Melanoma risk after ovarian stimulation for in vitro fertilization

Transcription

1 Human Reproduction, Vol.30, No.5 pp , 2015 Advanced Access publication on March 4, 2015 doi: /humrep/dev023 ORIGINAL ARTICLE Reproductive epidemiology Melanoma risk after ovarian stimulation for in vitro fertilization M. Spaan 1, A.W. van den Belt-Dusebout 1, M. Schaapveld 1, T.M. Mooij 1, C.W. Burger 2, F.E. van Leeuwen 1, *, on behalf of the OMEGA-project group 1 Department of Epidemiology, The Netherlands Cancer Institute, Plesmanlaan 121, Amsterdam, CX 1066, The Netherlands 2 Department of Obstetrics and Gynaecology, Erasmus Medical Center, Postbus 2040, Rotterdam, CA 3000, The Netherlands *Correspondence address. Department of Epidemiology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands. Tel: ; Fax: ; f.v.leeuwen@nki.nl Submitted on October 16, 2014; resubmitted on December 19, 2014; accepted on January 22, 2015 study question: Do women treated with ovarian stimulation for IVF have an increased risk of melanoma? summary answer: Ovarian stimulation for IVF does not increase risk of melanoma, even after a prolonged follow-up. what is known already: Although exposure to ultraviolet radiation is the major risk factor for melanoma, associations between female sex steroids and melanoma risk have also been suggested. The results of available studies on fertility drugs and melanoma risk are inconclusive since most studies had several methodological limitations such as short follow-up, a small number of cases and no subfertile comparison group. study design, size, duration: In 1996, a nationwide historic cohort study (the OMEGA-cohort) was established to examine the risk of cancer after ovarian stimulation for IVF. After a median follow-up of 17 years, cancer incidence was ascertained through linkage with the Netherlands Cancer Registry. Melanoma risk in the cohort was compared with that in the general population and between the IVF group and non-ivf group using multivariable Cox regression analyses. participants/materials, setting, methods: The cohort comprises women who received IVF between 1983 and 1995 and a comparison group of 5950 women who underwent subfertility treatments other than IVF. Detailed IVF-treatment data were obtained from the medical records and complete information on parity and age at first birth was obtained through linkage with the Dutch Municipal Personal Records Database. main results and the role of chance: In total, 93 melanoma cases were observed. The risk of melanoma was not elevated among IVF-treated women, neither when compared with the general population (standardized incidence ratio ¼ 0.89; 95% confidence interval (CI): ), nor when compared with the non-ivf group (adjusted hazard ratio (HR) ¼ 1.27; 95% CI: ). A higher number of IVF cycles was associated with apparent but statistically non-significant risk increases (5 6 cycles HR ¼ 1.92; 7 cycles HR ¼ 1.79). However, no significant trend emerged. In women with more follicle stimulating hormone/human menopausal gonadotrophin ampoules comparable nonsignificant risk increases were found. A longer follow-up did not increase melanoma risk. Nulliparous women did not have a significantly higher melanoma risk than parous women (HR ¼ 1.22; 95% CI: ). However, women who were 30 years of age or older at first birth had a significantly higher melanoma risk than women who were younger than 30 years at first birth (age: years HR ¼ 4.57; 95% CI: ,.34 years HR ¼ 2.98; 95% CI: ). limitations, reasons for caution: Despite our large cohort, the number of melanoma cases was rather small, especially in our comparison group, which hampered subgroup analyses. wider implications of the findings: Our results are reassuring for women who underwent IVF or are contemplating to start IVF. Since our cohort study is one of the largest published so far, with long-term follow-up, a subfertile comparison group, and detailed IVF-treatment data, our results add important information to the available evidence. study funding/competing interest: This study was supported by grants from the Dutch Cancer Society (NKI ), the Health Research and Development Counsel ( ) and the Dutch Ministry of Health. OMEGA-project group: see Appendix. & The Author Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved. For Permissions, please journals.permissions@oup.com

2 Melanoma risk after ovarian stimulation 1217 Key words: in vitro fertilization / melanoma / ovarian stimulation / parity / infertility Introduction During the last 50 years the incidence of melanoma has increased steadily in most Western countries (Karim-Kos et al., 2008; MacKie et al., 2009; Tucker, 2009; Hollestein et al., 2012). In the Netherlands, melanoma is the fifth most common cancer in women, with 2731 new cases diagnosed in 2012 (5.6% of all cancer cases) (IKNL, 2013). The main reason for the increase in melanoma incidence is assumed to be greater exposure to ultraviolet (UV) radiation (Titus-Ernstoff, 2000; MacKie et al., 2009). Besides UV radiation, other associations between female sex steroids and melanoma risk have been reported. Oestrogens are known to influence melanocyte proliferation (Thornton, 2002; Leslie and Espey, 2005; Zouboulis et al., 2007) which can lead to hyperpigmentation in women, as observed in some oral contraceptive (OC) users and in pregnant women (Snell and Bischitz, 1960; Thornton, 2002; Zouboulis et al., 2007). In support of a role of female sex hormones, melanoma is more common during a woman s reproductive life span and occurs more frequently in