Association between arthritis treatments and ovarian reserve: a prospective study.

At T 0 , participants in the arthritis group were older than the controls (median 35 years, interquartile range [IQR] 29–41, vs. 30 years, IQR 25–36; P = 0.001) ( Figure 1 ). Median BMI and BMI classes were similar between the arthritis and control groups (27.73 kg/m 2 , IQR-23.62–33.74, vs. 26.56 kg/m 2 , IQR 23.05–32.49; P =0.25) ( Table S2 ). The median AMH concentration was lower in the arthritis group than in the control group (1.21 ng/mL, IQR 0.48–2.71, vs. 2.50 ng/mL, IQR 1.44–4.01; P = 0.001). Median age-adjusted AMH concentrations were significantly lower in the arthritis group than in controls (2.21 ng/mL, 95% confidence interval [CI] 1.88–2.54, vs. 2.78 ng/mL, 95% CI 2.52–3.05; P = 0.009). Locally weighted scatter plot smoothing ( Figure 1 ) revealed that median AMH was lower in the arthritis group than in the control group at all ages, though the difference in the slopes of the curves was not significant. No included patient in either population endorsed being diagnosed with PCOS (polycystic ovarian syndrome). Of those enrolled in the prospective arthritis cohort, 129 women ages 18–45 years were included in this study ( Table 1 ). The majority of women had RA, and most were on methotrexate (MTX), TNFα-antagonist, or both. Those on another regimen were taking steroids, hydroxychloroquine, disease-modifying antirheumatic drugs such as leflunomide, sulfasalazine, joint injections, chlorambucil, monoclonal antibodies such as rituximab, abatacept, or azathioprine. Among the 74 women on an MTX-containing drug regimen, the majority were receiving 15–20 mg MTX subcutaneously each week. The proportion of patients on MTX-containing regimens per diagnosis was as follows: 34.8% (8/23) of JIA patients; 51.8% (44/85) of RA patients; 19.0% (4/21) of SpA patients. At enrollment, patients had a median duration of disease since diagnosis of 5.63 years (IQR 1.7–12.76 yrs). Of the included women, a majority had previous live births, did not use hormonal contraception at enrollment, and had low to moderate disease severity activity per the extrapolated DAS28 index. Further details on demographics, history of arthritis (including date of diagnosis), disease severity, current treatment at enrollment, and medical, obstetric, menstrual, and conception history are provided in Table 1 . At T 0 , 59 (45.7%) women with arthritis met the criterion for DOR (< 1.1 ng/mL AMH). At highest risk for DOR were women with history of tubal sterilization (odds ratio [OR] 2.30, 95% CI: 1.00–5.25), other gynecologic surgery (OR 3.03, 95% CI 1.46–6.27), or over age 35 (OR 8.24, 95% CI: 1.40–41.47). Black women with arthritis were more likely to have DOR than their white counterparts (OR 2.29, 95%CI: 1.01–5.16). At lowest risk for DOR were those with menarche at age 13 or older (OR 0.41, 95% CI: 0.20–0.86) ( Table 2 ). Because AMH was not measured before women began taking medications to treat their arthritis, we did not calculate an association between AMH at time of enrollment and the amount of time on or type of medication. Amongst the patients in the arthritis group, the median time between T 0 and T 1 was 1.72 years (IQR 1.12–2.53). The median AMH at T 1 was significantly lower than that at T 0 (0.98 ng/mL, IQR 0.29–2.66, vs. 1.21 ng/mL, IQR 0.48–2.71; P < 0.001). Amongst women in the arthritis group, those over age 35 (OR 4.75, 95%CI: 1.50–15.09) and those who had sought care for infertility (OR 3.02, 95%CI: 1.12–8.12) had the greatest odds of being in the highest quartile for annual percentage AMH decrease ( Table 3 ). The highest quartile included those with over 28% annual decrease in AMH concentration. There was no association between weekly MTX dose and annual percentage loss in AMH. The odds of attaining the highest quartile of ovarian reserve loss ( P = 0.29), and the prevalence of DOR (after accounting for age) ( P = 0.50) did not differ significantly between the autoimmune arthritis diagnoses (JIA, RA, or SpA). Given that seeking care for infertility is a possible consequence of large annual percentage AMH decrease, we did not adjust for this variable in our multivariate analysis. It is not a confounder of an association between the treatment regimen and annual AMH decrease. After controlling for age, patients on MTX alone or MTX plus TNFα-antagonists were less likely than those on no medications to be in the highest quartile for annual percentage AMH decrease ( Table 3 ).

