A critical analysis of the accuracy, reproducibility, and clinical utility of histologic endometrial dating in fertile women

Transcription

1 FERTILITY AND STERILITY VOL. 81, NO. 5, MAY 2004 Copyright 2004 American Society for Reproductive Medicine Published by Elsevier Inc. Printed on acid-free paper in U.S.A. Received July 28, 2003; revised and accepted November 21, Supported by the Nova Carta Foundation at the University of North Carolina, Chapel Hill. Presented in part at the American Society for Reproductive Medicine 58th Annual Meeting, Seattle, Washington, October 2002 and in part at the United States and Canadian Academy of Pathologists 92nd Annual Meeting, Washington, DC, March Reprint requests: Michael J. Murray, M.D., Northern California Fertility Medical Center, 406 1/2 Sunrise Avenue, Suite 310, Roseville, California (FAX: ; a Division of Reproductive Endocrinology and Infertility, Departments of Obstetrics and Gynecology, Pathology and Laboratory Medicine, and Biostatistics, University of North Carolina. b Present address: Northern California Fertility Medical Center, Roseville, California. c Department of Pathology, M.S. Hershey Medical Center, Pennsylvania State University, Hershey, Pennsylvania. d Present address: Division of Reproductive Endocrinology and Infertility, Greenville Hospital Systems, Greenville, South Carolina. e Present address: Wilmington Pathology Associates, P.A., Wilmington, North Carolina. f Present address: InCyte Pathology, Spokane, Washington. g Department of Pathology, Oregon Health Sciences /04/$30.00 doi: /j.fertnstert A critical analysis of the accuracy, reproducibility, and clinical utility of histologic endometrial dating in fertile women Michael J. Murray, M.D., a,b William R. Meyer, M.D., a Richard J. Zaino, M.D., c Bruce A. Lessey, M.D., Ph.D., a,d Debra B. Novotny, M.D., a,e Karen Ireland, M.D., f,g Donglin Zeng, Ph.D., a and Marc A. Fritz, M.D. a University of North Carolina, Chapel Hill, North Carolina Objective: To refine or redefine the traditional histologic criteria used to date the secretory phase endometrium. Design: Randomized, observational study. Setting: Academic clinical research center. Patient(s): One hundred and thirty healthy, regularly cycling, fertile volunteers, aged 18 to 35 years. Intervention(s): Patients were randomized to undergo endometrial sampling and measurement of serum estradiol and progesterone 1 to 14 days after the midcycle urinary luteinizing hormone surge. Three gynecologic histopathologists objectively scored each tissue specimen for 32 distinct histologic features and dated the endometrium using traditional histologic criteria. Main Outcome Measure(s): The 32 features were evaluated for [1] temporally dependent variation, [2] the amplitude of variations in score observed across the secretory phase, and [3] interobserver variability. Additionally, traditional dating criteria were analyzed. Result(s): The traditional endometrial histologic dating criteria are much less temporally distinct and discriminating than originally described, due to considerable intersubject, intrasubject, and interobserver variability. Neither traditional dating criteria nor any combination of the best performing histologic features identified by our objective and systematic analyses could reliably distinguish any specific cycle day or narrow interval of days. Conclusion(s): Histologic endometrial dating does not have the accuracy or the precision necessary to provide a valid method for the diagnosis of luteal phase deficiency or to otherwise guide the clinical management of women with reproductive failure. (Fertil Steril 2004;81: by American Society for Reproductive Medicine.) Key Words: Endometrium, endometrial biopsy, luteal phase deficiency, secretory phase University, Portland, Oregon. Over five decades have elapsed since criteria for dating the endometrial biopsy were established by Noyes, Hertig, and Rock (1). Their seminal work was the leading article published in the inaugural issue of Fertility and Sterility and has remained among the most often cited articles from the journal (2). More importantly, the histologic dating criteria described by Noyes et al. (1) have endured as the gold standard among methods in clinical practice for the evaluation of luteal function and for diagnosis of luteal phase deficiency (LPD). Jones (3, 4) was the first to hypothesize that delayed endometrial maturation resulting from inadequate corpus luteum progesterone (P) production might be a cause of infertility and early pregnancy loss. Whereas estimates of the prevalence of LPD have ranged from 3.5% to 60%, a prevalence between 5% and 10% in infertile women and approximately 10% to 25% in those with a history of recurrent early pregnancy loss is generally accepted (5 8). Since 1950, advances in reproductive endocrinology, molecular biology, and statistical anal- 1333

2 TABLE 1 Exclusion criteria. Prior history of infertility or recurrent spontaneous abortions (three or more) Previous uterine surgery (other than cesarean section, postpartum dilation and curettage, or elective pregnancy termination) Abnormal uterine bleeding Leiomyomata in or near the uterine cavity (as judged by transvaginal ultrasound examination) Cervical stenosis Persistent or recurrent abnormal cervical cytology Previous cervical conization or loop electrocautery excision procedure Adrenal or ovarian endocrinopathy Known or suspected endometriosis Use of an intrauterine device or any form of hormonal contraception during the 6 months preceding study Current lactation or efforts to conceive Chronic diseases that may or may not require chronic medications (e.g., systemic lupus erythematosus and diabetes mellitus) except for: Minor depression and treatment with a single prescribed antidepressant (n 9) Hypothyroidism with documented euthyroid indices while receiving thyroxine replacement therapy (n 5) Hypertension