clinical or statistically significant differences for any of the markers and therefore this variable was not included in the models. RESULTS In total 2946 patients were included in the study, of whom 2880 of these patient had valid AMH measurements, 1810 had measurement of AFC and 2377 had FSH samples The mean and median age of patients were 32.8 (4.5) and 33.2 (29.5; 36.5), respectively and the distribution of patients according to age categories ethnicity, BMI, endometriosis and the causes of infertility is shown in the Table 1. The summary statistics for the markers of ovarian reserve were as follows: AMH mean 17.5 (5.01), median 14 2 ( ); AFC mean 13.9 (6.3), median 13 (10-17) and FSH mean 7.9 (7.2), median 7 ( ) As expected, chronological age was found to be a significant determinant of all markers of ovarian reserve. We observed in average 5% decline in AMH levels, 2% decline in AFC and 1% increase in FSH measurements per year (Table 2-4). Out of 2946 patients 2021 had data on BMI measurements and in 925 BMI was not available. Table 5 describes age, AMH, AFC and FSH according to the availability of data on BMI. Distribution of patients by their ethnicity and an availability of data on BMI is provided in Table 6. Similarly, patient distribution by diagnosis and availability of data on BMI can be found in Table

148 Ethnicity The multivariable regression excluding BMI (Table 2) showed that woman of Black ethnicity and the group defined as Other ethnicity had significantly lower AMH measurements when compared to that of White (-.25 %; p=0.013 and -19%; p=0.047) and the overall ethnicity was a significant predictor of AMH (p=0.007) However, inclusion of BMI in the model reduced these effects and none of the groups had a statistically significant difference in AMH levels compared to that of White and the overall effect of ethnicity did not reach statistical significance (p=0.08). AFC was significantly reduced in Pakistani and women of Other ethnicities (Table 3); although the effect sizes were small (10-14%) and the overall effect of ethnicity was significant in the models with and without BMI (p=0.04 and p=0.03). None of the groups showed statistically significant differences in FSH (Table 4), although women of Other Asian ethnicity appear to have lower FSH measurements (12%) which was close to the level of statistical significance (-12%; p=0.07). BMI Obese women had 16% higher measurements of AMH (p=0.035) and overall effect of the BMI was significant (p=0.03). No interaction were detected between BMI and ethnicity, causes of infertility or diagnosis of endometriosis suggesting that effect of BMI was independent of these factors (Table 2). In the analysis of the effect of BMI on AFC measurements, we did not observe any differences that were statistically significant (Table 3). However FSH results showed that there is a modest association between BMI and FSH with both overweight (Table 4) and obese women having significantly lower FSH measurements compared to lean women (-5%; p=0.003 and -10%; p=0.003). Endometriosis In the absence of endometrioma, endometriosis was associated with lower AMH measurements, although this did not reach statistical significance 148

149 (Table 2). Neither AFC nor FSH was significantly different in the endometriosis group compared to those without endometriosis (Table 3-4). In contrast we observed around 31% higher AMH levels in the patients with endometrioma (p=0.034), although this reduced to 21% and did not reach statistical significance (p=0.10), in the smaller subset after adjustment for BMI (Table 2). AFC and FSH did not show any statistically significant association with endometrioma (Table 3-4). Causes of Infertility There were no differences in the AMH measurements between patients diagnosed with unexplained infertility compared to those with diagnosis; except the analysis that did not include BMI as a covariate, which found a weakly positive correlation (10%; p=0.03). Similarly, the estimation of the effect of a diagnosis of unexplained infertility on AFC as well as FSH showed that there were weak association between the markers and a diagnosis of unexplained infertility (Table 2-4). There were no significant differences in AMH, AFC and FSH in women with mild and severe tubal infertility in the models which included all covariates; other than weakly negative correlation between FSH and mild tubal factor (Table 2-4). Women diagnosed with male factor infertility had significantly higher AMH and lower FSH measurements; the effect sizes of which increased with the severity of the diagnosis. We did not find any significant difference in AFC between patients with and without male factor infertility (Table 2-4). DISCUSSION This is first study investigating the effect of demographic, anthropometric and clinical factors on all three markers of ovarian reserve using a large cohort of women of reproductive age. Furthermore the statistical analysis adjusted for relevant covariables using multivariable linear regression models. 149

150 Ethnicity Our study found that, amongst the main British ethnic groups, the effect of ethnicity on ovarian reserve measured using AMH, AFC and FSH is fairly weak; and can be accounted for by differences in BMI between the ethnic groups. Recently studies have been published on the relationship of ethnicity and markers of ovarian reserve, all of which compared North American populations. One study, which assessed a relatively small number of women (n=102) at late reproductive age did not find a difference in AMH levels between White and African American Women OR 1.23 (0.56, 2.71, P=0.70) (Freeman et al 2007). In contrast, Seifer et al reported that Black (n=462) women had around 25% lower AMH measurements (P=0.037) compared to that of White (n=122) (Seifer et al 2009), which is not consistent with our findings. The main differences of this study compared to our study were; a) a majority were HIV infected women, b) women were older (median 37.5 years of age), c) the analysis did not control for possible confounders related to PCO, reproductive pathology and surgery. Furthermore, unlike our results, the study did not find a correlation between BMI and AMH levels. Similarly, Shuh-Huerta and colleagues reported that, African American women (n=200) had significantly lower AMH levels (P= ) compared to that of White (n=232); Mean AMH pmol/l and pmol/l respectively (Shuh-Huerta et al 2012b). Although the group used very stringent selection of patients and statistical analysis, BMI was not included in the regression model. Indeed, our analysis without BMI in the model found similar results (Table 2). But controlling for BMI has revealed no significant difference in the AMH levels between White and Black ethnic groups. With regard to AFC measurements Shuh Huerta et al reported no difference in the follicle counts between White (n=245) and African American (n=202) women, which supports our findings (Shuh-Huerta et al 2012b). Again similar to our results, the authors reported that FSH results of these ethnic groups provided comparable results (Shuh-Huerta et al 2012a). Although our results do not support some of previously published data, in view of above arguments, we believe the ethnicity does not appear to play a major role in determination of ovarian reserve. However, in view of the discrepant findings of the currently available studies we suggest further studies 150

151 in large and diverse cohorts should be carried out in order to fully understand the role of ethnicity. BMI Our results show that BMI has direct correlation with AMH and AFC and negative correlation with FSH; suggesting women with increased BMI are likely to have higher ovarian reserve. The effect of this association was significant in the analysis of AMH and FSH; obese women appear to have approximately 16% higher AMH and 10% lower FSH measurements when compared to women with normal BMI. Although the difference in AFC measurements did not reach statistical significance, there was direct correlation between AFC and BMI. Published data on the effect of BMI to AMH levels provide conflicting results compared to our study; given the studies reported either no association (Buyuk et al 2011; Freeman et al 2007; Su et al 2008) or a negative correlation between these factors (Seifer et al 2008; Sahmay et al 2012; Skalba et al 2011). However, most of these studies assessed peremenopausal women or that of late reproductive age. Indeed the studies evaluated the effect of BMI to AMH measurements in women of reproductive age demonstrated that lower AMH levels in obese women were due to age rather than increased BMI (La Marca et. al. 2012; Streuli et. al. 2012). Furthermore, most of these studies either employed univariate analysis or multivariate regression models that did not included all relevant explanatory factors. In addition, these studies had significantly smaller numbers of samples, ranging from 10 to 809, compared to our study (n=1953). Indeed, other large study (n=3302) with multivariate analysis supports our findings on the effect of BMI on ovarian reserve, indicating obese women have significantly lower FSH levels (Randolph et. al. 2004). Endometriosis Here we present data on the measurement of all three main markers of ovarian reserve in women with endometriosis. We observed that women with endometriosis without endometrioma did not have significantly different AMH, AFC or FSH measurements compared to women that do not have this pathology. Intriguingly, women who had endometriosis with endometriomata 151

152 tended to have higher AMH levels. Given the data was collected retrospectively, we did not have full information on laparoscopic staging of endometriosis in all patients and therefore an analysis according to severity or staging of endometriosis was not feasible. However, the analysis controlled for the important variables mentioned above and importantly only included the patients without previous history of ovarian surgery. We therefore we believe the study provides fairly robust data on relationship of endometriosis and the markers of ovarian reserve. Although it is generally believed that endometriosis has a damaging effect on ovarian reserve, published literature provides conflicting views ranging from no correlation between these factors to a significant negative effect of endometriosis. As mentioned above, most studies were small and used matched comparison of patients with endometriosis to control group using retrospectively collected data. Carvalho et al compared women with endometriosis (n=27) and to that of male factor infertility (n=50) and reported there was no difference in basal AMH and AFC levels, whilst FSH levels of women with endometriosis was lower. Another small study, which used similar methodology where an endometriosis group (n=17) was compared to patients with tubal factor infertility (n=17), reported opposite results suggesting endometriosis was associated with lower AMH measurements and there was no correlation between the pathology and FSH or AFC (Lemos et al 2007). Shebl et al compared AMH results of women with endometriosis (n=153) to a matched group that did not have the pathology (n=306) and reported that women with mild endometriosis had similar AMH levels, whereas the group with severe endometriosis had significantly lower AMH compared to the control group (Shebl et al 2009). Although using well-matched control groups is a robust study design; direct comparison of the two groups without controlling for other important covariables, may result in inaccurate results. Indeed the study that used multivariate regression analysis was able to demonstrate that there are number of factors that can affect AMH results and indeed following controlling for these factors there was no difference between AMH results of women with endometriosis compared to that of without disease (Streuli et al 2012). In view of above considerations, we believe the effect of endometriosis to ovarian reserve is poorly understood and warrants further investigation. 152

153 Regarding the effect of endometrioma on AMH levels, current evidence is conflicting. Using univariate analysis without controlling for confounders, Uncu et al reported that women with endometrioma (n=30) had significantly lower AMH and AFC measurements compared to control (n=30) women (Uncu et al 2013). Similarly, Hwu et al reported that women with endometrioma (n=141) had significantly lower AMH measurements compared to that of without pathology (n=1323) pathology (Hwu et al 2013). However, the study population appears to have a disproportionately higher number of women with history of previous and current history of endometrioma (319/1642) compared to any published studies and therefore the study may have been subject of selection bias. Kim et al., reported lower AMH measurements in women with endometrioma (n=102) compared to control group (102), mean±sem, 2.9±0.3 ng/ml_vs. 3.3±0.3_ng/mL, although this did not reach statistical significance (P=0.28). In our view, the most robust data on measurement of AMH in women with endometriosis was published by Streuli et al, which compared AMH levels of 313 women with laparoscopically and histologically confirmed endometriosis to 413 women without pathology (Streuli et al 2009). The group with endometriosis consisted of women with superficial peritoneal endometriosis (n=35), deep infiltrating endometriosis (n=183) and ovarian endometrioma (n=95) and relevant factors such as age, parity, smoking and previous ovarian surgery were adjusted for using multivariate regression analysis. In keeping with our findings, women with endometriosis did not have lower AMH levels; except for patients with previous history of surgery for endometrioma. Most interestingly the findings of Streuili et al and this study suggest that women with ovarian endometrioma do not have low AMH levels. In contrast, according to our data, the presence of endometrioma may be associated with a significant increase in serum AMH levels. Given that an endometrioma is believed to cause significant damage to ovarian stroma this is an interesting finding. Increased AMH levels in the presence of endometrioma may be due to acceleration in the rate of recruitment of primordial follicles and/or increased expression of AMH in granulosa cells. Furthermore, increased AMH levels in these patients may be due to expressions of AMH in endometriotic cells. A study by Wang et al suggested that AMH is secreted by human endometrial cells in-vitro (Wang et al 2009). This was the first report of 153

154 non-ovarian secretion of AMH and suggested that AMH may play important role in regulation of the function of the human endometrium. Subsequently the findings of Wang et al. were independently confirmed by two different groups. Carrarelli et al assessed expression of AMH and AMH type II receptor (AMHRII) in specimens of endometrium obtained by hysteroscopy from patients with endometriosis (n=55) and from healthy (n=45) controls (Carrarelli et al. 2014). The study also assessed specimens from patients with ovarian endometriosis (n=29) and deep peritoneal endometriosis (n=26). The study found that both AMH and AMHRII were expressed in endometrium. Interestingly, ectopic endometrium obtained from patients with endometriosis had significantly higher AMH and AMHRII levels compared to that of healthy individuals. Furthermore, the specimens collected from ovarian and deep endometriosis had highest AMH and AMHII mrna expression. These findings confirm that AMH as well as AMHRII are expressed in human endometrium and AMH may play a role in pathophysiology of endometriosis. A further study conducted by Signorile et al also confirmed expression of AMH and AMHRII in human endometriosis glands. Furthermore, the study found that treatment of endometriosis cells with AMH resulted in a decrease in cell growth suggesting that AMH may inhibit the growth of endometriotic cells. This suggests that further studies to understand the role of AMH in pathophysiology of endometriosis are warranted. Causes of infertility Unlike the above-mentioned factors, the association of the various causes of infertility and the markers of ovarian reserve are poorly studied. Therefore, our study appears to provide only available data on AMH, AFC and FSH levels in women with three main causes of infertility; unexplained, tubal and male factor. In our study, AMH levels of women with unexplained infertility did not differ from those with a diagnosis. Similarly, the effect of a diagnosis on AFC and FSH measurements were weak. Women with unexplained infertility do not have any significant pathology that can account for their infertility. However understanding the role of ovarian reserve in these patients is important. Our study suggests that women with unexplained infertility have comparable AMH levels to other infertile women. 154

