Premature Ovarian Insufficiency in Childhood Cancer Survivors: A Report From the St. Jude Lifetime Cohort

Abstract

Context:

Long-term follow-up data on premature ovarian insufficiency (POI) in childhood cancer survivors are limited.

Objective:

To describe the prevalence of POI, its risk factors, and associated long-term adverse health outcomes.

Design:

Cross-sectional.

Setting:

The St. Jude Lifetime Cohort Study, an established cohort in a tertiary care center.

Patients:

Nine hundred twenty-one participants (median age, 31.7 years) were evaluated at a median of 24.0 years after cancer diagnosis.

Main Outcome Measure:

POI was defined by persistent amenorrhea combined with a follicle-stimulating hormone level >30 IU/L before age 40. Multivariable Cox regression was used to study associations between demographic or treatment-related risk factors and POI. Multivariable logistic regression was used to study associations between POI and markers for cardiovascular disease, bone mineral density (BMD), and frailty. Exposure to alkylating agents was quantified using the validated cyclophosphamide equivalent dose (CED).

Results:

The prevalence of POI was 10.9%. Independent risk factors for POI included ovarian radiotherapy at any dose and CED ≥8000 mg/m². Patients with a body mass index ≥30 kg/m² at the time of the St. Jude Lifetime Cohort assessment were less likely to have a diagnosis of POI. Low BMD and frailty were independently associated with POI.

Conclusion:

High-dose alkylating agents and ovarian radiotherapy at any dose are associated with POI. Patients at the highest risk should be offered fertility preservation whenever feasible. POI contributes to poor general health outcomes in childhood cancer survivors; further studies are needed to investigate the role of sex hormone replacement in improving such outcomes.

We report on the prevalence, risk factors, and consequences on general health of premature ovarian insufficiency in a cohort of 921 long-term survivors of childhood cancers.

Female childhood cancer survivors (CCSs) are at an increased risk of developing premature ovarian insufficiency (POI) because of the vulnerability of the ovaries to gonadotoxic treatment modalities such as pelvic radiotherapy and alkylating agent chemotherapy (1–7). Due to the intricate relationship between the hormone-producing granulosa cells and the oocyte, POI causes disruption of both endocrine and reproductive functions of the ovary. Puberty may be delayed or interrupted. Primary or secondary amenorrhea may occur depending on the pubertal stage at the time of cancer treatment. POI can occur early, during, or immediately following the completion of cancer treatment (2) or, more commonly, in the years that follow the completion of cancer treatments but prior to age 40 (3). Continuous improvements and increased availability of female fertility preservation techniques (8), together with the importance of sexual and reproductive outcomes on overall patient health and quality of life, mandate that we have a better understanding of POI and its risk factors (4, 9).

In the general population, POI has been associated with an increased risk of cardiovascular disease and low bone mineral density (BMD) (10, 11). The contribution of POI to adverse health outcomes and markers of poor physical condition such as frailty in CCSs has not been well established (12, 13). The current study reports on the overall prevalence of POI in a large cohort of long-term CCSs and examines associations between host and treatment factors and the development of POI. We also evaluated associations between POI and risk factors of cardiovascular disease (abdominal obesity, hypertension, abnormal glucose metabolism, dyslipidemia), as well as low BMD and frailty.

Patients and Methods

SJLIFE Study

Participants were enrolled in the previously described Institutional Review Board–approved St. Jude Lifetime Cohort (SJLIFE) protocol (14, 15). Eligibility criteria included ≥10 years postdiagnosis of childhood cancer, treatment at St. Jude Children’s Research Hospital, and age ≥18 years. Eligible subjects were invited to return to St. Jude Children’s Research Hospital for clinical evaluations (performed from 2007 to 2012) augmented with a core assessment battery and questionnaires that detail demographic, medical/reproductive histories, quality of life, and health habits.