women than in men (Hollestein et al., 2012; IKNL, 2013). Moreover, oestrogen receptors have been identified in both normal human skin cells and melanoma cells (Walker et al., 1987; de Giorgi et al., 2009). The combination of these findings suggests that female sex hormones are likely involved in the aetiology of melanoma in women. Several studies examined the potential associations between melanoma risk and female sex hormones and reproductive factors, such as age at menarche, parity, age at first birth, age at menopause, use of OCs and use of hormone replacement therapy (HRT) (Gandini et al., 2011; Kvaskoff et al., 2011; Tang et al., 2011). A meta-analysis concluded that there is no association between the use of OC or HRT and melanoma risk while there seem to be protective effects from parity and a younger age at first birth (Gandini et al., 2011). However, a large cohort study of subfertile women recently showed that parous women who were treated with IVF are at increased risk of melanoma. Therefore, results regarding parity and age at first birth remain inconclusive (Stewart et al.,2013). The use of ovarian-stimulating drugs has increased over the past decades. Since ovarian-stimulating drugs cause a temporary increase of oestrogen and progesterone levels, several investigators reported on the association between these drugs and melanoma risk (Brinton et al., 1989; Rossing et al., 1995; Venn et al., 1995; Modan et al., 1998; Young et al., 2001; Althuis et al., 2005; Hannibal et al., 2008; Calderon- Margalit et al., 2009; Silva Idos et al., 2009; Yli-Kuha et al., 2012; Stewart et al., 2013). However, most published studies have limitations such as relatively short follow-up, a low number of cases, imprecise information about type of fertility drugs (FDs) used, and lack of an appropriate comparison group. Therefore, the aim of this study was to investigate longterm melanoma risk in a large Dutch cohort of subfertile women treated with ovarian stimulation for IVF. Patients and Methods Study population In , we identified a nationwide historical cohort of subfertile women who received at least one IVF cycle with ovarian stimulation between 1983 and 1995 in one of the 12 IVF-treatment centres in the Netherlands (Fig. 1). The non-ivf comparison group consisted of 7515 women whose subfertility was diagnosed in the four participating clinics that had a computerized registry of all subfertile women evaluated during We attempted to frequency match the non-ivf comparison group according to the distribution of subfertility diagnoses in the IVF group. Most women in the non-ivf group registered in the 1980s (before IVF became a routine procedure) underwent tubal surgery, intrauterine insemination (IUI) without stimulation or hormonal treatments (e.g. clomiphene or stimulated IUI). The majority of those who registered after 1990 withdrew from the waiting list for IVF, became pregnant or decided to refrain from IVF. From the women selected into the non-ivf comparison group 911 women subsequently received IVF in a second centre. In the description of the cohort these women are included in the IVF group (Supplementary data, Table SI) (see also the Statistical analysis section) (van Leeuwen et al., 2011). Ethical approval The institutional ethics committees of all IVF clinics approved the study procedures, which have been described in detail previously (Klip, 2002; De Boer et al., 2003). In short, based on names, birth dates and addresses at the time of subfertility treatment, all cohort members were traced through the municipal population offices that fully cover the Netherlands. From the initial women, 4.2% were not approached (Fig. 1)(van Leeuwen et al., 2011). Risk factor questionnaire Between 1997 and 1999, women were sent a risk factor questionnaire, a study information letter, and a brochure. Each participant was asked written informed consent for future linkage with disease registries. The 23-page questionnaire ascertained information on the women s reproductive histories, subfertility treatment, use of exogenous hormones, lifestyle factors and family history of cancer. In total, women returned the questionnaire(response rate: 65.2%) (Fig. 1). The response rate was substantially lower in the non-ivf group (48.7%) than in the IVF group (71.1%). One percent (n ¼ 264) of the women actively refused to participate in the study. Medical records Trained abstractors collected information on subfertility causes and all subfertility treatments from the medical records. Cause of subfertility was classified as tubal, male factor, endometriosis, ovarian disorders, cervical factor, uterine abnormalities or unexplained. For each IVF and IUI cycle we recorded date, dosage and type of FDs used in each phase of the menstrual cycle [human menopausal gonadotrophin (hmg), follicle stimulating hormone (FSH), clomiphene, human chorionic gonadotrophin (hcg), gonadotrophin-releasing hormone (GnRH) and progesterone], number of oocytes collected and outcome. For FDs used prior to inseminations/ivf, we also coded date, dosage and type of FDs used per cycle. We also collected information on subfertility treatments provided outside the participating IVF clinics, from intake forms and physician letters. Besides information about FD use also information about tubal surgery and other surgical procedures to improve fertility was collected from the medical records. Due to limited funding in 2000, we could only complete medical record abstraction for 9 of 12 centres, i.e women (Klip, 2002; De Boer et al., 2003).

