Levels of exposure to ionizing radiation are increasing for women worldwide due to the widespread use of CT and other radiologic diagnostic modalities. Exposure to ionizing radiation as well as increased levels of estradiol and other sex hormones are acknowledged breast cancer risk factors, but the effects of whole-body radiation on serum hormone levels in cancer-free women are unknown. This study examined whether ionizing radiation exposure is associated with levels of serum hormones and other markers that may mediate radiation-associated breast cancer risk. Serum samples were measured from cancer-free women who attended biennial health examinations with a wide range of past radiation exposure levels (N = 412, ages 26–79). The women were selected as controls for separate case-control studies from a cohort of A-bomb survivors. Outcome measures included serum levels of total estradiol, bioavailable estradiol, testosterone, progesterone, prolactin, insulin-like growth factor-1 (IGF1), insulin-like growth factor-binding protein 3 (IGFBP-3), and ferritin. Relationships were assessed using repeated-measures regression models fitted with generalized estimating equations. Geometric mean serum levels of total estradiol and bioavailable estradiol increased with 1 Gy of radiation dose among samples collected from postmenopausal women (17%1Gy, 95% CI: 1%–36% and 21%1Gy, 95% CI: 4%–40%, respectively), while they decreased in samples collected from premenopausal women (−11%1Gy, 95% CI: −20%–1% and −12%1Gy, 95% CI: −20%– −2%, respectively). Interactions by menopausal status were significant (P = 0.003 and P < 0.001, respectively). Testosterone levels increased with radiation dose in postmenopausal samples (30.0%1Gy, 95% CI: 13%–49%) while they marginally decreased in premenopausal samples (−10%1Gy, 95% CI: −19%–0%) and the interaction by menopausal status was significant (P < 0.001). Serum levels of IGF1 increased linearly with radiation dose (11%1Gy, 95% CI: 2%–18%) and there was a significant interaction by menopausal status (P = 0.014). Radiation-associated changes in serum levels of estradiol, bioavailable estradiol, testosterone and IGF1 were modified by menopausal status at the time of collection. No associations with radiation were observed in serum levels of progesterone, prolactin, IGFBP-3 or ferritin.
INTRODUCTION
Exposure to ionizing radiation via diagnostic imaging, particularly CT scans, continues to increase worldwide 1. Numerous studies have proven the association between exposure to ionizing radiation and breast cancer 2. Moreover, data from the A-bomb survivors indicate that the breast is particularly susceptible to ionizing radiation exposure, with an attributable fraction for breast cancer of 27% for all women exposed to more than 5 mGy 3 [a radiation dose roughly twice the yearly background dose 4]. Breast cancer has also been associated with levels of endogenous hormones, particularly estradiol (E2) 5–7. Estrogens stimulate cell proliferation 8 and possibly exert a direct genotoxic effect on breast epithelial cells 9.
Other circulating hormones and proteins, including non-sex hormone-binding globulin (SHBG)-bound estradiol, testosterone 7 and insulin-like growth factor 1 (IGF1), have been associated with breast cancer 10. Additional serum markers have been hypothesized to be associated with breast cancer, including IGFBP-3 11, progesterone 12, prolactin 13 and ferritin. Ferritin levels and ionizing radiation are associated with increased levels of free radicals, and high ferritin levels have been reported to result in cell transformation and immortalization in mammary glands 14.
The question of whether ionizing radiation exposure modifies the underlying levels of risk-related circulating markers has not been addressed, however, and could be important in mechanistic studies of breast cancer or other hormone-related cancers after ionizing radiation exposure. The primary purpose of this project is to examine whether serum levels of risk-related markers vary with past radiation exposure in a representative sample of cancer-free female A-bomb survivors.