Ovarian dosimetry and quantification of alkylating agent and glucocorticoid exposures

Radiation fields potentially involving the ovaries included the following areas: pelvis, lumbar or sacral or whole spine, flank or abdomen if extending below the iliac crest, inverted Y, para-aortic, iliac, bladder or vaginal, in addition to total lymphoid or total body irradiation (9). Ovarian radiotherapy dose was quantified by reconstructing each patient’s individual radiation treatment on a computed tomography (CT)–based phantom with an organ library for calculation of normal organ radiation doses, as performed in other analyses in the SJLIFE protocol. The average location of the ovaries in our CT-based phantom was determined based on an averaging of organ locations from 10 patient CT data sets and was found to be similar to other literature on ovary location (16). The minimum dose to each ovary was used in our analysis. Dosimetry modeling did not account for oophoropexy or other location changes due to pelvic surgery; in these instances, survivors were excluded from analyses. Exposure to alkylating agents was quantified using the validated cyclophosphamide equivalent dose (CED) (17). Glucocorticoid exposure (dexamethasone and prednisone) was quantified by the cumulative prednisone equivalent dose received (1 mg prednisone = 0.15 mg dexamethasone) and expressed in percentiles based on the dose distribution in the cohort.

Diagnosis of POI

POI was defined by persistent amenorrhea with evidence of a primary ovarian origin before the age of 40 years. The diagnosis of POI was based on the medical history provided by the patients in regards to puberty, menarche, menstrual cycles, pregnancies, childbirth, use of hormonal therapies, including contraception, and timing of menopause and supplemented by clinical and laboratory data from the SJLIFE evaluation. A blood sample was collected from all patients for the measurement of luteinizing hormone, follicle-stimulating hormone (FSH), and estradiol using electro-chemiluminescent immunometric assays (Roche Cobas 6000 analyzer; Roche Diagnostics, Indianapolis, IN). In amenorrheic women <40 years old who were not on sex hormone replacement therapy or oral contraceptive pills, estradiol <17 pg/mL with FSH >30 IU/L was considered indicative of POI. In individuals receiving sex hormone replacement therapy or oral contraceptives, the diagnosis of POI was based solely on historical medical information. Women taking oral contraceptives to prevent a pregnancy, regulate cycles, or treat polycystic ovarian syndrome were assumed not to have POI at the time of the study. Subjects with hypogonadotropic hypogonadism (deficiency in luteinizing hormone and FSH) were excluded given their inability to sufficiently raise their plasma FSH levels and allow the diagnosis of potentially co-occurring POI. The diagnosis was based on medical/historical data in individuals who were being treated with sex hormone replacement therapy or oral contraceptives at the time of this study. Amenorrheic patients <40 years who were not receiving such treatments and had plasma estradiol levels <17 pg/mL and FSH <11.2 IU/L at the time of the SJLIFE assessment were also considered to have hypogonadotropic hypogonadism (18). Patients whose hormonal deficiencies could not be attributed with certainty to a primary ovarian vs pituitary origin and those with missing outcomes data were excluded (Fig. 1).

Figure 1.

Flow diagram.

Cardiovascular

Weight, height, waist circumference, and resting blood pressure were measured in all SJLIFE participants. A waist circumference ≥88 cm or a waist-to-height ratio of more than 0.5 were considered suggestive of abdominal obesity (19). Body mass index (BMI) was calculated as weight/height² (kg/m²). Blood pressure was evaluated after 5 minutes of quiet sitting; hypertension was defined by systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, and/or treatment with antihypertensive medications. Enzymatic spectrophotometric assays (Roche Modular P Chemistry Analyzer; Roche Diagnostics) were used to measure glucose, triglycerides, and lipids. Abnormal glucose metabolism was defined by impaired fasting glucose (blood glucose level ≥100 mg/dL) and/or treatment with medications used for diabetes mellitus management. Dyslipidemia was defined by having one or more of the following abnormalities: fasting total cholesterol ≥200 mg/dL, high-density lipoprotein <50 mg/dL, low-density lipoprotein ≥130 mg/dL, triglycerides ≥150 mg/dL, and/or treatment with lipid-lowering medications (20).

BMD

Quantitative CT with GE VCT LightSpeed 64-detector (GE Health Care, Waukesha, WI) and Mindways quantitative CT calibration phantoms and software (Mindways, Austin, TX) were used to assess BMD. Age- and sex-specific z scores were calculated using average volumetric trabecular BMD for lumbar vertebrae L1 and L2. Low BMD was defined as a z score < –2.0 (21).

