Main findings
Our study demonstrates that female AYA survivors of non-gynecologic cancer have an increased risk of subsequent POI diagnosis, compared to unexposed individuals. The risk of POI is elevated in survivors of leukemia, breast cancer, NHL, and HL in both age groups, and thyroid cancer in the older group (30–39 years). No increased risk was identified in melanoma survivors. Moreover, our study demonstrates that the risk of POI varies by age, with a higher risk in the age group 30–39 years than in those 15–29 years.
Interpretation
In comparison to our exposed cohort of 21,666 patients, the largest previous study thus far included only 2819 eligible childhood cancer survivors [5]. Additionally, other studies in the field have included much narrower age ranges, with most studies thus far focusing on cancer survivors < 21 years old, neglecting patients 21–39 years old [4, 5, 9,10,11,12,13,14]. The inclusion of young adults in this study was imperative, as several studies have revealed that an older age at cancer diagnosis is a risk factor for the development of POI [4, 9, 13], although one study found no association [8]. On the other hand, using data from the California Cancer Registry reported the opposite relationship, such that the risk of experiencing early menopause was increased at a younger age of cancer diagnosis in patients with HL, NHL and gastrointestinal malignancies [19]. However, this was only true for patients whose menses resumed within one year of diagnosis. The risk of experiencing acute ovarian failure was increased at an older age at diagnosis. In relation to the impact of age at cancer diagnosis, our study corroborates the former research, as we have demonstrated an increased risk of POI in young adults (age 30–39) compared with those aged 15–29 years. This increased risk is maintained across all types of cancer studied, excluding melanoma, with a higher risk of POI at an increased age of diagnosis for survivors of leukemia, breast cancer, NHL, HL, and thyroid cancer. Our study also demonstrates that this risk is applicable beyond the comparison of cancer treatment before and after pubertal onset, used in previous studies [4, 13], as 96% of girls achieve menarche by age 15 [20], the youngest age in our cohort. As well, our study supports the finding of the Reproductive Window Study, a cross-sequential study of the ovarian function of 763 female AYA survivors in California and Texas, which reported that younger age at cancer treatment was associated with higher long-term trajectories of ovarian function as measured by levels of the anti-Müllerian hormone (AMH) [15].
The results of this study may need to be interpreted differently than other studies in the field, based on the use of varying methods to define a POI diagnosis. While we identified a diagnosis of POI based on physician billing codes for menopause (ICD-9 627) in women < 40 years, other studies used different definitions. In the Childhood Cancer Survivor Study, the largest prior analysis in the field, self-report questionnaires were used to identify women with POI [5]. Here, patients were considered menopausal if they failed to experience spontaneous menses for a minimum of 6 months and other causes such as pregnancy and injectable hormone use were excluded. This study reported a prevalence of non-surgically induced POI of 8%, and a rate ratio of 13.21 for survivors of various childhood cancers, compared to controls [5]. This higher prevalence and risk may be explained by response biases in self-reported data. Several other studies also relied on self-reporting and questionnaires to identify women with POI. These included a study based in Ontario, Canada, which identified a prevalence of 8.8% [4], a US study which identified a prevalence of 31.42% [9], a small study of childhood sarcoma survivors which identified a prevalence of 49% [21], and a French study which identified a prevalence of 2.1% [13]. Importantly, these studies included patients diagnosed as early as 1945 and no later than 1998, thus reflecting older treatments that have changed considerably over time. Other smaller studies have relied on hormone measurements, including serum FSH levels, to diagnose POI. These studies tended to report a higher prevalence of POI, with values of 17% in a 100-participant large prospective Danish cohort [12], 31.25% in a study of 32 British participants with a history of HL [10], and 57.1% in a cohort of 21 participants in France who received high-dose chemotherapy and autologous bone marrow transplantation without radiation [11] . One larger prospective study with 921 participants from the St. Jude Lifetime Cohort also used serum FSH levels to diagnose POI, and found a prevalence of 10.9% for survivors of various childhood cancers [8].
Our study identified a prevalence of POI in AYA survivors of leukemia of 21%, while it was 7.5% in survivors of HL and NHL. In a case–control study of 2819 survivors of childhood cancer from the multicenter Childhood Cancer Survivor Study, Sklar et al. found that in survivors of childhood leukemia the occurrence of self-reported POI through survey data was 14% for leukemia, 56% for HL, and 3% for NHL [5]. In AYA survivors, using the California Cancer Registry, Letourneau et al., conducted a retrospective survey study of 1041 women diagnosed with cancer between the ages of 18 and 40 years, which identified an early menopause (< 45 years) prevalence of 37% (age 20) and 16% (age 35) for HL and 56% (age 20) and 16% (age 35) for NHL survivors [19]. These differences may be attributed to the use of self-reported data to ascertain a diagnosis of POI, and selection bias in prior survey studies (i.e., patients with POI might have been more interested in responding the survey than patients without POI). In addition, Letourneau et al. included patients’ with early menopause which will result in a higher proportion of patients meeting this diagnosis [19]. Differences among studies can also be explained by different treatment protocols in pediatric versus the AYA population, and therapies with different gonadotoxic potential.