effectively controlled with one prescribed antihypertensive agent (n 2) Acne under treatment with a single prescribed oral antibiotic (n 1) ysis have fueled a number of challenges to the dogma of endometrial dating, all based on perceived methodological flaws in the original study. The first and most fundamental criticism of the classic endometrial histologic dating criteria is that the normal standard was based on an analysis of tissue specimens obtained not from normally cycling parous women, but from infertile women (1). The original study population was therefore abnormal by definition. Second, the postovulatory day on which Noyes et al. (1) obtained tissue specimens was estimated in retrospect, by assuming that the secretory phase is uniformly 14 days in duration and subtracting the number of days that elapsed between biopsy and onset of the next menses. As the investigators themselves correctly acknowledged, since there is some variation in the secretory phase, our dates are somewhat arbitrary (1). Numerous later studies demonstrated that the normal secretory phase does indeed vary in duration (9 13). Retrospective estimates based on the onset of next menses correlate poorly with the time of ovulation (65%) and also ignore any effect that biopsy itself may have on when menses next begins, or when it is perceived to start (12, 14, 15). The method used to establish the original endometrial dating criteria was the best available at the time, but clearly inferior to the contemporary alternatives offered by modern hormone assay methodology and ultrasound technology. Observed and expected histologic dates correlate much better when the days of ovulation and tissue collection are defined prospectively, using the midcycle urinary luteinizing hormone (LH) surge (85%) or collapse of the dominant follicle as observed by serial transvaginal ultrasound examinations (96%) as a reference point (14, 16, 17). A third important weakness of the traditional histologic criteria relates to the inherent subjectivity of endometrial dating. In fact, immediately after their original publication, the dating criteria proposed by Noyes et al. (1) were criticized by high authority in an editorial by Novak (18), and Noyes felt obliged to defend the accuracy of endometrial dating in a second study (19). Over the decades since, based on the results of numerous studies and analyses of intraobserver and interobserver variations in histologic interpretation, dating differences within a 2-day interval generally have come to be considered normal and acceptable (1, 4, 19 22). Unfortunately, even this seemingly minor degree of variation is great enough to affect diagnosis and management in 20% to 40% of individual women (23 26). A fourth source of controversy involves the optimal time to obtain a tissue sample. Traditionally, biopsy for evaluation of endometrial maturation is performed within 3 days of the anticipated onset of menses, at a time when sampling might best reflect the cumulative effects of corpus luteum function and dating can be based on a greater number of histologic features (15, 27 29). Others have argued that earlier sampling during the midsecretory phase is more relevant, because it coincides with the putative window of embryo implantation and may identify abnormalities of endometrial maturation that might otherwise go undetected (30 32). Our study was conducted to reexamine critically the accuracy and clinical utility of histologic endometrial dating. We designed our study to address each of the fundamental criticisms and methodological weaknesses of the classic study by Noyes et al. (1) and to identify which specific histologic features have the greatest discriminatory value for dating purposes. Our ultimate goal was to refine or redefine the histologic dating criteria used in the clinical evaluation of luteal function and for the diagnosis of LPD. MATERIALS AND METHODS Study Design and Volunteers Study candidates were recruited via public advertisements within the region. All volunteers provided written informed consent and were compensated for their participation in accordance with the study protocol and compensation schedule approved by the Committee for the Protection of Human Subjects at the University of North Carolina at Chapel Hill. All 130 participating women were in good health, aged 18 to 35 years, had regular menstrual cycles at 25 to 35 day intervals, and were proven fertile (at least one previous live birth). Exclusion criteria are listed in Table Murray et al. Endometrial dating revisited Vol. 81, No. 5, May 2004

3 Tissue Collection All tissue specimens were collected and processed at the University of North Carolina at Chapel Hill between 1997 to Study participants monitored their urinary LH excretion daily using a commercial ovulation predictor kit (Ovu- Quick One-Step Ovulation Predictor; Quidel, San Diego, CA) from cycle day 10 until the midcycle urinary LH surge was identified. Cycle day 1 was defined by the onset of normal menstrual flow for the individual and testing was performed between the hours of 10:00 AM and 3:00 PM daily. On the day of urinary LH surge detection, the women were assigned to return 1 to 14 days later according to a computerized block randomization scheme. Accepting that the colorimetric reactions observed during monitoring of urinary LH excretion may not remain reliably stable over time, test results were reviewed and confirmed by the investigators on the day of urinary LH surge detection or endometrial biopsy in an effort to ensure compliance and improve accuracy. On the assigned luteal day (LD), a transvaginal ultrasound examination was performed immediately before collection of the endometrial biopsy to exclude women having any obvious uterine cavity lesion or deep intramural leiomyomata. Endometrial biopsies were performed using an endometrial suction aspiration device (GynoSampler; Gynétics Medical