2.1 Participants

Our study is part of the national population-based Fex-Can Childhood project addressing the prevalence and predictors of sexual dysfunction and fertility-related distress in young adult childhood cancer survivors.¹⁰ It included all survivors in the National Quality Registry for Childhood Cancer in Sweden who were 19 to 40 years of age and residents in Sweden at the time of enrolment. The registry included patients diagnosed with cancer before the age of 18 years; however, during the 1980s and part of the 1990s only patients diagnosed before the age of 16 were treated at the child oncology units and, hence, included in the registry. Patients were excluded from the study if they were unable to read/write in Swedish or had cognitive impairment that compromised study participation. The survivors were identified through the National Quality Registry for Childhood Cancer. A total of 2030 possible female participants were identified, whereof 41 were excluded due to unknown address (n = 23), living outside Sweden (n = 1), cognitive dysfunction (n = 14) and administrative failure (n = 3). Finally, of the 1989 eligible participants, 1333 (67%) participated in the study. They had been diagnosed with cancer 1981 to 2017. There were no significant differences in demographic or clinical characteristics between female participants and nonparticipants.¹¹

2.2 Procedure

Clinical data about primary cancer diagnoses, date of diagnosis, cancer treatments, disease relapses and second malignancies were retrieved from the National Quality Registry for Childhood Cancer. Information about TBI was not available. Abdominal irradiation possibly affecting the gonads and cranial irradiation included only the local irradiation doses. Diagnoses were classified according to the International Classification of Childhood Cancer, 3rd revision.¹² Each child's treatment was categorized by a pediatric oncologist using the Intensity of Treatment Rating scale (ITR-3.0), a psychometrically validated measure of treatment intensity of current treatment protocols in pediatric oncology.¹³ Different disease and/or treatment modalities were classified according to one of the four intensity levels, from level 1 (minimally intensive; including treatments such as surgery only or chemotherapy for Wilms' tumor Stages 1 and 2) to level 4 (most intensive; including treatments such as HSCT or chemotherapy for acute myeloid leukemia).

Sociodemographic data, medication use and information on menstruation and fertility were retrieved from a survey consisting of validated self-report instruments and study-specific items. The survey was sent to all eligible participants together with a letter describing the study.¹⁰ The survey could be completed on paper, Web or by telephone interview. Two reminders were sent out to the nonresponders. Data was collected August 2019 to February 2020.

2.3 Measures

The outcome measures in our study were self-reported need of induced puberty, ongoing use of medication with ERT, attempt to conceive, having undergone fertility investigations and having a biological child. This data was retrieved from the following five study-specific questions in the survey: (a) Have you needed medication to induce puberty (no/I do not know or remember/yes)? (b) During the past 2 years, have you used female sex hormone to replace estrogen during a minimum of 1 month (no/I do not know or remember/yes)? (c) Have you had difficulties becoming pregnant (no/some difficulty/yes/have not tried)? (d) Have you undergone any investigations due to difficulties becoming pregnant, for example, blood sample or ultrasound (no/I do not know or remember/yes)? (e) Do you have a biological child (yes/no)? Participants had the possibility to also write an open-ended response.

For background characteristics, the following questions from the survey were included: Have you ever cohabitated or been married to a partner (no/yes/do not want to specify)? Have you ever needed injections with growth hormone (no/I do not know or remember/yes)? During the past 2 years, have you used medication for thyroid function/diabetes during a minimum of 1 month (no/I do not know or remember/yes)? What is your menstrual status right now?

Of those 1297 females who indicated their current menstrual status, 41 (3.2%) considered themselves to be menopausal, 177 (13.6%) reported that they had no menstruation the past 6 months, 300 (23.1%) had irregular menses and 779 (60.1%) regular menses. However, based on a review of the participants multiple-choice responses and written comments, for example, based on the names of medications the participants had stated, there were discrepancies in the responses. Some who reported no menstruation were pregnant or lactating or had no menstruation due to hormonally induced amenorrhea. After this review, ERT was considered the best indicator of ovarian failure. However, also in ERT there were some contradictory information, why 37/1295 (2.9%) responses were re-classified. In most of these cases (n = 21) the survivor had indicated the use of ERT, or ERT and contraception, to avoid pregnancy. The re-classified cases were revised to “no current use of ERT” if the patient indicated hormone replacement therapy but described medication for in vitro fertilization, nongonadal hormonal deficiencies, endometriosis, polycystic ovary syndrome or contraceptive implant to avoid pregnancy.