MATERIALS AND METHODS
Subjects
Study subjects were selected from the Adult Health Study (AHS), a longitudinal cohort study of A-bomb survivors initiated in 1958 15. The AHS features biennial clinical examinations with serum collection and well-characterized radiation doses for numerous body organs 16. Serum samples were selected in the context of concurrent case-control studies wherein controls were selected by incidence density sampling from all subjects who met eligibility requirements and were at risk for a first primary cancer of any type at the time an index case was diagnosed. Eligibility was defined as women who were free from various medical conditions (oophorectomy, current exogenous hormone treatment, prior malignancies, pregnant), were cancer-free for at least 2 years after serum collection, and had estimated radiation organ doses. Hormone-related diseases/conditions were based on self-reports or, in the case of extreme measured values, a medical chart review. Samples were matched to an index case on age at collection (±2 years), sample collection year (same decade), city (Hiroshima or Nagasaki), and counter-matched using breast radiation dose strata (low, medium and high with cut-points at 5 mGy and 790 mGy) 17. Counter-matching assures a wide exposure range while still providing an appropriate sample for inference of the exposure effects within the underlying cancer-free cohort 18. Because it is not known which organ system might be responsible for changes to serum markers, we used doses to the colon for purposes of analyses. The colon is a centrally located organ, and colon doses have been used as representative doses in other studies (e.g. for overall solid cancer risk estimates), although the results should be relatively insensitive to the organ chosen since the doses to all organs are highly correlated. Radiation doses to the colon were adjusted to account for random errors in the individual radiation dose estimates 19. Menopausal status at the time of sample collection was assigned using a two-step process. First, we used a conservative age range to assign menopausal status. If a woman was less than 43 years of age at the time of sample collection, she was designated premenopausal (N = 218); if she was older than 55 years, she was designated postmenopausal (N = 139). For women in the intermediate age group (43–55 years old), FSH (follicular stimulating hormone) levels were measured and used to assign menopausal status primarily based upon the manufacturer's test guidelines (premenopausal cutoff: FSH < = 15 mIU/ml, postmenopausal cutoff: FSH > = 23 mIU/ml) 20. Of those in the intermediate age range and assigned by FSH values (N = 152), 75 were designated premenopausal, 60 were postmenopausal, and the others were in between the two cutoff values and were designated perimenopausal (N = 17). Due to the paucity of samples collected from perimenopausal women, they were excluded. The final data set included 492 samples collected from 412 women (293 premenopausal and 199 postmenopausal samples). Of the 80 subjects with two separate samples, 63 had samples that were both collected prior to menopause, 6 subjects had samples that were both collected after menopause, and 11 had one sample collected before and one after menopause.
All blood samples were obtained with informed consent, and the Human Investigation Committee of the Radiation Effects Research Foundation (RERF) approved the study protocol.
Sample Storage and Serum Assays
Sera were collected in the Hiroshima and Nagasaki laboratories between 1969 and 1992 and stored in vials at −80°C. Samples for this study were pulled from the freezers simultaneously and assayed for total E2, bioavailable E2, testosterone, progesterone, prolactin, IGF-1, IGFBP-3, ferritin, and total protein and protein fractions (albumin and globulin). All assays were performed blinded to age, city of collection and radiation exposure but were not performed in duplicate due to the limited serum amounts available. Commercial laboratories with internal quality assurance systems performed all assays. Supplementary Table 1 ( http:dx.doi.org/10.1667/RR2631.1.S1) lists the assays and their precision as reported by the kits' manufacturers.
TABLE 1
Characteristics of Study Cohort and Collected Samples by Menopausal Status
Serum levels of IGF-1, IGFBP-3, FSH, prolactin, testosterone, progesterone and ferritin were measured at a single laboratory (SRL Laboratory, Hachioji, Japan). Total protein concentration and protein fractions were measured by SRL using the Biuret method and cellulose acetate membrane electrophoresis, respectively. Serum levels of total E2 and percentage bioavailable E2 were measured at a single laboratory (Mitsubishi Kagaku Bio-clinical Laboratories, Tokyo, Japan). Percentage bioavailable E2 (the fraction of serum E2 that is free or loosely bound to albumin) was determined by ammonium sulfate precipitation of SHBG-bound E2 21. Total E2 was multiplied by this percentage to obtain the concentration of bioavailable E2. Serum ferritin was measured by the conventional chemiluminescence enzyme immunoassay (CLEIA). When the aliquots of samples were transported to outside laboratories for testing, they were packed in appropriate protective shipping boxes with sufficient amounts of dry ice and checked at the time of receipt by the measuring laboratories.