155 We did not find any significant differences in AMH, AFC or FSH measurements of women diagnosed with tubal factor infertility compared to infertile women without tubal disease. Pelvic inflammatory disease and endometriosis are well known causes of tubal pathology and our regression model has controlled for the effect of endometriosis in these analyses. Our results suggest that despite having damaging effect to the tubes, pelvic infection does not reduce ovarian reserve. In contrast, our analyses showed that women with mild and severe male factor infertility have significantly increased AMH and lower FSH measurements, which demonstrates that these women have better ovarian reserve compared to general infertility population. Strengths and Limitations of the study The study is based on retrospectively collected data and therefore was subject to the issues related to this methodology. However, we believe that we have overcome most problems and improved the validity of our results by using a robust methodology for data collection, large sample size and careful analysis. We included all women who presented during the study period and met the inclusion criteria of the study. Importantly, women with previous history of PCO, chemotherapy, radiotherapy, tubal surgery or ovarian surgery have been excluded from the study given these factors may have significant acute impact on ovarian reserve, effect of which may be difficult to control for. The analysis showed an interaction between BMI and ethnicity, which could not be explored fully due to missing data on BMI (Tables 6). Therefore, analyses with and without BMI in models have been performed (Tables 2-4) and the distribution of patients according to availability of data on BMI has been obtained (Tables 5-7). I suggest further studies with sufficient data should explore this interaction. I was not able to establish the patients that meet Rotterdam criteria for diagnosis of PCOS, given data on menstrual history and biochemical assessment of androgenemia were not available. Therefore ultrasound diagnosis of PCO was used to categories patients with polycystic ovaries and all patients with PCO were excluded from analysis. It is important to note that measurement of AMH using Gen II assay may provide erroneous results (Rustamov et al., 2012a). Therefore only samples 155

156 obtained using DSL assay have been included in the study. The DSL assay appears to provide more reproducible results than the Gen II assay (Rustamov et al., 2011 and Rustamov et al., 2012a) and therefore we believe the estimates in this study reflect the relationship between circulating AMH and the above factors. SUMMARY Our data suggests that there is no strong association between ethnicity and AMH, AFC or FSH, whilst women with increased BMI appear to have higher ovarian reserve. There was no evidence of reduced ovarian reserve in women with endometriosis who do not have a previous history of ovarian surgery. In contrast women with current history of endometrioma may have higher AMH levels, which warrants further investigation. Women with a history of unexplained infertility do not appear to have reduced ovarian reserve, as measured with AMH, AFC and FSH, compared to general infertile women. Similarly, women with tubal factor infertility have comparable ovarian reserve with women who do not have tubal disease. In contrast, women with male factor infertility have significantly higher ovarian reserve compared to infertile women who do not have male factor infertility. This study has elucidated the effect of demographic, anthropometric and clinical factors on all commonly used markers of ovarian reserve and demonstrated that some of these factors have significant impact on ovarian reserve. 156

157 References Buyuk E, Seifer DB, Illions E, Grazi RV, and Lieman H, Elevated body mass index is associated with lower serum anti-mullerian hormone levels in infertile women with diminished ovarian reserve but not with normal ovarian reserve. Fertility and Sterility_ Vol. 95, No. 7, June 2011 Carrarelli P. Rocha A, Belmonte, Zupi E, Abrao M, Arcuri F, Piomboni P and Petraglia F Increased expression of antimullerian hormone and its receptor in endometriosis. Fertil Steril_ 2014; 101: de Carvalho BR, Rosa-e-Silva AC, Rosa-e-Silva JC, dos Reis RM, Ferriani RA, de Sá MF.Increased basal FSH levels as predictors of low-quality follicles in infertile women with endometriosis. International Journal of Gynecology and Obstetrics 110 (2010) Dólleman M, Verschuren W. M. M., Eijkemans M. J. C., Dolle M. E. T., Jansen E. H. J. M., Broekmans F. J. M., and van der Schouw Y. T.. Reproductive and Lifestyle Determinants of Anti-Mullerian Hormone in a Large Population-based Study. J Clin Endocrinol Metab, May 2013, 98(5): Freeman EW, Gracia CR, Sammel MD, Lin H, Lim LC, Strauss JF 3rd. Association of anti-mullerian hormone levels with obesity in late reproductiveage women. Fertil Steril 2007; 87: Halawaty S, ElKattan E, Azab H, ElGhamry N, Al-Inany H. Effect of obesity on parameters of ovarian reserve in premenopausal women. J Obstet Gynaecol Can. 2010; 32: Hansen KR, Knowlton NS, Thyer AC, Charleston JS, Soules MR, Klein NA. A new model of reproductive aging: the decline in ovarian non-growing follicle number from birth to menopause. Hum Reprod 2008;23: Hansen KR, Hodnett GM, Knowlton N, Craig LB Correlation of ovarian reserve tests with histologically determined primordial follicle number. Fertil Steril 2011, 95: Hwu Y, Wu FS, Li S, Sun F, Lin M and Lee RK. The impact of endometrioma and laparoscopic cystectomy on serum anti-mullerian hormone levels. Reproductive Biology and Endocrinology 2011, 9:80 Kelsey TW, Wright P, Nelson SM, Anderson RA, Wallace WHB (2011) A Validated Model of Serum Anti-Müllerian Hormone from Conception to Menopause. PLoS ONE 6(7): e Kim MJ, Byung Chul Jee, Chang Suk Suh, and Kim SH. Preoperative Serum Anti-Mullerian Hormone Level in Women with Ovarian Endometrioma and Mature Cystic Teratoma. Yonsei Med J, Volume 54 Number 4 July 2013 La Marca A, Sighinolfi G, Papaleo E, Cagnacci A, Volpe A, et al. (2013) Prediction of Age at Menopause from Assessment of Ovarian Reserve May Be 157

158 Improved by Using Body Mass Index and Smoking Status. PLoS ONE 8(3): e Lemos NA, Arbo E, Scalco R, Weiler E, Rosa V, Cunha-Filho JS. Decreased anti-müllerian hormone and altered ovarian follicular cohort in infertile patients with mild/minimal endometriosis. Fertil Steril May; 89(5): Luborsky JL, Meyer P, Sowers MF, Gold EB, Santoro N. Premature menopause in a multi-ethnic population study of the menopause transition. Hum Reprod 2002;18: Maheshwari A, Stofberg L, Bhattacharya S. Effect of overweight and obesity on assisted reproductive technology a systematic review. Hum Reprod Update 2007;13: Michelmore K, Balen A, Dunger D, Vessey M. Polycystic ovaries and associated clinical and biochemical features in young women. Clin Endocrinol (Oxf) 1999;51: Nelson SM, Messow MC, Wallace AM, Fleming R, McConnachie A. Nomogram for the decline in serum antimullerian hormone: a population study of 9,601 infertility patients. Fertil Steril 95: e Chen X, Yang D, Mo Y, Li L, Chen Y, Huang Y. Prevalence of polycystic ovary syndrome in unselected women from southern China. Eur J Obstet Gynecol Reprod Biol 2008; 139: Randolph JF, Sowers M, Gold EB, Mohr BA, Luborsky J, Santoro M, et al. Reproductive hormones in early menopausal transition: relationship to ethnicity, body size, and menopausalstatus. J Clin Endocrinol Metab Apr; (4): [PubMed: ] Sahmay S, Usta T, Erel CT, Imamoğlu M, Kuuk M, Atakul N, Seyisoğlu H. Is there any correlation between amh and obesity in premenopausal women? Arch Gynecol Obstet Sep; 286(3): Seifer DB, Baker VL and Leader B. Age-specific serum anti-m ullerian hormone values for 17,120 women presenting to fertility centers within the United States. Fertility and Sterility_ Vol. 95, No. 2, February 2011 Seifer DB, Golub ET, Lambert-Messerlian G, Benning L, Anastos K, Watts H, Cohen MH, Karim R, Young MA, Minkoff H, and Greenblatt RM. Variations in Serum Mullerian Inhibiting Substance Between White, Black and Hispanic Women. Fertil Steril November; 92(5): Shebl O, Ebner T, Sir A, Schreier-Lechner E, Mayer RB, Tews G,Sommergruber M. Age-related distribution of basal serum AMH level in women of reproductive age and a presumably healthy cohort. Fertil Steril 2011; 95:

159 Shebl O, Ebner T, Sommergruber M, Sir A, Tews G. Anti muellerian hormone serum levels in women with endometriosis: a case-control study. Gynecol Endocrinol 2009;25: Schuh-Huerta SM, Johnson NA, Rosen MP, Sternfeld B, Cedars MI, Reijo Pera RA. Genetic variants and environmental factors associated with hormonal markers of ovarian reserve in Caucasian and African American women. Hum Reprod (2012a), 27: Schuh-Huerta SM, Johnson NA, Rosen MP, Sternfeld B, Cedars MI, Reijo Pera RA. Genetic markers of ovarian follicle number and menopause in women of multiple ethnicities. Hum Genet (2012b) 131: Signorile P, Petraglia F, Baldi A. Anti-mullerian hormone is expressed by endometriosis tissues and induces cell cycle arrest and apoptosis in endometriosis cells. Journal of Experimental & Clinical Cancer Research 2014, 33:46 Skałba P Cygal A, Madej P, Dabkowska-Huc A, Sikora J, Martirosian G, Romanik M, Olszanecka-Glinianowicz M. Is the plasma anti-mullerian hormone (AMH) level associated with body weight and metabolic, and hormonal disturbances in women with and without polycystic ovary syndrome? European Journal of Obstetrics & Gynecology and Reproductive Biology 158 (2011) Streuli I, de Ziegler D, Gayet V, Santulli P, Bijaoui G, de Mouzon J and Chapron C. In women with endometriosis anti-mullerian hormone levels are decreased only in those with previous endometrioma surgery. Human Reproduction, Vol.27, No.11 pp , 2012 Su IH, Sammel MD, Freeman EW, Lin H, DeBlasis T, Gracia, C. Body size affects measures of ovarian reserve in late reproductive age women. Menopause. 2008; 15(5): Uncu G, Kasapoglu I, Ozerkan K, Seyhan A, Oral Yilmaztepe A, Ata B. Prospective assessment of the impact of endometriomas and their removal on ovarian reserve and determinants of the rate of decline in ovarian reserve. Hum Reprod Aug; 28(8): Wang J, Dicken C, Lustbader JW, and Tortoriello DV. Evidence for a Mullerian-inhibiting substance autocrine/paracrine system in adult human endometrium. Fertility and Sterility_ Vol. 91, No. 4, April 2009 Wallace WHB, Kelsey TW (2010) Human Ovarian Reserve from Conception to the Menopause. PLoS ONE 5(1): e87 159

160 Table 1. Distribution of patients AMH AFC FSH n Mean (SD) n Mean (SD) n Mean (SD) All Ethnicity White (Reference) Other White Black Indian Other Asian Pakistani Other ethnic Not disclosed Data not available Patients with BMI Normal (Reference) Underweight Overweight Obese Data not available Diagnosis Unexplained Mild tubal Severe tubal Mild male Severe male Endometriosis endometrioma Endometriosis + endometrioma

161 Table 2. Regression models for AMH AMH (Log) BMI included n=1952 BMI excluded n=2816 Β 95% CI P β 95% CI P Age , , age , , Ethnicity Other White , , Black , , Indian , , Other Asian , , Pakistani , , Other ethnic , , Not disclosed , , BMI Underweight , Overweight , Obese , Diagnosis Unexplained , , Mild tubal , , Severe tubal , , Mild male , , Severe male , , Endometriosis , , Endometrioma , , _cons , ,

162 Table 3. Regression models for AFC AFC (Log) BMI Included n=1589 BMI Excluded n=1810 Β 95% CI P Β 95% CI P Age , , age , , Ethnicity Other White , , Black , , Indian , , Other Asian , , Pakistani , , Other ethnic , , Not disclosed , , BMI Underweight , Overweight , Obese , Diagnosis Unexplained , , Mild tubal , , Severe tubal , , Mild male , , Severe male , , Endometriosis , , Endometrioma , , _cons , ,

163 Table 4. Regression models for FSH FSH (Log) BMI Included n=1772 BMI Excluded n=2343 Β 95% CI P Β 95% CI P age , , age , , Ethnicity Other White , , Black , , Indian , , Other Asian , , Pakistani , , Other ethnic , , Not disclosed , , BMI Underweight , Overweight , Obese , Diagnosis Unexplained , , Mild tubal , , Severe tubal , , Mild male , , Severe male , , Endometriosis , , Endometrioma , , _cons , ,

164 Table 5. Distribution of patient characteristics by availability of data on BMI The number of observations and mean (SD) of the markers of ovarian reserve (Age, AMH, AFC and FSH) described according to an availability of data on BMI. BMI (+) BMI (-) Total n Mean (SD) n Mean (SD) n Mean (SD) Age AMH AFC FSH *BMI (+): Record of BMI available *BMI (-): Record of BMI not available *Total: All available observations 164

165 Table 6. Distribution of ethnicity by availability of data on BMI Distribution of the number of observations by ethnicity and availability of data on BMI AMH AFC FSH BMI (+) BMI (-) Total BMI (+) BMI (-) Total BMI (+) BMI (-) Total White 1, ,833 1, ,222 1, ,556 Other White Black Indian Other Asian Pakistani Other ethnic Not disclosed Data not available Total 1, ,880 1, ,810 1, ,377 *BMI (+): Record of BMI available *BMI (-): Record of BMI not available *Total: All available observations 165

