Anti-Müllerian Hormone Level Decline in Patients Undergoing Hysterectomy With and Without Oophorectomy Compared With Natural Menopause.

The Study of Women’s Health Across the Nation (SWAN) is a multi-ethnic, multi-site community-based study of 3302 women at baseline (1996-1997) followed through the menopausal transition. Study design has been previously described( 3 ). The present study was approved by the Colorado Multiple IRB (COMIRB). Briefly, approximately 450 women were recruited for the longitudinal cohort at each of seven clinical sites. In addition to White women, each site recruited women from one specified minority group as one of the goals of SWAN was to examine the menopausal experience of women of representative racial and ethnic groups across the United States. Cohort participants were required to be aged 42–52 years, have an intact uterus, have had at least one menstrual period and not be pregnant or lactating or have used reproductive hormones in the previous 3 months. Baseline and approximately annual follow-up assessments included interviewer-administered and self-administered questionnaires about health, lifestyle and psychosocial factors, plus anthropometric measurements. Age, race and ethnicity were obtained at the baseline examination; incident hysterectomies, smoking status, exogenous hormone use, and menopausal status were determined at each follow-up visit. The final menstrual period (FMP) was recorded as a calendar date for all women with an observed FMP, either at an annual visit or on a monthly menstrual calendar. At each follow-up visit, each participant was asked if, since her previous visit, she had “a hysterectomy (an operation to remove the uterus or womb).” Patients reporting an incident hysterectomy were asked for consent to have medical records requested; pertinent information was abstracted, including pre and post-operative notes and pathology reports, presenting symptoms, pre-operative diagnosis, histologic findings, uterine weight, record of uterine and ovary removal, and any surgical complications. Presenting symptoms were only available for 130/232 participants and included abnormal bleeding and chronic pain. Evidence of fibroids, adenomyosis, and endometriosis were verified by histology when available ( 4 ). At each visit, a fasting blood draw was scheduled between 8:00 and 10:00 am on menstrual cycle days 2 through 5 to measure AMH and other reproductive hormones. If 60 days passed without a collection, blood was collected at any time during the next 30 days. Patients who had undergone hysterectomy were sampled at any time within the 90-day window. Serum was processed promptly, frozen, and stored at −80°C until thawed for measurements. Participants who underwent natural menopause were included in a prior publication describing change in AMH with reproductive aging( 1 ). Additional samples from women who had undergone hysterectomy were accessed from the SWAN Repository and measured using a 2-site ELISA (MenoCheck picoAMH ELISA, Ansh Labs) with intra- and interassay CVs ranging from 2.5% to 5.1% and 3.4% to 4.9%, respectively, at levels of 91 and 290 pg/mL. The LOD was 1.85 pg/mL( 1 ). AMH was measured at Ansh Labs under the joint supervision of the Ansh Laboratory Director and the Director of Special Chemistry, Clinical Pathology Core Laboratory at Massachusetts General Hospital. Participants with a BSO or a hysterectomy with conservation of one (USO) or both (NSO) ovaries prior to an observed natural menopause (menstrual bleeding followed by 12 consecutive months of amenorrhea, no prior menopausal HT use) were included if there was at least one pre-surgical and at least one post-surgical specimen available in the SWAN Repository. AMH was measured on the pre-surgical specimen closest to surgery, and on all available post-surgical specimens in order to follow levels until they became undetectable. These participants were subdivided by the number of ovaries retained. The natural menopause cohort was comprised of all women who had natural menopause and at least one sample for AMH determination while pre- or early perimenopausal and for whom FMP date could be determined (e.g., no gap in data collection or use of exogenous hormone therapy that might mask FMP date). AMH was measured on all available pre-FMP specimens and on the first available post-FMP specimen. Participant characteristics were compared for the two cohorts as well as by the degree of ovarian conservation using analysis of variance for continuous variables and chi-square testing for categorical variables. AMH was log transformed due to right skewness, after random interpolation of values below the lower limit of detection. For each of the four groups separately, nonparametric local weighted scatterplot smoothing (LOESS) regression( 5 ) was used to examine patterns of AMH with respect to years before or after surgery (hysterectomy cohort) or FMP (natural menopause group), including possible knots where rate of change in AMH differed. Optimal knot placement was determined for each group separately using piecewise linear mixed modeling( 6 ) with model fit indicated by the Akaike Information Criterion (AIC). Statistical significance of between-group differences in segment-specific slopes was assessed by including all four groups in the same model and testing group × segment-time interactions. Kaplan-Meier plots, log-rank testing, and Cox proportional hazard modeling were used to compare groups regarding age at first undetectable AMH, before and after adjusting for baseline age, BMI, race/ethnicity, study site, self-reported fibroids, and smoking. Age at first undetectable AMH was considered to be: right-censored at age at last AMH measurement for participants never observed to have undetectable AMH; left-censored for participants whose first measured AMH was undetectable; and interval-censored – due to annual measurements – for all other participants( 7 , 8 ). In supplemental analyses, we adjusted instead for age at FMP, observed for the natural menopause group, and multiply imputed ( 18 ) using multivariate sequential regression in SAS-compatible IVEware( 9 ) for the hysterectomy groups as a function of characteristics associated with age at FMP ( 10 , 11 ); imputed age at FMP can be interpreted as “counterfactual,” or predicted timing of the FMP in the absence of surgery. All analyses were conducted in SAS 9.4 [SAS Institute, Cary, NC].