Methods of Analysis
Due to the skewed distributions of the markers, analyses were performed using log-transformed values for all of the markers except IGF1 and IGFBP-3, which did not require transformation. Effects of radiation exposure were estimated with adjustment only for other factors that were significantly related to marker level for statistical efficiency. Explanatory factors considered were: age at time of sample collection, year of birth, menopausal status (pre or post), total protein, body mass index [weight (kg)/height (m2)] and its square, history of exogenous hormone use (yes or no), age at time of radiation exposure (including dichotomy at 15 years as a surrogate for pre- or postmenarche at time of exposure), number of full-term pregnancies, age at first full-term delivery, and radiation dose. Decisions for inclusion of factors other than radiation dose were based on a backward selection process where the least significant explanatory factor was removed until all factors were significant at a level of P ≤ 0.05. The variable selection process was conducted separately for each marker. Analyses were performed using a regression method for correlated (repeated) measurements (generalized estimating equations; GEE) 22. P values and 95% confidence intervals were based on the Wald statistic.
Radiation effects were assessed using linear dose–response functions. Dose responses were estimated with and without regard to effect modification by menopausal status at the time of blood collection and ionizing radiation exposure before or after age 15. For markers that were transformed prior to analyses, the radiation effects were reported on the untransformed scale by exponentiating the regression coefficient obtained using the log-transformed data and represent the association between 1 Gy radiation dose and the percentage change in geometric mean of the marker level. IGF1 and IGFBP-3, which were not transformed, were also reported on a relative scale by dividing the dose coefficient by the regression constant. To test the strength of the observed radiation effects, we identified and removed the most influential observations using the size-adjusted criterion 2/sqrt(n) for the t-statistic-like standardized deletion diagnostic DFBETAS of Belsley et al. 23. To alleviate concerns about misclassification of menopausal status at the time of sample collection due to possible nonspecificity of the FSH results, we conducted a sub-analysis after restricting samples to those collected from women at ages less than 43 or greater than 55 (135 samples omitted) and compared results to the full analyses.
Statistical comparisons between samples collected pre- and postmenopause were performed using χ2 tests (for counts), t tests for equality (for means) and Wilcoxon rank-sum tests for independent samples (for medians). Interactions of radiation effects by menopausal status or exposure before or after age 15 were tested using the likelihood ratio. To compare fits of statistically significant non-nested interaction models, the quasi-likelihood/independence criterion (QIC) was used 24. All results were obtained using Stata 25.
RESULTS
Table 1 shows the demographics of the study cohort and marker values for samples separated by menopausal status at the time of collection. The distribution of the radiation dose is broader than what would be expected from a random sampling of all cancer-free women since they were counter-matched (low, medium, high) against the dose status of the case to ensure a broad representation of doses. The average year of sample collection is somewhat later for postmenopausal samples because the AHS study is a fixed cohort; more women were postmenopausal in later calendar years.
Table 2 shows the agreement of assays for the 63 pairs of samples collected from women who were premenopausal at each collection. Comparisons of multiple samples collected from the same women while postmenopausal were not reported (N = 6 pairs). The earlier sample was collected at average age 39.7 years and the average interval between sample collections was 2.8 years. Correlation coefficients ranged from 0.11 for progesterone to 0.75 for ferritin. A paired t test comparing the difference of the paired values with zero was not significant for any marker except for serum prolactin, where the earlier value proved to be significantly higher than the second value.