166 Table 7. Distribution of diagnosis by availability of data on BMI Distribution of number of observations in each diagnosis group tabulated by availability of data on BMI. AMH AFC FSH BMI (+) BMI (-) Total BMI (+) BMI (-) Total BMI (+) BMI (-) Total Unexplained Mild tubal Severe tubal Mild male Severe male Endometriosis endometrioma Endometriosis + endometrioma *BMI (+): Record of BMI available *BMI (-): Record of BMI not available *Total: All available observations 166

167 THE EFFECT OF SALPINGECTOMY, OVARIAN CYSTECTOMY AND UNILATERAL SALPINGOOPHERECTOMY ON OVARIAN RESERVE Oybek Rustamov, Monica Krishnan, Stephen A. Roberts, Cheryl Fitzgerald To be submitted to: Gynecological Surgery

168 Title Effect of salpingectomy, ovarian cystectomy and unilateral salpingooopherectomy on ovarian reserve. Authors Oybek Rustamov a, Monica Krishnan b, Stephen A. Roberts c, Cheryl Fitzgerald a, Institutions a Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH, UK; b Manchester Royal Infirmary, Central Manchester University Hospitals NHS Foundation Trust, Manchester M13 9WL, UK; c Centre for Biostatistics, Institute of Population Health, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 9PL, UK; Corresponding author & reprint requests Dr. Oybek Rustamov, Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH Grants or fellowships: No funding was sought for this study Disclosure summary: There were no potential conflicts of interest. Clinical Trial registration number: Not applicable. Word count: 2904 Acknowledgement The authors would like to thank colleagues Dr Greg Horne (Senior Clinical Embryologist), Ann Hinchliffe (Clinical Biochemistry Department) and Helen Shackleton (Information Operations Manager) for their help in obtaining datasets for the study. 168

169 Declaration of authors roles OR prepared the dataset, conducted statistical analysis and prepared all versions of the manuscript. MK assisted in data extraction, contributed in discussion and the review of the manuscript. SR and CF oversaw and supervised preparation of dataset, statistical analysis, contributed in discussion and reviewed all versions of the manuscript. 169

170 ABSTRACT Objective To estimate the effect of salpingectomy, ovarian cystectomy and unilateral salpingo-oopherectomy on ovarian reserve. Design Single centre retrospective cross-sectional study. Setting Women referred to secondary and tertiary level referral centre for management of infertility. Participants A total of 3179 patients were included in the study. The AMH measurements of 66 women were excluded due to haemolysed samples or delay in processing the samples, leaving 3113 women for analysis. There were 138 women who had unilateral or bilateral salpingectomy, 36 women with history of unilateral salpingo-oopherectomy, 41 women with history of cystectomy for ovarian cysts that other than endometrioma and 40 women had cystectomy for endometrioma. Interventions Serum AMH, AFC and basal FSH measurements. Main outcome measure Serum AMH, basal serum FSH and basal AFC measurements. Results The analysis did not find any significant differences in AMH (9%; p=0.33), AFC (-2%; p=0.59) and FSH (-14%; p=0.21) measurements between women with a history of salpingectomy and those without history of surgery. Women with history of unilateral salpingo-oopherectomy were found to have significantly lower AMH (-54%; p=0.001) and AFC (-28%; p=0.34) and increased FSH (14%; p=0.06). The study did not find any significant 170

171 association between a previous history of ovarian cystectomy that was for conditions other than endometrioma and AMH (7%; p=0.62), AFC (13%; p=0.18) or FSH. (11%; p=0.16). The analysis of the effect of ovarian cystectomy for endometrioma showed that women with history of surgery had around 66% lower AMH (p=0.002). Surgery for endometrioma did not significantly affect AFC (14%; p=0.22) or FSH (10%; p=0.28). Conclusions Salpingo-oopherectomy and ovarian cystectomy for endometrioma have a significant detrimental impact on ovarian reserve. Neither salpingectomy nor ovarian cystectomy for cysts other than endometrioma has an appreciable effect on ovarian reserve. Key Words Salpingectomy, Ovarian cystectomy, Salpingo-oopherectomy, ovarian reserve, AMH, AFC, FSH. 171

172 INTRODUCTION Human ovarian reserve is determined by the size of oocyte pool at birth and decline in the oocyte numbers thereafter. Both of these processes are largely under the influence of genetic factors and to date no effective interventions are available to improve physiological ovarian reserve (Shuh- Huerta et al 2012). However, various other environmental, pathological and iatrogenic factors appear to play a role in the determination of ovarian reserve and consequently it may be influenced either directly or indirectly. Evidently, the use of chemotherapeutic agents, certain radio-therapeutic modalities and surgical interventions that damage ovarian parenchyma can cause substantial damage to ovarian reserve (Nielsen et al 2013, Somigliana et al 2012). Estimation of the effect of each of these interventions is of paramount importance in ascertainment of lesser ootoxic treatment modalities and safer surgical methods. Age is the main determinant of the number of non-growing follicles, accounting for 84% of its variation, and used as marker of ovarian reserve (Hansen et al 2008). However biomarkers that allow direct assessment of the dynamics of growing follicles, AMH and AFC, may provide more accurate estimation of ovarian reserve. Although these markers only reflect folliculogenesis of already recruited growing follicles, there appears to be a good correlation between their measurements and histologically determined total ovarian reserve (Hansen et al 2011). Thus the biomarkers can effectively be utilized for estimation of the effect of above adverse factors on the primordial oocyte pool. Surgical interventions that lead to disruption of the blood supply to ovaries or involve direct damage to ovarian tissue may be expected to lead to a reduction in the primordial follicle pool. Indeed, a number of studies have reported an association between surgical interventions to ovaries and reduction in ovarian reserve (Somigliana et al 2012). However given both underlying disease and surgery may affect ovarian reserve, disentanglement of the individual effects of these factors may be challenging and requires robust research methodology. Here we present a study that intended to estimate the effect of tubal and ovarian surgery on ovarian reserve independent of underlying disease. 172

173 METHODS The effect of salpingectomy, ovarian cystectomy and unilateral salpingooopherectomy on ovarian reserve were studied using serum AMH, AFC and FSH measurements in a large cross sectional study. Population All women between the ages of 20 to 45 who were referred to the Women s Outpatient Department (WOP) and the Reproductive Medicine Department (RMD) of Central Manchester University Hospitals NHS Foundation Trust for management of infertility between 1 September 2008 and 16 November 2010 and had an AMH measurement using the DSL assay (DSL, Active MIS/AMH ELISA; Diagnostic Systems Laboratories, Webster, Texas) were included. We excluded patients referred for fertility preservation (e.g. prior to, or after. treatment for a malignant disorder) and those with a diagnosis of polycystic ovaries (PCO) on transvaginal ultrasound scan which was defined as volume of one or both ovaries more than 10ml. Patients with haemolysed AMH and/or FSH samples were not included in the analysis of these markers. Non-smoking is an essential criteria for investigation prior to assisted conception and therefore, to our best knowledge, our population consisted of non-smokers. Measurement of AMH Blood samples for AMH were taken without regard to the day of women s menstrual cycle and serum samples were separated within two hours of venipuncture in the Biochemistry laboratory of our hospital. All samples were processed strictly according to the manufacturer s recommendations and frozen at -20C until analysed in batches using the enzymatically amplified twosite immunoassay (DSL, Active MIS/AMH ELISA; Diagnostic Systems Laboratories, Webster, Texas). The working range of the assay was up to 100pmol/L and a minimum detection limit was 0.63pmol/L. The intra-assay coefficient of variation (CV) (n=16) was 3.9% (at 10pmol/l) and 2.9% (at 56pmol/l). The inter-assay CV (n=60) was 4.7% (at 10pmol/l) and 4.9% (at 56pmol/l). In patients with repeated AMH measurements, the first AMH of the patients were selected. 173

174 Measurement of FSH Patients had measurement of basal FSH, LH and oestradiol levels (E2) during the early follicular phase (Day 2-5) of their menstrual cycle as a part of their initial work up. Blood samples were transported to the Biochemistry Laboratory within two hours of venipuncture for sample processing and analysis. Specific immunoassay kits (Cobas, Roche Diagnostics, Mannheim, Germany) and an autoanalyser platform was used (Roche Modular Analytics E170, Roche, USA) for analysis of FSH. The intra-assay CV was 6.0% and inter-assay CV was 6.8%. The FSH measurements in the samples with high E2 levels (>250pmol/L) were excluded from the analysis given these samples are likely to have been taken outside of early follicular phase of menstrual cycle. In patients with repeated FSH measurements, measurements conducted on the same day as first AMH were selected. If the patient did not have FSH measurement on the day of AMH sampling, the measurement with the closest date to first AMH sample was selected. Measurement of AFC Measurement of AFC is conducted in patients referred for assisted conception during their initial work up. Our department uses a stringent protocol for the assessment of AFC and qualified radiographers who have undergone specific training on measurement of AFC. The methodology consists of counting of all antral follicles measuring 2-6mm in longitudinal and transverse cross sections of both ovaries using transvaginal ultrasound scanning at early follicular phase (Day 0-5) of the menstrual cycle. The AFC measurement with the closest date to first AMH sample was selected. Data collection Data was extracted from electronic clinical data management systems and from information held in written hospital notes for each patient. Data on AMH and FSH measurements were obtained from the Biochemistry Department and validated by checking the results documented in the hospital case notes of randomly selected 50 patients against the results obtained from electronic clinical data management system (Clinical Workstation) finding 100% concordance. Information on AFC, BMI, the causes of infertility, the duration of infertility, the history of reproductive pathology and reproductive 174

175 surgery were obtained from the hospital case notes. The ethnicity of the patients was established using a patient questionnaire and data were extracted from the hospital database for the patient demographics (PAS). Definitions and groups First the datasets were merged using a unique patient identifier (hospital number). Validation of the merger using additional patient identifiers (NHS number, name, date of birth) revealed existence of duplicate hospital numbers in patients transferred from secondary care infertility services of our hospital to IVF Department. We established that, in our datasets, combination of the patient s first name, surname and date of birth in a continuous string variable could be used as a unique identifier. Hence, we used this identifier to merge all datasets achieving a robust merger of all independent datasets into a combined final dataset. Following creation of an anonymised a unique study number for each patient, all patient identifiers were dropped and the anonymised combined dataset was used for the analysis. Body mass index (BMI) of patients was categorized using standard NHS cut-off reference ranges: Underweight (<18.5), Normal ( ), Overweight ( ) and Obese (30-40) (The Office for National Statistics, 2011.). Causes of infertility were established by searching the hospital notes including the referral letters, clinical notes and letters generated following clinic consultations. Patients with history of bilateral tubal block, which was confirmed by laparoscopic dye test, and patients with history of bilateral salpingectomy were categorized as having severe tubal factor infertility. Patients with unilateral tubal patency or unilateral salpingectomy were categorized as having mild tubal factor infertility. Severe male factor infertility was defined as azoospermia or severe oligospermia (<1mln sperm sample) Patients with abnormal sperm count but do not meet above criteria were classified as having mild male factor infertility. Patients with reproductive surgery were categorized as having history of salpingectomy, cystectomy for endometrioma, cystectomy for ovarian cysts other than endometrioma or unilateral salpingo-oopherectomy. First measurement of AMH, AFC and FSH following surgery was selected for the study. 175

176 Statistical analysis A multivariable regression model that included age, ethnicity, BMI, endometriosis, presence of endometrioma, the causes of infertility, tubal and ovarian surgery was fitted for each of the ovarian reserve markers: AMH, AFC and FSH. Difference between the groups were considered significant at p Preliminary analysis of AMH, AFC and FSH indicated that logarithmically transformed values with a quadratic age term provided adequate fits The precise age on the day measurement of each of the marker of ovarian reserve (AMH, AFC and FSH) was included in the model as a quadratic function following centering to 30 years of age. Interactions between all explanatory variables were tested at a significance level of We observed significant interaction between BMI and other covariates. This may be due to biological complexity in the relationship of BMI and other factors (e.g. ethnicity) in determination of ovarian reserve. However, given data on BMI was not available in considerable number of patients, the observed interactions may be due to limitation of our dataset. Therefore, in order to assist in interpretation of the results, analyses with and without BMI in the models were conducted. RESULTS In total 3179 patients were included in the study. The AMH measurements of 66 women were excluded due to haemolysed samples or delay in processing the samples, leaving 3113 women for analysis of patients had measurement of AFC and 2580 had FSH samples that met inclusion criteria The mean age, AMH, AFC and FSH of patients were 32.8±4.5, 17.3±14.8, 13.9±6.2, 8.0±7.5 respectively. There were 138 women who had unilateral or bilateral salpingectomy, 36 women with history of unilateral salpingo-oopherectomy, 41 women with history of cystectomy for ovarian cysts that other than endometrioma and 40 women had cystectomy for endometrioma (Table 1). The results of regression analysis on the effect of reproductive surgery on AMH, AFC and FSH measurements are shown in Table 2. The analysis did not find any significant differences in AMH (9%; p=0.33), AFC (-2%; p=0.59) and FSH (-14%; p=0.21) measurements in women with history of salpingectomy compared to women without history of 176