3 1218 Spaan et al. Figure 1 Identification of the OMEGA-study cohort. (a) According to the treatment centre where these women were identified. (b) Women who were selected into the non-ivf group but subsequently received IVF (n ¼ ¼ 952); these women contributed person-time to both the non-ivf group and IVF group. (c) Women in this category contributed person-time from the first IVF treatment or first visit gynaecologist until date of death. (d) For unknown reason these women had originally not been identified as belonging in the IVF group; the women did not contribute person-time to the non-ivf group because IVF treatment was administered before Updated information on IVF treatment and parity In 2012, we were able to abstract subfertility treatment data from the medical records of an additional 1200 women from one centre that had not been abstracted in , leading to a total of medical records abstracted. We also updated IVF-treatment data with electronically available data for the period from all IVF clinics regarding treatment cycles that women received after questionnaire completion. When updating treatment data, we identified 41 additional women in the non-ivf group who received IVF after questionnaire completion, leading to a total of 952 women who were first selected in the non-ivf group and subsequently received IVF (see also the Statistical analysis section) (van Leeuwen et al., 2011). In addition, in 2013 we obtained authorization to link our cohort members (both responders and nonresponders) with the Dutch Municipal Personal Records Database to complete and update information regarding all births, resulting in complete information until end of the follow-up for parity and age at first birth for 99% of the cohort. Incidence of melanoma and vital status assessment Cancer incidence in the period was ascertained through linkage with the population-based Netherlands Cancer Registry (NCR) according to a validated record linkage protocol (van den Brandt et al., 1993). The NCR granted us permission to link not only responders who gave permission, but also non-responders and deceased women, under additional privacy regulations. Only women who completed the questionnaire but explicitly refused future linkage with disease registries (n ¼ 1091; 4.3% of all women) were excluded from NCR linkage (but included in the analysis, see below). For each melanoma we received information on date of diagnosis and morphology. Women diagnosed with cancer before entering the cohort or before 1989 (start NCR), were excluded from the analysis (71 women of whom 8 had melanoma). Deceased women who had been linked with the NCR (n ¼ 90) were included in the analytical cohort. Vital status until July 2009 was obtained by linkage with the Central Bureau for Genealogy, which keeps computerized records of all deceased persons in the Netherlands since Statistical analysis The analytic study cohort consisted of women; women in the IVF group and 5950 women in the non-ivf group (Fig. 1). Because the NCR did not fully cover the Netherlands before 1989, the observation time for each participant started on 1 January 1989 or the date of first IVF treatment (IVF group), or first clinic visit for subfertility evaluation (non-ivf group), whichever came last. Person-years of observation were calculated to the

4 Melanoma risk after ovarian stimulation 1219 date NCR follow-up ended (30 June 2009), date of diagnosis of first malignancy (melanoma or other malignancy), date of death or date of questionnaire completion if the woman refused linkage with the NCR, whichever came first. Women selected into the non-ivf comparison group who subsequently received IVF (n ¼ 952) contributed person-time to the non-ivf group until the date of first IVF treatment, and switched to the exposed group after this date (Breslow and Day, 1987). Subfertility cause(s) and treatments were preferably based on the medical records, and only derived from the women s questionnaires if the medical records had not been abstracted (see the Medical records section). For non-responding women information from hospital databases was added when available. Due to the small numbers of cases in the group of women with a hormonal cause of subfertility and in women with other causes of subfertility, those groups were combined in the statistical analyses. Information on reproductive factors was derived from the Dutch Municipal Personal Records Database, since these variables could change after IVF-treatment and questionnaire completion. First, we compared melanoma incidence in the IVF group and non-ivf group with incidence in the general population. We determined the standardized incidence ratio (SIR) as the ratio of the observed (O) and expected (E) number of cancers in the cohort. Expected numbers were based on age- and calendar period-specific melanoma incidence rates derived from the NCR (IARC, 2008). Secondly, Cox proportional hazards models, with time-dependent and fixed covariates and age as the timescale, were used to compare cancer risk between the IVF and non-ivf groups, adjusting for potential confounders such as age at menarche, parity, multiple pregnancies, age at first birth, OC use, body mass index (BMI), total number of IUI cycles and subfertility cause. Dates of IVF cycles and birthdates of children were included in the model as time-varying variables. We identified confounders as those factors which led to a 10% or more change in the risk estimate for the exposure of interest in a model including the potential confounder(s) and the variable of interest. Confounders were tested in models restricted to women who responded to the questionnaire since information on a number of potential confounders (age at menarche and OC use) was missing for nonresponders. The selected confounders were subsequently included in all analyses including responders and non-responders. Dose response associations with subfertility treatments were investigated by examining associations with the number of IVF cycles, total number of FSH/hMG ampoules used in IVF treatments as well as the number of oocytes collected in the first IVF-treatment cycle. Although levels of oestrogen and progesterone are much higher in IVF-stimulated cycles than in IUI cycles we also investigated