2.4 Statistical analyses

For continuous variables, sociodemographic characteristics are presented as median (range), as the data was not normally distributed. Categorical variables are presented as frequencies and percentages. Group comparisons of the background variables were undertaken with the Mann-Whitney U-test and Fisher's exact χ²-test.

Five binary logistic regression analyses were undertaken to examine predictors associated with the outcome variables: the need of induced puberty, current use of ERT, attempt to conceive, having undergone fertility investigations and having a biological child. Independent variables were treatment-related variables (chemotherapy, irradiation potentially exposing the gonads, central nervous system [CNS] irradiation, HSCT; yes/no), age at diagnosis, age at study and partner status (if the participant had ever been married or cohabitated with a partner; yes/no). These predictors were chosen a priori based on their theoretical importance. Employment status (yes/no) was initially added as a predictor but was removed since it did not add predictive value. Cancer diagnosis or ITR-3.0 were not used in the regression analyses since they are strongly correlated with the cancer treatment variables.

To obtain a cutoff value for the CNS and abdominal irradiation doses, the outcome variables were subjected to AUC analysis of ROC curves for the participants with a known irradiation dosage. For AUC, 0.8 is considered acceptable.¹⁴ Hence, only significant AUC analyses with AUC ≥ 0.8 are presented. Statistical calculations were carried out using IBM SPSS Statistics 28.0. All tests of significance were two-sided (P < .05).

3.1 Sociodemographic characteristics

Demographic characteristics are presented in Table 1. The survivors were a median of 28 years at the time of the study. The majority of survivors (86.0%) were working or studying full- or part-time or were on parental leave. The most common education was a university degree. Approximately two-thirds had at some point cohabitated with a partner or been married.

3.2 Clinical characteristics

Outcome variables are presented by diagnosis and treatment modality in Table 2. The most common diagnoses were hematological cancers in 45.3% of the survivors and the most common treatment modality was moderately intensive.

Of the 1333 female participants, one had undergone hysterectomy and was surgically sterile. Additionally, 10 females had undergone a unilateral oophorectomy. Of the 10 females who had undergone oophorectomy (median age at study 29.5, range 19-37 years), only one had ongoing ERT. Further, six indicated that they had not tried to become pregnant while four had biological children without having undergone fertility investigations.

3.3 Premature ovarian insufficiency

Of those 1234 females who had indicated their menarchal status at the time of diagnosis, 967 (78.4%) female patients were premenarchal before cancer diagnosis. Of these, spontaneous menarche post-treatment was reported by 898 participants (92.9%) at a median age of 13.0 (range 8-18) years. Spontaneous menarche was seen in all females with lymphoma, nonhigh stage neuroblastoma or renal tumors, hepatic tumors, retinoblastoma and in survivors with minimally intensive treatment (Table 2). Hormone-induced puberty was reported in 69 (7.1%) of the females who were premenarchal at the time of diagnosis. In the whole population the prevalence was 5.3%. Induced puberty was common after myeloid leukemia, high-stage neuroblastoma or renal tumors, as well as in survivors with the most intensive treatments (Table 2).

The need of hormone-induced puberty among females who were premenarchal before cancer diagnosis was significantly predicted by all treatment variables included in the logistic regression model, χ² (5) = 122.31, P < .001 (Table 3). The ROC analysis showed significant predictive power of the abdominal irradiation dose to predict the need for induced puberty (n = 13, AUC 0.925, P = .013). The analysis included only the local irradiation doses possibly affecting the gonads. Information about TBI was not available. A local abdominal irradiation dose of 36.0 Gy had high sensitivity (100%) and specificity (80%) to predict the need of induced puberty. A cutoff at 24.0 Gy had lower sensitivity (63%) but better specificity (100%). The ROC analysis with local CNS irradiation dose was not significant.