177 surgery and we did not observe marked change in the estimates in a smaller subset where BMI was included in the model (Table 2). Women with history of unilateral salpingo-oopherectomy were found to have significantly lower AMH (-54%; p=0.001) and AFC (-28%; p=0.34) and increased FSH (14%; p=0.06) measurements, where effect on AMH reached the level of statistical significance. Similarly the analysis of the model that included BMI showed significantly lower AMH and AFC and higher FSH measurements in surgery group where both AMH and FSH analysis were statistically significant (Table 2). The study did not find a significant association between previous history of ovarian cystectomy that was for disease other than endometrioma and measurement of AMH (7%; p=0.62), AFC (13%; p=0.18) or FSH. (11%; p=0.16), which did not change noticeably following adding BMI in the model (Table 2). The analysis of the effect of ovarian cystectomy for endometrioma showed that women with history of surgery had around 66% lower AMH (p=0.002) measurements. The effect of surgery for endometrioma was not significant in assessment of AFC (14%; p=0.22) and FSH (10%; p=0.28). However in the model with BMI, association of the surgery with both AMH (- 64%; p=0.005) and FSH (24%; p=0.015) were found to be significant (Table 2). DISUCUSSION Salpingectomy The blood supply to human ovaries is maintained by the direct branches of aorta, ovarian arteries, which form anastomoses with ovarian and tubal branch of uterine arteries in mesovarium and mesosalpynx. In salpingectomy often tubal branches of uterine arteries are excised alongside mesosalpynx and hence it is believed disruption to blood supply to ovaries may lead to a reduction of ovarian reserve. However, in our study we did not observe an appreciable association between salpingectomy and any of the biomarkers of ovarian reserve suggesting this surgery does not appreciably affect ovarian reserve. These findings are supported by study that assessed the effect of tubal 177

178 dissection to AMH, AFC, FSH levels (n=49) using longitudinal data (Erkan et al 2012). There were no differences between preoperative and 3 month postoperative measurements with median AMH (1.5 vs. 1.4; p=0.07), AFC ( vs ; p=0.09), FSH ( vs ; p=0.10). da Silva et al assessed the effect of tubal ligation (n=52) in longer term postoperative period (1 year) and reported that median AMH (1.43, IQR vs. and 1.30 IQR ; p=0.23) and mean AFC ( 8, IQR vs. 11, IQR 7-15; p=0.12) measurements did not change significantly. Our results and on other published evidence, suggest that salpingectomy or tubal division does not have an adverse effect to ovarian reserve. Unilateral salpingo-oopherectomy Although salpingo-oopherectomy is rare in women of reproductive age, significant ovarian pathologies and acute diseases such as ovarian torsion may necessitate unilateral salpingo-oopherectomy. There is a plausible causative relationship between this surgery and ovarian reserve, although to our knowledge there is no previous published evidence. We found that women with a history of unilateral salpingo-oopherectomy have significantly lower AMH (-54%) and higher FSH (13%) measurements suggesting the surgery has considerable negative impact to ovarian reserve. Important clinical question in this clinical scenario is Do these patients have comparable reproductive lifespan or experience accelerated loss of oocytes resulting premature loss of fertility?, as this would allow appropriate pre-operative counseling of patients regarding long term effect of the surgery to fertility and age at menopause. Considering, our data had relatively small number of patients with a history of salpingo-oopherectomy we were not able to obtain reliable estimates on agerelated decline of ovarian reserve in this study. We suggest that studies with larger number of patients, preferably using longitudinal data, should address this research question. Ovarian cystectomy In women with a history of ovarian cystectomy for ovarian cysts other than those due to endometrioma, we did not observe any significant association between the surgery and markers of ovarian reserve. However, women that had ovarian cystectomy for endometrioma appear to have 178

179 significantly lower AMH (-66%) measurements compared to those without history of surgery. During the last few years a number of studies have assessed the effect of cystectomy on AMH levels in patients with endometrioma (Chang et al 2010; Erkan et al 2010; Lee et al 2011). The studies have been summarised by a recent systematic review, which concluded that cystectomy results in damage to ovarian reserve (Somigliana et al 2012). Further studies evaluated the mechanism of damage and these suggest that coagulation for purpose of hemostasis as well as stripping of the cyst wall may cause direct damage to ovarian reserve. Sonmezer et al compared the effect of diathermy coagulation (n=15) for hemostasis compared to use of hemostatic matrix (n=13) in a randomized controlled trial and reported that use of diathermy coagulation is associated with significantly lower AMH measurements (1.64 ± 0.93 vs ± 1.49 ng/ml) in the first postoperative month. Similarly, stripping of the cyst wall also appears to have detrimental effect of ovarian reserve due to inadvertent removal of ovarian tissue (Donnez et al 1996). Using histological data, Roman et al. demonstrated that normal ovarian tissue was removed in 97% specimens of surgically removed endometriomata (Roman et al 2010). Furthermore, it appears that ovarian cortex containing endometrioma appears to have significantly reduced density compared to normal ovarian cortex and therefore loss of oocyte containing normal ovarian cortex may be unavoidable in cystectomy for endometrioma (Sanchez et al 2014). Matsuzaki et al conducted histological assessment of cystectomy specimens and found that normal ovarian tissue adjacent to cyst wall was found in 58% (71/121) of patients with endometrioma, whereas normal ovarian tissue was excised in 5.4% (3/56) following cystectomy for other benign cyst (Matsuzaki et al 2008). Similarly, in our study women with a history of cystectomy for endometrioma had significantly lower AMH measurements, whilst those had cystectomy for other benign cysts do not appear to have lower AMH measurements. In view of our findings and other published research evidence, it seems clear that cystectomy for endometrioma results in significant reduction in ovarian reserve and women undergoing surgery should be counseled regarding the adverse effect of surgery. 179

180 Strengths and Limitations The published studies have used longitudinal data comparing biomarkers before and after cystectomy and provide reliable estimates on the effect of the intervention on ovarian reserve. However data on the effect of salpingectomy and unilateral salpingoophorectomy is lacking. In addition to reevaluation of the effect of cystectomy, this is study has assessed the impact of salpingectomy and unilateral salpingoophorectomy on the markers of ovarian reserve. In contrast to published studies this study employed analysis of cross sectional data. Given a robust adjustment for all relevant factors has been conducted, our analysis of the cross sectional data should provide reliable estimates of the effects of various intervention on the markers of ovarian reserve. Furthermore, the effect of surgery on all the main biomarkers of ovarian reserve has been assessed which improves our understanding of the clinical value of each test in the assessment of patients with history of tubal or ovarian surgery. In addition, the analyses adjusted for other relevant factors that may affect ovarian reserve. In patients with history of cystectomy for endometrioma, we estimated independent effects of pathology and surgery providing important data for preoperative counseling. It is important to note that, the study evaluated The effect of surgery using retrospective data which has limitations due variation in recording of surgical history and missing data. In addition, given BMI results for around one third of patients were not available we were not able to fully explore the effect of BMI. However, data on the analyses with and without BMI in the model have been provided to evaluate the effect of this factor. The study employed the data obtained using first generation DSL AMH assay which is no longer in use. However, the paper describes the effects of the interventions in percentage terms and therefore the results are interpretable in any AMH assay measurement Important to note although the effects are significant in population level, there is considerable variation between individuals which is evident from the fact there is overlap between median and interquartile ranges of the groups (Figure 1). This indicates that clinicians should exercise caution in predicting the effect of surgery to ovarian reserve of individual patients. Nevertheless, given I used a robust methodology for data extraction and conducted careful analysis I think that the study provides fairly reliable estimates on the effect of surgery to ovarian reserve. 180

181 CONCLUSION This multivariable regression analysis of retrospectively collected crosssectional data suggests that, neither salpingectomy nor ovarian cystectomy for cysts other than endometrioma has an appreciable effect on ovarian reserve determined by AMH, AFC and FSH. In contrast, salpingoophorectomy and ovarian cystectomy for endometrioma have a significant detrimental impact to ovarian reserve. On the basis of findings of this study and other published studies women undergoing reproductive should be counseled with regards to the effect of the surgery on their ovarian reserve. 181

182 References Biacchiardi CP, Piane LD, Camanni M, Deltetto F, Delpiano EM, Marchino GL, et al. Laparoscopic stripping of endometriomas negatively affects ovarian follicular reserve even if performed by experienced surgeons. Reprod Biomed Online 2011;23: Chang HJ, Han SH, Lee JR, Jee BC, Lee BI, Suh CS, et al. Impact of laparoscopic cystectomy on ovarian reserve: serial changes of serum anti- Mullerian hormone levels. Fertil Steril 2010;94: Dogan E, Ulukus EC, Okyay E, Ertugrul C, Saygili U, Koyuncuoglu M. Retrospective analysis of follicle loss after laparoscopic excision of endometrioma compared with benign nonendometriotic ovarian cysts. Int J Gynaecol Obstet 2011;114: Ercan CM, Sakinci M, Duru NK, Alanbay I, Karasahin KE, Baser I (2010). Antimullerian hormone levels after laparoscopic endometrioma stripping surgery. Gynecol Endocrinol 2010;26: Ercan CM, Duru NK, Karasahin KE, Coksuer H, Dede M, Baser I (2011). Ultrasonographic evaluation and anti-mullerian hormone levels after laparoscopic stripping of unilateral endometriomas. Eur J Obstet Gynecol Reprod Biol. 2011;158: Hansen KR, Hodnett GM, Knowlton N, Craig LB. Correlation of ovarian reserve tests with histologically determined primordial follicle number. Fertil Steril. 2011;95: Hachisuga T, Kawarabayashi T. Histopathological analysis of laparoscopically treated ovarian endometriotic cysts with special reference to loss of follicles. Hum Reprod 2002;17: Hirokawa W, Iwase A, Goto M, Takikawa S, Nagatomo Y, Nakahara T, et al. The post-operative decline in serum anti-mullerian hormone correlates with the bilaterality and severity of endometriosis. Hum Reprod 2011;26: Hwu YM, Wu FS, Li SH, Sun FJ, Lin MH, Lee RK. The impact of endometrioma and laparoscopic cystectomy on serum anti-mullerian hormone levels. Reprod Biol Endocrinol 2011;9:80. Iwase A, Hirokawa W, Goto M, Takikawa S, Nagatomo Y, Nakahara T, et al. Serum anti-mullerian hormone level is a useful marker for evaluating the impact of laparoscopic cystectomy on ovarian reserve. Fertil Steril 2010;94: Kitajima M, Khan KN, Hiraki K, Inoue T, Fujishita A, Masuzaki H. Changes in serum anti-mullerian hormone levels may predict damage to residual normal ovarian tissue after laparoscopic surgery for women with ovarian endometrioma. Fertil Steril 2011;95: e1. Kitajima M, Defr_ere S, Dolmans MM, Colette S, Squifflet J, van 182

183 Langendonckt A, et al. Endometriomas as a possible cause of reduced ovarian reserve in women with endometriosis. Fertil Steril 2011;96: Lee DY, Young Kim N, Jae Kim M, Yoon BK, Choi D. Effects of laparoscopic surgery on serum anti-m ullerian hormone levels in reproductiveaged women with endometrioma. Gynecol Endocrinol 2011;27: Matsouzaki S, Houlle C, Darcha S, Pouly JL, Mage G, Canis M. Analysis of risk factors for the removal of normal ovarian tissue during laparoscopic cystectomy for ovarian endometriosis. Hum Reprod 2009; 24: Muzii L, Bianchi A, Croc_e C, Manci N, Panici PB. Laparoscopic excision of ovarian cysts: is the stripping technique a tissue-sparing procedure? Fertil Steril 2002;77: Office for National Statistics (ONS), Social Trends 41, Health, Roman H, Tarta O, Pura I, Opris I, Bourdel N, Marpeau L, et al. Direct proportional relationship between endometrioma size and ovarian parenchyma inadvertently removed during cystectomy, and its implication on the management of enlarged endometriomas. Hum Reprod 2010;25: Romualdi D, Franco Zannoni G, Lanzone A, Selvaggi L, Tagliaferri V, Gaetano Vellone V, et al. Follicular loss in endoscopic surgery for ovarian endometriosis: quantitative and qualitative observations. Fertil Steril 2011;96: Rustamov O, Smith A, Roberts SA, Yates AP, Fitzgerald C, Krishnan M, Nardo LG, Pemberton PW. Anti-Mullerian hormone: poor assay reproducibility in a large cohort of subjects suggests sample instability. Hum Reprod 2012; 27: Rustamov O, Smith A, Roberts SA, Yates AP, Fitzgerald C, Krishnan M, Nardo LG, Pemberton PW. Reply: reproducibility of AMH. Hum Reprod 2012;27: Sanchez A, P. Viganò P, Somigliana E, Panina-Bordignon P. Vercellini and Candiani M. The distinguishing cellular and molecular features of the endometriotic ovarian cyst: from pathophysiology to the potential endometrioma-mediated damage to the ovary, Hum. Reprod. Update (March/April 2014) Shi J, Leng J, Cui Q, Lang J. Follicle loss after laparoscopic treatment of ovarian endometriotic cysts. Int J Gynaecol Obstet 2011;115: Tsolakidis D, Pados G, Vavilis D, Athanatos D, Tsalikis T, Giannakou A, et al. The impact on ovarian reserve after laparoscopic ovarian cystectomy versus three-stage management in patients with endometriomas: a prospective randomized study. Fertil Steril 2010;94:71 7. Vicino M, Scioscia M, Resta L, Marzullo A, Ceci O, Selvaggi LE. Fibrotic tissue in the endometrioma capsule: surgical and physiopathologic considerations from histologic findings. Fertil Steril 2009;91(4 Suppl):

184 Figure 1. Box plots of AMH by various groups. Upper panel shows the raw data and the lower panel the AMH measurement (in pmol/l) adjusted for age, ethnicity, BMI, causes of infertility, endometriosis, endometrioma and surgery. Groups (left to right): 1) Endometrioma without history of cystectomy (endoma-no surg), 2) Cystectomy for endometrioma (endoma+surg), 3) Endometriosis without endometrioma (endsisonly), 4) Without endometriosis or any surgery (No end+no surg), 5) Oopherectomy (oe), 6) Cystectomy for cyst other than those for endometrioma (other cyst), 7) Salpingectomy (se). 184