Our study also identified that AYA survivors of thyroid cancer have an increased risk of POI. This differs from previous literature which found no difference in the age at menopause, up to 47 years old, between differentiated thyroid cancer survivors and controls [22]. Differences with our study may be attributed to the older study population, sole inclusion of treatment with radioiodine-131, and use of older data. These discrepancies should be evaluated in future research. In fact, in our population-based study on the risk of infertility in survivors of AYA cancer in Ontario we identified an increased risk of infertility in AYA survivors of thyroid cancer (1.20, 95% CI 1.10, 1.30) compared with an age-matched cohort of individuals without cancer [23]. As well, others have reported a decreased overall pregnancy rate in women with thyroid cancer (SIR 0.79; 95% CI 0.72, 0.86) and a decreased cumulative incidence of first pregnancy in nulliparous women (HR 0.69; 95% CI 0.59, 0.81) [24]. The impact of thyroid cancer on reproductive function needs further investigation. Finally, our study identified no increased risk of POI in survivors of melanoma. Nonetheless, our prior research has identified a small increased risk of infertility in melanoma survivors (RR 1.17, 95% CI 1.01–1.35), thus further research is needed on the reproductive impact of female patients with melanoma [23].
Overall, our findings highlight an increased risk of POI in AYA cancer survivors, contributing to the growing body of research underlining reproductive health after cancer. The ovarian reserve plays an important role in the physiopathology of POI after specific cancer diagnosis. The ovarian reserve declines with age, as documented by studies modeling the trajectories of AMH levels throughout the reproductive lifespan [25, 26]. Age at time of treatment, and hence ovarian reserve, will determine POI. For breast cancer and hematological malignancies, systemic cancer therapy, and pelvic radiation contribute the most to POI. For thyroid cancer, if an association with POI is further confirmed, the mechanisms need to be investigated given that thyroid cancer is not generally treated with chemotherapy, or radiotherapy to a field that involves the ovaries. These results should be incorporated into the counseling of female AYA cancer survivors by their primary care providers, cancer care providers, and fertility specialists, to improve patient understanding about their risk of experiencing POI and its health implications. Counseling should include a discussion on the risks of infertility and prompt referral for fertility preservation prior to treatment initiation if desired. These recommendations support the statements by current clinical practice guidelines [27,28,29]. In particular, the American Society of Clinical Oncology (ASCO) guidelines have evolved from recommending infertility risk discussion with cancer patients in 2006 [30], to emphasizing the importance of addressing gonadotoxicity and fertility preservation in all patients with reproductive potential, including the pediatric population in 2013 [31], to providing current guidance regarding fertility preservation options for people with cancer anticipating treatment in 2018 [27]. Yet, fertility preservation counseling and referral rates remain low. [32, 33] Thus, our research further advocates for increased efforts in knowledge translation and improvements in interdisciplinary coordination to overcome barriers to fertility preservation referral, as well as long-term surveillance of reproductive function in survivors of AYA cancer. In terms of surveillance, assessment of pre-treatment ovarian function, in particular through AMH levels, in premenopausal women with a diagnosis of breast cancer or haematological malignancy is recommended to predict post-treatment recovery of ovarian function [28].
Strengths and limitations
Strengths of this study include the population-based study design, large sample size, and inclusion of adolescents and young adult cancer survivors. Study limitations included possible nondifferential misclassification of the study outcome. Such misclassification is likely to result in an attenuation of our risk estimates. In fact, when validated against FSH levels > 25 IU/L available in a subset of this cohort, the use of a single ICD-9 627 code as a diagnosis of POI, resulted in low sensitivity 30.1% (95% CI 29.1–31.2), but high specificity of 97.0% (95% CI 97.0–97.1). It is therefore likely that some misclassification resulted in an underestimate of the effect size. On the other hand, ovarian function may be lost directly following cancer treatment, which is a separate entity termed Acute Ovarian Failure (AOF), and can be subsequently restored [34]. Data suggests that much of the recovery occurs early on, with up to 50% having resumption of menses 12-months following treatment [34]. For most of the cancers included in our study the mean age at cancer diagnosis and at POI diagnosis differed by at least 3 years, except for Leukemia (2.5 years), and breast cancer (1.4 years), decreasing the likelihood of AOF being the diagnosis instead of POI. Another limitation was the inability to account for other potential confounding variables including family history of POI, ethnicity, and smoking status, as they are not available in the administrative databases used. Finally, this study did not evaluate the impact of specific cancer treatments (e.g., alkylating agents, immunotherapy) or medications to decrease the impact of gonadotoxic treatment (e.g., GnRH agonists), which were not recorded in the databases included in our study. Future studies in the AYA population are needed in this regard.