dose response associations with total number of IUI cycles and with the total number of IUI and IVF cycles since both women in the non-ivf and IVF groups could have been treated with IUI. Furthermore, we examined the independent association with parity and age at first birth and the modifying effect of parity on the association between IVF and melanoma risk. In all analyses, missing values were included as a separate category. Analyses were performed with the use of the STATA software, version 11. Results Population characteristics The analysis included IVF-treated women and the 5950 women not treated with IVF. Women in the non-ivf group had a slightly longer follow-up duration than women in the IVF group (19.4 versus 16.3 years) and they were also older at the end of follow-up (median age: 51.2 versus 49.6 years). These differences reflect the initial inclusion criteria for the IVF and non-ivf groups, with an over-representation of women in the non-ivf group seeking subfertility treatment in the years before IVF treatment became a routine procedure. In the IVF group 32% had one to two stimulated cycles, 33% had three to four stimulated cycles and 22% had five or more stimulated cycles. IVF stimulation regimens used in the cohort have been described in detail previously (De Boer et al., 2004; van Leeuwen et al., 2011). In the IVF group more women remained nulliparous (36%) compared with the non-ivf group (27%). A full comparison of the groups is shown in Supplementary data, Table SI. In total, 93 cases of invasive melanoma were observed, 72 in the IVF group and 21 in the non-ivf group (Table I). Of the 93 melanomas 64 (69%) were superficial spreading melanomas, 9 (10%) were nodular melanomas, 3 (3%) melanomas had other morphologies and 17 (18%) melanomas were not specified (data not shown). Women with melanoma more often had an unexplained cause of subfertility and less often a hormonal (and other) cause of subfertility compared with non-cases. Furthermore, melanoma cases were more often nulliparous (45%) than non-cases (34%). Comparisons with external reference rates The SIRs for melanoma in both the IVF group and the non-ivf group were not elevated compared with the incidence of melanoma in the Dutch general population (SIR ¼ 0.89; 95% CI: and SIR ¼ 0.75; 95% CI: , respectively) (Table II). Melanoma risk did not significantly increase with a higher number of IVF cycles (all SIRs varied between 0.75 and 1.09) (P for trend: 0.09) or with longer follow-up duration ( 20 years: SIR ¼ 0.78; 95% CI: ). In women with missing data on the number of IVF cycles melanoma risk was significantly decreased. Women who were classified as having a hormonal or other cause of subfertility (other than male, tubal and unexplained) had a decreased melanoma risk (SIR ¼ 0.46, 95% CI: ). The lower risk was mainly driven by women with hormonal causes because no melanoma cases were observed in this group. Parous women also had a decreased melanoma risk (SIR ¼ 0.72; 95% CI: ), while the SIR for nulliparous women did not show a significant increase (SIR ¼ 1.13; 95% CI: ). Women who were younger at first birth (,30 years) had a significantly lower melanoma risk than expected (SIR ¼ 0.29; 95% CI: ). Time-dependent Cox regression analysis Ovarian stimulation for IVF and melanoma risk When directly comparing the IVF group with the non-ivf group (Table III), an age-adjusted hazard ratio (HR) for melanoma of 1.31 (95% CI: ) was found. Adjustment for confounding factors (age at menarche and cause of subfertility) made no significant difference (HR ¼ 1.27; 95% CI: ). Although the HRs for 5 6 and 7 cycles were apparently increased but not statistically significant (HRs of 1.92 and 1.79, respectively), no clear trend emerged with a higher number of IVF cycles (P ¼ 0.14). Accounting for the number of administered IUI cycles did not materially change our IVF risk estimate (adjusted HR without IUI cycles ¼ 1.27, adjusted HR with IUI cycles ¼ 1.19) (data not shown). With a higher number of FSH/hMG ampoules or a higher number of IUI cycles, risks were also non-significantly elevated. Also, when combining the number of IUI cycles with the number of IVF

5 1220 Spaan et al. Table I Population characteristics by melanoma status. Melanoma cases, N 5 93 Non-cases, N Total, N n (%) n (%) n (%) Median years of the follow-up Exposure IVF 72 (77.4) (76.3) (76.3) Non-IVF 21 (22.6) 5929 (23.7) 5950 (23.7) Age at treatment or first visit gynaecologist (years) a 26 6 (6.5) 2573 (10.3) 2579 (10.3) (19.4) 4214 (16.9) 4232 (16.9) (25.8) 6235 (24.9) 6259 (24.9) (20.4) 5832 (23.3) 5851 (23.3) (28.0) 6161 (24.6) 6187 (24.6) Start of IVF treatment or first visit gynaecologist a, (8.6) 1244 (5.0) 1252 (5.0) (15.1) 4404 (17.6) 4418 (17.6) (40.9) 8996 (36.0) 9034 (36.0) (35.5) (41.5) (41.4) Age at the end of follow-up (years) b,45 59 (63.4) 4668 (18.7) 4727 (18.8) (20.4) 7886 (31.5) 7905 (31.5) (12.9) 8121 (32.5) 8133 (32.4).54 3 (3.2) 4340 (17.4) 4343 (17.3) Total no. of IVF cycles 0 cycles 21 (22.6) 5929 (23.7) 5950 (23.7) 1 2 cycle(s) 25 (26.9) 6104 (24.4) 6129 (24.4) 3 4 cycles 24 (25.8) 6198 (24.8) 6222 (24.8) 5 6 cycles 12 (12.9) 2566 (10.3) 2578 (10.3) 7 cycles 7 (7.5) 1680 (6.7) 1687 (6.7) Missing 4 (4.3) 2538 (10.1) 2542 (10.1) Total no. of IUI cycles c 0 cycles 39 (41.9) (43.5) (43.5) 1 6 cycle(s) 23 (24.7) 4447 (17.8) 4470 (17.8) 7 cycles 11 (11.8) 2718 (10.9) 2729 (10.9) Missing 20 (21.5) 6970 (27.9) 6990 (27.8) Total no. of IVF and IUI cycles combined c 0 cycles 8 (8.6) 2307 (9.2) 2315 (9.2) 1 2 cycle(s) 9 (9.7) 3267 (13.1) 3276 (13.1) 3 4 cycles 17 (18.3) 3725 (14.9) 3742 (14.9) 5 6 cycles 13 (14.0) 2405 (9.6) 2418 (9.6) 7 8 cycles 8 (8.6) 1835 (7.3) 1843 (7.3) 9 10 cycles 6 (6.5) 1343 (5.4) 1349 (5.4).10 cycles 12 (12.9) 2934 (11.7) 2946 (11.7) Missing 20 (21.5) 7199 (28.8) 7219 (28.8) Total no. of FSH/hMG ampoules for IVF 0 ampoules 22 (23.7) 6393 (25.6) 6415 (25.6) 1 40 ampoule(s) 15 (16.1) 3206 (12.8) 3221 (12.8) ampoules 14 (15.1) 3587 (14.3) 3601 (14.3) ampoules 9 (9.7) 2283 (9.1) 2292 (9.1).120 ampoules 12 (12.9) 2344 (9.4) 2356 (9.4) Continued