185 Table1. Distribution of patients BMI excluded BMI Included Age AMH AFC FSH AMH AFC FSH Mean (SD) N Mean n Mean (SD) N Mean (SD) n n N Non-surgery 32.8± ± ± ± Oophorectomy 32.4± ± ± ± Salpingectomy 33.1± ± ± ± Cystectomy Other 33.6± ± ± ± Cystectomy Endometrioma 32.7± ± ± ±

186 Table 2. Multivariable regression analysis. Adjusted for age, ethnicity causes of infertility, endometriosis (without endometrioma), endometrioma and reproductive surgery. BMI(+) BMI(-) N Coeff. 95% CI P N Coeff 95% CI P Oophorectomy AMH , , AFC , , FSH , , Salpingectomy AMH , , AFC , , FSH , , Cystectomy Other AMH , , AFC , , FSH , , Cystectomy Endometrioma AMH , , AFC , , FSH , ,

187 ASSESSMENT OF DETERMINANTS OF OOCYTE NUMBER USING RETROSPECTIVE DATA ON IVF CYCLES AND EXPLORATIVE STUDY OF THE POTENTIAL FOR OPTIMIZATION OF AMH- TAILORED STRATIFICATION OF CONTROLLED OVARIAN HYPERSTIMULATION Oybek Rustamov, Cheryl Fitzgerald, Stephen A. Roberts 6 187

188 Title Assessment of determinants of oocyte number using large retrospective data on IVF cycles and explorative study of the potential for optimization of AMH-tailored stratification of controlled ovarian stimulation Authors Oybek Rustamov a, Cheryl Fitzgerald a, Stephen A. Roberts c Institutions a Department of Reproductive Medicine, St Mary s Hospital, Central Manchester University Hospital NHS Foundation Trust, Manchester Academic Health Science Centre (MAHSC), Manchester M13 0JH, UK; b Centre for Biostatistics, Institute of Population Health, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 9PL, UK; Word count: 7520 Grants or fellowships: No funding was sought for this study. Disclosure summary: There were no potential conflicts of interest. Clinical Trial registration number: Not applicable. Acknowledgement Authors would like to thank Dr Monica Krishnan (Foundation Trainee, Manchester Royal Infirmary) for her assistance in data extraction. We would also like to thank colleagues Dr Greg Horne (Senior Clinical Embryologist), Ann Hinchliffe (Clinical Biochemistry Department) and Helen Shackleton (Information Operations Manager) for their help in obtaining datasets for the study. 188

189 Declaration of authors roles OR prepared the study protocol, prepared the dataset, conducted statistical analysis and prepared all versions of the manuscript. SR and CF oversaw and supervised preparation of dataset, statistical analysis, contributed to the discussion and reviewed all versions of the manuscript. 189

190 ABSTRACT Objectives 1) To determine the effect of age, AMH, AFC, causes of infertility and treatment interventions on oocyte yield 2) To explore potential for optimization of AMH-tailored individualisation of ovarian stimulation Design Retrospective cross sectional study using multivariable regression analysis. First, the effect of a set of plausible factors that may affect the outcomes have been established; including assessment of the effect of age, AMH, AFC, causes of infertility, attempt of IVF ICSI cycle, COH protocol changes, gonadotrophin preparations, operator for oocyte recovery, pituitary desensitisation regime and initial daily dose of gonadotrophins. Then, the regression models that examined the effect of gonadotrophin dose and regime categories on total and mature oocyte numbers have been developed. Setting Tertiary referral centre for management of infertility, St Mary s Hospital, Central Manchester University Hospitals NHS Foundation Trust. Participants Women without ultrasound features of polycystic ovaries who underwent IVF ICSI cycle using pituitary desensitisation with GnRH long agonist or GnRH antagonist regimes and had previous measurement of AMH with the DSL assay. In total of 1847 IVF or ICSI cycles of 1428 patients met the inclusion criteria for the study. AMH measurements of all cycles and AFC measurements for 1671 cycles (n=1289 patients) were available. In the analysis of total oocytes 1653 cycles were included and the analysis of metaphase II oocytes comprised of 1101 ICSI cycles. Interventions None (observational study) 190

191 Main outcome measures Total oocyte number. Metaphase II oocyte number Results After adjustment for all the above factors, age remained a negative predictor of oocyte yield, whereas we observed a gradual and significant increase in oocyte number with increasing AMH and AFC values, suggesting all these markers display an independent association with oocyte yield. Compared to 1 st IVF cycles those with 2 nd (8%; p=0.01) and particularly 3 rd attempt (24%; p=0.001) had considerably higher total oocytes. The effect of attempt on mature oocyte yield was not significant (p=0.45). Similarly there was significant between-operator variability in total oocyte number when oocyte recovery practitioners were compared (p=0.0005). However, the effect of oocyte recovery practitioner on mature oocyte yield did not reach statistical significance (p=0.058). Comparison of the effect of gonadotrophin type showed that rfsh was associated with higher total oocyte yield compared to that of HMG (p=0.008), although the numbers of mature oocytes were not significantly different between the groups (p=0.26). After adjustment for all above factors, compared to a reference group (Agonist with IU hmg/rfsh), none of the regime and dose categories provided higher total oocyte yield and Antagonist with IU hmg/rfsh (-36%; p=0.0005) provided significantly less total oocyte. With regards to the mature oocyte yield, Antagonist with IU rfsh/hmg (43%, p=0.05) and Antagonist 375 IU rfsh/hmg (47%, p=0.02) were associated with significantly higher oocyte number compared to that of above reference group. This implies that compared to long Agonist down regulation, Antagonist regime is associated with higher mature oocyte yield. Following adjustment for all above variables, we did not observe significant increase in oocyte number with increasing gonadotrophin dose categories. 191

192 Conclusions Given there was no expected increase in oocyte number with increasing gonadotrophin dose categories; we believe there may not be significant direct dose-response effect. Consequently strict protocols for tailoring the initial dose of gonadotrophins may not necessarily improve ovarian performance in IVF treatment. It is important to note, our COS protocols instructed the use of cycle monitoring with ultrasound follicle tracking and oestradiol levels and corresponding adjustment of daily dose of gonadotrophins during ovarian stimulation, which may undermine the effect of initial dose of gonadotrophins. However, further analysis with adjustment for the total gonadotrophin dose and dose adjustment during the stimulation did not have significant impact on oocyte yield. Nevertheless further time series regression analysis, with full parameters of cycle monitoring and the dose adjustments in the model, should be conducted in order to ascertain the role of AMH in tailoring the dose of gonadotrophins in cycles of IVF. Key Words Ovarian reserve; AMH; AFC; IVF; Controlled ovarian stimulation; AMHtailored ovarian stimulation; Individualisation of ovarian stimulation 192

193 INTRODUCTION According to the HFEA, around 12% of IVF cycles in the UK are cancelled due to poor or excessive ovarian response in the UK, which highlights the importance of the provision of optimal ovarian stimulation in improving the outcomes (Kurinczuk et al 2010). Traditionally, patient s age and basal FSH measurements were used for the assessment of ovarian reserve, with subsequent tailoring of the initial dose of gonadotrophins and regime for pituitary desensitisation for controlled ovarian stimulation in IVF. Studies on the prognostic value of markers of ovarian reserve show that AMH and AFC are the best predictors of ovarian response in cycles of IVF (Broer et al 2011). Furthermore, unlike most other markers, AMH has potential discriminatory power due to significantly higher between-patient (CV 94%) variability compared to its within-patient (CV 28%) variation (Rustamov et al 2011), which allows stratification of patients into various degrees of (e.g. low, normal, high) ovarian reserve. Consequently, development of optimal ovarian stimulation protocol for each band of ovarian reserve using AMH may be feasible. Controlled ovarian stimulation (COS) based on tailoring the pituitary desensitisation and initial dose of gonadotrophins to AMH measurements is known under various names; individualisation of ovarian stimulation, AMHtailored stratification of COS, personalization of IVF are the most commonly used. This strategy is believed to be effective and has been widely recommended (Nelson et al 2013, Dewailly et al 2014, La Marca et al 2014). Although AMH based assessment of ovarian reserve, with pituitary down regulation in patients with extremes of ovarian reserve, may improve the outcomes of ovarian response compared to conventional ovarian stimulation protocols (Nelson et al 2009; Yates et al 2011), there is no robust data on AMH-tailored individualisation of ovarian stimulation. To establish individualisation of ovarian stimulation, the studies should ideally assess various pituitary desensitisation regimes and initial doses of gonadotrophins in patients across the full range of ovarian reserve. For instance in AMH-tailored individualisation of pituitary desensitisation regime, studies should evaluate the effect of both GnRH Agonist and GnRH Antagonist regimes for the groups for each band of AMH levels (e.g. low, normal, high), necessitating 6 comparison groups (Figure 1). In individualisation of the initial dose of 193

194 gonadotrophins, the groups of each band of AMH should be treated with the range of doses of gonadotrophins (e.g. low, moderate, high dose), which requires 9 treatment groups (Figure 2). Consequently, to evaluate the individualisation of both the stimulation regime and the initial dose of gonadotrophin across the full range of AMH measurements in a single study, ideally 18 comparison groups are needed. Indeed, the study should have a large enough sample to adjust for the confounders and obtain sufficient power for the estimates of each treatment group. In addition, assessment of ovarian reserve should be based on reliable AMH measurements with minimal sampleto-sample variation, which appears to be an issue at present (Rustamov et al 2013). Finally evidence on AMH-tailored individualisation of ovarian stimulation should, ideally, be based on randomized controlled trials, given in this context AMH is being used as a therapeutic intervention. At present there is no single RCT that assessed AMH-tailored individualisation of ovarian stimulation and most quoted research evidence appear to have been based on two retrospective studies (Nelson et al 2009, Yates et al 2011). Both studies display a number of methodological issues, including small sample size and centre-dependent or time-dependent selection of cohorts. Therefore, the role of confounding factors on the obtained estimates of these studies is unclear. The first study on AMH-tailored individualisation ovarian stimulation compared outcomes of the cohorts who had IVF cycles in two different IVF centers (Nelson et al 2009). In this case control study, the patients in the 1 st centre (n=370) had minimal tailoring of dose of gonadotrophins and were offered mainly GnRH agonist regime for pituitary desensitisation; except patients with very low AMH (<1.0pmol/L) who had GnRH antagonist regime. In patients undergoing treatment in the 2 nd centre (n=168), the daily dose of the gonadotrophins was tailored on the basis of AMH levels and GnRH antagonist based protocol employed for women with low (1-5 pmol/l) and high (>15 pmol/l) AMH levels, whereas patients with normal (5-15 pmol/l) AMH levels had standard long GnRH agonist regimen. In addition, the patients with very low AMH (<1.0 pmol/l) had modified natural cycle IVF treatment in 2 nd centre. The study reported that, the group that had significant tailoring of both mode and degree of stimulation to AMH levels (2 nd centre) had higher pregnancy rate and less cycle cancellation. However, given the methodological weaknesses, the findings of the study ought to be interpreted with caution. First, the study compared the outcomes of small number of 194

195 patients who had treatment in two different centers, suggesting that differences in the outcomes may be due to variation in the characteristics of patient populations and/or performance of two different centers. Moreover, both cohorts had some degree of tailoring of pituitary desensitisation regimens as well as the daily dose of gonadotrophins to AMH levels, suggesting estimation of the effect of AMH tailoring to the outcome of treatment may not be reliable. A subsequent study attempted to address the above issues by assessing a somewhat larger number of IVF cycles from the same fertility centre (Yates et al 2011). The study compared IVF outcomes of the cohorts that underwent ovarian stimulation using chronological age and serum FSH (n=346) with women that had AMH-tailored (n=423) treatment cycles (Yates et al 2011). The study found that the group that had AMH-tailored ovarian stimulation had significantly higher pregnancy rate, less cycle cancellation due to poor or excessive ovarian response and had significantly lower treatment costs. However, this study also has appreciable weaknesses given, that it was based on retrospective data that compared outcomes of treatment cycles that took place over two year period. During this period, apart from introduction of AMH-tailored stimulation protocols, other new interventions were introduced, particularly in the steps involved in embryo culture. Although the outcomes of the ovarian response to stimulation could have mainly been due to performance of the stimulation protocols, downstream outcomes such as clinical pregnancy rate may be associated with the introduction of new interventions in embryo culture techniques. Nevertheless, the study demonstrated that tailoring of ovarian stimulation protocol to AMH levels could reduce the incidence of cycle cancellation, OHSS and the cost of treatment, supporting the need for more robust studies on the use of AMH in the individualisation of ovarian stimulation in IVF. It appears, despite a lack of good quality evidence that AMH-tailored individualisation has been widely advocated and has been introduced in clinical practice in a number of fertility units. In the absence of good quality evidence, we decided to obtain more reliable estimates on the feasibility of AMH-tailored ovarian stimulation using more robust methodology. Availability of the data on a large cohort of patients with AMH measurements who, subsequently underwent IVF treatment cycles in a single centre may allow us to obtain more reliable estimates on the effectiveness of AMH-tailored COS. Furthermore due 195