Of 232 women who had a hysterectomy with or without ovarian conservation prior to a natural menopause and who had sufficient serial AMH measurements to be included in the study, there were 159 women with BSO, 13 with USO, and 60 with NSO. Of 1536 naturally menopausal participants, 108 were excluded due to HT use, gaps in data collection or pre-FMP gynecologic surgery rendering the FMP date imprecise. The remaining 1428 participants were included in analyses. The analytic sample of 1660 women provided a total of 8989 AMH measurements ( Figure 1 ). The mean number of AMH measurements per participant was 5.42 (range 1-14) overall, 7.3 (range 1-14) in the hysterectomy cohort, 5.1 (range 1-12) in the natural menopause group. At baseline, naturally menopausal women were older, more likely to be Japanese or Chinese and have a lower BMI, and were less likely to be diagnosed with fibroids or to have ever undergone a dilatation and curettage (D&C) or other uterine procedure ( Table 1 ). Patients with natural menopause had an FMP at a mean age of 52.1±0.2 years, significantly earlier than the counterfactual FMP of the women with hysterectomy (52.9±0.2 years, p=.002). Patients undergoing BSO were slightly older than the USO or NSO groups at baseline ( Table 2 ). Race or ethnicity and BMI did not differ by hysterectomy group. Of the USO and BSO groups, 80-90% had an abdominal procedure, versus 54% in the NSO group. Presenting symptoms did not differ significantly by surgical subgroup except for a greater proportion of prolapse reported by women who underwent hysterectomy and NSO. Figure 2 depicts fitted log AMH from piecewise (3-segment) linear mixed models in relation to years before/after surgery date for women with hysterectomies and in relation to years before/after FMP date for women with a natural menopause. Figure 3 presents the same fitted lines as in Figure 2 , along with the observed AMH values, stratified by group and indicating all data points. Patients undergoing a BSO experienced an immediate reduction in AMH to undetectable after surgery, consistent with complete removal of both ovaries( Table 3 ). All other groups’ second-segment slopes differed significantly from the BSO group’s slope (p<.001for all three pairwise comparisons). In contrast, the second segment’s per-year decline was least steep for the NSO group, followed by the natural menopause group and the USO group; all of these pairwise between-group differences were statistically significant (p<.0127). AMH declined in the first segment (pre-surgery/pre-FMP) for all four groups, with the smallest per-year drop in the BSO group (p<.02 for BSO versus NSO and versus natural menopause). For all four groups, third-segment declines were relatively flat and significantly more gradual than in the corresponding second segment, and not significantly different from 0 for BSO and natural menopause. The continuing AMH decline in the third segment was largest for the NSO group, consistent with conservation of both ovaries, although it differed significantly only from the BSO group (p<.001); no other pairwise differences were statistically significant. Among the surgical groups, the knot dividing the second and third segments, i.e., the location of the change from the steeper post-surgery decline to a flatter trajectory, was closest to surgery for BSO, farthest from surgery for NSO, and intermediate for USO. Figure 4 presents Kaplan-Meier curves for age at first undetectable AMH. Participants in the NSO and natural menopause groups had a similar age at undetectable AMH (p=.31), which was significantly later than in women with a BSO (p<.001) or with a USO (p=.02 versus USO, p=.09 versus natural menopause; p=.91 for BSO versus USO). Table 4 presents corresponding hazard ratios from Cox proportional hazards modeling. Prior to covariate adjustment, compared with natural menopause, age at first undetectable AMH was significantly lower in BSO and USO and later but not significantly different in NSO women. Adjustment for baseline covariates attenuated the natural menopause versus USO difference (no longer statistically significant), consistent with a higher proportion of smoking in the latter group. Adjustment for age at FMP had little impact on hazard ratios for USO and NSO groups, reflecting similar ages at FMP for the three groups, but slightly increased the natural menopause versus BSO difference, reflecting the approximately 1-year higher (counterfactual) age at FMP in the latter group. In all models, BSO and USO groups did not differ, nor did NSO and natural menopause women differ. All other pairwise differences were statistically significant in all models (p<.05).