TABLE 2
Comparison of Assay Values from Women Donating Multiple Samples
The first column of Table 3 shows the associations of radiation with marker levels with no allowance for effect modification but with adjustment for other significant factors. No significant radiation associations were observed in any of the markers when ignoring menopausal status. Columns two through five in Table 3 show the radiation associations separately by menopausal status at the time of sample collection or by age at time of bombing (≤15 or >15 years; ATB15). Because menopausal status at blood donation was highly correlated with ATB15 in this study sample (particularly for a woman's first/only serum sample), the latter two columns largely mirror the results displayed in the former two. Total and bioavailable E2, testosterone and IGF-1 demonstrated statistically significant associations with radiation dose in one or both periods.
TABLE 3
Relative Radiation Associations with Adjusted Mean Marker Values and Effect Modification By Menopausal Status and Age at Radiation Dose
Total E2 (Fig. 1) decreased with radiation dose in the premenopausal period (relative change 0.89 at 1 Gy; 95% CI: 0.80, 0.99; P = 0.027), or in other words, the level of total E2 changed as k*0.89dose, where k is the level of total E2 when the dose was zero (results from other log-transformed variables are interpreted in the same way). The level of total E2 increased with dose in the postmenopausal period (relative change 1.17 at 1 Gy; 95% CI: 1.01, 1.36; P = 0.037). Effect modification by menopausal status was highly significant (P = 0.003) and remained significant after removing the most influential values (P = 0.038). The association of total E2 with radiation dose was similar when categorizing subjects based on ATB15 (rather than menopausal status at the time of collection); relative change for a 1-Gy dose at age less than 15 years was 0.86 (95% CI: 0.76, 0.96; P = 0.010) and relative change for exposure after age 15 was 1.12 (95% CI: 0.99, 1.26; P = 0.063). Models used to estimate the radiation effects on total E2 included menopausal status, age at sample collection, total protein, BMI and a BMI*postmenopause interaction. The QIC values for the two models were very similar, but the value for the ATB15 interaction model was slightly lower (i.e. better fit) than for the model with interaction by menopausal status.
Bioavailable E2 (Fig. 2) showed a significant decrease with radiation dose in the premenopausal period, with a relative change of 0.88 at 1 Gy (95% CI: 0.80, 0.98; P = 0.022) and a significant increase with radiation dose in the postmenopausal period with a relative change of 1.21 at 1 Gy (95% CI: 1.04, 1.40; P = 0.011). Effect modification by menopausal status was highly significant (P < 0.001) and remained so after exclusion of influential observations (P = 0.003). As with total E2, bioavailable E2 also demonstrated similar radiation effects according to whether exposure was before or after age 15 (0.87 at 1 Gy; 95% CI: 0.77, 0.98; P = 0.019 and a relative increase of 1.13 at 1 Gy; 95% CI: 1.00, 1.27; P = 0.045, respectively). The model used to estimate the radiation effects on bioavailable E2 included menopause status, age at sample collection and total protein. Again the QIC value was slightly lower for the model with ATB15 interaction.
Testosterone (Fig. 3) exhibited an increase with radiation dose in the postmenopausal period (relative change 1.30 at 1 Gy; 95% CI: 1.13, 1.49; P < 0.001) and a marginally significant decrease with radiation dose in the premenopausal period (0.90 at 1 Gy; 95% CI: 0.81, 1.00; P = 0.049). Estimates of effect modification by menopausal status were highly significant when estimates were derived using all observations (P < 0.001) and after excluding outliers (P = 0.003), whereas no effect modification was evident with ATB15. The testosterone model included menopause status, age at sample collection, total protein and a total protein*postmenopause interaction, and BMI.