Patient data from women with arthritis were collected from a prospective cohort study conducted between December 2008 and March 2014. This study enrolled 300 premenopausal females age 4 to 50 years visiting Adult and Pediatric Rheumatology clinics at Barnes-Jewish Hospital in St. Louis, MO. Enrolled patients had a diagnosis of Juvenile Idiopathic Arthritis (JIA), Rheumatoid Arthritis (RA), or Spondyloarthritis (SpA) and underwent blood sampling every 3–6 months during follow-up visits with their rheumatologists. Women who were pregnant, menopausal, cognitively impaired, or had systemic JIA were excluded from enrollment. If a patient became pregnant during the study period, she was not removed from follow-up. Instead, blood draws ceased until after delivery. The Washington University School of Medicine in St. Louis Institutional Review Board approved the study protocol, and written informed consent was obtained before inclusion. At baseline and follow-up visits, participants provided information on sociodemographic characteristics (age, race, body mass index [BMI], smoking history) and medical history (current and past treatment regimens for arthritis, disease severity measures, obstetric history, contraception history, menstrual history, and inability to conceive after 12 months of trying). Patients were evaluated at each visit, including providing serum. Full data for the study were not previously analyzed or published. For the current study, all women ages 18–45 years without a history of oophorectomy and with at least two serum samples were included (n=131). We measured serum collected at the first visit (T 0 ) and last visit (T 1 ), which was up to 4.5 years later. Two women with insufficient serum to measure AMH were excluded, resulting in a final sample size of 129. ( Figure S1 ). We used data and serum samples from an existing population of healthy women without arthritis or infertility recruited for a previous study aimed at characterizing associations between lifestyle factors and obesity ( Bedrick et al. 2020 ). The parent study was approved by the Washington University in St. Louis Institutional Review Board, and all participants provided written informed consent. Women (n=198) were between ages 18 and 44 years and were recruited from the St. Louis, MO, metropolitan area between May 2014 and May 2018. Women were excluded from the parent study if they were pregnant, had a major chronic disease (e.g., diabetes mellitus, hypertension, or autoimmune disease), or a history of ovarian surgery or infertility. Participants provided serum samples at one time point. All women in the control group had regular menstrual cycles (from 25 to 35 days), and 48% had one or more live births. Serum samples from both populations were stored at −80°C between time of collection and analysis. All samples from both the arthritis and control groups were measured in the same lab. Serum AMH concentrations (ng/mL) were measured with an Elecsys® AMH immunoassay (Roche Diagnostics), which has a linear range of 0.03–23 ng/mL, limit of detection of 0.01 ng/mL, and reported percent coefficient of variation of 1.0–1.6%. For samples with concentrations below 0.03 ng/mL, the AMH level was entered as 0.03. For those in the arthritis group, disease activity at enrollment was assessed by using two measures: number of objective joints involved in arthritis (according to the rheumatologist-supplied 28-joint index) and an extrapolated Disease Activity Score in 28 joints (DAS28) ( Scott and Houssien 1996 , Inoue et al. 2007 , Tamhane et al. 2013 ). The calculated DAS28 was scored by using the following inputs: the tender and swollen joint count, erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP) concentration, and a global health score (GHS) as assessed by the strongly-correlated patient Health Assessment Questionnaire (HAQ) Disability index ( Markenson et al. 2013 ). The HAQ score ( X HAQ ) was converted to a GHS score as follows: if X HAQ = 0, then GHS = 0 (Mild Disability); if 0.1 ≥ X HAQ X HAQ < 2, then GHS = 7.5 (Severe Disability); if X HAQ ≥ 2, then GHS = 10 (Very Severe Disability). DAS28-ESR or DAS28-CRP values were scored as indicating disease remission, low disease activity, moderate disease activity, or high disease activity as described previously ( Inoue et al. 2007 , Tamhane et al. 2013 ). To compare AMH concentrations across ages between the control and arthritis groups, the statistical package STATA, version 16.1, was used to model AMH concentrations by locally weighted scatter plot smoothing regression. The statistical package SPSS Statistics, version 24 (IBM), was used for all other analyses. Differences between the control and arthritis groups were calculated by using the Mann-Whitney U-test for continuous variables. Univariate analysis of covariance was used to adjust for age differences between control and arthritis groups. Nonparametric tests and the Wilcoxon matched-pair signed-rank test were used to compare intra-patient AMH concentrations at T 0 to AMH concentrations at T 1 (non-normal). To assess the association between sociodemographic characteristics and diminished ovarian reserve (DOR, defined as AMH < 1.1 ng/mL at T 0 ) in the arthritis group, univariate logistic regression was performed. Two-sided P -values less than 0.05 were considered statistically significant. The percent change in AMH concentration per year between T 0 and T 1 was calculated as: [AMH(T 1 )-AMH(T 0 )] / [AMH (T 0 )] / [T 1 -T 0 in years]. Patients in the arthritis group were dichotomized according to their annual decrease in AMH into two groups: >75 th percentile and ≤ 75 th percentile (defined as the referent group). To identify the sociodemographic and clinical variables associated with the highest annual percent decrease in AMH (>75 th percentile), univariate and multivariable logistic regression were carried out. In this way, significant variables associated with highest annual percent decrease in AMH were identified. Given the small sample size, Firth’s penalized logistic regression was used to examine the association between the covariates (found to be significant in the univariate logistic regression described above) in patients with the highest annual percent decrease in AMH concentration. Firth’s logistic regression uses a penalized likelihood estimate of the logistic model that reduces small-sample bias, which can be observed with standard maximum likelihood measures ( Firth 1993 ).

