Incidence and factors associated with premature ovarian insufficiency: a population-based cohort study

Premature ovarian insufficiency (POI) is defined as loss of ovarian function before 40 years of age ( Panay et al. , 2024 ). Recent estimates suggest a global POI prevalence of 3.7% (95% CI 3.1, 4.3), with higher prevalence reported in low- and middle-income countries and increasing rates over time ( Golezar et al. , 2019 ; Li et al. , 2023 ). The etiology of POI is heterogeneous and includes genetic, autoimmune, infectious, and iatrogenic causes. However, in ∼70% of non-iatrogenic cases, the underlying cause remains unknown ( Jiao et al. , 2017 ). Genetic conditions account for an estimated 7–13% of POI cases ( Jiao et al. , 2017 ; Rossetti et al. , 2017 ). The most common chromosomal disorders among women with POI are Turner Syndrome and Fragile X premutation, both involving X-chromosome abnormalities ( Rossetti et al. , 2017 ). Autoimmune diseases, including Type 1 diabetes mellitus, hypothyroidism, Grave’s disease, Hashimoto thyroiditis, Addison’s disease, systemic lupus erythematous, and rheumatoid arthritis—are estimated to contribute to 4–30% of POI cases ( Kirshenbaum and Orvieto, 2019 ; Sharif et al. , 2019 ). Iatrogenic POI results from follicular destruction related to ovarian surgery, chemotherapy, radiotherapy, and lifestyle exposures such as cigarette smoking ( Jiao et al. , 2017 ; Panay et al. , 2024 ). Infectious conditions, including mumps and HIV, have also been associated with POI ( Li et al. , 2023 ). Despite growing recognition of POI as an important reproductive and long-term health condition ( Velez et al. , 2019 ; Price et al. , 2021 ; Panay et al. , 2024 ), existing studies estimating its incidence and prevalence are limited by reliance on self-reported age at natural menopause, cross-sectional study designs, small sample sizes, and limited ability to evaluate factors associated with POI ( Golezar et al. , 2019 ; Li et al. , 2023 ). To address these limitations, we conducted a population-based cohort study to estimate the incidence of POI and identify associated factors among women in Ontario, Canada, using linked administrative health data from a universal, single-payer healthcare system.

Between 1995 and 2019, a total of 3 635 702 women aged 16–39 years were eligible for inclusion. Using linked administrative health databases, participants were followed for a median of 11 years (IQR 6–18) years. During follow-up, 168 173 women were diagnosed with POI from any etiology ( Fig. 1 ). Cohort creation.   a Diagnosis types are mutually exclusive, patients are included in only a single type based on medical history, exposures, and surgical interventions prior to date of diagnosis. b 1 or more physician visit for menopause before the age of 40 years (OHIP code ICD 9-627) in the absence of surgical POI, radiotherapy, or chemotherapy. c Records of surgical menopause prior to the age of 40 years (1 CCP or CCI code for bilateral oophorectomy, 2 occurrences of a code for unilateral oophorectomy, or 1 code indicating removal of remaining ovary). OHIP, Ontario Health Insurance Plan; POI, premature ovarian insufficiency; ICD 9, International Classification of Diseases, 9th Revision; CCP, Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures; CCI, Canadian Classification of Health Interventions. Overall IRs were 351.4 per 100 000 person-years for non-iatrogenic POI, 30.2 per 100 000 person-years for surgically induced POI, 5.5 per 100 000 person-years for POI following chemotherapy and/or radiotherapy, and 2.3 per 100 000 person-years for POI following surgery in combination with chemotherapy and/or -radiotherapy. Baseline characteristics of the study population are shown in Table 1 . The median age at POI diagnosis was 31 years (IQR 25–36). Most women with POI were Canadian born (94.9%) and resided in urban areas (89.3%). Women with POI were relatively evenly distributed across neighborhood income quintiles, ranging from 20.1% in the lowest to 18.9% in the highest. The prevalence of obesity, tobacco use or substance use, diabetes, and hypertension was similar among women with and without POI. With the exception of endometriosis (SD 0.11), no other comorbidities demonstrated clinically meaningful