A prospective cohort study of infertility and cancer incidence.

The Nurses’ Health Study II (NHSII) is an ongoing prospective cohort study of female nurses in the United States. In 1989 (baseline), 116,429 women aged 25–42 years were enrolled.( 19 ) Mailed or electronic questionnaires have been sent to participants biennially to update health information. The cumulative response rate exceeds 85% ( Table S1 ). The present study was approved by the Institutional Review Boards of Brigham and Women’s Hospital and the Harvard T.H. Chan School of Public Health, and those of participating registries as required. Return of questionnaires implied informed consent. Infertility was self-reported biennially from 1989 to 2001 and quadrennially thereafter through 2009, when the youngest participant was 45 years old. At baseline, participants were asked ‘Have you ever tried to become pregnant for more than one year without success?’ If yes, they were asked to report the age when first experienced infertility and to specify the underlying one or more cause(s) of infertility (tubal blockage, ovulatory disorder, endometriosis, cervical mucus factors, factors related to their spouse, no investigation done, cause not found, or other). This definition is in accordance with infertility definitions during the time of study.( 20 ) In follow-up questionnaires (1991–2009), participants were asked about infertility and infertility diagnosis since the questions were last asked. The validity of self-reported infertility has been confirmed in this cohort in a randomly selected subset of 100 of women who had reported primary ovulatory infertility (e.g., infertility with no prior live birth).( 21 ) Among 90 women who provided medical records, 84 (93%) reported a confirmatory diagnostic test (abnormality of basal body temperature charts, progesterone assay, endometrial biopsy specimen, or other confirmatory test) or treatment (clomiphene, menotropins, human chorionic gonadotropin, or other confirmatory treatment). Incident invasive cancer cases were identified through self-report on follow-up questionnaires, and then confirmed through review of medical records. Only cases confirmed in medical records were counted. Cancer was further categorized into obesity-related cancer (colorectal, gallbladder, kidney, multiple myeloma, thyroid, pancreatic, esophageal, gastric, liver, endometrial, ovarian, post-menopausal breast) and non-obesity related cancer (all other cancer, not including basal cell carcinoma and squamous cell carcinoma) based on the classification from the International Agency for Research on Cancer (IARC).( 22 ) Incident cancers ascertained during follow-up are summarized by cancer site in Table S2 . Women were asked about their date of birth, race/ethnicity, age at menarche, height, weight at age 18, oral contraceptive use before age 18, and marital status on the baseline questionnaire. Weight, pregnancy history, menopausal status, post-menopausal hormone use, smoking status (cigarettes [cigs]/day), family history of cancer in first-degree relatives (breast, ovarian, endometrial, colorectal, prostate, pancreatic, and melanoma), history of hysterectomy/ultrasound confirmed uterine fibroids, breastfeeding history, and cancer screening (i.e., breast exam, mammography, or sigmoidoscopy/colonoscopy in the past two years) were queried on biennial questionnaires since baseline. Current usual menstrual cycle length (when not ovulating or lactating) was assessed in 1993. History of ovulation induction treatment (clomiphene and/or gonadotrophins) was ascertained on biennial questionnaires starting 1993. Number of in vitro fertilization (IVF) cycles were assessed in 2009. If women were currently pregnant, BMI was carried forward from the most recent non-pregnant follow-up cycle. Physical activity was reported at baseline and quadrennially thereafter. Diet was assessed every four years using a validated food frequency questionnaire.( 23 , 24 ) Diet quality was characterized by the Alternate Healthy Eating Index-2010 (AHEI-2010)( 25 ) (for which higher scores indicate a healthier diet). Participants were eligible for this study if they had ever been pregnant or reported a history of infertility at baseline, or if they became pregnant or had newly reported infertility during follow-up. We excluded women who withdrew from the cohort (n=17), who had a prevalent cancer diagnosis at baseline (n=1,466), or who never returned follow-up questionnaires (n=1,353), leaving 103,080 women for analysis. We considered a woman to have infertility throughout follow-up after a report of infertility, regardless of responses to subsequent questionnaires. Women with newly reported infertility after age 40 years were considered unexposed in the main analysis, as new infertility after age 40 is rare and primarily driven by advanced maternal age.