IGF-1 (Fig. 4) displayed a significant increase with radiation dose in the premenopausal period (relative change 1.11 per Gy; 95% CI: 1.02, 1.18; P = 0.024). Since the values from this assay were not log-transformed, the results are interpreted as the percentage change in marker levels per gray. No significant change with dose was observed in the postmenopausal period (relative change 0.91 per Gy; 95% CI: 0.75, 1.04; P = 0.18). Similar effects were seen when categorizing the women by age at exposure. Effect modification by menopausal status was significant (P = 0.014) and remained so after removing the most influential observations (P = 0.017). Comparing QIC values indicated that the model with menopausal interaction model was preferable to the one with ATB15 interaction. Estimates of the radiation effects were adjusted for age at sample collection, year of birth, total protein*premenopause interaction, and number of full-term pregnancies.
The other markers (IGFBP-3, progesterone, prolactin and ferritin) demonstrated no associations with radiation dose, either with or without effect modification. No effect modification of the radiation effects was observed in relation to the sample storage period, time since exposure, or the number of full-term pregnancies, although E2 levels did increase at a marginally significant level for each additional full-term pregnancy (results not shown). Model fits generally improved when including the number of full-term pregnancies (as either an effect modifier or a main effect), but the estimated coefficients for the number of pregnancies were never statistically significant and therefore the results were not reported. No improvements in model fit were detected by the addition of a quadratic dose term. Point estimates were generally unchanged by the removal of samples retrieved when women were aged 43 to 55.
DISCUSSION
The present study was designed to investigate the associations of ionizing radiation exposure with sex and growth hormones in cancer-free A-bomb survivors using stored sera. Increasing levels of ionizing radiation exposure led to disparate changes by menopausal status in a number of serum markers. Levels of E2 and bioavailable E2 increased in postmenopausal samples while they decreased in premenopausal samples. Caution must be used when interpreting the results in premenopausal samples, especially E2 levels, which are highly variable in premenopausal women due to the menstrual cycle. Table 2 illustrates the high variability of serum marker levels among premenopausal women. Controlling for the day of the menstrual cycle was not possible, because the clinical examination protocol did not attempt to record these values. It should be noted, however, that the point estimates for the radiation findings were largely unchanged after separately removing the most influential measurements (e.g., those associated with the E2 peak at ovulation) and excluding those in the perimenopausal age range 43–55. Also, it is unlikely that there is any association with the timing of blood sample collection and radiation dose, eliminating most concerns of confounding. Nevertheless, chance findings for the premenopausal hormone associations cannot be completely ruled out.
The magnitude of change in total E2 levels was about 10% at 1 Gy of radiation (negative in premenopausal women and positive in postmenopausal women). It is not clear that either of these are clinically significant changes, particularly among premenopausal women where there is so much variation. However, if E2 levels were assigned to quartiles in samples taken from postmenopausal women, a 10% change in any particular result would generally be sufficient to increase its quartile assignment by one category. Interpolating the results from Fig. 1 of the meta-analysis performed by Key et al. 7, this increase in total E2 would roughly correspond to a 25% increase in breast cancer risk.
The significance of IGF-1 and IGFBP-3 levels in pre- or postmenopausal breast cancer has been under debate. The most recent evidence indicates that high levels of IGF-1 confer an increased risk of breast cancer 10. In our study, radiation exposure appeared to increase the level of IGF-1 in premenopausal women but had no effect on levels of IGF-1 in postmenopausal women. Again, however, large variations were observed in the premenopausal IGF-1 levels. IGFBP-3 was not associated with radiation exposure.
If the results are correct, questions arise regarding the radiation-related biological pathways that may affect sex hormone levels, particularly the most potent hormone E2. In premenopausal women, E2 is produced primarily by the ovaries whereas after ovarian involution, estrogen is primarily produced in adipose tissue by converting androgen to estrogen via aromatase. A number of speculative mechanisms for an association between radiation and sex steroid alterations can be suggested, including damage to primordial ovarian follicles leading to decreased ovarian aromatase and depressed E2 production as observed in the samples taken from women who were exposed at less than 15 years of age 26, 27. Ataya et al. reported similar findings with decreased ovarian size and long-term decreased serum E2 levels in rhesus monkeys with ovarian exposure to ionizing radiation 28. Larsen et al. reported decreased ovarian volume in spontaneously ovulating women who had received chemotherapy and radiotherapy for cancers diagnosed at mean age 5.4 years; however, E2 levels were reported to be higher in those receiving (any type of) therapy compared to a control group of similar age 29.