6 Melanoma risk after ovarian stimulation 1221 Table I Continued Melanoma cases, N 5 93 Non-cases, N Total, N n (%) n (%) n (%) Missing 21 (22.6) 7202 (28.8) 7223 (28.8) Subfertility cause d Male factor 17 (18.3) 4856 (19.4) 4873 (19.4) Tubal 30 (32.3) 7084 (28.3) 7114 (28.3) Unexplained 24 (25.8) 4118 (16.5) 4142 (16.5) Hormonal and other factors e 8 (8.6) 4012 (16.0) 4020 (16.0) Missing 14 (15.1) 4945 (19.8) 4959 (19.8) Parity f Nulliparous 42 (45.2) 8481 (33.9) 8523 (34.0) 1 birth 32 (34.4) 7588 (30.3) 7620 (30.4) 2 births 15 (16.1) 6343 (25.4) 6358 (25.3) 3 births 4 (4.3) 2346 (9.4) 2350 (9.4) Missing ( ) 257 (1.0) 257 (1.0) Age at first birth (years) g,25 3 (5.9) 2809 (17.3) 2812 (17.2) (9.8) 3567 (21.9) 3572 (21.9) (56.9) 5694 (35.0) 5723 (35.1) (27.5) 4021 (24.7) 4035 (24.7) Missing ( ) 186 (1.1) 186 (1.1) no., number; IUI, intrauterine insemination. a Start indicates start of IVF treatment for the IVF group and first visit at gynaecologist for controls. b The follow-up ended at date of any cancer diagnosis, date of death, date of completeness of cancer registry (30 June 2009) or date of questionnaire completion for women who refused linkage with the NCR (n ¼ 1091), whichever came first. c Includes both stimulated and non-stimulated IUI cycles. d Defined as the major cause of subfertility, based on information from medical records if available and on information from questionnaires if no medical record information was available. e Hormonal factors includes factors such as ovulation disorders, polycystic ovary syndrome and premature menopause, other factors includes factors such as endometrioses and cervical factors. f Based on the Dutch Municipal Personal Records Database and questionnaires. g Only among parous women (n ¼ ). cycles, women with more cycles had an apparent but statistically nonsignificantly increased melanoma risk (HRs varying between 1.68 and 2.08 for women with 3 4 to.10 cycles), but no clear trend was observed. In a sensitivity analysis, which also included in situ melanomas (n ¼ 21), we also found apparent but statistically non-significantly increased risks, although to a lesser extent, for number of IVF-treatment cycles (adjusted HRs ranging between 1.26 and 1.37) (data not shown). Women who ever used clomiphene did not have a higher melanoma risk compared with women who never used clomiphene (HR ¼ 0.91; 95% CI: ) (data not shown). When considering the follow-up time, the risk of melanoma did not differ significantly either in the first 10 years of the follow-up (HR ¼ 1.52; 95% CI: ) or after 10 or more years of the follow-up (HR ¼ 1.10; 95% CI: ) (data not shown). Among IVF-treated women, there was no difference in melanoma risk between women who had four or more oocytes collected during the first IVF-treatment cycle compared with women who had less than four oocytes collected. Also, the type of luteal phase support, number of FSH/hMG ampoules and use of GnRH did not affect melanoma risk (Table III). In an analysis stratified by parity, parous women treated with IVF had no increased melanoma risk (HR ¼ 0.97; 95% CI: ), while nulliparous women treated with IVF had a non-significantly increased melanoma risk (HR ¼ 2.17; 95% CI: ) (data not shown). However, the interaction term between parity and IVF was not significant (P ¼ 0.10). Reproductive factors and melanoma risk Nulliparous women had an apparent but statistically non-significantly increased risk of melanoma compared with parous women (age-adjusted HR ¼ 1.22; 95% CI: ) (Table IV). Parous women who were 30 years or older at first birth had a significantly increased risk compared with women who were younger than 30 years at first birth (30 34 years: age-adjusted HR ¼ 4.57; 95% CI: ,.34 years: age-adjusted HR ¼ 2.98; 95% CI: ). Nulliparous women also had an increased risk compared with women who had their first birth before age 30 (age-adjusted HR ¼ 3.20; 95% CI: ). Age at menarche, OC use and cause of subfertility did not appear to increase melanoma risk. Timing of parity with respect to IVF treatment appeared to influence risk of melanoma (Table V). A lower melanoma risk was found among women who already had children before the start of their subfertility

7 1222 Spaan et al. Table II Melanoma risk compared with the general Dutch population. Comparison with general population... Person-years Obs Exp SIR 95% CI Exposure Non-IVF IVF Total no. of IVF cycles 1 2 cycle(s) cycles cycles cycles Missing Total no. of IUI cycles a 0 IUI cycles IUI cycle(s) IUI cycles Missing Total no. of IVF and IUI cycles combined a 0 cycles cycle(s) cycles cycles cycles cycles cycles Missing Total no. of FSH/hMG ampoules for IVF 0 ampoules ampoule(s) ampoules ampoules ampoules Missing Follow-up interval (years) Follow-up interval IVF group (years) Follow-up interval non-ivf group (years) Continued

8 Melanoma risk after ovarian stimulation 1223 Table II Continued Comparison with general population... Person-years Obs Exp SIR 95% CI Year of first treatment or visit gynaecologist b Subfertility cause c Male factor Tubal Unexplained Hormonal and other factors d Missing Parity Nulliparous Parous Missing Age at first birth (years) e, Missing Obs, observed; Exp, expected; SIR, standardized incidence ratio; CI, confidence interval; no., number; IUI, intrauterine insemination. a Includes both stimulated and non-stimulated IUI cycles. b Start indicates start of IVF treatment for the IVF group and first visit at gynaecologist for controls. c Defined as the major cause of subfertility, based on information from medical records if available and on information from questionnaires if no medical record information was available. d Hormonal factors includes factors such as ovulation disorders, polycystic ovary syndrome and premature menopause, other factors includes factors such as endometrioses and cervical factors. e Only among parous women (n ¼ ). treatment (age-adjusted HR ¼ 0.42; 95% CI: ) compared with women who remained nulliparous, independent of exposure status (IVF and non-ivf groups together). Additional adjustment for age at first birth did not materially affect this finding (data not shown). In women who were nulliparous at the start of their treatment but became parous after subfertility treatment such a decreased risk was not found (age-adjusted HR ¼ 0.85; 95% CI: ). When stratifying for subfertility treatment (IVF and non-ivf), risk estimates were comparable but no longer statistically significant. Discussion