Abstract
Previous research suggests there may be a hormonal influence on glioma risk as evidenced by lower rates in females, change in incidence rates around ages at menarche and menopause and presence of hormone receptors in glial tumors. Using the large San Francisco Bay Area Adult Glioma Study, we investigated whether reported reproductive factors and hormone use were associated with gliomas overall or with histologic subtypes among female cases (n=619) and controls (n=650). We found that reproductive factors were generally not associated with gliomas. Weak to moderately elevated odds ratios were observed for self-reported later age at menarche (14+ years old versus 12–13 years old: adjusted odds ratio (AOR) = 1.39, 95% confidence interval (CI): 1.02 –1.89), particularly for non-glioblastoma histologies (AOR = 1.64, 95% CI: 1.11–2.43). Inverse associations were observed for ever self-reported use of exogenous hormones (oral contraceptive use: AOR = 0.72, 95% CI: 0.53–0.99; postmenopausal hormone use: AOR = 0.56, CI: 0.37–0.84). However, cumulative hormone exposure defined multiple ways demonstrated no clear pattern of association. The results of this study suggest that any protective effect of hormones on gliomas may be limited to exogenous hormones, but a more detailed history of exogenous hormone use are needed to confirm findings.
Keywords: reproductive factors, gliomas, case-control study, exogenous hormone use
Introduction
Hormones play a critical role during brain development in-utero;1,2 appear to work as a neuroprotectant via estrogen-receptor independent pathways in brain degeneration,3 and may reduce the risk of death and degree of disability in cases of traumatic brain injury.4 Whether similar hormonal pathways are involved in the etiology or progression of brain cancers is unknown, but cell cultures and animal studies demonstrate evidence of estrogen directly killing glioma cells5 and inhibiting the growth of gliomas,6,7 the most common and deadly brain tumor. Plausible mechanisms by which sex hormones may play a role in gliomagenesis have been postulated. For example, estrogen can inhibit cell cycle entry by increasing mitogen-activated protein kinase (MAPK) levels in astrocytes and stimulating the AKT/PIP pathway to assist in cell cycle control.8 Other plausible mechanisms include an estrogen-astrocyte-TGF-β1 pathway9 and melatonin-driven neuroprotection via aromatase.10,11
Data from human studies appear to support a hormonal influence on gliomas, including reports of consistent excess of gliomas among males, hormone receptors in glial tumors, and changes in incidence rates around ages at menarche and menopause. However, epidemiologic studies using reproductive and menstrual history as surrogates of hormone exposure have been generally inconclusive due, in part, to small sample sizes, different histology groupings among glioma classification systems, or the combining of heterogeneous histologic subtypes.12–22 We used data from the largest population-based case control study of adult brain tumors to date, the San Francisco Bay Area Adult Glioma Study, to estimate the effects of reproductive factors and hormone use on gliomas among women. The large sample size allowed a valuable opportunity for analysis by glial tumor subtype and further exploration of cumulative female hormone using ages at menarche and menopause and years of exogenous hormone use to more completely characterize dose-response relationships and improve on previous studies using reproductive variables indicative of only one point in time.
Materials and Methods
Study participants
Details of subject recruitment and participation rates for the study series 1 and 2 have been described previously23–25 and protocols for the most recent series 3 were the same as previously used. Briefly, cases diagnosed with histologically confirmed incident gliomas (International Classification of Diseases for Oncology, morphology codes 9380–9481) between August 1991 to April 1994 (Series 1), May 1997 to August 1999 (Series 2), and November 2001 to September 2004 (Series 3) were identified using the Northern California Rapid Case Ascertainment program. Eligible cases had pathologically confirmed glioma, and resided in the 6 county San Francisco Bay Area (Alameda, Contra Costa, Marin, San Mateo, San Francisco, and Santa Clara). Controls were selected from the same residential area as cases were identified using random digit dialing and frequency matched to cases on age (in 5 year age groups), gender, and race/ethnicity (White, Black, Hispanic, Asian, or other).