Physical fitness assessment and frailty

Physiologic frailty was defined as including three or more of the following items: low muscle mass, self-reported exhaustion, low energy expenditure, slow walking speed, and muscle weakness (Supplemental Table 1 ^{(18.6KB, docx)} ) (12, 22). Appendicular mass (kg) was measured using dual x-ray absorptiometry and divided by height squared (m²) to estimate muscle mass (23). Exhaustion was scored using the Vitality subscale of the Medical Outcomes Survey Short Form-36 (24). Energy expenditure was calculated using the National Health and Nutrition Examination Survey Physical Activity Questionnaire (25). Walking speed was calculated based on the time to cover a distance of 15 feet and was adjusted for height (22). Muscle strength was assessed using hand grip strength (kg) measured using a Jamar dynamometer (Preston-Sammons, Nottinghamshire, United Kingdom) with adjustment to BMI (22).

Statistical analysis

Descriptive statistics included point prevalence, defined as the proportion of survivors with POI at the time of their clinical assessment. Unadjusted associations of POI with ethnicity, educational attainment, age at primary cancer diagnosis or exposure to pelvic radiotherapy or alkylating agents, history of oophoropexy, ovarian radiation dose, CED, exposure to any pelvic radiotherapy, exposure to any alkylating agent, exposure to both pelvic radiotherapy and alkylating agents, use of tobacco, alcohol or illicit drugs, and BMI at the time of the SJLIFE assessment were assessed using χ² or Fisher’s exact test. Variable groupings were selected based on clinical relevance and to assure adequate numbers of persons within group for statistical power. Variables with P values ≤ 0.10 from the unadjusted analysis were included in a Cox regression model using age at POI or age at the SJLIFE assessment (for those without POI) as the time variable. Persons without POI who were 40 years of age or older at the SJLIFE assessment were censored at age 40 years given that POI is not a possible outcome past that point. Results are presented as adjusted hazard ratios (HRs) and 95% confidence intervals (CIs). Associations between POI and abdominal obesity (defined by increased waist circumference or waist/height ratio), hypertension, abnormal glucose metabolism, dyslipidemia, low BMD, and frailty were tested, first in unadjusted and then in logistic multivariable regression models. The covariates considered in the unadjusted analysis for all outcomes were: age at SJLIFE, age at primary cancer diagnosis, ethnicity, BMI (<30 or ≥30 kg/m²), educational level attained, smoking, heavy alcohol drinking, and illicit drug use. In addition, ovarian radiotherapy dose (none, <10 Gy, and ≥10 Gy) was included for hypertension while prednisone equivalent dose (none, 0 to 25th percentile, >25th and ≤75th percentile, and >75th percentile) and treatment with total body irradiation (yes/no) were included for low BMD. The multivariable analysis examined the effect of POI after adjustment for all factors with a P value ≤ 0.1 in the unadjusted analysis. Results for this analysis are presented as adjusted odds ratios (ORs) and 95% CI. SAS version 9.2 (Cary, NC) was used for all analyses.

Results

Point prevalence of POI

Of 1644 potentially eligible participants, 988 (60.1%) were available for study (Fig. 1). Participants were more likely to have received cranial radiation (P = 0.03) and alkylating agents (P < 0.0001) than nonparticipants; there were no significant differences in patient demographics, age at cancer diagnosis, or age at study. After applying the exclusions related to difficulties in assessing ovarian status, 921 participants were retained for the analysis (Fig. 1). The survivor characteristics at enrolment in SJLIFE are summarized in Table 1. The majority of patients were survivors of hematological malignancies (61.1%); cancer and treatment history data are summarized in Table 2. Patients were assessed at a median age of 31.7 years (range, 19.0 to 60.6) and at a median of 24.0 years (range, 10.2 to 48.1) after the initial cancer diagnosis. Treatment exposures included pelvic radiotherapy in 153 patients (13.3%) and alkylating agents in 542 (58.8%). One hundred patients were diagnosed with POI (10.9%). At the time of the SJLIFE assessment, only 31 of 100 patients with POI were receiving sex hormone replacement therapy.

Table 1.