These data demonstrate that women who underwent unilateral or bilateral ovarian conservation (USO or NSO) have a similar AMH trajectory immediately postoperatively to women undergoing natural menopause, contrary to our hypothesis. Interestingly, we also found that women undergoing hysterectomy regardless of ovarian conservation status had a somewhat older age at counterfactual FMP than those undergoing natural menopause, meaning that had they not undergone a hysterectomy their FMPs would have occurred almost 10 months later than women with natural menopause. Patients who have had a bilateral oophorectomy have a very rapid postoperative decline in AMH, consistent with complete removal of the ovaries. Our data diverge from those of others, which suggest that surgical menopause with ovarian conservation hastens the time to complete ovarian failure. Prior studies such as the PROOF (Prospective Research on Ovarian Function) study indicated an increased risk of ovarian failure after hysterectomy with ovarian conservation( 12 , 13 ) using FSH as a biomarker. However, the large variation in FSH across the menstrual cycle and the inability to time sampling in women with hysterectomy may have affected findings. Moreover, there is not a universally agreed-upon FSH concentration that defines menopause. A more recent PROOF publication that examined AMH levels before and after hysterectomy found a greater decrease in women undergoing hysterectomy with ovarian conservation versus controls. However, an enzyme-linked immunosorbent assay (ELISA) with a limit of detection of 100pg/mL was used, which may have been insufficiently sensitive to detect differences in AMH at the end of reproductive life, a time when many hysterectomies are performed( 14 ). Others have followed AMH or FSH as markers of ovarian failure for up to a year postoperatively ( 12 ). However, based on our findings, one year may be insufficient time to assess a post operative change in the slope of AMH decline, which we observed as long as 4.25 years after ovarian conserving surgery. The women reported herein, with a mean age of 45-46 at study entry, were all relatively late in their reproductive life spans compared to women studied in the PROOF cohort( 12 ), who had a mean age of 40. It is likely that ovarian reserve is compromised differentially as a function of age. Our data using piecewise linear regression indicates that prior to the FMP, women experience an acceleration in the rate of AMH decline reflecting changes related to the menopausal transition, and this has not been factored into other study designs. In LOESS plots, the pre-surgery AMH decline was least steep in women who met criteria for premenopause at the last pre-surgery blood draw, intermediate for those who were early perimenopausal at the last pre-surgery blood draw, and steepest for those in the late perimenopause at last pre-surgery blood draw (data not shown). Sample sizes did not permit dividing women by menopausal stage and surgery. It is also challenging to find an ideal way to compare women before and after natural menopause to those underoing surgery, and we therefore analyzed our data in several different ways to try to address this issue. However, it is likely that menopausal stage at the time of surgery is a factor in determining the rate of AMH decline and may account for some of the differences in segment slopes we observed between the NSO and USO groups. While women undergoing NSO did not have an earlier age at undetectable AMH, those in the USO group, even after adjustment, had an almost 2-fold increase in risk of earlier age at undetectable AMH, although the confidence interval was relatively wide (1.01-3.69, Table 4 ) and the sample size is relatively small (N=13). We therefore cannot rule out the possibility of an effect of USO on AMH depletion. Prior work addressing the impact of different surgical strategies on ovarian function have generally been limited to small samples with relatively brief follow-up and may not have taken into account the evolution of minimally invasive techniques which may reduce ovarian devitalization or may have included surgical techniques that lead to excessive collateral tissue damage. Cho et al., in a report on 90 women undergoing hysterectomy, noted that a higher proportion of women who underwent laparoscopic procedures had evidence of decline in AMH at 2 and 6 months compared to those undergoing hysterectomy with an abdominal approach ( 15 ) and posited that laparoscopic procedures cause more thermal injury to adjacent tissues and thereby may be more likely to reduce ovarian reserve. These authors did not follow AMH beyond 6 months post-operatively and may have missed postsurgical ovarian recovery. Another study of 86 women who underwent hysterectomy without adnexal surgery indicated that women undergoing laparoscopic procedures had a greater immediate postoperative (Day 3) decline in AMH than those undergoing abdominal hysterectomy( 16 ). On the other hand, Tavana et al., have recently reported opposite findings of increased postoperative decline in AMH among women undergoing abdominal hysterectomy in a smaller sample (33 women per group)( 17 ). Notably, in the latter report, women were aged 40-50, baseline AMH levels were overall low (400 pg/ml