Of the 1905 eligible cases in the parent study, full interviews were obtained for 1408 cases (74%) and 15 completed an abbreviated questionnaire form (1%). The remaining cases were not included because either they could not speak English (n=24), were too ill (n=1), did not receive physician permission (n=18), did not have sufficient specimen to confirm glioma diagnosis (n=26), were not able to be contacted (n=63), or otherwise refused (n=267). Of the 27,285 phone numbers randomly generated to contact potential controls, 18% never answered the phone after 10 attempts and 13% were business phone numbers. For those phone contacts, nearly 16% refused to give information to assess eligibility and 5% were either too ill or did not understand English. Approximately 7% (n=1908) were eligible controls and 73% (n=1396) agreed to participate and complete the interview. The other eligibles were dropped because either they refused (n=415), did not speak English (n=14), were not located (n=49), or otherwise could not be used (n=43). Eligible cases and controls for this study were restricted to females 20 years and older, leaving 619 cases and 650 controls available for analysis. Among women, the participation rates (defined as number of cases (or controls) consenting to full interview divided by number of eligible cases (or controls)) were 79% for cases and 79% for controls.
Exposure Assessment
Reproductive, menstrual, and exogenous hormone histories were collected via in-person or phone interview with cases (or their proxies) and controls. On average, self-reporting cases were interviewed within 5 months and proxies within 8 months from cases’ diagnosis (range: 3 days – 36 months). Proxy respondents were used in 37% of cases and 0.3% of controls. Questions asked included age at first menstrual period, age menstrual periods stopped completely, number of live births, number of pregnancies not surviving to term (question asked only in series 1), age at each birth, oral contraceptive use (ever/never, never/former/current) and duration (years or months), and hormone replacement therapy use (ever/never, never/former/current) and duration (years or months). Menopausal status was determined by whether they were still menstruating (choices included: yes; yes, but pregnant or postpartum; yes, but infrequently, probably perimenopausal; yes, but taking menopausal estrogen now) or were no longer menstruating (choices included no, went through natural menopause; no, had a hysterectomy; no, other). If they reported having a hysterectomy, further inquiry was made into type of surgery (womb removed, womb and both ovaries removed, womb and one ovary removed, one or both ovaries removed, type unknown). Types of surgery were evaluated independently in association with gliomas, but were ultimately combined into a single variable (surgical menopause) due to small numbers in most strata. Women reporting partial hysterectomies (ovaries remained) were categorized as premenopausal if younger than 50 years of age; women reporting a partial hysterectomy at 50 years of age or older or bilateral oopherectomy were categorized as postmenopausal. In the instance where younger women (<45 years of age) reported not having periods due to unknown or sundry reasons (eg. thyroid disorders, polycystic ovarian syndrome), we coded as premenopausal. For analysis of postmenopausal hormone use, we excluded those women with indeterminate menopausal status (n=32 cases and 3 controls). Like previous studies, cumulative hormonal exposure was calculated using combinations of age at menopause, age at menarche, exogenous hormone use, and parity to crudely account for variability of hormone fluctuations throughout reproductive life.26–28 Other covariates of interest included cigarette smoking, ethnicity, highest degree earned (education), income, and age at diagnosis.
Statistical Analysis
Category boundaries for continuous variables were informed by univariate quartile distributions among controls but cutpoints were primarily determined by known or suspected biological significance provided strata contained more than five persons within each cell. Adjusted odds ratios and 95% confidence intervals were calculated for each reproductive, menstrual, and exogenous hormone factor independently using unconditional logistic regression models. We identified age, ethnicity, and education for potential confounding using causal directed acyclic graphs29 and further evaluated confounding by stratified analysis and change in the beta coefficient after removal from models (>10% change). Initial models adjusted for ethnicity, education, and age coded using tail-restricted quadratic splines. Menstruation years was additionally adjusted for parity. Since adding study series to the model containing other predictors of glioma did not notably change the odds ratio for all reproductive factors, series was not included in the final model. In addition to analyzing both proxy and self-respondents, we restricted models to self-report females only. Age at diagnosis or interview, menopausal status and smoking were selected a priori as potential effect modifiers for selected reproductive factors and tested using likelihood ratio tests.