Patient Characteristics at Study

	Survivors (N = 921)
	n	%
Race/ethnicity
Non-Hispanic white	770	83.60
Non-Hispanic black	129	14.01
Hispanic	16	1.74
Other	6	0.65
Education level
No high school or GED	78	8.47
High school or GED	220	23.89
Some college	240	26.06
Bachelor’s degree or higher	381	41.37
Unknown	2	0.22
Age at study
18–25 years	202	21.93
26–35 years	420	45.60
36–45 years	241	26.17
>45 years	58	6.30
Body mass index
<18.5 kg/m2	43	4.67
18.5–24.9 kg/m2	338	36.70
25–29.9 kg/m2	224	24.32
≥30 kg/m2	316	34.31
Current tobacco use
No	735	79.80
Yes	185	20.09
Unknown	1	0.11
Excessive alcohol consumptiona
No	894	97.07
Yes	17	1.85
Unknown	10	1.09
Illicit drug use
Never	838	90.99
1–9 times	45	4.89
≥10 times	25	2.71
Not reported	13	1.41
Therapy
No alkylating agent/no ovarian RT	320	34.74
Alkylating agent only	401	43.54
Ovarian RT only	59	6.41
Alkylating agent and ovarian RT	141	15.31
Pregnant times
0	298	32.36
1–2 times	333	36.16
3–4 times	162	17.59
≥5 times	32	3.47
Not reported	96	10.42
Live/still birth history
Not pregnant	298	32.36
1–2 times	366	39.74
3–4 times	101	10.97
≥5 times	5	0.54
Not reported	151	16.40

Abbreviations: GED, general educational diploma (high school equivalency diploma); RT, radiotherapy.

^a Defined by consuming more than four drinks per day for males or more than three drinks per day for females.

Table 2.

Primary Cancer Diagnoses and Cancer Treatment Information

	n	%
Diagnosis
Leukemia	398	43.21
Lymphoma	165	17.92
Central nervous system tumor	52	5.65
Embryonal tumors	178	19.33
Bone and soft tissue sarcoma	105	11.40
Carcinomas	12	1.30
Other	11	1.19
Age at diagnosis
0–4 years	370	40.17
5–9 years	206	22.37
10–14 years	208	22.58
≥15 years	137	14.88
Oophoropexy
No	863	93.70
Yes	58	6.30
Hypothalamic/pituitary radiation dose
None	630	68.40
<1000 cGy	0	0
1000–1499 cGy	16	1.74
1500–2999 cGy	219	23.78
≥3000 cGy	56	6.08
Ovarian radiation dose
None	721	78.28
<100 cGy	53	5.75
100–999 cGy	53	5.75
1000–1999 cGy	32	3.47
≥2000 cGy	27	2.93
Unknown	35	3.80
CED
0 mg/m2	379	41.15
<4000 mg/m2	85	9.23
≥4000 and <8000 mg/m2	166	18.02
≥8000 and <12,000 mg/m2	164	17.81
≥12,000 and < 20,000 mg/m2	86	9.34
≥20,000 mg/m2	41	4.45

Risk factors for POI

The multivariable analysis (Table 3) showed independent associations between POI and ovarian radiotherapy at any dose [<1000 cGy (HR = 13.85; 95% CI, 6.50 to 29.51) or ≥1000 cGy (HR = 132.34; 95% CI, 62.88 to 278.53)] and CED ≥8000 mg/m² [8000 to 11,999 mg/m² (HR = 2.77; 95% CI, 1.18 to 6.51), 12,000 to 19,999 mg/m² (HR = 3.90; 95% CI, 1.80 to 8.43), or ≥20,000 mg/m² (HR = 4.13; 95% CI, 1.63 to 10.50)]. Patients with BMI ≥30 kg/m² at the time of the SJLIFE evaluation were less likely to have been diagnosed with POI (HR = 0.43; 95% CI, 0.22 to 0.86).

Table 3.