Of 1295 females, 121 (9.3%) indicated current use of ERT. The need of ERT was common after myeloid leukemia, high-stage neuroblastoma as well as in survivors with the most intensive treatments including relapse therapy and therapy for second malignancy. Accordingly, more than half of the survivors who had undergone HSCT or abdominal irradiation possibly affecting the gonads had ongoing ERT (Table 2). Of the 121 females with ERT, 27 (22.3%) had one additional hormone deficiency and 30 (24.8%) had two or more additional hormone deficiencies (thyroid, growth hormone) or diabetes. ERT together with two or more hormone deficiencies was most prevalent in patients with high-grade CNS tumors (12/121, 9.9%), patients treated with CNS irradiation (19/121, 15.7%) or HSCT (10/121, 8.3%). In the regression model, χ² (6) = 182.17, P < .001, ongoing ERT was significantly predicted by all the treatment variables included as well as older age at diagnosis, but not with age at study (Table 3). However, ROC analyses were not significant and, hence, cutoffs for local CNS or abdominal irradiation doses could not be estimated.

3.4 Chance of pregnancy

Of 1321 respondents, 626 (47.4% of responders) reported that they had tried to become pregnant. Possible relationships between the wish to conceive and treatments, particularly CNS irradiation through its effect on sexual functions and partnering, were explored. In a logistic regression, χ² (7) = 598.99, P < .001, survivors who had never cohabitated with a partner, who were younger at assessment and who had undergone CNS irradiation were less likely to have tried to conceive (Table 3). Chemotherapy, abdominal irradiation, HSCT and age at diagnosis were not significantly associated with the attempt to conceive in the regression analysis.

Of the 626 women who had tried to conceive, 251 women (40.1%) reported that they had at least some difficulties to become pregnant and 158 (25.2%) that they had undergone fertility investigations. Need of fertility investigations was common after high-stage lymphoma and neuroblastoma. Further, more than a third of the survivors who had needed a fertility investigation had received abdominal irradiation, HSCT and/or had undergone therapy for a second malignancy (Table 2). In a logistic regression, χ² (7) = 32.93, P < .001, women who had undergone HSCT had 5.4 times the odds of needing fertility investigations. No other treatment-related variable or factors in the current situation were significant predictors for the need of fertility investigations.

Finally, of the 626 women who had tried to conceive, 451 women (72.0%) had a biological child. High frequencies of pregnancies were reported in survivors of histiocytosis, nonhigh stage renal tumors and hepatic tumors. It was less common to have a biological child after high-stage neuroblastoma, retinoblastoma or after the most intensive treatment modality (Table 2). Similarly, a lower proportion of women with subsequent need of ERT had a biological child. This proportion was 10.7% (13/121) among all women with ongoing ERT and 29.5% (13/44) in women with ERT who had tried to conceive. In a logistic regression model, χ² (7) = 108.94, P < .001, having had a partner, older age at assessment and non-HSCT treatment predicted having a biological child (Table 3).

4.1 Methodological considerations

Our study presents a large national sample with a high response rate. No systematic response bias was detected, and, hence, the sample seems representative for the population. However, self-reports were in some cases observed not to be reliable. Some survivors mixed up medications and the indications for these medications, or menstrual cycle and bleeding for cyclic ERT, leading to contradictory information. Since comprehensive information was available, data could in these cases be corrected. In future survey studies, study-specific questions need to be constructed in a very precise manner taking this into account. Moreover, the study lacked laboratory data to confirm menopausal status.

Due to the cross-sectional design, timing of the outcome variables could not be assessed. Moreover, since the sample consists of young adults many have not yet tried to conceive a child. However, the desire of the survivors to become pregnant was known. Hence, only females who had stated that they had tried to conceive were included in analyses concerning infertility investigations and having a biological child.

Concerning the treatment variables used, large registry data was available. Nevertheless, we did not have information on which patients treated with HSCT had received TBI. Hence, we could not analyze the effect of TBI based conditioning regimens compared to non-TBI based regimens. All irradiation variables were presented as local irradiation only. Further, specific doses for alkylating agents were not available in the National Quality Registry for Childhood Cancer. Hence, drug- or dose-specific risks could not be assessed. However, based on the available data survivors could be categorized into four different groups based on the intensity of treatment.