196 to changes on COS protocol, combination of various regime and initial dose of gonadotrophin were used for patients in each band of ovarian reserve. This may facilitate development of predictive models for both regime and dose for the whole range of AMH measurements. In addition, as a part of the study we decided to establish the role of patient and treatment related factors in determination of ovarian response in cycle of IVF. I believe that understanding the effect of various factors on ovarian performance in COS will improve the methodology of the study and can be used as a guide for identification of confounders in future studies. The first step in such an analysis is to develop a statistical model to describe the relationship between ovarian response and patient and treatment factors. This can then be utilized to explore the effects of treatment on outcome and potentially to allow optimal treatments to be identified for given patient characteristics and ovarian reserve. METHODS Objective The objectives of the study were 1) to determine the effect of age, AMH, AFC, causes of infertility and treatment interventions on oocyte yield and 2) to explore potential for optimization of AMH-tailored individualisation of ovarian stimulation. Population Women of years of age undergoing ovarian stimulation for IVF ICSI treatment using their own eggs at the Reproductive Medicine Department of St Mary s Hospital, Manchester from 1 st October 2008 to 8 th August 2012 were included. Patients with previous AMH measurements using DSL assay were included and patients that had AMH measurement with only Gen II assay were excluded, given the observed issues with this assay (Rustamov et al 2012). The patients with ultrasound features of PCO, previous history of salpingectomy, ovarian cystectomy and/or unilateral salpingoophorectomy have been excluded from the analysis. Similarly, cycles with ovarian stimulation other than GnRH agonist long down regulation or Short GnRH antagonist cycles were not included in the study. 196

197 Dataset The dataset for the study was prepared using a protocol for the data extraction, management, linking and validation, which is described in Chapter 4. In short, first the data contained in clinical data management systems were obtained on patient demography, AMH measurements and IVF treatment cycles. Then, data not available in electronic format were collected from the patient case notes, which includes causes of infertility, previous history of reproductive surgery, AFC and folliculogram for monitoring of ovarian stimulation. Each dataset was downloaded in original Excel format into Stata 12 Data Management and Statistics Software (StataCorp LP, Texas, USA) and analysis datasets were prepared in Stata format. All IVF cycles commenced during the study period were identified and the combined study dataset was created by linking all datasets in cycle level using the anonymised patient identifiers and the dates of interventions. All steps of data handling have been recorded using Stata Do files to ensure reproducibility and provide a record of the data management process. Categorization of diagnosis Patients with history of unilateral tubal occlusion or unilateral salpingectomy were categorized as mild tubal factor infertility and patients with blocked tubes bilaterally or with history of bilateral salpingectomy were allocated to severe tubal disease. Severe male factor infertility was defined if the partner had azoospermia, surgical sperm extraction or severe oligospermia, which necessitated Multiple Ejaculation Resuspension and Centrifugation test (MERC) for assisted conception. Mild male factor was defined as abnormal sperm count that do not above meet criteria for severe male infertility. Diagnosis of endometriosis was based on a previous history of endometriosis confirmed using Laparoscopy. Diagnosis of endometrioma was established using transvaginal ultrasound scan prior to IVF treatment. In couples without a definite cause for infertility following investigation, the diagnosis was categorized as unexplained. Women with features of polycystic ovaries on transvaginal ultrasound were categorized as PCO and excluded from analyses. 197

198 Measurement of AMH and AFC AMH measurements were performed by the in-house laboratory, Clinical Assay Laboratory of Central Manchester NHS Foundation Trust, and the procedure for sample handling and analysis was based on the manufacturer s recommendations. Venous blood samples were taken without regard to the day of women s menstrual cycle and serum samples were separated within two hours of venipuncture. Samples were frozen at -20C until analysed in batches using the enzymatically amplified two-site immunoassay (DSL, Active MIS/AMH ELISA; Diagnostic Systems Laboratories, Webster, Texas). The intra-assay coefficient of variation (CV) (n=16) was 3.9% (at 10pmol/l) and 2.9% (at 56pmol/l). The inter-assay CV (n=60) was 4.7% (at 10pmol/l) and 4.9% (at 56pmol/l). Haemolysed samples were not included in the study. In patients with repeated AMH, the measurement closest to their IVF treatment cycle was selected. The working range of the assay was up to 100pmol/L and a minimum detection limit was 0.63pmol/L.The results with minimum detection limit were coded as 50% of the minimum detection limit (0.31 pmol/l) and the test results that are higher than the assay ranges were coded as 150% of the maximum range (150 pmol/l). In our department, the measurement of AFC is conducted as part of initial clinical investigation before first consultation with clinicians and prior to IVF cycle. Qualified radiographers performed the assessment of AFC during early follicular phase (Day 0-5) of menstrual cycle. The methodology of measurement of AFC consisted of the counting of all antral follicles measuring 2-6mm in longitudinal and transverse cross sections of both ovaries using transvaginal ultrasound scan. The AFC closest to the IVF cycle was selected for the analysis. Description of COS Protocols On the basis of their AMH measurement, patients were stratified into the treatment bands for ovarian stimulation using COS protocols. During the study two different COS protocols were used in our centre and in addition three minor modifications were made in the 2 nd protocol. Time periods, AMH bands, down regulation regimes, initial dose of gonadotrophins and adjustment of daily dose of gonadotrophins of the protocols are described in Table 1. Similarly the management of excessive ovarian response was tailored to 198

199 pretreatment AMH measurements, although mainly based on the results of oestradiol and scan monitoring the cycle stimulation (Table 2). Assessment of transvaginal ultrasound guided follicle tracking and serum oestradiol levels in specific days of the stimulation were used for monitoring of COS (Table 2). The criteria for the cycle cancellation for poor ovarian response were same across all protocols; fewer than 3 follicles >15mm in size on Day 10 of ovarian stimulation. In patients undergoing their first IVF cycle, AMH measurement obtained at the initial assessment was used for determination of which band of COS the patient would be allocated. In the patients with repeated IVF cycles, AMH measurements were obtained prior to each IVF cycle, unless a last measurement performed within 12 months of period was available. During the study period two different assay methods for measurement of AMH was used in our centre: DSL Assay (1 October November 2010) and Gen II Assay (17 November August 2012). Correspondingly, during the study period two different COS Protocols were used: 1 st Protocol (1 October December 2010) and 2 nd Protocol (1 January August 2012). Consequently, allocation into the ovarian reserve bands of the patients of 1 st protocol were based on DSL assay samples, whereas the stratification of patients of 2 nd protocol was based either on DSL assay or Gen II assay samples. Specifically, the patients with recent DSL measurements (<12 months old) who had IVF treatment during the period of 2 nd Protocol had stratification on the basis of their DSL measurements. In these patients in order to obtain equivalent Gen II value, the DSL result was multiplied by 1.4 in accordance with the manufacturer s recommendation at the time. In the patients without previous or recent (<12 months old) DSL measurements stratification into ovarian reserve bands was achieved using their most recent Gen II measurements. Therefore, DSL measurements presented in this study may or may not have been used for formulation of the treatment strategies for individual patients. In fact, in this study DSL measurements have been included in order to understand the role of AMH in determination of ovarian response in IVF cycles, rather than an evaluation of AMH-tailored COS protocols. In addition to introduction of 2 nd protocol, further modifications were made to the protocol and therefore 2 nd protocol comprised of 4 different versions (Table 1-2). These changes in the protocols allowed us to compare the effect of the various modifications to COS protocols on oocyte yield. 199

200 Pituitary desensitisation regimes Selection of pituitary desensitisation regime was based on the patient s AMH according to the COH protocol at the time of commencement of IVF cycle (Table 1). Long agonist regime involved daily subcutaneous injection of 250 g or 500 g of the GnRH agonist Buseralin acetate (Supercur, Sanofi Aventis Ltd., Surrey, UK) from the mid-luteal phase (Day 21) of preceding menstrual cycle, which continued throughout ovarian stimulation. Women treated with Antagonist regime had daily subcutaneous administration of GnRH antagonist Ganirelex (Orgalutran, Organon Laboratories Ltd., Cambridge, UK) from Day 4 post-stimulation until the day of HCG/GnRH agonist trigger. Ovarian stimulation was achieved by injection of daily dose of hmg, Menopuir (Ferring Pharmaceuticals, UK) or rfsh, Gonal F (Merck Serono) as per AMH-tailored protocols (Table 1). Oocyte maturation was triggered using 5000 international units of HCG (Pregnyl, Organon Laboratories Ltd., Cambridge, UK) and the criteria for timing of HCG injection was consistent across all protocols: one (or more) leading follicle measuring >18mm and two (or more) follicle >17mm. Oocyte collection Oocyte collection was conducted hours following injection of HCG for follicle maturation. An Ultrasound Guided Oocyte Recovery (USOR) was conducted by experienced clinicians under sedation. The names of practitioners were anonymised and the practitioner with the largest number of oocyte recovery was categorized as a reference group. Practitioners with a small number (<10) of oocyte collection were pooled (group J). If the cycle was cancelled before oocyte recovery, it was categorized under the practitioner who was on-call for oocyte recovery session on the day of cycle cancellation. In cycles with pre-usor cancellation for excessive ovarian response, total oocyte number was coded as 27 and Metaphase II oocyte number was coded as 19. This was based on mean oocyte number in the patients who had post-usor cancellation for excessive ovarian response or OHSS. Quantitative assessment of total oocytes were conducted immediately post-usor by an embryologist. In patients undergoing ICSI, the assessment of the quality of oocytes were conducted 4-6 hours post-usor and the 200

201 oocytes assessed as in Metaphase II stage (MII) of maturation were categorized as mature oocytes. Statistical analysis The total number of collected oocytes in all cycles and the number of mature oocytes in the subset of ICSI cycles were used as outcome measures for the study. Oocyte was selected as the primary outcome measure for assessment of ovarian performance as this provides an objective measure which is largely determined by effectiveness of ovarian stimulation regimens. In contrast downstream measures such as clinical pregnancy and live birth are influenced by factors related to management, gametes and embryos. Statistical analysis was conducted using multivariable regression models and the process of model building included following steps: 1) Analyses of distribution of the groups and variables, 2) Univariate analysis to establish the factors that likely to affect total oocyte number, 3) Evaluation of representation of continuous variables, 4) Analysis of interaction between explanatory variables, 5) Sensitivity analysis. First the distribution of patients, the ovarian reserve markers, interventions and the outcomes were explored using cross tabulation, histograms, Box Whisker and scatter plots. Then, in order to establish the factors that likely to affect the oocyte number, univariate analyses of Age, AMH, AFC, PCO status, attempt of IVF/ICSI, ethnicity, BMI, protocol, regime, USOR practitioner and initial dose of gonadotrophins were conducted. Following this, all these explanatory variables were run as part of initial multivariable regression model. Adjustment for confounders related to the modifications of the protocols and unknown time-dependent changes conducted by inclusion of the COS protocol categories in the regression model. Evaluation of representation of oocyte number, Age, AMH, AFC, initial dose of gonadotrophins were conducted by establishing best fit on the basis of Akaike and Bayesian Information Criteria. In addition, interpretability of the data and clinical applicability of the results (e.g. cut off ranges) were used as a guide for selection of optimal representation. Given the oocyte number was not normally distributed, it was represented in logarithmic scale (log(oocyte number+.5). To establish best representation for AMH, AFC and initial dose 201

202 the models in following scales were run for each variable: Linear, quadratic, cubic, 4 th order polynomial, linear (log), quadratic (log), cubic (log), 4 th order polynomial (log), cut-off ranges according to distribution. Age adjustment in quadratic scale following centering it to 30 years of age was found to provide the most parsimonious representation. AMH was found to be best represented using following cut-off ranges: 0-3, 4-5, 6-8, 9-10, 11-12, 13-15, 16-18, 19-22, and The best representation for AFC was found to be cut-off ranges of 0-7, 8-9,10-11,12-14, 15-19, and Initial dose of gonadotrophins were categorized as following: IU, IU, 300IU, 375IU, 450IU. Subsequently, interactions between explanatory variables were tested at significance level of p<0.01, which revealed there were significant interaction between PCO status and other covariables. Given, these interactions were found to be complex and not easily computable we decided to restrict the regression analysis to the non-pco group. We observed significant interaction between regime and initial dose and therefore these variables were fitted with interaction term in the model. Finally sensitivity analyses of final regression models were conducted. Significance of the results was interpreted using p value (<0.05), effect size and clinical significance. For assessment of feasibility of individualization of stimulation regime and initial dose, visual representation of data was achieved using plots for observed and fitted values (Figure 1-4). RESULTS Description of data A total of 1847 IVF or ICSI cycles of 1428 patients met inclusion criteria for the study. AMH measurements of all cycles and AFC measurements for 1671 cycles (n=1289 patients) were available. In the analysis of total oocytes 1653 cycles were included and the analysis of MII oocytes comprised of 1101 ICSI cycles. Mean AMH was found to be 17.8 ( 12.5), mean AFC was , mean number of total oocytes was and mean number of mature oocytes was The distribution of the cycles according to patient characteristics and interventions is shown in Tables