( 2 , 26 ) Follow-up time for each eligible participant was calculated from the date of return of the questionnaire on which a pregnancy or infertility was first reported until the date of cancer diagnosis, death, or the end of follow-up (June 1, 2015), whichever occurred first. For patients reporting multiple primary cancers, follow-up ended with the first report. We tested the proportional hazards assumption by running models that included interaction terms between each primary exposure and age in months. Crude survival curves were estimated by the Kaplan-Meier method, and statistical significance was assessed using the log-rank test. We used Cox proportional hazard models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for cancer incidence in relation to infertility status and age at infertility diagnosis, stratified by age (months) and calendar year. Trend analysis was conducted using a Wald test treating indicator levels as continuous variables. We fit separate models using total, obesity-related, non-obesity-related, and site-specific cancers as dependent variables. Subsequently, we examined the associations between causes of infertility and cancer risk. Multivariable Cox models were adjusted for a priori selected known predictors of cancer, including family history of cancer (yes, no), race/ethnicity (White, non-White), age at menarche (≤10, 11, 12, 13, or ≥14 years), BMI at age 18 (continuous), and oral contraceptive use before age 18 (never, <2 months, 10+ months). We set the cutoff of age 18 to qualify covariates as potential confounders and facilitate causal inference. In a secondary analysis, we further adjusted for covariates measured in adulthood, including time-varying smoking status (never, past: 1–34, past: 35+, current: 1–34, current: 35+ cigs/day), physical activity (0, 0.1–1.0, 1.1–2.4, 2.5–5.9, or ≥6 h/week), BMI (continuous), AHEI-2010 (quintiles), post-menopausal hormone use (pre-menopausal, post-menopausal: never/past/current users in joint groups, unknown), recent cancer screening (breast exam, mammogram, sigmoidoscopy, and colonoscopy), marital status (yes/no). Additional models adjusted further for 1) gravidity (0, 1, 2, 3+), or 2) gravidity and age at first pregnancy (nulligravid, 1/2+ previous pregnancy and age at first pregnancy <25y/≥25y in joint groups), or 3) parity and age at first birth (nulliparous, 1–2/3+ previous live births and age at first birth <25/≥25 years in joint groups). Missingness of all variables was <5%. Indicator variables were used for any missing covariate information for categorical variables. Missing values of continuous variables were replaced with the median.( 27 ) We applied causal mediation analysis methods to examine the proportion of the association between infertility and cancer that was mediated by parity and age at first birth (joint group) and breastfeeding history (≤3, 4–12, >12 months). The proportion mediated on a risk difference scale was calculated as the indirect effect divided by the total effect, and tests were conducted for evidence of mediation using a publicly available macro.( 28 – 30 ) We used cross-product terms to test for effect modification by BMI at 18, time-varying BMI, smoking status, gravidity, and use of ovulation induction therapy. The statistical significance of multiplicative interaction was tested by the likelihood ratio test. The additive interaction was assessed by calculating the relative excess risk due to interaction (RERI) considering the interaction with respect to all levels of the (main effect) variables.( 31 ) We conducted several sensitivity analyses. First, to ensure that the age at infertility evaluation was not a proxy for age at first pregnancy attempt, we conducted an analysis matching age at infertility (among those who reported primary infertility) with age at first birth (among those who did not report infertility during follow-up). Second, we fit models with linear and quadratic terms for BMI. Third, we included women whose infertility was entirely attributed to their spouse in the reference group. Fourth, because several gynecological disorders are known to be common causes of infertility, to reduce potential misclassification, we excluded women who did not directly report a history of infertility but who had a self-reported diagnosis of endometriosis, uterine fibroid (confirmed by ultrasound or hysterectomy), or having menstrual cycle length >39 days (as a proxy for polycystic ovarian syndrome (PCOS))( 32 , 33 ) from the analysis. Fifth, we considered women with newly reported infertility after age 35 as fertile to minimize misclassification resulting from age-related fecundity decline. Sixth, we expanded our exposure definition to include all women reporting 12-months of infertility, even those newly occurring after age 40. Seventh, we allowed all otherwise eligible women in the analysis regardless of pregnancy history. Eighth, we additionally adjusted for history of ovulation induction and number of IVF cycles. All data were analyzed using SAS 9.4 (SAS Institute Inc). All statistical tests were two-sided, and a P value of <0.05 was considered statistically significant.