differences between women with and without POI. Characteristics of 3 635 702 women in Ontario between 1 January 1995 and 31 December 2019, according to premature ovarian status (POI) status. All data are presented as a number (%) unless specified otherwise. Includes Type 1 diabetes, rheumatoid arthritis, thyroid disorders (hypothyroidism, Graves’ disease, Hashimoto), lupus, Crohn’s, Addison. For diabetes diagnoses prior to April 2003 Type 1 vs Type 2 cannot be differentiated and all diagnoses are included. Includes prior cancer diagnosis, radiotherapy, and chemotherapy. Overall incidence declined between 2002 and 2008, followed by a sustained increase through 2019, driven primarily by non-iatrogenic POI ( Fig. 2 , Supplementary Table S3 ). The incidence of surgically induced POI decreased from 60 per 100 000 person-years in 1995 to 20 per 100 000 person-years in 2019. In contrast, POI related to chemotherapy and/or radiotherapy, with or without surgery, increased steadily over the study period. Age- and year-specific IRs demonstrated the most pronounced increase among women 36- to 39-year age group ( Fig. 3 , Supplementary Table S4 ); Linear regression analyses showed a statistically significant upward trend in this age group ( P  < 0.001), whereas trends in younger age groups were not statistically significant. Incidence rate (95% CI) of premature ovarian insufficiency (POI) per 100 000 person-years by year of diagnosis and type ((A) overall, (B) non-iatrogenic, (C) surgical, (D) chemotherapy and/or radiotherapy, (E) surgery and chemotherapy, and/or radiotherapy). Incidence rate (95% CI) of premature ovarian insufficiency (POI) per 100 000 person-years by year of diagnosis and age categories. Age-adjusted IRRs for factors associated with POI are presented in Table 2 . The strongest association was observed for endometriosis (IRR 3.45, 95% CI 3.39, 3.52), followed by cancer diagnosis with chemotherapy or radiotherapy (IRR 2.01, 95% CI 1.89, 2.14), tobacco or substance use (IRR 1.59, 95% CI 1.51, 1.67), chromosomal disorders (IRR 1.50, 95% CI 1.42, 1.59), and autoimmune disease (IRR 1.19, 95% CI 1.17, 1.22). In subgroup analyses by age at diagnosis, the highest incidence was observed among women aged 36–39 years in 2019 (802.2 per 100 000 person-years; 95% CI 765.2, 840.5), with peak incidence occurring in 2019 across all age categories ( Fig. 3 , Supplementary Table S4 ). Medical history and its association with premature ovarian insufficiency. Adjusted for age at index. Includes Type 1 diabetes, rheumatoid arthritis, thyroid disorders (hypothyroidism, Graves’ disease, Hashimoto), lupus, Crohn’s, Addison. For diabetes diagnoses prior to April 2003 Type 1 vs Type 2 cannot be differentiated and all diagnoses are included. Includes prior cancer diagnosis, radiotherapy, and chemotherapy. Table 3 presents POI diagnosis subtype according to comorbidities associated with POI. Among women with autoimmune disease, chromosomal disorders, or tobacco or substance use, non-iatrogenic POI was the most common diagnosis subtype. Among women with endometriosis, 55.8% had non-iatrogenic POI and 43.2% had surgically induced POI. Among women with a history of cancer, 44.2% had non-iatrogenic POI, while 48.4% had POI related to chemotherapy and/or radiotherapy, with or without surgery. Comorbidities known to be associated with premature ovarian insufficiency (POI) in relation to type of diagnosis. Data presented as number (row percent). Includes Type I1 diabetes, rheumatoid arthritis, thyroid disorders (hypothyroidism, Graves’ disease, Hashimoto), lupus, Crohn’s, Addison. For diabetes diagnoses prior to April 2003 Type 1 vs Type 2 cannot be differentiated and all diagnoses are included. Includes prior cancer diagnosis, radiotherapy, and chemotherapy. In sensitivity analysis using serum FSH levels, 252 276 women had at least one FSH value available ( Fig. 4 ; Supplementary Table S5 ). IRs using laboratory-defined POI were consistent with those based on administrative diagnosis, demonstrating an upward trend across all age groups from 2007 to 2019. Incidence rate (95% CI) of premature ovarian insufficiency (POI) per 100 000 person-years by year of diagnosis and age categories, using FSH levels between 4 October 2007 and 31 December 2019.