Products N.V., Hamont-Achel, Belgium). A venous blood sample was obtained on the day of biopsy from all women. Sera were obtained, aliquoted, and stored at 80 C for later measurement of estradiol (E 2 ) and P. Histologic Evaluation The majority of each endometrial tissue specimen obtained was fixed in 10% buffered formalin and paraffin embedded before sectioning. For each specimen, three consecutive 4- to 5-mm sections were cut, mounted, and stained with hematoxylin and eosin for subsequent histologic examination. Three independent observers (K.I., D.B.N., and R.J.Z.), all recognized experts in gynecologic histopathology and having more than 75 years of combined experience, evaluated the endometrium. One of each trio of sections from each tissue specimen was distributed to the three pathologists, who were blinded to the day on which the specimen was obtained. Because our study was designed to critically analyze and refine or redefine the current gold standard (the criteria of Noyes et al. [1]) and no other histologic standard has wide acceptance, it was not possible to validate formally our descriptive scoring system or chosen histologic criteria before conducting the study. We included the eight histologic features subjectively described by Noyes et al. (1) and all other characteristics judged to have potential value for dating the endometrium. The histologic features to be evaluated and the scoring system used to describe them were established by consensus among the three pathologists before the study began. Endometrial tissue specimens were first examined to determine whether the functionalis layer was sufficient for evaluation. Thirty-two individual characteristics of the endometrial glands, epithelium, and stroma were evaluated separately and assigned an ordinal score (Table 2). Tissue specimens were also dated using the traditional criteria of Noyes et al. (1). Endometrial development was considered delayed if at least two pathologists assigned a postovulatory date (POD) more than 2 days behind the actual LD on which the biopsy was performed; otherwise, histologic development was regarded as normal. Five tissue specimens were assigned a POD by at least two pathologists that was advanced more than 2 days beyond the actual LD and were considered normal; the LD on which these five specimens were obtained varied widely (LD 1, 1; LD 6, 2; LD 8, 1; and LD 11, 1). For standardization and to facilitate a valid direct comparison to the histologic dating criteria established in the Noyes study, ovulation was assumed to have occurred on the day after the urinary LH surge was detected. Our assumption in this regard is consistent with observations in previous studies that have examined the temporal relationship between the urinary LH surge and collapse of the dominant ovarian follicle as observed by serial transvaginal ultrasound examinations during the late follicular phase (14, 33, 34). Therefore, the day of ovulation defined as LD 0 and LD 1 corresponds to what Noyes et al. (1) described as cycle day 15 or POD 1. Steroid Radioimmunoassay Serum E 2 and P levels were determined by highly specific solid-phase RIA using commercial kits in accordance with the manufacturer s instructions (Coat-A-Count; Diagnostic Products Corporation, Los Angeles, CA). The sensitivities of the hormone assays were 8 pg/ml and 0.02 ng/ml for E 2 and P, respectively. In this study, the intra-assay and interassay coefficients of variation were 7.4% and 8.2% (E 2 ), and 5.4% and 5.0% (P), respectively, all values consistent with those described by the manufacturer. Statistical Analysis To be included in our analyses, the functionalis layer of the endometrial tissue specimen must have been judged adequate for evaluation by all three pathologists. All data derived from any individual woman were excluded from analysis if the tissue specimen was obtained on LD 4 or later, but corroborative evidence of ovulation by both histologic (secretory) and endocrinologic criteria (P 3.0 ng/ml) was lacking. To determine which histologic features had the greatest value for discriminating specific luteal days, three separate statistical analyses were performed. The operational premise was that those criteria most useful for endometrial dating would have the following characteristics: [1] a dynamic, temporally dependent pattern of variation across the secre- FERTILITY & STERILITY 1335

4 TABLE 2 Histologic features of the endometrium functionalis that were scored individually. Variance attributable to histologic change across time a Agreement among pathologists b Histologic feature c R 2 (%) Rank r Rank 1. Gland/Stroma ratio d Gland characteristics Gland configuration 2. Straight Coiled Serrated Gland luminal diameter 5. Narrow Straight Gland luminal contours 7. Smooth communal borders Convex borders/apical blebs Irregular, concave, or frayed borders Epithelial characteristics 10. Epithelial mitoses e Tall columnar Low columnar Cuboidal Secretions 14. Subnuclear cytoplasmic vacuoles Supranuclear/apical cytoplasmic vacuoles Fine cytoplasmic vacuolization Luminal secretions Luminal secretions loose Luminal secretions inspissated Nuclear location 20. Pseudostratified Linear/midline Linear/basal Stroma characteristics 23. Stromal mitoses e Edema f Spiral arterioles e Predecidual changes 26. Periarteriolar e Subcapsular e Remaining stroma e Infiltrate 29. Lymphocyte e Neutrophil e Breakdown 31. Gland Stroma Note: For purposes of statistical comparisons all histologic features were normalized on a scale from 0 to 1. The actual scoring schemas used by the three pathologists for a given histologic feature are listed in notes. a R 2 was determined by one-way ANOVA and reflects the proportion of day-to-day variation that can be explained by histologic changes. The effect of differences between women on day-to-day variation can be calculated from the formula 1 R 2. b r was determined using the mixed model and represents the agreement (correlation) among the three pathologists, where r 1 suggests complete agreement, and r 1suggests complete disagreement. c Unless otherwise indicated, features were scored on an ordinal scale as follows: 0 0%, 1 1% to 25%, 2 26% to 50%, 3 51% to 75%, and 4 76% to 100%. d Ordinal scale: 1 1:1, 2 1:1, and 3 1:1. e Ordinal scale: 1 absent, 2 sparse/focal, and 3 diffuse/numerous. f Ordinal scale: 1 absent, 2 sparse, 3 variable, and 4 diffuse Murray et al. Endometrial dating revisited Vol. 81, No. 5, May 2004