Our results demonstrate that AMH concentrations in premenopausal women with several types of autoimmune arthritis are lower than those of healthy controls. It remains to determine whether the difference between the two groups is clinically relevant. Consistent with studies in other populations, we found that women with arthritis who had DOR were more likely than those without DOR to be over age 35 years and to have a history of tubal sterilization or prior gynecologic surgery ( Ye et al. 2015 ). As in other studies, later menarche (after age 13) correlated with the lowest odds of DOR ( Weghofer et al. 2013 ). Those in the highest quartile of annual AMH decrease were more likely than those in the referent group to be over age 35 and report a history of difficulty conceiving. Finally, we found that women with arthritis taking methotrexate had lower odds of being in the highest quartile of annual AMH decrease than women on other drug regimens or no medications. Together, our findings suggest that long-term methotrexate is not harmful to ovarian reserve as measured by AMH. There have been conflicting reports about the impact of autoimmune arthritis on ovarian reserve, and most have been based on one-time clinical assessments of patients. In our study of patients evaluated at two time points, the designation of DOR (AMH < 1.1 ng/mL) and annual decrease in AMH did not correlate with disease severity, specific arthritis diagnosis, disease duration, race/ethnicity, current smoking status, current use of hormonal contraception, number of previous live births, age at last pregnancy, or BMI. Our findings are in line with those of Henes et al., who found that patients with RA and SpA had significantly lower AMH concentrations than healthy controls ( Henes et al. 2015 ). However, our results contradict those of Brouwer et al., who concluded that AMH concentrations of women with early RA were comparable to those in healthy controls ( Brouwer et al. 2013 ).The difference in findings may reflect duration of disease. Whereas Brouwer et al. studied women with joint symptoms that were present for less than 12 months, Henes et al. and our study included women with median durations of disease of 5.9 and 5.6 years, respectively ( Brouwer et al. 2013 , Henes et al. 2015 ). However, we also found that the annual percent decrease in AMH did not correlate with disease severity, which is consistent with previous findings ( Brouwer et al. 2013 ). Earlier onset of menopause has been reported in RA patients, and one study reported that early menopause was an independent predictor of rheumatoid arthritis ( Del Junco et al. 1989 , Pikwer et al. 2012 ). Fertility is also reduced in women with RA ( Smeele and Dolhain 2019 ). One possibility is that the chronic inflammation in arthritis patients has a negative impact on ovarian reserve. Although it is unclear whether inflammation directly reduces ovarian reserve, our findings suggest that reducing inflammation with long-term low-dose methotrexate is not harmful to ovarian reserve. It seems likely that control of inflammation with pharmacologic therapies will not harm ovarian reserve and may even benefit ovarian reserve in patients with autoimmune arthritis. It is also likely that factors outside ovarian reserve contribute to subfertility in RA. In fact, some reviews suggest systemic inflammation in arthritis may impact endometrial receptivity to an implanting blastocyst, and some speculate that there may be a relationship between sexual dysfunction and subfertility in the younger RA population.( Provost et al. 2014 ) One weakness of our study was that serum samples were stored at −80 °C for up to 10 years before analysis. However, several studies have shown that long-term storage of serum samples at −80 °C and freeze-thawing have negligible impact on AMH concentration measurements with current methods ( Morse M 2016 ). Additionally, given that all samples were subjected to the same conditions, it is unlikely that long-term freezing affected the outcomes of the study. Another weakness of the study was that our control group was not accrued at the same time as the study population, and the control participants only provided one serum sample. It is unlikely, however, that the differences in recruitment timing influenced our comparison of median age and AMH levels between the two populations. In addition, the number of patients on methotrexate was small, so confirmation of our findings in a larger sample size is necessary. Finally, although we assessed many variables, unmeasured confounding variables or misclassification of variables (such as precise dates of medication dose adjustments) may have influenced the results. An important advantage of our study was that we measured AMH concentration in serum collected from arthritis patients at two time points, most over one year apart. Another strength was the detailed