This population-based cohort study included all women residing in Ontario who were ≤39 years between 1 January 1995 and 31 December 2019. Participants were followed from cohort entry until diagnosis of POI, age 40 years, death, loss of Ontario Health Insurance Plan (OHIP) eligibility, or the end of follow-up on 31 December 2021, whichever occurred first. Individuals were excluded if records contained invalid birth date, death date, sex, or ICES Key Number (IKN), or if geographic information was missing. This study used linked, patient-level administrative health data held at ICES ( www.ices.org.ca ), an independent, non-profit research institute whose legal status under Ontario’s health information privacy legislation allows it to collect and analyze de-identified health and demographic data for health system evaluation and improvement. Ontario has a single payer, universal healthcare system that covers physician care and hospital services provided to all Ontario residents through the OHIP. The OHIP database is the physician fee-for-service claims file; it is the mechanism according to which physicians are paid in Ontario. This database includes codes for both the services rendered and patient diagnoses. The OHIP databases use International Classification of Diseases and Related Health Problems, Ninth Revision, Canada (ICD-9) for coding of diseases. Detailed descriptions of data sources, cohort creation, outcome definitions, and covariates are provided in Supplementary Tables S1 and S2 . The primary outcome was a diagnosis of POI occurring before age 40 years, identified using a minimum of one of two data sources: (i) Surgical menopause, defined as one Canadian Classification of Health Interventions (CCI) or Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (CCP) code for bilateral oophorectomy, two occurrences of unilateral oophorectomy codes, or one code indicating removal of the remaining ovary; or (ii) Physician-diagnosed menopause, defined as one or more OHIP physician billing claims coded with International Classification of Diseases, Ninth Revision (ICD-9) code 627. POI was classified as iatrogenic if identified following surgical menopause or after initiation of chemotherapy or radiotherapy. Non-iatrogenic POI was defined as POI diagnosed in the absence of prior surgical menopause, chemotherapy, or radiotherapy. An exception was made for women diagnosed with HIV prior to these exposures, given the higher baseline prevalence of POI reported in this population ( Li et al. , 2023 ). Baseline characteristics assessed at cohort included age, neighborhood income quintile, rurality, immigration status, parity, obesity, tobacco and drug dependence, chromosomal disorder, autoimmune disease, cancer, endometriosis, and receipt of chemotherapy or radiotherapy. Neighborhood income quintile (lowest [1] to highest [5]) was derived using residential postal code linked to the Postal Code Conversion File and Canadian census data. Rurality was defined using the Rurality Index of Ontario (RIO <40 vs ≥40). Immigration status (refugee immigrant, non-refugee immigrant, or Canadian-born) was determined using the Immigration, Refugees and Citizenship Canada Permanent Residents (IRCC-PR) database. Parity was ascertained using the MOMBABY dataset. Obesity (ICD-9 278), tobacco abuse (305), drug dependence (304), and chromosomal disorders (758) were identified based on the presence of two or more OHIP diagnostic codes during the lookback period. Autoimmune diseases known to be associated with POI—including type 1 diabetes mellitus, rheumatoid arthritis, thyroid disorders (hypothyroidism, Graves’ disease, Hashimoto thyroiditis), systemic lupus erythematosus, Crohn’s disease, and Addison’s disease—were identified using diagnostic codes. Pre-existing diabetes and chronic hypertension were identified using the Ontario Diabetes Dataset (ODD) and Ontario Hypertension Dataset (HYPER), respectively. Endometriosis was defined using the presence of an OHIP ICD-9 code 617 or ICD-10 code N80 recorded during an inpatient admission or surgical encounter and was further classified as surgically or medically diagnosed. Full coding definitions are provided in Supplementary Table S2 . Baseline characteristics were summarized using proportions for categorical variables and means with SDs or medians with interquartile range (IQR) for continuous variables. Standardized differences were calculated to compare characteristics across POI subtypes, with values >0.10 considered clinically meaningful. Annual incidence rates (IRs) of POI were calculated using census data, with the Ontario female population aged ≤39 years as the denominator and stratified by 5-year age categories (16–20, 21–25, 26–30, 31–35, 36–39 years). Temporal trends in IRs were assessed using linear regression. Associations between clinical factors and POI were estimated using age-adjusted modified Poisson regression models to derive incidence rate ratios (IRRs). Each clinical factor was evaluated in a separate age-adjusted model. Comparisons of medical history and comorbidities across POI subtypes were similarly adjusted for age. In an additional subgroup analysis stratified by age at diagnosis, POI cases were categorized into the following age groups: 16–20, 21–25, 26–30, 31–35, and 36–40 years. This subgroup analysis was performed to evaluate whether temporal trends in POI incidence varied across age groups. A sensitivity analysis was performed in which POI incidence was re-estimated using laboratory data from the Ontario Laboratory Information System (OLIS), available since 2007, defining POI as one or more FSH measurements within the menopausal range (25.8–134.8 IU/l). This sensitivity analysis was performed to assess the consistency of POI incidence estimates using a laboratory-based definition of POI. This study was reviewed and approved for ethical compliance by the Queen’s University Health Sciences & Affiliated Teaching Hospitals Research Ethics Board.