5 tory phase, [2] a relatively large difference between the minimum and maximum scores observed, and [3] relatively little interobserver variability. For statistical comparisons, all histologic feature scores were normalized on a scale of 0 to 1. The raw scoring schema used by the three pathologists for each histologic feature is detailed in Table 2. To compare the extent of day-to-day (temporally dependent) variation among the 32 histologic features examined, each LD was treated as a separate category and one-way analysis of variance (ANOVA) was used to analyze the relationship between the mean scores of the three pathologists for each feature by LD. Factors having potential influence on day-to-day variation are interobserver variability, differences between women, and the extent to which each histologic feature changes across time. In this analysis, the use of the mean score effectively eliminates interobserver variability, leaving only the latter two factors as a source of variance. The R 2 represents the proportion of variation attributable to histologic changes across time; greater R 2 values suggest greater discriminating value. The effect of differences between women on day-to-day variation can be calculated from the formula 1 R 2. To facilitate comparison, features were ranked according to their corresponding R 2 values (see Table 2). To identify which features exhibited the largest absolute changes in score across the secretory phase, a mixed model (multivariate analysis) was employed. Each individual pathologist s score was considered, as were other potentially relevant variables including age, body mass index, and parity. To define the pattern for each feature across the secretory phase, LD was treated as continuous time, and cubic polynomial curves were fitted. The values used to compare features in this analysis were the ranges in score (maximum minus minimum score) observed across time; greater ranges suggest greater discriminating value. Curves were generated for each of the 32 histologic features examined, and those corresponding to the features described by Noyes et al. (1) provide the means for direct comparison of our objectively derived curves with the subjective illustrations that appeared in their original publication (Fig. 1). Interobserver variability for each of the 32 histologic features was analyzed using the same mixed model approach. In this third analysis, r values represent the extent of agreement (correlation) among the histologic scores of the three pathologists for each feature across the secretory phase, where r 1 suggests complete agreement and r 1 suggests complete disagreement. To facilitate comparison, features were ranked according to their corresponding r values (see Table 2). All statistical analyses were performed with SAS software programs (SAS Institute Inc., Cary, NC). RESULTS Among the 153 women originally recruited to participate, 130 completed the study. Thirteen women withdrew from the study before the biopsy was obtained, and another 10 were excluded. Eight women failed to detect a midcycle urinary LH surge. Cervical stenosis prevented tissue sampling in one woman and another conceived before the biopsy was performed. Among the 130 endometrial tissue specimens obtained, 9 failed to meet the criteria for analysis. Of these nine, five were excluded because one or more of the three pathologists judged the endometrial functionalis inadequate for evaluation. The remaining four tissue specimens excluded from analysis were removed because histologic examination and the serum P concentration in the corresponding serum samples failed to confirm that ovulation had occurred, most likely reflecting errors in interpretation of test results during urinary LH monitoring. Among the remaining total of 121 endometrial tissue specimens analyzed, the numbers obtained on each LD ranged between 6 and 10 (LD 0, 10; LD 1, 6; LD 2, 10; LD 3, 8; LD 4, 8; LD 5, 10; LD 6, 10; LD 7, 7; LD 8, 10; LD 9, 7; LD 10, 9; LD 11, 8; LD 12, 9; LD 13, 9). Four of the corresponding sera samples were hemolyzed or inadequate, leaving 117 serum samples available for analysis. Study Population Demographics Data describing the study population are expressed as mean SD, followed by range in parentheses. Age was years (20 to 35), body mass index was kg/m 2 (18.3 to 54.9), gravidity was (1 to 8), and the number of living children was (1 to 5). Among those participants who underwent tissue sampling, 101 out of the 130 (78%) had delivered a liveborn child within the 4 years immediately before entering the study. The subjects reported menstrual cycle length before entering the study was days (25 to 35). In study cycles, the follicular phase (days from onset of menses to detection of the midcycle urinary LH surge) was days (9 to 21) in duration. Analyses of Data Resulting from Scoring of Histologic Features Figure 1 compares the fitted polynomial curves generated from our data with the original curves drawn by Latour and modified by Noyes et al. (1). Our fitted polynomial curves provide an objective and accurate graphic representation of the temporal patterns we observed, in contrast with the original subjective illustrations. Whereas our curves generally and sometimes closely resemble those of Noyes et al., they also reveal that these selected histologic features are much less temporally distinct than was originally described. The histologic features having the greatest amounts of their variations in score attributable to changes across time were pseudostratified nuclear location (rank 1) and epithelial FERTILITY & STERILITY 1337