203 Effect of patient and treatment related factors on oocyte yield Age, AMH, AFC Table 4a and 4b show that there was a significant negative association of oocyte yield with age and oocyte number following adjustment for AMH, AFC, causes of infertility, attempt of IVF ICSI cycle, USOR practitioner, COS protocol, pituitary desensitisation regime, type of gonadotrophin preparation and initial daily dose of gonadotrophins (Table 4a). With each increase of age by 1 year, we observed approximately a 3% reduction in total oocyte (p=0.0005) and a 2% decrease in mature oocyte number (p=0.006), which was independent of age and other covariables. In the analysis of AMH, there was significant gradual increase in total oocyte as well as mature oocyte number with increasing AMH following adjustment for all covariables (Figure 1 and 2). Compared to an AMH range of 0-3 pmol/l, there was increase of 25% in the range of 4-5 pmol/l (p=0,07), 36% in 6-8 pmol/l (p=0.008), 60% in 9-10 pmol/l (p=0.0005), 65% in pmol/l (p=0.0005), 77% in pmol/l (p=0.0005), 83% in pmol/l (p=0.0005), 80% in pmol/l (p=0.0005), 95% in pmol/l (p=0.0005) and 112% in the range of pmol/l (p=0.0005) in total oocyte number (Table 4a). Similar, but less marked, increase in MII oocyte number was observed with increasing AMH. The data on AFC also showed that there was gradual increase in total oocyte number with increasing AFC following adjustment of all covariables (Table 4a). Compared to an AFC of 0-7, there was increase of 14% in the range of (p=0.03), 22% in AFC of (p=0.001), 26% in AFC of (p=0.0005), 34% in AFC of (p=0.0005) and 40% in AFC of >25 (p=0.005). However, there was no increase in total oocyte number in AFC range of 8-9 compared to that of 0-7. AFC-related Increase in MII oocytes was less marked compared to that of total oocytes (Table 4a). Causes of infertility We did not observe any significant associations between the causes of infertility and number of retrieved oocytes. However, women diagnosed with unexplained infertility appear to have marginally higher (10%; p=0.02) total number of oocytes compared to women whose causes of infertility were 203

204 known. Diagnosis of severe tubal (-37%; p=0.19) and severe male (-37%; p=0.35) factor infertility was found to be associated with lower number of MII oocytes, compared to other causes of infertility. However neither of these parameters reached statistical significance. Similarly, there was no significant association between oocyte number and diagnosis of endometriosis with or without endometriomata compared to women that were not diagnosed with the disease (Table 4a). Attempt Analysis of total number of oocytes showed that women who had their 2 nd attempt of IVF ICSI cycle had slightly higher (8.5%; p=0.01) and those that had their 3 rd or 4 th attempt of treatment had significantly higher total oocyte yield (24%; p=0.001) compared to women undergoing their 1 st attempt of IVF ICSI cycle (Table 4a). Similarly overall effect of attempt on total oocyte yield was significant (p=0.001). However, we did not observe any association between the attempt and MII oocyte number in the analysis of the subset of ICSI cycles (p=0.45). USOR practitioner, COS protocol and gonadotrophin preparation There was a significant association (p=0.0005) between total oocyte yield with USOR practitioner (Table 4b). However, the association of USOR practitioner with MII oocyte number did not reach statistical significance (p=0.058). We observed significant association between the COS protocols in the analysis of total number of oocytes. 1 st version of 2 nd Protocol (-18%; p=0.0005), 2 nd & 3 rd versions of 2 nd Protocol (-14%; p=0.05) and 4 th version of 2 nd Protocol (-24%; p=0.009) provided significantly lower number of total oocytes compared to 1 st Protocol. However the effect of the COS Protocol changes to MII oocyte number was not significant (p=0.24). Compared to hmg, ovarian stimulation using rfsh provided 13% higher total oocytes (p=0.008). In the analysis of Metaphase II oocytes, there was no significant difference in oocyte yield between hmg and rfsh (0.26). 204

205 Regime and Initial dose of gonadotrophins The regression analyses of the regimes for pituitary desensitisation and initial dose categories were conducted in comparison to the reference group (Agonist with IU hmg/rfsh). IVF ICSI cycles where Antagonist with IU of hmg/rfsh (-36%; p=0.0005) was used provided significantly lower total oocyte yield, whereas cycles with Agonist and 300IU hmg/rfsh (15%; p=0.05) provided marginally higher total oocyte number. In the analysis of MII oocytes, cycles using Antagonist with IU of hmg/rfsh (43%; p=0.05), Agonist with 300IU of hmg/rfsh (25%; p=0.16) and Antagonist with 375IU hmg/rfsh (47%; p=0.02) yielded higher number of oocytes. Use of Agonist with 375IU hmg/rfsh (-18%; p=0.5) and Agonist with 450IU of hmg/rfsh (-28%; p=0.2) was associated with lower mature oocyte number, although the analysis did not reach statistical significance. AMH-tailored individualization of COS The overall effect of initial gonadotrophin dose to total oocyte yield was found to be significant (p<0.001). However, other than the lowest dose category with Antagonist regime, the analysis did not show any consistent dose-response effect on total oocyte number with increasing gonadotrophin dose (Table 4b, Figure 3a, Figure 3b, Figure 4a and Figure 4b). In the analysis of MII, compared to reference group of IU of initial daily gonadotrophins, we observed increased oocyte yield in the categories of IU (43%; p=0.05) and 375 IU (47%, p=0.02) of gonadotrophins. However both of these groups had Antagonist regime for pituitary desensitisation, compared to that of Agonist in the reference group, and therefore the observed effect may be related to the regime of COS rather than daily dose of gonadotrophins. 205

206 DISCUSSION In this study we explored the effect of age, AMH, AFC, causes of infertility, attempt of IVF ICSI treatment and interventions of COS on ovarian performance using a retrospective data on large cohort of IVF ICSI cycles of non-pco patients. To our knowledge, this is largest study to have conducted a detailed analysis of the effect of AMH and AFC on ovarian performance in IVF ICSI cycles. The study utilized a dataset that was prepared using a robust protocol for data extraction and handling. Similarly, the statistical analysis was based on a systematic exploration of the effect of all relevant factors, followed by adjustment for all relevant factors and finally careful analysis. With regards to the outcome measures, the quantitative response of ovaries were measured using total collected oocytes in IVF ICSI cycles and the MII oocyte number in the subset of ICSI cycles were used as a measurement of quantitative response of ovaries to COS. Arguably oocyte number is the best outcome measure for determination of ovarian response to COS, given it is mainly determined by patient s true ovarian reserve, the quality of assessment of ovarian reserve and treatment strategies for ovarian stimulation. In contrast downstream outcomes, such as clinical pregnancy and live birth are subject to additional clinical and interventional factors, which may not always be possible to adjust for using retrospective data. Indeed, large observational studies suggest that achieving optimal ovarian response is one of the most important determinants of success of IVF ICSI cycles and recommend to use oocyte number as a surrogate marker for live birth (Sunkara et al 2011). It appears around total oocytes or 3-4 mature oocytes provide optimal chance for a one live birth in IVF ICSI cycles (Sunkara et al 2011; Stoop et al 2012). Therefore, oocyte number appears to be most useful marker for assessment of ovarian response to COS as well as in prediction of live birth in cycles of IVF ICSI. 206

207 Effect of patient and treatment related factors on oocyte yield Age, AMH, AFC After adjusting for AMH AFC, the patient characteristics and above mentioned treatment interventions, age remained as an independent predictor of ovarian response to COS. Our data showed approximately 3% (p=0.0005) decrease in total oocyte and 2% (p=0.006) reduction in mature oocyte number with increase of age by factor of 1 year (Figure 3b and Figure 4b). Interestingly, the effect of AMH was also found to predict oocyte yield independently of age; with an effect actually more pronounced compared to that of age. After adjusting for age and all other factors there was gradual increase in total oocyte number with increasing AMH, which were both clinically (25-110%) and statistically (p=0.07-p=0.0005) significant (Table 4a). We observed, a largely, similar effect of AMH in the analysis of mature oocytes. It is important to note that due to the issues with Gen II AMH assay (Rustamov et al 2012), in this study we included only measurements obtained with the DSL assay. Consequently, presented cut-off ranges may not be applicable with current assay methods. We suggest that future studies should revisit the optimality of the cut-off ranges once a reliable assay method has been established. Similarly, after adjusting for all factors, the effect of AFC on total oocytes remained significant (14-40%; p<0.03). However the effect of AFC appears to be less marked compared to AMH. It is important to note that, the AFC assessment in this study is based on the measurement of 2-6mm antral follicles using two-dimensional transvaginal ultrasound scan. The cut-off ranges may not be applicable in centers where AFC measurement is obtained using different criteria. Our analysis suggests that age, AMH and AFC are independent determinants of total and MII oocyte number in IVF ICSI cycles and can be used as predictors of ovarian performance irrespective of patient and treatment characteristics. However, assessment of oocyte number is the quantitative response of ovaries to COS and may not necessarily reflect qualitative outcome. 207

208 Causes, Endometriosis, Endometrioma The causes of infertility do not appear to make a significant contribution in determining total oocyte number after controlling for age, AMH, AFC, the attempt and treatment interventions. Although in the analysis of MII oocytes we observed reduced oocyte yield in women with severe tubal (-37%) and severe male (-37%) infertility, this was not statistically significant. The analysis of MII oocytes only included the subset of ICSI cycles consisting of women with male factor infertility. Therefore the effect of severe male factor infertility may have been more marked in this model. We did not observe a significant difference in total or MII oocyte number in women with a history of endometriosis with or without endometriomata. Current understanding of the effect of endometriosis in the outcomes of IVF treatment suggests that the disease has detrimental effect on IVF outcomes (Barnhart et al 2007; Barnhart et al 2002). However, some argue that no association is observed if the analysis conducted using proper adjustment for all relevant confounders (Surrey 2013). Our data suggests that after adjustment for all relevant factors there is no measurable association with endometriosis (with or without endometriomata) and oocyte number. Some suggest that using ultra-long down regulation using depot GnRH analogue up tp 3-6 months prior to ovarian stimulation improves ovarian performance in patients with endometriomata. Our dataset did not have information on pituitary desensitisation prior IVF treatment cycles and we are therefore unable to assess the effect of this intervention. Attempt Our study found that, 2 nd and 3 rd cycles were associated with 8% (p=0.01) and 24% (p=0.001) higher total oocytes compared to that of 1 st IVF cycle. However, the effect of the attempt on MII oocytes was not significant. In our centre only patients with a previously unsuccessful IVF treatment are offered subsequent cycles and therefore compared to the patients with repeated attempts the group with first cycle may be expected to have better oocyte yield. However when controlled for all relevant confounders, including adjustment of treatment interventions, 1 st IVF cycle does not appear to provide better oocyte yield. In keeping with our findings, a recent study demonstrated independence of attempts of IVF cycles in terms of outcomes (Roberts SA and 208

209 Stylianou C 2012). Increased total oocyte yield with progressed attempts is likely to be due to the adjustment of COS on the basis of information on the ovarian response in previous cycles. It is important to note that, in this study we assessed oocyte yield as the outcome measure and this may not necessarily translate into live birth, which is desired outcome for the couples. Therefore, availability of data on the attempt-dependency of live birth in IVF cycles is important and we suggest future studies should explore it. USOR practitioner To our knowledge this is the first study that explored the effect of an oocyte recovery practitioner on oocyte yield adjusting for all relevant confounders. We observed a considerable operator-dependent effect on total oocyte yield, which may be due to a variation of patients across the days of the week (p=0.0005). The practitioners were allocated to the sessions of oocyte recovery using a specific rota template according to the day of the week. Given in our centre we do not conduct oocyte recovery at weekends, there may be day-dependent variation in selection of patients. For instance, the patients who are likely to have maturation of leading follicles during the weekend may have been scheduled slightly earlier. Similarly, the patients with confirmed maturation of leading follicles whose oocyte recovery would have fallen on weekends may have been scheduled after the weekend allowing maturation of additional follicles. Therefore practitioners conducting the sessions of oocyte recovery in extremes of weekdays may not necessarily have similar patients compared to that of other days, which may have introduced some bias in estimating the outcomes of individual practitioners. Nevertheless, given the statistical analysis adjusted for age, ovarian reserve and treatment interventions we think there is considerable true between-operator variability on total oocyte number. We suggest that future studies should assess it further by including adjustment for follicle number and size on the day of HCG. Interestingly overall effect of the operator did not reach statistical significance in the analysis of MII oocytes in ICSI subset (p=0.058). This may suggest, irrespective of total oocyte yield, aspiration of only follicles of larger than a certain size provides oocytes with potential for fertilization. 209

210 COS Protocol Controlled ovarian hyperstimulation in IVF is conducted using a predefined protocol, which contains the policy on selection of regime for pituitary desensitisation, the initial daily dose of gonadotrophins, the monitoring of ovarian response, the adjustment of daily dose of gonadotrophins, the policy for cancellation due to poor or excessive ovarian response and criteria for HCG trigger for final maturation of oocytes. Determination of the optimal treatment regime and the initial dose of gonadotrophins for each patient is frequently achieved by stratification of patients into various bands of ovarian reserve on the basis of the assessment of ovarian reserve. The assessment of ovarian reserve prior to IVF cycle is performed using biomarkers which usually consist of one or combination of following: Age, AMH, AFC and FSH. In our centre stratification of patients into the bands of ovarian reserve was determined on the basis of the patient s AMH measurements. For instance, the patients deemed to have lower ovarian reserve were allocated to the treatment band with higher daily dose of gonadotrophins and vice versa (Table 1). The study found that the 2 nd protocol was associated with 14-24% lower total oocyte yield compared to the 1 st protocol. The differences in the interventions between the protocols are described in Table 1 and Table2. Compared to the 1 st protocol, the 2 nd protocol had a) some patients allocated to COS bands using Gen II assay measurements, which later was found to provide inaccurate measurements, b) more AMH cut-off bands for COS bands, c) strict monitoring of ovarian response and corresponding adjustment of daily dose of gonadotrophins and d) strict criteria for cycle cancellation for excessive response. Therefore our data suggests that the COS protocols with broader AMH cut-off bands with less strict criteria for adjustment of daily gonadotrophins may provide higher oocyte yield. However, given it is retrospective analysis, the limitation of the study should be recognized and we recommend more robust prospective studies on optimization of AMH tailored protocols should be conducted. Gonadotrophin type The study showed that rfsh was associated with higher total oocyte number (13%; p=0.008). Interestingly analysis of MII oocyte showed a larger confidence interval and did not reach statistical significance, suggesting the 210