The incidence of POI in Ontario, Canada has increased in recent years, driven primarily by non-iatrogenic POI, with endometriosis emerging as a major associated factor. Concurrently, the incidence of surgically induced POI has declined substantially since 1995, likely reflecting widespread adoption of ovarian-preserving gynecologic surgical practices. Improved survival among individuals exposed to gonadotoxic cancer therapies has also contributed to increasing POI incidence. These findings underscore the need for improved risk stratification, counselling, and further studies to elucidate underlying mechanisms, modifiable risk factors, and the long-term health implications of POI and its different subtypes.

This population-based cohort study provides contemporary estimates of the incidence of POI among women in Ontario, Canada between 1995 and 2019, followed by an evaluation of factors associated with POI diagnosis. Overall, POI incidence declined between 2002 and 2008, followed by a sustained increase through 2019, driven primarily by non-iatrogenic POI. In contrast, surgically induced POI decreased markedly between 1995 and 2007 and subsequently plateaued, while POI related to chemotherapy and/or radiotherapy increased steadily over the study period. Among associated factors, endometriosis demonstrated the strongest association with POI. Recent meta-analyses have reported global POI prevalence estimates of 3.5% (95% CI 3.0, 4.0) and 3.7% (95% CI 3.1, 4.3) ( Golezar et al. , 2019 ; Li et al. , 2023 ). In our study, the incidence of POI in Ontario was 397.8 per 100 000 person-years. We observed a U-shaped temporal pattern, with declining incidence between 1999 and 2008 followed by rising incidence after 2011. This pattern differs from that reported among Finnish women, where incidence remained relatively stable between 1988 and 2017 ( Silvén et al. , 2022 ). This discrepancy may reflect differences in case ascertainment, diagnostic practices, and healthcare system factors rather than true biologic variation. Differences in surgical practices, cancer survival, population demographics, and age-specific diagnostic patterns between jurisdictions may further contribute to observed variations in incidence trends. A key finding of this study is the strong association between endometriosis and POI. Ovarian impairment has long been linked to endometriosis, although the relative contribution of disease-related mechanisms versus surgical treatment remains debated. Potential mechanisms include damage from ovarian endometriomas—present in ∼44% of women with endometriosis ( Tan et al. , 2023 ),—as well as inflammation and fibrosis related to pelvic disease. Prior studies have demonstrated reduced anti-Müllerian hormone levels and decreased antral follicle counts in women with endometriosis, particularly in affected ovaries, suggesting diminished ovarian reserve independent of surgery ( Muzii et al. , 2018 ; Moreno-Sepulveda et al. , 2022 ). In our cohort, 55.8% of POI cases among women with endometriosis were classified as non-iatrogenic, while 43.2% were associated with surgical diagnosis or intervention. Consistent with this, our prior population-based studies of endometriosis surgery in Ontario demonstrated widespread use of ovarian-preserving surgical approaches, suggesting that contemporary surgical management alone is unlikely to fully explain the elevated POI risk observed in this population ( Bougie et al. , 2021 , 2025b ). This finding underscores the likelihood that both disease-related and treatment-related pathways contribute to ovarian dysfunction. Proposed mechanisms include activation of primordial follicles following surgical injury and local inflammatory processes that may accelerate follicular depletion or apoptosis ( Uri-Belapolsky et al. , 2014 ). Taken together, the association between endometriosis and POI further supports the growing recognition of endometriosis as a chronic condition with important reproductive and long-term consequences ( Velez et al. , 2022 ; Blom et al. , 2023 ; Thiel et al. , 2024 ; Bougie et al. , 2025a ; Milne et al. , 2026 ). Tobacco and substance use were also associated with higher POI incidence, consistent with prior literature linking these exposures to earlier age at menopause ( Whitcomb et al. , 2018 ). Although the proportion of women with chromosomal or autoimmune