6 FIGURE 1 Comparison of the fitted polynomial curves generated from scoring data with the traditional curves drawn by Noyes et al. (1). The curves on the left represent the predicted mean and 95% confidence intervals calculated from the scores assigned by three pathologists masked to the luteal day. The curves on the right are copied from the original drawings in the Noyes study. mitoses (rank 2). Both were clearly prominent during the early secretory phase (see Fig. 1) and also ranked well when judged by the difference between minimum and maximum scores (ranks 5 and 1, respectively; ranking data not shown). However, the extent of agreement among pathologists on absolute scores for pseudostratified nuclear location (r 0.51, rank 10) and epithelial mitoses (r 0.58, rank 7) was relatively poor. Based on these observations, when more than 25% of nuclei are pseudostratified and epithelial mitoses are numerous, one can reliably assign an endometrial sample to the early secretory phase. Unfortunately, the relatively high interobserver variation observed among assigned scores for the two features prevents a narrower or more specific interpretation. During the interval corresponding to the midluteal phase, serrated gland configuration and linear/basal nuclear location were the best performing and thus most discriminating features (see Table 2). Unfortunately, whereas both features first appear during this phase of the cycle, both also remain relatively prominent for the duration of the cycle (data not shown) and thus have little value for defining the phase of 1338 Murray et al. Endometrial dating revisited Vol. 81, No. 5, May 2004

7 FIGURE 2 Endometrial specimens obtained from two different proven fertile women on luteal day 8. (A) In one tissue specimen, subnuclear vacuoles (arrow) are abundant; (B) in the other specimen, subnuclear vacuoles are altogether absent. If judged according to traditional histologic criteria (1), maturation is abnormally delayed in specimen A and normal in specimen B. the cycle, much less for defining a specific LD or interval of days. Classically, the appearance of pseudodecidual reaction in the stroma defines the late secretory phase endometrium. However, our data demonstrate that pseudodecidual changes often may be observed much earlier, beginning in the early stages of the midsecretory phase and becoming increasingly prominent thereafter. Agreement among pathologists regarding the location and extent of pseudodecidual changes also was relatively poor (see Table 2; see Fig. 1). Moreover, we observed that persistence of subnuclear (basal) vacuoles (heretofore regarded as a feature unique to the early secretory phase) frequently persist well beyond LD 4 (see Fig. 1; Fig. 2), in sharp contrast with the pattern originally described by Noyes et al. (1). Result of Histologic Dating Using the Traditional Criteria of Noyes et al. (1) The interobserver variation among the three pathologists and intersubject variability observed when traditional endometrial dating criteria were applied are shown in Figure 3. With the exception of LD 7, the mean assigned dates for the three pathologists were delayed by approximately 1 day when compared with the actual LD. Interobserver variability among the three expert pathologists rarely exceeded 1 day and none of the pathologists was observed to consistently assign a more advanced or delayed date. The shaded area shown in Figure 3 corresponds to the SD around the mean assigned POD (mean line not shown) among the three pathologists. The mean assigned POD was calculated from the individual results of the three pathologists for each specimen obtained on each LD, thereby eliminating interobserver variability. The shaded area in Figure 3 thus represents the variability one might expect to observe among fertile women on any given LD (intersubject variability). The relatively narrow shaded region observed during the early secretory phase implies that intersubject histologic variability is rather low when compared with that which may be observed during the middle and late secretory phases (wider shaded area). Beyond LD 7, intersubject histologic variability (SD) was as great as 7 full days. The prevalence of delayed endometrial histology in our population of normally cycling fertile women, when traditional dating criteria (1) were applied, is shown in Figure 4. In the periovulatory phase (LD 0 to 2) all specimens were judged normal (assigned POD no more than 2 days behind the actual LD) with 100% agreement among pathologists, primarily because many of the specimens exhibited the histologic characteristics of proliferative endometrium. Although secretory phase histologic characteristics were consistently observed in specimens obtained on or after LD 3, agreement among the interpretations of the three pathologists (normal vs. delayed) was less than 100% (data not shown). When the biopsy was performed during the middle and late luteal phases (LD 5 to 13), the interpretations of all three pathologists agreed for 79% of specimens defined as normal and for 81% of specimens judged as delayed. The overall mean serum E 2 and P concentrations observed across the luteal phase were consistent with those expected, exhibiting the patterns characteristically observed during the luteal phase of the menstrual cycle (data not shown). Between Cycle Variation (Intrasubject Variability) A second endometrial biopsy was obtained from 49 women in a subsequent menstrual cycle on the same LD as FERTILITY & STERILITY 1339