Multivariable Cox Regression Model Fits for POI

	POI (n = 100)
Characteristic	N	na	%	HR	CI	P Value
Age at cancer diagnosis (years)b
Mean (SD)	8.10 (5.57)			0.97	0.92–1.02	0.28
Oophoropexy
No	863	80	9.27	1.00
Yes	58	20	34.48	1.33	0.70–2.53	0.39
Body mass index
≥18.5–24.9 kg/m2	338	42	12.43	1.00
<18.5 kg/m2	43	12	27.91	1.52	0.71–3.23	0.28
25–29.9 kg/m2	224	29	12.95	0.93	0.54–1.61	0.80
≥30 kg/m2	316	17	5.38	0.43	0.22–0.86	0.02
Ovarian radiation dose
None	72c	10	1.39	1.00
>999 cGy	86c	29	27.36	13.85	6.50–29.51	<0.001
≥1000 cGy	79c	44	74.58	132.34	62.88–278.53	<0.001
CED
0 mg/m2	379	19	5.01	1.00
< 8000 mg/m2	251	32	12.75	1.55	0.77–3.11	0.10
8000–11,999 mg/m2	164	16	9.76	2.77	1.18–6.51	0.02
12,000–19,999 mg/m2	86	24	27.91	3.90	1.80–8.43	0.001
≥20,000 mg/m2	41	9	21.95	4.13	1.63–10.50	0.003

^a Number of patients with POI with each characteristic.

^b Variable was included in the model as continuous.

^c Radiation dose data missing on 35 patients.

To test for a synergistic effect between alkylating agents and ovarian radiotherapy exposure on POI risk, we added a combined variable of these exposures (none, either, or both) to the multivariable analysis (Table 4). In this model, POI was significantly associated with ovarian radiotherapy exposure alone (HR = 71.7; 95% CI, 16.50 to 311.58) and with the combination of alkylating agents and ovarian exposure to radiotherapy (HR = 95.56; 95% CI, 23.30 to 391.93). Patients with BMI ≥30 kg/m² at the time of SJLIFE evaluation were less likely to have been diagnosed with POI (HR = 0.36; 95% CI, 0.20 to 0.65). The percentage of participants on oral contraceptives or hormone replacement therapies was not statistically different between obese and nonobese individuals (38.29% vs 42.48%; P = 0.22). In women who were not on such treatments (n = 518), the FSH levels (mean ± standard deviation [SD]) did not differ between premenopausal obese and nonobese participants (5.9 ± 4.8 vs 6.6 ± 6.1; P = 0.23) but were lower in obese survivors who experienced nonsurgical menopause when compared with those who were not obese (43.10 ± 32.0 vs 72.1 ± 43.6; P = 0.003) (Supplemental Table 2 ^{(12.7KB, docx)} ).

Table 4.

Multivariable Cox Regression Model Fits for POI After Combining Treatment Modalities

	POI (n = 100)
Characteristic	N	n a	%	HR	CI	P Value
Age at cancer diagnosis (years)b
Mean (SD)	8.10 (5.57)			1.02	0.98–1.06	0.41
Oophoropexy
No	863	80	9.27	1.00
Yes	58	20	34.48	0.72	0.42–1.23	0.23
Body mass index
≥18.5–24.99 kg/m2	338	42	12.43	1.00
<18.5 kg/m2	43	12	27.91	1.87	0.97–3.59	0.06
25.0–29.9 kg/m2	224	29	12.95	0.92	0.56–1.52	0.74
≥30 kg/m2	316	17	5.38	0.36	0.20–0.65	0.001
Treatment exposure
No alkylating agent nor ovarian radiotherapy	318c	2	0.63	1.00
Alkylating agent only	400c	8	2.00	2.98	0.63–14.06	0.17
Ovarian radiotherapy only	59c	17	28.81	71.70	16.50–311.58	<0.001
Both	141c	73	51.77	95.56	23.30–391.93	<0.001

^a Number of patients with POI with each characteristic.

^b Variable was included in the model as continuous.

^c Information incomplete on n = 3 patients.

POI and adverse health outcomes

In multivariable analyses female survivors with POI had an increased odds of low BMD (OR = 5.07; 95% CI, 1.97 to 13.05) and frailty (OR = 3.51; 95% CI, 1.78 to 6.93) compared with females without POI (Table 5).

Table 5.