211 effect of rfsh was not a strong determinant of mature oocytes. Perhaps observation of higher total oocytes in rfsh cycles compared to that of HMG and yet comparable mature oocyte number in the study suggest that regardless of total oocyte yield, only follicles with a potential for maturation will achieve a stage of metaphase II. Ovarian stimulation in cycles for IVF is largely achieved by two different analogues of follicle stimulating hormone; human menopausal gonadotrophin (hmg) and recombinant follicle stimulating hormone r(fsh). Although purified, hmg contains more luteinising hormone compared to rfsh, which is believed to assist endometrial maturation and improve odds of implantation in cycles of IVF. Furthermore, the LH component of hmg is believed to assist in maturation of oocyte with subsequent improvement in live birth. On the other hand historically rfsh was believed to have less batch-to-batch variation compared to that of HMG, which allows administration of more precise daily dose of gonadotrophins. To date a number of studies have been published comparing these two forms of gonadotrophin preparations, which provide conflicting findings. However, systematic review that compared of the effect of these types of gonadotrophins on live birth rate suggests that there is no significant difference on live birth rate (van Wely et al 2011). This supports our findings on that irrespective of total oocyte yield clinically useful, mature, oocyte number is comparable between the groups. Regime and dose of gonadotrophins The study found that compared to the reference group (Agonist IU), none of the combination of the regime and gonadotrophin dose provided a higher total oocyte yield. Women that were in Antagonist regime group with an initial daily dose of IU gonadotrophins produced approximately 36% fewer total oocytes (p=0.0005). However, comparison of MII oocytes of these groups did not reach statistical significance and the effect size was much smaller (-19%; p=0.23). This and reference groups represent the patients with high ovarian reserve who had milder ovarian stimulation because of risk of excessive ovarian response and OHSS. Lower total oocyte yield and comparable mature oocyte number in the Antagonist regime may explain why this regime is reported to be associated with reduction in the risk of OHSS and 211

212 yet comparable live birth in patients with high ovarian reserve (Yates et al 2012). In the analysis of MII oocytes, Antagonist with IU of gonadotrophin and Antagonist with 375 IU of gonadotrophin provided around 43% (p=0.05) and 47% (p=0.02) more oocytes compared to that of the reference group (Agonist IU). Interestingly, total oocytes of these groups were comparable to that of reference group, suggesting that using Antagonist protocol may be associated with improvement in oocyte maturation compared to Long Agonist regime. Perhaps in addition to the effect of exogenous HCG, endogenous LH may play role in oocyte maturation in IVF/ICSI cycles and shorter desensitisation of pituitary using Antagonist regime may allow secretion of LH during COS in lower quantities. AMH-tailored individualisation of COS Given that we did not observe a significant dose-dependent effect on oocyte number, we were not able to develop AMH or AFC tailored individualisation protocols for COS. Although the initial dose of gonadotrophin is believed to be one of the main determinants of oocyte yield, our study suggests that the association between these variables is weak. Consequently strict protocols for tailoring the initial dose of gonadotrophins may not necessarily improve ovarian performance in IVF treatment. It is important to note that our COS protocols recommended close monitoring of ovarian response and corresponding dose adjustment starting from 3 rd day of COS, which may have masked the effect of initial dose. However, further analysis with adjustment for the total gonadotrophin dose and dose adjustment during the stimulation did not have significant impact on oocyte yield. Nevertheless further time series regression analysis, with full parameters of cycle monitoring and the dose adjustments in the model, should be conducted in order to ascertain the role of AMH in tailoring the dose of gonadotrophins in cycles of IVF. 212

213 Strengths of the study Here we presented the largest study on assessment of the role of patient and treatment related factors on oocyte yield and exploration of optimization of AMH-tailored COS using a validated dataset. Statistical analysis included systematic assessment of the effect possible confounders on measured outcome; including of age, AMH, AFC, causes of infertility, attempt of IVF treatment, USOR practitioner, type of gonadotrophin, pituitary desensitisation regime and initial dose of gonadotrophins. On the basis of above analysis a robust multivariable regression models for assessment of the effect all above factors on total and mature oocyte number have been developed. Prior to conducting this study, previous projects explored the performance of AMH assay methods. The studies found that Gen II assay may yield highly non-reproducible measurements compared to that of DSL assay (Rustamov et al., 2012a). Therefore in this study only DSL AMH assay measurements were included. Furthermore previous projects (Chapter 5 and 6) explored the effect of various patient related factors on AMH, AFC and FSH measurements and found that some of the factors had measurable impact on ovarian reserve. These findings were used in establishing which patient related factors ought to be explored in the building of regression models for this study. However the DSL assay is no longer available and most clinics are mainly using Gen II AMH assay in formulation of COS in IVF. Given its observed instability, AMH-tailoring based on Gen II samples may lead to erroneous allocation of patients to the band that is significantly inconsistent with patient s ovarian reserve. Subsequently this may result in the extremes of ovarian response to COS, including severe OHSS and cycle cancellation. Weaknesses of the study The main weakness of the study is that the analysis is based on retrospectively collected data. The methodology included an extensive exploration for possible confounders and adjustment for the ones that were found to be significant. However, there are may be unmeasured factors that that might have affected the estimates. In addition, the study included only patients that did not have PCO appearance on ultrasound scan. The analysis in all patients showed that, interaction of PCO status with other covariables was complex which could introduce bias in estimation of the effects of other 213

214 factors. Therefore, analyses of the groups with and without PCO were run separately. Subsequently results of non-pco group was presented in the thesis, given it had the largest number of cycles. Compared to non-pco analysis, we did not observe significant difference in the results of PCO model. The study assessed ovarian response using oocyte yield only. Other outcomes of ovarian response, such as duration of ovarian stimulation, total dose of gonadotrophins, cycle cancellation due to poor or excessive ovarian response and OHSS have not been analysed. Therefore, it is important to interpret the findings of this study in the context of ovarian response determined by oocyte yield. Specifically, the study should not be used to interpret cycle cancellation for excessive ovarian response. As described in the methodology of the study, the oocyte number in the cycles with cancellation of oocyte recovery due to excessive response were recoded with comparable values with cycles that were cancelled following oocyte recovery for OHSS. Given the main desired outcome of IVF treatment is live birth, the overall success of a treatment cycle should reflect this outcome measure. This study does not assess the effect of above factors to overall success of IVF treatment. However the study provides a robust data on research methodology in assessment of IVF outcomes, which can assist in the assessment of other outcome measures in future studies. SUMMARY After adjustment for all the above factors, age remained a negative predictor of oocyte yield, whereas we observed a gradual and significant increase in oocyte number with increasing AMH and AFC values, suggesting all these markers display an independent association with oocyte yield. IVF attempt, oocyte recovery practitioner, type of gonadotrophin were found to have significant effect on total oocyte yield. However, the effect of these factors on mature oocyte number did not reach statistical significance. Whilst total oocyte number was comparable between pituitary desensitisation regimes, GnRH antagonist cycles were found to provide significantly higher mature oocytes compared to that of long GnRH agonist regime. In terms of the effect of initial dose on oocyte yield, following adjustment for all above variables, we did not observe significant increase in 214

215 oocyte number with increasing gonadotrophin dose categories. Therefore, strict protocols for tailoring the initial dose of gonadotrophins may not necessarily improve ovarian performance in IVF treatment. However, further time series regression analysis, with full parameters of cycle monitoring and the dose adjustments in the model, should be conducted in order to ascertain the role of AMH in tailoring the dose of gonadotrophins in cycles of IVF. This study demonstrates complexity of the factors that determine ovarian response in IVF cycles. Therefore assessment of AMH-tailored individualisation of ovarian stimulation should be based on a robust methodology, preferably using a large randomized controlled trial. Furthermore measurement of AMH ought to be based on a reliable assay method, which is currently not available. In the meantime, the limitations of available evidence on AMH-tailored individualisation of ovarian stimulation should be taken into account in the management of patients. 215

216 References Broer, S.L., et al., AMH and AFC as predictors of excessive response in controlled ovarian hyperstimulation: a meta-analysis. Hum Reprod Update (1): p Barnhart K, Dunsmoor-Su R, Coutifaris C. Effect of endometriosis on in vitro fertilization. Fertil Steril 2002;77: Dechaud H, Dechanet C, Brunet C, et al. Endometriosis and in vitro fertilization: a review. Gynecol Endocrinol 2009;25: Dewailly D, Andersen CY, Balen A, Broekmans F, Dilaver N, Fanchin R, Griesinger G, Kelsey TW, La Marca A, Lambalk C, Mason H, Nelson SM, Visser JA, Wallace WH, Anderson RA. The physiology and clinical utility of anti-mullerian hormone in women. Hum Reprod Update Kurinczuk JJ et. al. Fertility Treatment in A Statistical Analysis. HFEA, La Marca A. and Sunkara S. K.. Individualization of controlled ovarian stimulation in IVF using ovarian reserve markers: from theory to practice. Human Reproduction Update, Vol.20, No.1 pp , 2014 Nardo LG, Gelbaya TA, Wilkinson H, Roberts SA, Yates A, Pemberton P and Laing I. Circulating basal anti-müllerian hormone levels as predictor of ovarian response in women undergoing ovarian stimulation for in vitro fertilization. Fertil Steril 2009; 92: Nelson SM, Yates RW and Fleming R. Serum anti-müllerian hormone and FSH: prediction of live birth and extremes of response in stimulated cycles-- implications for individualization of therapy. Hum Reprod 2007;22: Nelson, S.M., et al., Anti-Mullerian hormone-based approach to controlled ovarian stimulation for assisted conception. Hum Reprod, (4): p Nelson SM. Biomarkers of ovarian response: current and future applications. Fertil Steril 2013;99: Roberts SA, Stylianou C. The non-independence of treatment outcomes from repeat IVF cycles: estimates and consequences. Hum Reprod Feb;27(2): Rustamov O, Pemberton PW, Roberts SA, Smith A, Yates AP, Patchava SD and Nardo LG. The reproducibility of serum anti-müllerian hormone in subfertile women: within and between patient variability. Fertil Steril 2011;95: Rustamov O, Smith A, Roberts SA, Yates AP, Fitzgerald C, Krishnan M, Nardo LG and Pemberton PW. Anti-Mullerian hormone: poor assay reproducibility in a large cohort of subjects suggests sample instability. Hum 216

217 Reprod 2012a;27: van Rooij IA, Broekmans FJ, te Velde ER, Fauser BC, Bancsi LF, de Jong FH and Themmen AP. Serum anti-mullerian hormone levels: a novel measure of ovarian reserve. Hum Reprod 2002;17: Stoop D, Ermini B, Polyzos N.P, Haentjens, P, De Vos M, Verheyen G, and Devroey P Reproductive potential of a metaphase II oocyte retrieved after ovarian stimulation: an analysis of ICSI cycles. Human Reproduction, Vol.27, No.7 pp , 2012 Sunkara SK, Rittenberg V, Raine-Fenning N, Bhattacharya S, Zamora J, Coomarasamy A. Association between the number of eggs and live birth in IVF treatment: an analysis of treatment cycles. Hum Reprod 2011; 26: Sunkara SK, Coomarasamy A, Faris R, Braude P, Khalaf Y. Effectiveness of the GnRH agonist long, GnRH agonist short and GnRH antagonist regimens in poor responders undergoing IVF treatment: a three arm randomised controlled trial. (ESHRE) 2013,London, UK. Surrey.ES. Endometriosis and Assisted Reproductive Technologies: Maximizing Outcomes. Semin Reprod Med 2013;31: van Wely M1, Kwan I, Burt AL, Thomas J, Vail A, Van der Veen F, Al-Inany HG. Recombinant versus urinary gonadotrophin for ovarian stimulation in assisted reproductive technology cycles. Cochrane Database Syst Rev Feb 16;(2):CD Yates AP, Rustamov O, Roberts SA, Lim HY, Pemberton PW, Smith A, Nardo LG. Anti-Mullerian hormone-tailored stimulation protocols improve outcomes whilst reducing adverse effects and costs of IVF. Hum Reprod 2011;26:

218 Figure 1. Study groups for assessment of Individualisation of pituitary desensitisation regime Individualisation of pituitary desensitisation regimens can be studied comparing the effectiveness of AMH-tailoring in women of low, medium and high ovarian reserve. Low AMH (e.g. DSL assay pmol/l) GnRH Antagonist GnRH Agonist Individualisation of COS Regime Normal AMH (e.g. DSL assay pmol/L) GnRH Antagonist GnRH Agonist High AMH (e.g. DSL assay >28.8 pmol/l) GnRH Antagonist GnRH Agonist 218

219 Fiure 2. Study groups for assessment of individualisation of initial gonadotrophin dose Individualisation of daily dose of gonadotrophins can be studied comparing the effectiveness of AMH-tailoring in women of low, medium and high ovarian reserve. High Dose (e.g. HMG IU) Low AMH (e.g. DSL assay: pmol/l) Moderate Dose (e.g. HMG IU) Low Dose (e.g. HMG IU) Individualisation Gonadotrophin Dose Normal AMH (e.g. DSL assay: pmol/l) High Dose (e.g. HMG IU) Moderate Dose (e.g. HMG IU) Low Dose (e.g. HMG IU) High Dose (e.g. HMG IU) High AMH (e.g. DSL assay: >28.8 pmol/l) Moderate Dose (e.g. HMG IU) Low Dose (e.g. HMG IU) 219