8 FIGURE 3 Interobserver and between subject variation in histologic dating using the traditional criteria of Noyes et al. (1). The three plots illustrate the extent of interobserver variability. Individual points (,, and Œ) represent the mean histologic dates (postovulatory day) assigned by the three different pathologists for all tissue specimens obtained on the same luteal day. Between-subject variation is represented by the shaded area and corresponds to the standard deviation around the mean assigned histologic date among the three pathologists. The mean assigned date was calculated from the individual results for all specimens obtained on each luteal day, thereby eliminating interobserver variation from calculation of the standard deviation. in the first study cycle. The between-cycle variation observed in our study population of healthy, fertile volunteers is shown in Figure 5. To eliminate interobserver variability for this analysis, each tissue specimen was designated with a single POD, calculated by averaging the dates assigned by each of the three pathologists. Between-cycle variation was considered present when the mean POD for the woman s two tissue specimens differed by more than 2 days. Although the small number of women in this subset precludes confident conclusion, our data suggest that between-cycle variation among normal women is also quite common and may be observed in up to one-third of between-cycle comparisons. DISCUSSION Endometrial dating using the traditional criteria first described in the classic paper by Noyes, Hertig, and Rock (1) has remained the gold standard of methods for diagnosis of luteal phase deficiency and has served as the foundation for virtually all research on the development and function of the secretory phase endometrium for more than 50 years. As recently as 6 years ago, 62.5% of board-certified reproductive endocrinologists who responded to a national survey said they always or almost always obtained an endometrial biopsy to confirm ovulation and the quality of luteal function during a standard infertility evaluation (35). Despite historical importance and wide acceptance among clinicians and pathologists, the accuracy and utility of the traditional endometrial dating criteria have been repeatedly questioned (18, 36). The principal criticisms of endometrial dating relate to using a normal standard established with observations in infertile women, the inaccuracies inherent in defining the day of sampling by onset of the next menses assuming an unvarying 14-day luteal phase duration, and clinically significant intraobserver and interobserver variations in histologic interpretation. Despite these criticisms and lack of data validating the accuracy and utility of endometrial histologic dating, the practice has endured and remains relatively common in both clinical and research settings. Acknowledging both the traditions and controversies surrounding the practice of endometrial dating, we carefully designed our study to address each of the real and perceived limitations of the original (1). The most important elements of our study design included a study population consisting 1340 Murray et al. Endometrial dating revisited Vol. 81, No. 5, May 2004

9 FIGURE 4 The proportion (%) of endometrial specimens obtained in the periovulatory (luteal days [LD] 0 to 2), early (LD 3 to 5), middle (LD 6 to 10), and late (LD 11 to 13) luteal phase in which the histologic date (postovulatory day) assigned by two or three pathologists was delayed by more than 2 days from the actual LD, using traditional criteria (1). FIGURE 5 Between-cycle variation (%) in the periovulatory (luteal day [LD] 0 to 2), early (LD 3 to 5), mid (LD 6 to 10) and late (LD 11 to 13) luteal phase as determined in 49 women. A second endometrial biopsy was obtained from each woman in a subsequent menstrual cycle on the same LD as in the first study cycle. Using traditional criteria (1), each tissue specimen was designated with a postovulatory date (POD), calculated by averaging the dates assigned by each of the three pathologists. Between-cycle variation was considered present when the mean POD for the woman s two tissue specimens differed by more than 2 days. Using the average POD of the three pathologists eliminates interobserver variability in this comparison. entirely of normally cycling, proven fertile women, randomization of participants to a specific day for sampling as carefully defined by detection of the midcycle urinary LH surge in study cycles, and masking of the panel of three expert gynecologic histopathologists who examined the tissue specimens. In addition, our study involved the evaluation of 32 distinct histologic features, each of which was scored by the three pathologists using criteria established by consensus before the study was conducted. Lastly, we subjected our data to rigorous and objective analyses designed to identify those features having the greatest discriminating value for dating purposes. Over the decades since the original publication by Noyes et al. (1), numerous investigators have included many of the design modifications described above, but ours is the largest study to address all of the many cited limitations of the original study in a comprehensive, objective, and clinically relevant manner. Our systematic evaluation of 32 distinct histologic features and objective analyses of the resulting data yielded results that were similar to the original description of Noyes et al. (1) in many ways, but also quite different in others (see Fig. 1). The endometrium does indeed exhibit a consistent sequence of histologic maturational changes across the secretory (luteal) phase. Two of the most discriminating histologic features we identified, epithelial mitoses and nuclear pseudostratification, were also among the original dating criteria proposed by Noyes et al. (1). Others, including linear/basal nuclear location and serrated gland configuration, were not. Taken together, the best performing histologic features clearly do have value for differentiating