Multivariable Logistic Regression Model Fits for Risk Factors for Markers of Cardiovascular Disease, Low BMD, and Frailty

Outcome Variables	POI
	Absent [N (%)]	Present [N (%)]	Unadjusted Analysis		Multivariable Analysis
			OR a	95% CI	OR a	95% CI
Increased waist circumferenceb	308 (38.64)	23 (24.73)	0.52	0.32–0.85	1.20	0.58–2.51
Increased waist-to-height ratioc	434 (52.86)	41 (41.00)	0.66	0.43–1.02	0.95	0.56–1.61
Hypertensiond	170 (21.25)	39 (40.63)	2.54	1.63–3.94	1.44	0.71–2.94
Elevated glucosee	66 (8.70)	9 (10.71)	1.26	0.60–2.63	1.38	0.64–2.99
Dyslipidemiaf	70 (8.55)	15 (15.15)	1.91	1.05–3.49	1.73	0.91–3.29
Low BMDg	17 (2.38)	10 (14.08)	6.73	2.95–15.34	5.07	1.97–13.05
Frailtyh	39 (4.89)	15 (15.96)	3.69	1.95–6.99	3.51	1.78–6.93

An unadjusted analysis was performed first to select the covariates to be included into the model (P ≤ 0.1).

^a OR is calculated for POI yes vs no.

^b Increased waist circumference: Multivariable model was adjusted for age at study, BMI ≥30 kg/m², educational attainment, and smoking status.

^c Increased waist-to-height ratio: Multivariable model was adjusted for age at study, BMI ≥30 kg/m², educational attainment, and smoking status.

^d Hypertension: Multivariable model was adjusted for age at study, age at diagnosis, ovarian radiotherapy dose (0, >0 to 999, and ≥1000 cGy), and BMI ≥30 kg/m². Significant association was found between hypertension and ovarian radiotherapy dose ≥1000 cGy (OR, 2.23; 95%CI, 1.01 to 4.92).

^e Elevated glucose: Multivariable model was adjusted for age at study and BMI ≥30 kg/m².

^f Dyslipidemia: Multivariable model was adjusted for age at study, age at diagnosis, ethnicity/race, and BMI ≥30 kg/m².

^g Low BMD: Multivariable model was adjusted for BMI ≥30 kg/m² and exposure to total body irradiation.

^h Frailty: Multivariable model was adjusted for age at diagnosis, educational attainment, current smoking status, and illicit drug use.

Discussion

We report the prevalence of POI in the largest cohort of clinically assessed female CCSs to date. The increased availability of fertility preservation techniques over the past 5 years has heightened interest in predictive markers of POI in CCSs at risk (4). Mature oocyte cryopreservation is no longer deemed experimental; it is nevertheless a procedure that requires preparation and comes at a cost (8). It is therefore important to accurately identify individuals at risk for POI and those most likely to benefit from fertility preservation (4). Our study findings extend previous knowledge regarding associations between POI and treatment-related risk factors such as minimum ovarian radiotherapy dose and alkylating agents. This study is also unique in that we assessed associations between POI and several adverse long-term health outcomes (12, 26).

The prevalence of POI in SJLIFE participants, approaching 11%, was comparable to that previously observed in the SJLIFE cohort for those undergoing risk-based screening for POI based on cancer treatment exposures (11.8%) (14). This finding underscores the strong association between POI and known treatment-related risk factors (27–29). The reported prevalence is likely an underestimate, given the possible subsequent occurrence of menopause in at-risk women who were younger than 40 years old at the time of their participation in this research (4).

The pool of primordial follicles undergoes a natural age-related decline. Exposure to cancer therapies accelerates this natural decline in ovarian reserve (2–4). Pelvic irradiation is a major risk factor of POI and doses as low as 2 Gy have been estimated to deplete a patient’s follicular pool by as much as 50% (6). The findings of the present report show that exposure of the ovaries to any dose of radiotherapy increases the risk of POI. Patient age at the time of cancer diagnosis or treatment has been previously shown to be a significant risk factor for early-onset POI (2) and has been incorporated in risk prediction models (4). We were not able to demonstrate this association in the SJLIFE cohort likely because of the overriding effect of other risk factors, especially ovarian radiotherapy dose.