In this large subfertile population, we found that ovarian stimulation for IVF did not appear to affect melanoma risk, even after a follow-up period of more than 20 years. Larger numbers of IVF or IUI cycles appeared to be associated with increased risk but none of these associations was statistically significant. A later age at first birth was statistically significantly associated with increased melanoma risk. Our results are consistent with the results reported in a meta-analysis by Gandini et al. (2011) who also found no increased risk after use of FDs. However, their results were only based on three studies investigating FD use and melanoma risk. In a more recently published review on the association between FD use and melanoma risk the authors concluded that definitive answers cannot be given yet because most study results are inconsistent, likely due to a number of limitations in the studies (Tomao et al., 2014). Since our cohort study is one of the largest published so far, with long-term and complete follow-up, a subfertile comparison group and detailed IVF-treatment data, our results add important information to the available evidence. To date, 11 cohort studies reported on the association between FDs and melanoma risk (Brinton et al., 1989; Rossing et al., 1995; Venn et al., 1995; Modan et al., 1998; Young et al., 2001; Althuis et al., 2005; Hannibal et al., 2008; Calderon-Margalit et al., 2009; Silva Idos et al., 2009; Yli-Kuha et al., 2012; Stewart et al., 2013). However, only three of these directly assessed the effect of ovarian stimulation for IVF (Venn et al., 1995; Yli-Kuha et al., 2012; Stewart et al., 2013). In the first study, no increased melanoma risk was observed in women treated with IVF (Venn et al., 1995). However, this study included only 16 melanoma cases. Furthermore, in a Finnish cohort of women with 21 melanoma cases, of whom 9175 subfertile women were treated with

9 1224 Spaan et al. Table III Exposure to IVF and melanoma risk; time-dependent Cox regression analyses. Total no. No. of cancers Age-adjusted HR (95% CI) Adjusted HR a (95% CI) Exposure Non-IVF (reference) 1.0 (reference) IVF ( ) 1.27 ( ) Total no. of IVF cycles b Non-IVF (reference) 1.0 (reference) 1 2 cycle(s) ( ) 1.36 ( ) 3 4 cycles ( ) 1.47 ( ) 5 6 cycles ( ) 1.92 ( ) 7 cycles ( ) 1.79 ( ) Missing ( ) 0.56 ( ) Total no. of FSH/hMG ampoules for IVF 0 ampoules (reference) 1.0 (reference) 1 40 ampoule(s) ( ) 1.52 ( ) ampoules ( ) 1.28 ( ) ampoules ( ) 1.30 ( ).120 ampoules ( ) 1.70 ( ) Missing ( ) 1.11 ( ) Total no. of IUI cycles c 0 cycles (reference) 1.0 (reference) 1 6 cycle(s) ( ) 1.64 ( ) 7 cycles ( ) 1.41 ( ) Missing ( ) 0.71 ( ) Total no. of IVF and IUI cycles b,c 0 cycles (reference) 1.0 (reference) 1 2 cycle(s) ( ) 0.91 ( ) 3 4 cycles ( ) 1.70 ( ) 5 6 cycles ( ) 2.08 ( ) 7 8 cycles ( ) 1.68 ( ) 9 10 cycles ( ) 1.83 ( ).10 cycles ( ) 1.76 ( ) Missing ( ) 0.81 ( ) Within IVF-treated women (n ¼ ) No. of oocytes d,4 oocytes (reference) 1.0 (reference) 4 13 oocytes ( ) 1.12 ( ) 14 oocytes ( ) 1.04 ( ) Missing ( ) 0.56 ( ) GnRH No (reference) 1.0 (reference) Yes ( ) 1.96 ( ) Missing ( ) 1.26 ( ) Total no. of FSH/hMG ampoules for IVF 0 40 ampoules(s) (reference) 1.0 (reference) ampoules ( ) 0.89 ( ) ampoules ( ) 0.91 ( ).120 ampoules ( ) 1.21 ( ) Missing ( ) 0.81 ( ) Continued

10 Melanoma risk after ovarian stimulation 1225 Table III Continued Total no. No. of cancers Age-adjusted HR (95% CI) Adjusted HR a (95% CI) Luteal phase support No luteal phase support (reference) 1.0 (reference) hcg ( ) 0.98 ( ) Progesterone ( ) 0.61 ( ) Missing ( ) 0.60 ( ) no., number; HR, hazard ratio; CI, confidence interval; IUI, intrauterine insemination. a Additionally adjusted for age at menarche and cause of subfertility. b Due to the availability of dates, the number of women per IVF cycle category differs somewhat in the time-dependent data compared with the fixed data. c Includes both stimulated and non-stimulated IUI cycles. d No. of oocytes collected at first IVF-treatment cycle. Table IV Reproductive factors and melanoma risk; time-dependent Cox regression analyses. Total no. No. of Age-adjusted cancers HR (95% CI)... Parity a Parous (reference) Nulliparous ( ) Age at first birth (years) a, (reference) ( ) ( ) Nulliparous ( ) Age at menarche (years), (reference) ( ) ( ) Missing ( ) Subfertility cause b Male factor (reference) Tubal ( ) Unexplained ( ) Hormonal and ( ) other factors c Missing ( ) Duration of OC use Never use (reference) 5 years ( ).5 years ( ) Missing ( ) no., number; HR, hazard ratio; CI, confidence interval; OC, oral contraceptive. a Due to the availability of dates, the number of parous and nulliparous women differ somewhat in the time-dependent data compared with the fixed data. b Defined as the major cause of subfertility, based on information from medical records if available and on information from questionnaires if no medical record information was available. c Hormonal factors includes factors such as ovulation disorders, polycystic ovary syndrome and premature menopause, other factors includes factors such as endometrioses and cervical factors. IVF, melanoma risk was not increased compared with a control group of randomly selected women from the general population (Yli-Kuha et al., 2012). The most recently published study from Australia (Stewart et al., 2013) also reported no increased risk of melanoma in IVF-treated women compared with a subfertile control group. Melanoma risk did also not differ between women treated with two or more IVF-treatment cycles and women who only had one IVF-treatment cycle (Stewart et al., 2013). Our study is the first one able to investigate the dose response relationship for three or more IVF-treatment cycles. We observed apparent but statistically non-significantly increased melanoma risks (HRs ranging between 1.36 and 1.92), but no clear trend. This may be a chance finding, but alternatively, may point to a true risk increase, as our power for subgroup analyses was limited. Somewhat surprisingly, we observed a decreased risk of melanoma in women with missing data on the number of IVF cycles. However, these results are probably attributable to chance since all women were followed up through linkage with the NCR and we did not