Among 103,080 women, 26,208 women (25.4%) reported ever having a history of infertility over follow-up. The mean (standard deviation, SD) age at infertility diagnosis was 28.4 (5.1) years. Most participants were non-Hispanic White (95.7%), married (91.9%), and had never smoked (61.6%). Baseline characteristics of participants with and without a history of infertility were very similar ( Table 1 ) except gravidity and nulliparity. The most common underlying cause of infertility was ovulatory disorders (28.8%). During 2,149,385 person-years of follow-up, we documented 6,925 incident invasive cancer cases ( Table S2 ). The median age at end of follow-up was 56.0 years. The five leading primary sites of cancers were pre-menopausal breast, post-menopausal breast, melanoma, endometrial, and thyroid cancer, accounting for 62.8% of all cases. Women with a history of infertility had a higher incidence of invasive cancers during follow-up compared to gravid women without a history of infertility ( Figure 1 ) but had similar age of cancer diagnosis (mean (SD) years, with infertility: 51.2 (7.4); without infertility: 50.5 (7.7)). After adjusting for demographic, reproductive, and lifestyle risk factors, women with a history of infertility had a 7% higher risk (HR=1.07; 95% CI, 1.02–1.13) of developing cancer than women who did not experience infertility. However, this association was restricted to obesity-related cancers (HR=1.13; 95% CI, 1.05–1.22; non-obesity-related cancers, HR=0.98; 95% CI, 0.91–1.06, Table 2 , model 2). The association of infertility with obesity-related cancer was slightly attenuated but remained statistically significant after adjusting for gravidity and age at first pregnancy or parity and age at first birth ( Table 2 , models 3–4). Site-specific analysis revealed that the increased risk of obesity-related cancer was mainly driven by obesity-related gynecological (post-menopausal breast, endometrial, and ovarian) cancers ( Table S3 ). The overall risk of developing gynecological cancers (n=2,009 cases) was 17% higher (HR=1.17; 95% CI, 1.06–1.29) among women with a history of infertility than among women without a history of infertility after adjusting for demographic, reproductive, and lifestyle risk factors. The corresponding HR for incident endometrial or ovarian cancers (n=836 cases) was 1.17 (95% CI, 1.00–1.36). Mediation analysis showed that age at first birth and parity together explained 35.8% (95% CI, 16.1%−61.9%) of the increased risk in obesity-related gynecological cancer ( Table S4 ). The associations were not mediated by breastfeeding history. Compared to women without a history of infertility, women who first experienced infertility at or before age 25 years had a 19% (95% CI, 1.07–1.33) higher risk of obesity-related cancer, whereas the corresponding effect estimate for women first experiencing infertility after age 30 years was 7% (95% CI, 0.94–1.22) ( Table 3 ). This age-related dose-response pattern was more pronounced when we restricted the analysis to nulliparous and age-matched women ( Table S5 ). Similarly, the association between infertility and obesity-related cancer was only found among women with primary infertility ( Table S6 ). To examine whether any of the causes of infertility was driving the association between infertility and cancer risk, we conducted the analysis by self-reported causes of infertility among women who had a history of infertility. Spouse/partner factors and ‘other causes’ were associated with increased risk of obesity-related cancer compared to women who did not report infertility ( Table S7 , models 1–2), but did not differ significantly with other diagnoses ( Table S7 , model 3). We found no evidence that the association between infertility and obesity-related cancer differed by time-varying smoking status, BMI, or use of ovulation induction drugs ( Table S8 ). Notably, even women who had a BMI within the normal range (18.5–24.9 kg/m 2 ) at age 18 were at higher risk of developing obesity-related cancer (HR=1.16; 95% CI, 1.06–1.27; P value=0.001); only women who were underweight at age 18 did not have an increased obesity-related cancer risk among those with infertility (HR=0.92; 95% CI, 0.74–1.13; P value=0.41). Nulliparous women were at higher risk of developing obesity-related cancers after infertility than parous women ( P multiplicative interaction=0.006; P additive interaction=0.01; RERI (95% CI) =0.42 (0.15–0.70)). Results remained unchanged in analyses that adjusted for BMI using nonlinear terms, excluded women who reported endometriosis, uterine fibroids, and cycle length >39 days from the reference group, moved spouse-only infertility into the reference group, treated newly reported infertility after age 35 years as unexposed, included newly reported infertility after age 40 years, included all otherwise eligible women in the analysis regardless of their pregnancy history, or additionally adjusted for infertility treatment ( Table S9 ).