Effect measure modification of the ever oral contraceptive use-glioma association by ever postmenopausal hormone use was evaluated by comparing the model with only main effects with the model inclusive of cross-product terms ever oral contraceptive use and ever postmenopausal hormone use (likelihood ratio test p value < 0.10). Interaction on the additive scale was determined by comparing the observed odds ratio in the doubly exposed (oral contraceptive and postmenopausal hormone use) to the expected OR assuming additive risks.
Continuous variables for main effects were modeled to test for linear trend. Beta coefficients and p value of the Wald test were used to determine if slope differed from zero. Graphical representations using spline regressions were utilized to confirm monotonic trends. Analyses of associations for specific histologic types (grouped as glioblastoma and non-glioblastoma) compared each subtype to the entire control group. All analyses were conducted using SAS 9.1 software (SAS Institute, Inc., Cary, North Carolina).
Results
Baseline characteristics of the 619 cases and 650 controls were similar except that cases were less likely than controls to have a college education (Table 1). The average age at diagnosis was 56.3 (standard deviation = 17.1) and the average age at interview for controls was 55.4 (standard deviation = 16.8). Proxy respondents were much more common for cases (37%) than controls (0.3%) due to cases’ death or disability. Of the 229 proxies in this study, most were daughters (32%) and husbands (32%). The average ages at menarche and menopause was 12.6 (standard deviation = 1.7) and 48.8 (standard deviation = 6.3), respectively. Approximately 62% self-reported as postmenopausal, but a lower proportion (52%) was observed when excluding those with indeterminate menopausal status. Postmenopausal women were older in age and less likely to report ever using oral contraceptives (41%) compared to premenopausal women (77%), but were otherwise similar. The majority of tumors among cases were glioblastomas (60%). Missing information with regard to baseline characteristics was minimal and similarly distributed between cases and controls.
Table 1.
Selected characteristics of glioma cases and controls among female participants in the San Francisco Adult Glioma Study, 1991–2004
| Female participants | ||||
|---|---|---|---|---|
| Case (n=619) | Control (n=650) | |||
| n | (%) | n | (%) | |
| Age at interview (yrs) | ||||
| 20–44 | 173 | (28) | 192 | (30) |
| 45–54 | 95 | (15) | 118 | (18) |
| 55–64 | 121 | (20) | 117 | (18) |
| 65+ | 230 | (37) | 223 | (34) |
| mean (sd) | 56.3 (17.0) | 55.4 (16.8) | ||
| Ethnicity | ||||
| White | 494 | (80) | 525 | (81) |
| Black | 30 | (5) | 30 | (5) |
| Asian | 32 | (5) | 36 | (6) |
| Hispanic/Mexican | 44 | (7) | 40 | (6) |
| Other | 19 | (3) | 19 | (3) |
| Smoking 100 cigarettes | ||||
| Never | 325 | (53) | 332 | (51) |
| Past | 225 | (36) | 234 | (36) |
| Current | 66 | (11) | 83 | (13) |
| Missing | 3 | 1 | ||
| Education | ||||
| High school or less | 301 | (49) | 259 | (40) |
| College | 232 | (37) | 305 | (47) |
| Graduate school/Professional | 85 | (14) | 86 | (13) |
| Missing | 1 | 0 | ||
| Income | ||||
| <30,000 | 153 | (25) | 194 | (30) |
| 30,000–70,000 | 221 | (36) | 222 | (34) |
| 70,000+ | 210 | (34) | 219 | (34) |
| Missing | 35 | (6) | 15 | (2) |
| Marital Status | ||||
| Single | 75 | (12) | 91 | (14) |
| Married/spouse | 355 | (57) | 299 | (46) |
| Separated/divorced | 85 | (14) | 145 | (22) |
| Widowed | 104 | (17) | 114 | (18) |
| Missing | 0 | 1 | ||
| Histologic Tumor Type | ||||
| Glioblastoma | 370 | (60) | ||
| Astrocytoma | 115 | (19) | ||
| Oligodendroglioma/ Oligoastrocytoma |
80 | (13) | ||
| Other* | 54 | (9) | ||
Numbers do not always total to 100% due to rounding
other=ependymoma, subependymoma, juvenile pilocytic, astrocytoma- not classified, medulloblastoma, gangliogliomas, and not specified
Results of multivariable analyses are presented in tables separately for all women and for self-reporting respondents only. We discuss results of self-reporting women only (indicated by odds ratio among index respondents, ORindex) since risk estimates were generally not altered by proxy status and self-report is assumed to be more accurate than recall by proxy.