spend more efforts to complete the follow-up for patients with known cycle numbers. Larger cohort studies with prolonged follow-up are needed to provide more information on the melanoma risk after many IVF cycles. Several other studies only reported on the use of FDs outside the IVF setting (or in four studies, without information on IVF treatments) and melanoma risk (Brinton et al., 1989; Rossing et al., 1995; Modan et al., 1998; Young et al., 2001; Althuis et al., 2005; Hannibal et al., 2008; Calderon-Margalit et al., 2009; Silva Idos et al., 2009). Although most studies did not find an association between FD use and melanoma risk, some studies reported increased risks in subgroups of women according to type of FD used, i.e. for gonadotrophins (Hannibal et al., 2008), GnRH (Hannibal et al., 2008) and clomiphene (Rossing et al., 1995). In our study, we did not find higher melanoma risks in women who used clomiphene, gonadotrophins or GnRH. In our study, we also accounted for the number of IUI cycles, when assessing the effect of IVF cycles. Although levels of oestrogen and progesterone are much higher in IVF-stimulated cycles than in IUI cycles, it is important to take these IUI cycles into account since both women in the non-ivf and IVF groups could be treated with IUI. Previous studies did not include IUI cycles. Since the questionnaire did not distinguish between stimulated and non-stimulated IUI cycles, we probably overestimated the number of cycles given. However, medical record data

11 1226 Spaan et al. Table V Timing of parity and subfertility treatment and melanoma risk; Cox regression analyses. Total no. No. of cancers Age-adjusted HR (95% CI) Timing parity Nulliparous (reference) Parous before start treatment ( ) Parous after start treatment ( ) Exposure and timing parity Non-IVF and parous after start treatment (reference) Non-IVF and parous before start treatment ( ) Non-IVF and nulliparous ( ) IVF and nulliparous ( ) IVF and parous before start treatment ( ) IVF and parous after start treatment ( ) Timing parity within IVF-treated women a IVF and nulliparous (reference) Parous before IVF treatment ( ) Parous after IVF treatment ( ) no., number; HR, hazard ratio; CI, confidence interval. a Only in IVF-treated women, comparing parous women (with a maximum of one child) with nulliparous women. showed that 80% of the IUI cycles given were stimulated cycles, rendering substantial overestimation unlikely. Associations between cause of subfertility and melanoma risk were also investigated. Women who were classified as having a hormonal or other cause as their main reason for subfertility had a lower melanoma risk than women from the general population. Since this is a very heterogeneous group of women, comprising women with polycystic ovarian syndrome, early menopause and cervical problems, it is difficult to interpret this result. However, the decreased risk was mainly driven by the women with a hormonal cause of their subfertility; no melanoma cases were observed in this group. We also investigated the influence of reproductive factors on melanoma risk and found no increased melanoma risk in nulliparous women compared with parous women. However, this might be due to the relatively old age at first birth (median: 31 years) in our cohort, since we observed a significantly increased risk in nulliparous women when compared with parous women who gave birth before age 30. Furthermore, we observed that an older age at first birth increases melanoma risk. However, due to the rather small number of women within the early age at first birth categories risk estimates were very much dependent on the cut-off. Women who were 30 years or older at first birth had a significantly increased risk, while women in the category with a first birth from 31 to 34 years had a non-significantly increased melanoma risk (age-adjusted HR ¼ 1.50; 95% CI: ) compared with women who had their first birth before age 31 (data not shown). Although our result regarding age at first birth is in line with previously published data, an explanation for this effect remains unclear (Gandini et al., 2011). Possibly part of the association between melanoma risk and older age at first birth could be explained by residual confounding by lifestyle factors. Kaae et al. (2007) found decreased risks for parity and early age at first full-term pregnancy both in women and men. Therefore, they suggested that the effects may be due to other lifestyle factors. When we adjusted for BMI and educational level, our risk estimates did not change, rendering residual confounding by these factors unlikely. We also investigated whether timing of parity with respect to IVF treatment influenced melanoma risk and found that giving birth before the start of the subfertility treatment was associated with a decreased melanoma risk, irrespective of IVF treatment. Giving birth after treatment did not increase the risk. Stewart et al. (2013) compared, within IVF-treated women, parous women (with a maximum of one birth) who only gave birth before IVF and parous women who only gave birth after IVF with IVF-treated women who remained nulliparous and found in both groups of parous women significantly increased melanoma risks. We could not replicate this finding as we did not observe an increased risk in women who gave birth before (HR ¼ 0.92) or after IVF (HR ¼ 0.77). Our study has several strengths and limitations. Advantages are the large size of the cohort and prolonged follow-up. Selection bias could be ruled out since we were able to link 96% of our cohort (both responders and non-responders to the questionnaire) with the NCR. Detailed information about subfertility treatments and nearly complete data about parity and age at first birth were available. Therefore, we were able to investigate the association between ovarian-stimulating drugs used for IVF and melanoma risk but also to investigate in detail the association between the number of IVF treatments given and number of FSH/ hmg ampoules used. Furthermore, our study is one of the few studies (Rossing et al., 1995; Young et al., 2001; Althuis et al., 2005; Silva Idos et al., 2009) with a comparison group of subfertile women, in addition to a comparison group from the general population. Such a comparison group is important since IVF-treated women differ from the general population with regard to several risk factors for melanoma (cause of subfertility, parity and age at first birth). To our knowledge, our study is the first one investigating the separate and combined effects of IUI and IVF cycles. However, since IUI could also