In this large prospective cohort study of women followed for up to 26 years, we found that a history of infertility was significantly associated with an increased risk of invasive cancer, especially obesity-related cancers. Results were comparable when we applied different cut-off ages for new report of infertility. These associations were limited to infertility occurring early in reproductive life and were primarily driven by an elevated risk of gynecological cancers (post-menopausal breast, ovarian, and endometrial cancer) and were largely mediated by delayed childbearing and lower parity. The association between infertility and obesity-related cancer was stronger among nulliparous women but did not vary by BMI, smoking status, or exposure to ovulation induction treatment. Previous studies have yielded inconsistent results comparing the crude cancer risk of women reporting infertility with that of the general population. A retrospective study among 2,335 patients seeking infertility evaluation (1935–1964) found that the observed number of total cancers was nearly identical to the expected number derived from the Connecticut Cancer Registry (standardized incidence ratio [SIR]=1.00).( 8 ) An analysis of 10,358 IVF-treated Australian women with infertility (1978–1992) also showed similar cancer incidence rates compared to the general Australian population (SIR=1.01, 95% CI: 0.82–1.25).( 10 ) A hospital-based study in Israel (1964–1974) found no significant increase in cancer risk among 2,632 women who underwent fertility treatment compared to the national cancer incidence (SIR=1.1, 95% CI: 0.8–1.5)( 3 ) but found an elevated cancer incidence after extended follow-up through 1991 (SIR=1.2, 95% CI: 1.0–1.5).( 9 ) In contrast, another multicenter study in Israel (1981–1992) suggested a lower cancer incidence among women who underwent IVF treatment compared to national population data (SIR=0.76; 95% CI: 0.50–1.10).( 34 ) Discrepancies in these results reflect differences in the study populations and study designs, as well as some important methodological issues including low statistical power and the inability to control for important confounders. In addition, as most of the abovementioned studies were conducted decades ago, the dramatic changes in infertility screening, fertility treatments, and management cannot be ignored. More importantly, the existing literature limited cases to women who pursued fertility evaluation and treatment, which may not be representative of women with infertility, as only 65% of women who met the diagnosis criteria of infertility had a medical evaluation.( 35 ) Our finding of an association between a history of infertility with obesity-related gynecological cancers is consistent with previous reports of a positive association between infertility and risk of post-menopausal breast,( 36 , 37 ) endometrial, 9 and ovarian cancer.( 7 , 38 ) Nevertheless, our findings do not agree with our initial hypothesis that common pathophysiological and metabolic characteristics observed in women with infertility, such as hyperinsulinemia, hyperandrogenism, and chronic inflammation would result in an elevated risk of obesity-related cancers. Although the associations identified were with gynecologic cancers that also happened to be obesity-related, there was no observed association with non-gynecologic obesity-related cancers, such as colorectal and thyroid cancer. Instead, our results suggest that infertility results in changes in reproductive risk factors for reproductive cancers, such as delayed age at first birth and lower parity, which in aggregate result in a small elevation in risk of these malignancies.( 4 , 7 , 39 – 42 ) There are other possible mechanisms linking infertility to obesity-related gynecological cancers, such as shared hormonal disturbances and immune aberrations that we could not document in our study but worth additional research.( 12 – 18 ) However, the results should be interpreted with caution, because there may be some misclassification of obesity-relatedness based on the IARC evidence. For instance, although non-Hodgkin lymphoma was not classified as obesity-related cancer by IARC, accumulating evidence suggests that diffuse large B-cell lymphoma, the most common type of non-Hodgkin lymphoma, may be obesity-related.( 43 – 45 ) In contrast, serous ovarian cancer, the most common subtype of ovarian cancer, is not associated with BMI.( 46 ) We found that earlier onset of infertility and primary infertility were associated with higher cancer risk. This is biologically plausible, as the etiology of infertility in younger women differs from that in their older counterparts. Women who experience infertility at an earlier age are more likely to have primary infertility rather than secondary infertility, in which the shift in some reproductive risk factors such as age at first birth and parity would be more prominent.