Reproductive factors
After adjusting for age and ethnicity, gliomas were not associated with ever giving birth (ORindex = 0.99, 95% CI: 0.73–1.34), or gravidity (ORindex = 0.88, 95% CI: 0.46–1.69, Table 2). Categorical odds ratios for increasing age at first birth among parous women were near unity and the slope of the continuous trend estimate for a one-year change in age at first birth was essentially flat (β = −0.03, 95% CI: −0.03–0.03, p trend = 0.85). Parous women with more children had higher glioma incidence than those with fewer children, with a change in slope per child of β = 0.08 (95% CI: −0.05–0.21, p trend = 0.23).
Table 2.
Frequencies and adjusted odds ratios (95% CI) of reproductive factors among glioma cases and controls, and self-reporting participants only in the San Francisco Adult Glioma Study, 1991–2004
| Cases | ||||||||
|---|---|---|---|---|---|---|---|---|
| Reproductive factors | All women (n=619) |
Self- reporting women (n = 390) |
Controls (n=650) |
OR (95% CI)‡ All women |
OR (95% CI)‡ Self-reporting women only |
|||
| Parity | ||||||||
| Never | 139 | (22) | 116 | (30) | 164 | (25) | 1.00 (ref) | 1.00 (ref) |
| Ever | 480 | (78) | 274 | (70) | 486 | (75) | 1.08 (0.82–1.44) | 0.99 (0.73–1.34) |
| Missing | 0 | 0 | 0 | |||||
| Children | ||||||||
| None | 139 | (22) | 116 | (30) | 164 | (25) | 1.00 (ref) | 1.00 (ref) |
| 1 | 100 | (16) | 59 | (15) | 128 | (20) | 0.87 (0.61–1.25) | 0.76 (0.51–1.14) |
| 2 | 188 | (30) | 116 | (30) | 181 | (28) | 1.19 (0.86–1.64) | 1.06 (0.74–1.50) |
| 3+ | 192 | (31) | 99 | (25) | 177 | (27) | 1.15 (0.82–1.63) | 1.17 (0.79–1.74) |
| Gravidity γ | ||||||||
| Never | 38 | (19) | 27 | (26) | 37 | (18) | 1.00 (ref) | 1.00 (ref) |
| Ever | 164 | (81) | 77 | (74) | 172 | (82) | 0.85 (0.49–1.48) | 0.88 (0.46–1.69) |
| Missing | 1 | 0 | 0 | |||||
| Age at first birth | ||||||||
| Nulliparous | 139 | (23) | 116 | (30) | 164 | (25) | 1.00 (ref) | 1.00 (ref) |
| ≤20 years old | 100 | (16) | 56 | (14) | 92 | (14) | 1.05 (0.73–1.52) | 1.05 (0.68–1.60) |
| 21–25 years old | 171 | (28) | 88 | (23) | 167 | (26) | 1.09 (0.81–1.45) | 1.00 (0.72–1.41) |
| 26–30 years old | 128 | (21) | 80 | (21) | 141 | (22) | 1.20 (0.64–2.25) | 0.99 (0.47–2.05) |
| 30+ years old | 78 | (13) | 48 | (12) | 86 | (13) | 0.89 (0.45–1.75) | 0.98 (0.44–2.15) |
| Missing | 3 | 2 | 0 | |||||
| Age at menopauseω | ||||||||
| <45 years old | 68 | (23) | 29 | (18) | 64 | (19) | 1.06 (0.67–1.67) | 0.87 (0.50–1.54) |
| 45–50 years old | 117 | (39) | 67 | (43) | 143 | (43) | 1.00 (ref) | 1.00 (ref) |
| 50+ years old | 115 | (38) | 61 | (39) | 125 | (38) | 1.22 (0.84–1.75) | 1.15 (0.74–1.80) |
| Missing | 42 | 2 | 2 | |||||
| Menopausal type | ||||||||
| Natural | 239 | (72) | 119 | (75) | 254 | (76) | 1.00 (ref) | 1.00 (ref) |
| Surgical | 92 | (28) | 40 | (25) | 79 | (24) | 1.00 (0.68–1.46) | 1.12 (0.70–1.81) |