the early, middle, or late secretory phase endometrium. However, with the exception of stromal breakdown, which accurately heralds the onset of menstruation, no single feature or combination of features can reliably define the phase of cycle more narrowly. Some of the features included among the traditional dating criteria, notably subnuclear (basal) vacuoles and pseudodecidual reaction, actually have relatively little discriminating value, chiefly because these features are much less temporally distinct than was originally described. According to the traditional dating criteria, subnuclear vacuoles are most prominent between LD 2 and 4, and disappear altogether after LD 6. Our analysis revealed that subnuclear vacuoles are indeed most prominent in the early secretory phase, but they commonly persist well beyond that time, into the middle and even late secretory phase (see Figs. 1, 2). Similarly, pseudodecidual reaction is classically described as first appearing on LD 9, but we observed this feature as much as 4 days earlier, and interobserver variability in the scoring of this feature was also quite high (see Fig. 1; see FERTILITY & STERILITY 1341

10 Table 2). Other classic dating criteria, notably luminal secretions and stromal edema, exhibited essentially no discriminating value for dating purposes. The patterns we observed for the remaining two of the traditional criteria, stromal mitoses and leukocytic infiltration, were quite similar to those described by Noyes et al. (1) and performed reasonably well, except that interobserver variations in assessing these features were unacceptably high. In recent years, efforts to improve the results of assisted reproductive technologies and research aimed at defining the mechanisms of embryo implantation have focused new and greater attention on the finer elements of endometrial histology. Although classic endometrial dating was based on the most advanced histologic characteristics observed in either the glands or stroma (1, 37), separate dating of the two compartments has become increasingly common. This practice, not described or advocated by Noyes et al. (1), has spawned the concept of glandular-stromal dyssynchrony, regarded by many as evidence of endometrial dysfunction or as a subtle form of LPD (38). Our observations that individual histologic features clearly are far less temporally distinct than previously thought (1) strongly suggest that any perceived glandular-stromal dyssynchrony is merely a reflection of the spectrum of normal variations in secretory phase endometrial maturation. Evaluation of the tissue specimens obtained in our study, using the traditional dating criteria of Noyes et al. (1), yielded results that corroborate those of earlier studies in several ways. First, we found that abnormally delayed endometrial maturation (an assigned POD greater than 2 days behind the actual LD of sampling) is common, even in healthy, otherwise fertile women (see Fig. 4) (15, 39 43). Second, between-cycle variations in the histologic features of the secretory endometrium are observed just as commonly in fertile women (see Fig. 5) (44) as in infertile women (45). Third, even among a group of expert gynecologic pathologists using the same carefully defined criteria, interobserver variability in histologic interpretation is significant and could frequently change the clinical diagnosis and management (19, 23, 24, 26). Our original intent was to refine or redefine histologic dating criteria that might have clinical value for evaluating the quality of endometrial maturation and arriving at a more accurate diagnosis of LPD. Instead, the results of our study lead to the inevitable conclusion that histologic endometrial dating simply has neither the accuracy nor the precision necessary to be a valid method to diagnose LPD or otherwise guide the clinical management of women with reproductive failure. If histologic dating were indeed a valid diagnostic tool, our rigorously controlled study design should have demonstrated its value, but it did not. Our methods, which were far more rigorous than those employed in everyday clinical practice, were ultimately futile. Taken together, variations in the temporal sequence of endometrial histologic changes across the secretory phase among individuals, between cycles in individuals, and among the interpretations of different observers are simply too great to reliably define a specific LD or even a narrow interval of days. As numerous studies have repeatedly shown, variability in the length of the luteal phase is quite normal, even among fertile women (9 13). Similarly, our data indicate that substantial variability in the histologic characteristics of the secretory phase endometrium is also entirely normal. The results of our study do not deny that abnormal secretory endometrial maturation may adversely affect reproductive performance, or the possibility that other dating techniques using biochemical or molecular markers of endometrial function might prove useful. Rather, our results simply but conclusively demonstrate that traditional histologic dating of the endometrium is not a valid clinical diagnostic tool. References 1. Noyes RW, Hertig AW, Rock J. Dating the endometrial biopsy. Fertil Steril 1950;1: Key JD, Kempers RD. Citation classics: most-cited articles from Fertility and Sterility. Fertil Steril 1987;47: Jones GES. Some newer aspects of the management of infertility. JAMA 1949;141: Jones GS. The luteal phase defect. Fertil Steril 1976;27: Tho PT, Byrd JR, McDonough PG. Etiologies and subsequent reproductive performance of 100 couples with recurrent abortion. Fertil Steril 1979;32: Balasch J, Vanrell JA. Corpus luteum insufficiency and fertility: a matter of controversy. Hum Reprod 1987;2: Soules MR. Luteal phase deficiency: a subtle abnormality of ovulation. In: Keye WR, Chang RJ, Rebar RW, Soules MR, eds. Infertility: evaluation and