( 47 , 48 ) In addition, we found that even a normal BMI at age 18 (e.g., any BMI but an underweight value) was associated with an increased risk of obesity-related cancer later in life, suggesting that earlier life body leanness may in part mitigate the hormonal disturbances and immune aberrations linking infertility with obesity-related cancer. The strengths of this study include large sample size, more than 6,000 incident cancer cases after rigorous adjudication, and a longitudinal design with high response rates over two decades of follow-up. Moreover, we were able to control for prospectively collected information on important lifestyle and reproductive risk factors for cancer. We also managed to disentangle the effect of gravidity, parity, age at infertility evaluation, different causes of infertility, and infertility treatment apart from the aggregated infertility exposure itself. Our study has several limitations. First, infertility and its underlying causes were self-reported. Although self-reported ovulatory infertility diagnoses have been validated in this cohort,( 21 ) the possibility of exposure misclassification cannot be ruled out. We did not observe any meaningful association between any specific infertility diagnosis with cancer risk, but the means of distinguishing different causes of infertility was crude, and we do not have information about the underlying diagnoses of ‘other causes’. For instance, ovulatory disorders could encompass both women with diminished ovarian reserve and PCOS, which are etiologically distinct infertility diagnoses. Furthermore, changes in clinical practice over time may have introduced misclassification. Second, because this was an observational study, residual confounding likely remains. Third, the study primarily took the approach of comparing women reporting infertility with participants who were gravid but never reported a history of infertility. Many of the women who we excluded may have experienced infertility, as they never had a documented pregnancy or pregnancy attempt during follow-up. However, results were comparable when we included all participants regardless of pregnancy attempts. Fourth, this cohort is relatively homogenous regarding race (96% White), profession (nurses), and socioeconomic status, potentially limiting generalizability. Fifth, we do not have detailed information regarding infertility treatment relating to each pregnancy.

Results from this study showed that a history of infertility is associated with cancer risk later in life, but this is limited to risk of obesity-related gynecological cancers. In particular, women experiencing infertility earlier in their reproductive life had a higher risk of developing cancer. Our findings provide knowledge to women’s health primary care providers that women with a history of infertility may need increased screening for obesity-related gynecologic cancer. Health care providers should consider informing patients with history of infertility about their long-term health risks, maintaining regular follow up, and encouraging lifestyle changes to affect modifiable risk factors of cancer.

Infertility, defined as the inability to conceive after 12 months of regular, unprotected sex, reaches a lifetime risk of 20–25% in the United States.( 1 , 2 ) Infertility has been suggested as a risk factor for gynecological cancers, including ovarian and endometrial cancers, possibly due to lower parity and use of exogenous hormones during infertility treatment (e.g., multiparity reduces the number of lifetime ovulatory cycles, subsequently lowering risk of ovarian cancer).( 3 – 7 ) Attempts have been made to investigate the associations between infertility and other cancers, but results remain inconclusive, reflecting the limitations of a retrospective study design, restricting the study sample to women seeking medical evaluation for fertility, relatively short follow-up, low cancer incidence, or failure to control for important predictors of cancer risk such as lifestyle factors, hormonal therapy, and reproductive history.( 3 , 8 – 10 ) Infertility is a highly heterogeneous disease. Causes of infertility include endocrine, infectious, autoimmune, and psychological disorders, or a combination of multiple factors.( 11 ) Women experiencing infertility have some common underlying pathophysiological changes, including insulin resistance, hyperandrogenism, and chronic inflammation.( 12 – 15 ) These changes were also established risk factors for cancer, especially obesity-related cancers.( 15 – 18 ) Therefore, we hypothesized that women with a history of infertility might be at increased risk of developing cancer and that this association would be stronger for obesity-related cancers. To address these questions and overcome the shortcomings in the existing literature, we leveraged a large prospective cohort that has actively followed participants for over 26 years to investigate the association between history of infertility and subsequent risk of cancer. In addition, we examined whether age at first infertility diagnosis, the timing of infertility in relation to parity (primary vs. secondary infertility) and causes of infertility were associated with cancer risk.