Exposure to high-dose alkylating agents, quantified by CED ≥8000 mg/m² in SJLIFE, was associated with POI, further highlighting the ovarian toxicity of these agents (2–5, 9, 17, 30). The study findings support the synergistic effects of alkylating agents and ovarian exposure to radiotherapy as evidenced by the fact that those at highest risk for POI were exposed to both modalities (Table 4).

Patients who were obese at the time of SJLIFE assessment experienced significantly lower rates of POI compared with those who were not obese, a somewhat unexpected finding. The percentage of obese patients whose hormonal data could not be used to assess their menopausal status because they were taking oral contraceptive or hormonal treatments was not significantly different from that of nonobese patients. In women who were not on such treatments and who did not experience menopause due to surgery, the FSH values followed similar trends as in the general population, with similar values among obese and nonobese individuals prior to menopause and lower values among obese menopausal women compared with those who were not obese (31). Given that obesity during cancer treatment is a strong predictor of obesity in later life (32, 33) and the significance of the interaction between BMI at SJLIFE and POI, one can speculate that body composition (obesity or underweight) during treatment, whether due to host factors (genetics, lifestyle) or the primary cancer, may influence or predict ovarian vulnerability to treatment (pelvic radiotherapy and/or CED) toxicity. This finding requires further study and validation in other cohorts with the understanding that the full assessment of the relationship between body composition and fertility would require the incorporation of many more factors than those available in SJLIFE to take into account the changes in the hormonal milieu brought about by obesity or polycystic ovarian syndrome (e.g., the regularity of menstrual cycles, antral follicle counts, serum anti-Müllerian hormone, and androgen levels).

Significant associations were found in this study between POI and chronic health conditions known to affect quality of life and mortality in CCSs. Our results indicate that POI is a significant independent predictor of low BMD. This finding is consistent with observations of decreased BMD and POI in the general population (11). Our data also suggest a strong association between POI and frailty, a marker of decreased physical fitness and a phenotype that may predict early mortality (12). Despite the known associations between premature menopause and cardiovascular disease in the general population (10, 34), we were not able to identify independent associations between POI and risk factors of cardiovascular disease in our cohort, possibly because of the overriding effects of other cancer treatments such as abdominal radiotherapy or nephrotoxic chemotherapy on these outcomes (35).

Treatment with estrogen until the normal age of menopause, 45 years or older, has been suggested to decrease mortality from cardiovascular disease and possibly improve BMD in patients with POI in the general population (36). Nevertheless, only a minority of patients with POI (31%) in our cohort were receiving sex hormone replacement therapy at the time of their SJLIFE evaluation. The increased risk of secondary breast cancer in certain groups of CCSs such as patients treated with chest radiotherapy (37) may be a concern to patients requiring sex hormone replacement therapy or their medical providers. The high percentage of nontreated CCSs may also be due to other barriers to hormone replacement therapy such as difficulties with access to experienced providers. The potential obstacles to sex hormone replacement in CCSs and the risk-benefit ratio of this therapy deserve further investigation (28).

The current study has several limitations. Markers of follicular reserve, such as ultrasound-based follicle counts and anti-Müllerian hormone levels, which might have provided additional insight to the patients’ current status, were not available in SJLIFE (4). We may have underestimated the prevalence of POI by assuming that participants taking oral contraceptives to prevent a pregnancy or regulate menstrual cycles or for the treatment of polycystic ovarian syndrome did not have POI; this assumption may also have influenced the study’s risk association findings. Additional limitations include the study’s cross-sectional design, difficulties in assessing with certainty ovarian function in a subset of patients on sex hormone replacement therapy or oral contraceptives, and the need to retrospectively estimate ovarian radiation dose.

In summary, POI is frequent in CCSs and is strongly associated with known treatment-related risk factors such as exposure of the ovaries to radiotherapy and treatment with alkylating agents. Changes in therapy over time, interindividual variability in follicular reserve at the time of therapy and the adverse effects of various modalities create unique challenges for the accurate prediction of the risk of POI in individual CCSs. POI may be associated with adverse general health outcomes in CCSs; the benefits and risk of sex hormone replacement therapy in this population deserve further investigation.