treatment. Philadelphia: W.B. Saunders, 1995: Lessey BA, Fritz MA. Defective luteal function. In: Fraser JS, Jansen RPS, Lobo RA, Whitehead MI, eds. Estrogens and progestogens in clinical practice. Philadelphia: W.B. Saunders, 1998: Treloar AE, Boynton RE, Behn BG, Brown BW. Variation of the human menstrual cycle through reproductive life. Int J Fertil 1967;12: Vollman RF. The menstrual cycle. Major Probl Obstet Gynecol 1977; 7: Johannisson E, Parker RA, Landgren BM, Diczfalusy E. Morphometric analysis of the human endometrium in relation to peripheral hormone levels. Fertil Steril 1982;38: Lenton EA, Landgren BM, Sexton L. Normal variation in the length of the luteal phase of the menstrual cycle: identification of the short luteal phase. Br J Obstet Gynaecol 1984;91: Munster K, Schmidt L, Helm P. Length and variation in the menstrual cycle a cross-sectional study from a Danish county. Br J Obstet Gynaecol 1992;99: Shoupe D, Mishell DR Jr, Lacarra M, Lobo RA, Horenstein J, d Ablaing G, et al. Correlation of endometrial maturation with four methods of estimating day of ovulation. Obstet Gynecol 1989;73: Batista MC, Cartledge TP, Merino MJ, Axiotis C, Platia MP, Merriam GR, et al. Midluteal phase endometrial biopsy does not accurately predict luteal function. Fertil Steril 1993;59: Jordan J, Craig K, Clifton DK, Soules MR. Luteal phase defect: the sensitivity and specificity of diagnostic methods in common clinical use. Fertil Steril 1994;62: Guermandi E, Vegetti W, Bianchi MM, Uglietti A, Ragni G, Crosignani P. Reliability of ovulation tests in infertile women. Obstet Gynecol 2001;97: Novak E. Editorial. Obstet Gynecol Surv 1950;5: Noyes RW, Haman JO. Accuracy of endometrial dating: correlation of endometrial dating with basal body temperature and menses. Fertil Steril 1953;4: Andrews WC. Luteal phase defects. Fertil Steril 1979;32: Kurman RJ, Mazur MT. Benign diseases of the endometrium. In: Kurman RJ, ed. Blaustein s pathology of the female genital tract. Heidelberg, Germany: Springer-Verlag, 1994: Murray et al. Endometrial dating revisited Vol. 81, No. 5, May 2004

11 22. Duggan MA, Brashert P, Ostor A, Scurry J, Billson V, Kneafsey P, et al. The accuracy and interobserver reproducibility of endometrial dating. Pathology 2001;33: Scott RT, Snyder RR, Strickland DM, Tyburski CC, Bagnall JA, Reed KR, et al. The effect of interobserver variation in dating endometrial histology on the diagnosis of luteal phase defects. Fertil Steril 1988; 50: Gibson M, Badger GJ, Byrn F, Lee KR, Korson R, Trainer TD. Error in histologic dating of secretory endometrium: variance component analysis. Fertil Steril 1991;56: Scott RT, Snyder RR, Bagnall JW, Reed KD, Adair CF, Hensley SD. Evaluation of the impact of intraobserver variability on endometrial dating and the diagnosis of luteal phase defects. Fertil Steril 1993;60: Smith S, Hosid S, Scott L. Endometrial biopsy dating. Interobserver variation and its impact on clinical practice. J Reprod Med 1995;40: Huang KE. The primary treatment of luteal phase inadequacy: progesterone versus clomiphene citrate. Am J Obstet Gynecol 1986;155: Rosenfeld DL, Chudow S, Bronson RA. Diagnosis of luteal phase inadequacy. Obstet Gynecol 1980;56: Brodie BL, Wentz AC. An update on the clinical relevance of luteal phase inadequacy. Semin Reprod Endocrinol 1989;7: Gautray JP, de Brux J, Tajchner G, Robel P, Mouren M. Clinical investigation of the menstrual cycle. III. Clinical, endometrial, and endocrine aspects of luteal defect. Fertil Steril 1981;35: Li TC, Cooke ID. Evaluation of the luteal phase. Hum Reprod 1991; 6: Castelbaum AJ, Wheeler J, Coutifaris CB, Mastroianni L Jr, Lessey BA. Timing of the endometrial biopsy may be critical for the accurate diagnosis of luteal phase deficiency. Fertil Steril 1994;61: Miller PB, Soules MR. The usefulness of a urinary LH kit for ovulation prediction during menstrual cycles of normal women. Obstet Gynecol 1996;87: Nielsen MS, Barton SD, Hatasaka HH, Stanford JB. Comparison of several one-step home urinary luteinizing hormone detection test kits to OvuQuick. Fertil Steril 2001;76: Glatstein IZ, Harlow BL, Hornstein MD. Practice patterns among reproductive endocrinologists: the infertility evaluation. Fertil Steril 1997;67: American Society for Reproductive Medicine. A practice committee report: optimal evaluation of the infertile female. Birmingham, AL: ASRM, 2000: Ferenczy A. Anatomy and histology of the uterine corpus. In: Kurman RJ, ed. Blaustein s pathology of the female genital tract. Heidelberg, Germany: Springer-Verlag, 1994: Balasch J, Vanrell JA, Creus M, Marquez M, Gonzalez-Merlo J. The endometrial biopsy for diagnosis of luteal phase deficiency. Fertil Steril 1985;44: Aksel S. Sporadic and recurrent luteal phase defects in cyclic women: comparison with normal cycles. Fertil Steril 1980;33: Li TC, Dockery P, Cooke ID. Endometrial development in the luteal phase of women with various types of infertility: comparison with women of normal fertility. Hum Reprod 1991;6: Balasch J, Fabregues F, Creus M, Vanrell JA. The usefulness of endometrial biopsy for luteal phase evaluation in infertility. Hum Reprod 1992;7: Peters AJ, Lloyd RP, Coulam CB. Prevalence of out-of-phase endometrial biopsy specimens. Am J Obstet Gynecol 1992;166: Batista MC, Cartledge TP, Zellmer AW, Merino MJ, Nieman LK, Loriaux DL, et al. A prospective controlled study of luteal and endometrial abnormalities in an infertile population. Fertil Steril 1996;65: Davis OK, Berkeley AS, Naus GJ, Cholst IN, Freedman KS. The incidence of luteal phase defect in normal, fertile women, determined by serial endometrial biopsies. Fertil Steril 1989;51: Ordi J, Creus M, Quinto L, Casamitjana R, Cardesa A, Balasch J. Within-subject between-cycle variability of histological dating, alpha v beta 3 integrin expression, and pinopod formation in the human endometrium. J Clin Endocrinol Metab 2003;88: FERTILITY & STERILITY 1343