12 Melanoma risk after ovarian stimulation 1227 be used in the comparison group of women not treated with IVF, the proportion of women not receiving any type of ovarian stimulation was rather small. Unfortunately, our risk factor questionnaire did not include questions about skin type and sun exposure habits and therefore we could not examine the potential confounding effects of these variables. It has been hypothesized in other studies of FD use and melanoma risk that there may be some confounding by sun exposure, due to modest sunbathing habits among women with children. However, in a study from Kvaskoff et al. (2011) investigating several reproductive risk factors on melanoma risk, adjustment for ambient UV radiation did not alter the risk estimates. Furthermore, especially sunburns at younger ages seem to increase melanoma risk (Oliveria et al., 2006). Therefore, also in view of the late age at first birth in our cohort, it is unlikely that confounding by sun exposure influenced our estimates. Finally, despite our large cohort, the number of melanoma cases was rather small, especially in our comparison group, which hampered subgroup analyses according to the type of stimulation, parity and follow-up duration. In conclusion, our results suggest that ovarian stimulation for IVF does not increase melanoma risk, while reproductive factors do seem to play a role in the development of melanomas. Our results are reassuring for women treated with IVF and women who are contemplating to start IVF. Supplementary data Supplementary data are available at Acknowledgements The authors thank the participants of the OMEGA project, without whom this study would not have been possible. We thank the medical registries of the participating clinics for making patient selection possible, and all attending physicians for providing access to their patients medical files. We are especially grateful to H. Klip for all her efforts in the recruitment phase of the study. The authors also acknowledge the Netherlands Cancer Registry for providing the cancer incidence data and the Dutch Municipal Personal Records Database for parity data. Authors roles F.E.v.L. and C.W.B. designed the Ovarian stimulation and gynaecological disorders study and were principal investigators of the study. F.E.v.L. also coordinated statistical analyses, contributed to interpretation of the data and drafting of the manuscript. C.W.B. contributed to interpretation of the data and drafting of the manuscript. M.Sp. did data preparation and statistical analyses, contributed to the study design, the data collection, the interpretation of the data, and drafted the manuscript. A.W.vd.B. did data preparation and statistical analyses, contributed to data collection, the interpretation of the data, and drafting of the manuscript. M.Sc. contributed to the statistical analyses and interpretation of the data. T.M.M contributed to the study design, did data preparation, and drafting of the manuscript. All members of the OMEGA-project group contributed to the delivery of data. All authors contributed to critical revisions of the draft manuscript. All authors approved the final version of the article. Funding This study was supported by grants from the Dutch Cancer Society (NKI ), the Health Research and Development Counsel ( ) and the Dutch Ministry of Health. Conflict of interest None declared. Appendix The OMEGA-project group includes R. Schats (VU University Medical Center, Amsterdam), C.B. Lambalk (VU University Medical Center, Amsterdam), M. Kortman (University Medical Center, Utrecht), J.S.E. Laven (Erasmus Medical Center, Rotterdam), C.A.M. Jansen (Diaconessenhuis Voorburg), F.M. Helmerhorst (Leiden University Medical Center), B.J. Cohlen (Isala Clinics Zwolle), D.D.M. Braat (Radboud University Nijmegen Medical Center), J.M.J. Smeenk (St Elisabeth Hospital, Tilburg), A.H.M. Simons (University Medical Center, Groningen), F. van der Veen (Academic Medical Center, Amsterdam), J.L.H. Evers (Academic Hospital, Maastricht) and P.A. van Dop (Catharina Hospital, Eindhoven). References Althuis MD, Scoccia B, Lamb EJ, Moghissi KS, Westhoff CL, Mabie JE, Brinton LA. Melanoma, thyroid, cervical, and colon cancer risk after use of fertility drugs. Am J Obstet Gynecol 2005;193: Breslow NE, Day NE. Comparisons among exposure groups. Statistical Methods in Cancer Research, Volume II. Lyon: International Agency for Research on Cancer, 1987, Brinton LA, Melton LJ III, Malkasian GD Jr, Bond A, Hoover R. Cancer risk after evaluation for infertility. Am J Epidemiol 1989;129: Calderon-Margalit R, Friedlander Y, Yanetz R, Kleinhaus K, Perrin MC, Manor O, Harlap S, Paltiel O. Cancer risk after exposure to treatments for ovulation induction. Am J Epidemiol 2009;169: De Boer EJ, Den Tonkelaar I, te Velde ER, Burger CW, van Leeuwen FE. Increased risk of early menopausal transition and natural menopause after poor response at first IVF treatment. Hum Reprod 2003;18: De Boer EJ, van Leeuwen FE, Den Tonkelaar I, Jansen CA, Braat DD, Burger CW. [Methods and results of in-vitro fertilisation in the Netherlands in the years ]. Ned Tijdschr Geneeskd 2004; 148: de Giorgi V, Mavilia C, Massi D, Gozzini A, Aragona P, Tanini A, Sestini S, Paglierani M, Boddi V, Brandi ML et al. Estrogen receptor expression in cutaneous melanoma: a real-time reverse transcriptase-polymerase chain reaction and immunohistochemical study. Arch Dermatol 2009; 145: Gandini S, Iodice S, Koomen E, Pietro AD, Sera F, Caini S. Hormonal and reproductive factors in relation to melanoma in women: current review and meta-analysis. Eur J Cancer 2011;47: Hannibal CG, Jensen A, Sharif H, Kjaer SK. Malignant melanoma risk after exposure to fertility drugs: results from a large Danish cohort study. Cancer Causes Control 2008;19: Hollestein LM, van den Akker SA, Nijsten T, Karim-Kos HE, Coebergh JW, de Vries E. Trends of cutaneous melanoma in The Netherlands: increasing