conditions did not differ substantially between those with and without POI, age-adjusted IRRs demonstrated statistically significant associations. This aligns with existing literature, which reports wide variability in the contribution of autoimmune disease to non-iatrogenic POI and suggests that clustering of autoimmune conditions does not necessarily imply direct immune-mediated ovarian destruction ( Panay et al. , 2020 ). The marked decline in surgically induced POI observed in this study parallels broader changes in gynecologic surgical practice. In Canada, hysterectomy rates declined from 484 per 100 000 women in 1997 to 303 per 100 000 in 2017 ( Chen et al. , 2019 ), alongside decreasing rates of bilateral oophorectomy at the time of hysterectomy due to greater awareness of the long-term adverse consequences of premenopausal oophorectomy ( Salvador et al. , 2017 ). Our findings suggest that adoption of ovarian-preserving surgical approaches in Ontario has substantially reduced surgically induced POI over time ( Bougie et al. , 2021 , 2025b ). In subgroup analyses by age at diagnosis, the highest incidence of POI was observed among women aged 36–39 years, consistent with previous reports ( Silvén et al. , 2022 ). POI incidence increased modestly in the youngest age groups over the study period, including among women aged 16–20 years. Increased recognition and investigation of primary amenorrhea may contribute to higher diagnostic rates in this age group. This trend may also reflect improved survival among adolescent and young adult (AYA) cancer patients treated with gonadotoxic therapies, with current survival rates exceeding 80% ( Canadian Cancer Society, 2023 ). Consistent with this observation, our previous population-based studies have demonstrated higher rates of POI among AYA cancer survivors ( Rushton et al. , 2022 ; Flatt et al. , 2023 ; Luong et al. , 2024 ). This study has several strengths. It is the first Canadian population-based study to examine both incidence and associated factors of POI across all reproductive age groups. Use of linked administrative health data at ICES enabled longitudinal follow-up within a universal, publicly funded healthcare system. OHIP coverage is provided to all Canadian citizens, indigenous persons, landed immigrants or those who have applied for permanent residence (i.e. immigrants or refugees), or are a refugee who have made Ontario their primary residence, capturing the predominant portion of women who use Ontario’s healthcare system. Additionally, sensitivity analysis incorporating laboratory-defined POI using FSH measurements yielded consistent temporal and age-specific patterns, supporting the robustness of our findings. Limitations include reliance on administrative health data, which may result in outcome misclassification and limited availability of detailed clinical and demographic information. For example, patient-level demographic and clinical factors, including education, race, income, and family history, are not comprehensively captured within administrative databases and may influence both POI diagnosis and associated clinical factors ( Velez et al. , 2010 ). In addition, women who did not access healthcare services or were not formally diagnosed with POI may not have been captured in administrative databases, potentially resulting in under-ascertainment of POI incidence. Detailed chromosomal diagnoses—including fragile X premutation, Y-chromosome material, and rare syndromic causes—could not be individually identified due to aggregation within billing codes. Furthermore, information on endometriosis stage, laterality, and medical management was unavailable. We were also unable to identify women who underwent ovarian-preserving endometriosis surgery, such as endometrioma cystectomy, and determine how these interventions contributed to subsequent POI. As a result, some women with surgery-related negative impacts on ovarian function may have been classified as having non-iatrogenic POI, which may have partially contributed to the strong association observed between endometriosis and non-iatrogenic POI. Given that over half of POI cases among women with endometriosis were non-iatrogenic, further studies incorporating clinical detail are needed to better characterize disease severity, treatment pathways, and mechanisms linking endometriosis to POI.