or less) and salpingectomies were performed in these patients. The conflicting data suggest that the precise surgical approach and possibly an independent role of salpingectomy may influence postoperative ovarian reserve but effects may be transient and are incompletely understood. The indication for the hysterectomy may also play an important role in whether the ovaries are conserved and may be related to the later age at counterfactual FMP in the hysterectomy cohort. In the SWAN cohort, women with ovarian diseases such as endometriosis that would be likely to compromise their ovarian reserve( 18 ) would be more likely to have had their ovaries removed at the time of hysterectomy. This may have effectively enriched our sample of women with hysterectomy and ovarian conservation with those whose ovaries were relatively free of disease and were therefore destined to have a longer reproductive life span. Patients who undergo hysterectomy may also have fundamental differences in estrogen processing. The estrogen receptor alpha (ER alpha) PP genotype, which is more prevalent in Black vs White women, has been associated with uterine leiomyomata and greater tumor burden( 19 ) as well as a higher rate of hysterectomy( 20 ). Strengths of this study include the lengthy follow-up of AMH and its tracking to the point where it is undetectable , which has not been performed in prior studies. The use of a highly sensitive AMH assay also provides additional assurance that an undetectable AMH does, indeed, represent close proximity to menopause in a cohort of this age range. Adjudication of surgical cases included confirmation of the histological findings ( 4 , 21 , 22 ). The large and well characterized group of women who underwent observed natural menopause as controls also provides a strong comparison group. On the other hand, there are several aspects of the study that make it difficult to compare with others. These include lack of precise information on the surgical techniques used for each case, which may have obscured differences between hysterectomy subgroups. Our data did not allow for an estimation of the effect of salpingectomy, as it was not part of routine medical practice at the time that the hysterectomies were performed. The overall number of cases is also relatively small although consistent with most of the literature on this topic to date. The possibility that AMH varied across the menstrual cycles of the women studied may have influenced the findings( 23 ). In summary, these findings suggest that the onset of subsequent menopause in women who undergo hysterectomy with ovarian conservation is not necessarily earlier than women who undergo natural menopause, particularly in those retaining both ovaries, and it should therefore not be assumed that women without a uterus will undergo menopause at an earlier age than they would have had they not had a hysterectomy. However, women who undergo hysterectomy appear to have characteristics that would predict an even later age at FMP had the hysterectomy not occurred, although the age at hysterectomy and the precise indications and surgical approach may influence whether or not the surgery per se reduces their reproductive life span. Patients with BSO did not demonstrate any evidence of an ovarian remnant in this study, implying that ovarian remnants—even microscopic and clinically silent remnants--are relatively rare findings. Depending upon the clinical scenario, adjuvant testing of ovarian reserve may be helpful in determining whether a woman without a uterus has impending ovarian failure, as this may have long term implications for bone and cardiovascular health and may help substantiate the basis for menopausal symptoms.

Determining when ovarian function ceases is challenging in women who have had a hysterectomy because menopause is typically defined as the absence of menstrual cycles for 12 consecutive months. Therefore, an accurate assessment of ovarian reserve (quantity of oocytes within the ovary) is desirable in this post-surgical population. FSH, an indirect marker of ovarian function subject to sex steroid and protein hormone feedback, varies throughout the menstrual cycle and is of limited usefulness in determining the onset of menopause in women with hysterectomy. Anti-Müllerian hormone (AMH) is a product of granulosa cells and thus provides a more direct measure of ovarian function with minimal variation across the menstrual cycle. A newer ELISA with a limit of detection of 1.85 pg/ml was recently found to be predictive of the final menstrual period (FMP) in a cohort of more than 1500 women undergoing natural menopause from the Study of Women's Health Across the Nation (SWAN)( 1 ). The present study sought to determine time to undetectable AMH level in women in the SWAN cohort who underwent hysterectomy compared to those who underwent natural menopause with no hysterectomy. We hypothesized that hysterectomy with ovarian preservation would result in a shortened time to an undetectable AMH versus natural menopause without a hysterectomy. We also hypothesized that women undergoing bilateral oophorectomy would have an immediate reduction in AMH to undetectable levels and that failure to observe an undetectable AMH would be indicative of an ovarian remnant( 2 ).