For the results seen among older birth cohorts (i.e. postmenopausal samples), radiation-induced upregulation of inflammatory cytokines, specifically TNFα, can lead to increased estrogen biosynthesis 30–33; radiation can also alter lipid metabolism 34–36, possibly leading to hyperandrogenism through hypothalamus-pituitary-adrenal axis dysregulation 37. Either of these latter mechanisms would be consistent with the observed increased E2 and testosterone levels among the postmenopausal women.
No effect modification by the number of full-term pregnancies was evident in the radiation effects on any of the markers. The number of women reporting any history of exogenous hormone intake was less than 10%, and that variable was not a significant risk factor in any of the tested models. Therefore, we do not believe that hormone treatment had a meaningful impact on our findings. Interaction models categorizing exposures before or after age 15 more often had a better fit than models categorized by menopausal status at the time of serum collection. However, this may be due to misclassification of menopausal status as opposed to an important biological effect involving menarche status at the time of radiation exposure. McDougall et al. found no association of breast cancer risk and menarche status at the time of radiation exposure but did not consider serum marker levels 38.
Associations of radiation and markers modified by menopausal status were tested in the results of eight bioassays (the two parallel analyses of modification by either menopause status or ATB15 are effectively the same), which may raise concerns of finding associations due to multiple testing. Using the conservative Bonferroni correction for multiple testing (α = 0.05/8 = 0.00625), all of the tests for interaction remain significant with the exception of IGF-1 (P = 0.014).
The study boasts a number of strengths, including well-characterized radiation doses to a representative population of women, and data collected prospectively from cancer-free women as part of a long-term follow-up regimen, including medical history data on hormone-related conditions. However, there are also several weaknesses. The day of blood collection in relation to the menstrual cycle was not available, which is problematic for premenopausal estradiol measurements in particular. The use of FSH to assign menopausal status has been shown to be not fully specific 39. It is also likely that misclassification occurred when assigning menarche status by age. Due to the limited number of samples, duplicate assays for quality control were not attempted. A limited number of women did have multiple samples collected while premenopausal, for which the correlation between marker assays was generally poor, probably due to different times in the menstrual cycle. Despite these limitations, no major changes in the interactions between the measured radiation associations and menopausal status were detected after separately removing the most influential observations and those observations drawn from women in the perimenopausal age range, which helps confirm the robustness of the associations.
Conclusions
Previous studies have proven the potency of ionizing radiation as a carcinogen via direct damage effects to DNA while separate lines of research have established serum levels of E2, testosterone and IGF1 as risk factors for breast cancer. The results of this study suggest that ionizing radiation may lead to changes in serum levels of cancer-related hormones and proteins in cancer-free women and that these changes are dependent upon either the menopausal status at the time of collection or the menarche status at the time of exposure. Several mechanisms for these changes are suggested, but they are all speculative. Researchers of cancer etiology after whole-body ionizing radiation exposure should be aware of possible underlying changes in the levels of serum markers, which may confer independent carcinogenic risks.
ACKNOWLEDGMENTS AND FUNDING
We deeply thank the participating members of the AHS. The Radiation Effects Research Foundation (RERF), Hiroshima and Nagasaki, Japan is a private, non-profit foundation funded by the Japanese Ministry of Health, Labor and Welfare (MHLW) and the U.S. Department of Energy (DOE), the latter in part through the National Academy of Sciences. This publication was supported by RERF Research Protocol RP 06-02 and by US National Cancer Institute contract NO1-CP-31012 and the Japanese Ministry of Education, Culture, Sports, Science and Technology grants 14031227 and 15026220.