obstetric, gynecologic, medication, and past medical history available for all arthritis patients, including data on use of hormonal contraception, parity, and menarche. Although age at menarche was self-reported, our use of a dichotomous variable minimized the potential misclassification bias here, as women are reliably able to recall whether their age of at menarche was before or after 12 years of age (or within one year) ( Lundblad and Jacobsen 2017 ). Finally, of the potential study population, only 11 arthritis patients were excluded for having fewer than two serum samples collected, so the potential for selection bias due to this inclusion criteria was likely minimal. To our knowledge, this is the first study on the ovarian reserve effects of long-term, low-dose MTX in humans. Animal studies showed that high-dose MTX (5 g/m 2 ) could induce destruction of primordial follicles ( Gol et al. 2009 ), but recurrent low dose MTX (1 mg/kg) did not cause a statistically significant change in ovarian reserve as measured by AMH ( Benian et al. 2013 ). Here, we found no dosage-dependent association between the weekly MTX dose (7.5 to 20 mg) and the annual decrease in AMH concentration. In conclusion, our study suggests that long-term MTX is not associated with DOR as measured by AMH in arthritis patients. Instead, MTX use may result in lower odds of rapid ovarian reserve loss than no treatment or other drugs. A large, longitudinal study with appropriate controls is needed to definitively determine the impact of MTX on ovarian reserve in autoimmune arthritis patients.

Worldwide, autoimmune diseases affect approximately 8% of the population, and 78% of those affected are women ( Fairweather and Rose 2004 ). Many women of child-bearing age have a type of autoimmune arthritis such as rheumatoid arthritis (RA), juvenile idiopathic arthritis (JIA), or spondyloarthritis (SpA). Women with rheumatoid arthritis can have prolonged times to pregnancy ( Brouwer et al. 2015 ). Furthermore, other types of autoimmune disease can increase women’s risks of early menopause, and these diseases may accelerate loss of ovarian function ( Sammaritano 2012 ). However, it is unclear whether this loss in ovarian function is due to chronic inflammation or the medications required to treat these diseases ( de Araujo et al. 2014 , Henes et al. 2015 , Eudy et al. 2019 ). Moreover, it is unclear whether autoimmune arthritis, or the drugs used to treat it, affect ovarian reserve. Whereas Brouwer et al . found that ovarian reserve did not differ between those with and without early-onset RA, Henes et al. found that patients with RA or SpA had lower ovarian reserve than healthy controls ( Brouwer et al. 2013 , Henes et al. 2015 ). Patients with autoimmune arthritis are often managed with methotrexate (MTX) or tumor necrosis factor-α (TNF-α) antagonist for many years ( Köhler et al. 2019 , Ringold et al. 2019 , Ward et al. 2019 ). MTX has little effect on ovarian reserve or the follicle pool when women are given one to three doses ( Boots et al. 2016 ), but the effect of chronic MTX use on ovarian reserve has not been determined. Endogenous TNF-α, a proinflammatory cytokine, is implicated in many normal ovarian processes, such as folliculogenesis, ovulation, and atresia ( Field et al. 2014 ). Elevated expression of TNF-α is found in mice with impaired oocyte competence, and injection of TNF-α antagonists in these mice reduced the adverse apoptotic effects of stress on oocytes.( Zhao et al. 2020 ). Thus, TNF-α antagonists may protect the ovary, and one recent case report described normalization of AMH in a patient with psoriasis treated with anti-TNF-α agents ( Czyżyk A 2019 ). However, to our knowledge, no study has addressed whether or not these common medications affect ovarian reserve in patients with autoimmune arthritis. The primary objectives of this study were to determine the clinical profile of women with arthritis who have diminished ovarian reserve, and to examine the association between arthritis drug regimens and ovarian reserve. Additionally, we aimed to compare ovarian reserve among females with arthritis and a group of women without arthritis (hereafter referred to as controls). For this work, we used serum concentration of the glycoprotein hormone Anti-Müllerian Hormone (AMH) as a marker for ovarian reserve. Serum concentration of AMH, which is secreted by ovarian granulosa cells of preantral and small growing follicles, is a well-established marker of ovarian reserve( Tsepelidis et al. 2007 ). AMH can be reliably measured on any day in an individual’s menstrual cycle ( Tsepelidis et al. 2007 ), and an AMH concentration below the detection limit (< 0.4ng/mL) is correlated with premature ovarian failure ( Méduri et al. 2007 ).

Supplementary Figure S1. Flowchart for arthritis group participants Supplementary Table S2. Control participant demographics mmc1.docx