| Don’t know | 10 | 0 | 1 | |||||
| Missing | 1 | 0 | 0 | |||||
| Menstruation years† | ||||||||
| <25 years | 133 | (26) | 120 | (31) | 166 | (26) | 1.00 (ref) | 1.00 (ref) |
| 25–34 years | 179 | (35) | 136 | (35) | 209 | (32) | 1.31 (0.88–1.96) | 1.18 (0.76–1.81) |
| ≥35 years | 199 | (39) | 131 | (34) | 268 | (42) | 1.29 (0.84–1.99) | 1.11 (0.69–1.79) |
| Missing | 108 | 3 | 7 | |||||
Menstruation years = menopausal age (or current age, if premenopausal) – menarcheal age; additionally adjusted for parity
Adjusted for age (restricted quadratic splines), ethnicity, and education
obtained in series 1 only
among postmenopausal women
NOTE: Two controls were proxies and were accounted for in the adjusted OR for self-reporting respondents.
Menstrual factors
Using 12–13 year olds as the referent group due to the large proportion of controls with ages at menarche during this period, higher odds ratios were observed for later age at menarche (≥14 years versus 12–13 years old: ORindex = 1.39, 95% CI: 1.02–1.89). The odds ratio for later age at menopause (>50 years versus 45–50 years old: ORindex = 1.15, 95% CI: 0.74–1.80, Table 2) was consistent with no effect, as was menopausal status (postmenopausal vs. premenopausal: ORindex = 1.14, 95% CI: 0.81–1.62, data not shown). Postmenopausal women experiencing a later age at menarche (≥ 14 years) and early onset of menopause (<45 years old) (ORindex = 0.87, 95% CI: 0.34–2.20) demonstrated relatively comparable odds ratios as compared to postmenopausal women with earlier age (<11 years) and average or later age of menopause (ORindex = 0.74, 95% CI: 0.23–2.41, data not shown), but effect estimates were imprecise. Further examination of menstruation years adjusted for parity was not informative.
Stratification by histologic tumor type did not alter odds ratios, with the exception of age at menarche (table 3). An odds ratio of 1.64 (95% CI: 1.11–2.43) was calculated for the association between non-glioblastoma cases and later age at menarche (≥14 years) as opposed to glioblastoma cases (ORindex = 1.16, 95% CI: 0.77–1.73).
Table 3.
Frequencies and adjusted odds ratios (95% CI) of age at menarche and glioma by histologic subtype among all women and among self-reporting women only in the San Francisco Adult Glioma Study, 1991–2004
| Glioblastoma Cases | ||||||||
|---|---|---|---|---|---|---|---|---|
| All women (n=370) |
Self-reporting women (n = 180) |
Controls (n=650) |
OR (95% CI)‡ All women |
OR (95% CI)‡ Self-reporting women only |
||||
| Age at menarche | ||||||||
| ≤11 years old | 53 | (18) | 33 | (18) | 125 | (19) | 0.99 (0.68–1.45) | 0.95 (0.61–1.50) |
| 12–13 years old | 161 | (55) | 99 | (55) | 367 | (57) | 1.00 (ref) | 1.00 (ref) |
| ≥14 years old | 81 | (27) | 48 | (27) | 154 | (24) | 1.12 (0.80–1.56) | 1.16 (0.77–1.73) |
| Missing | 75 | 0 | 4 | |||||
| Non-glioblastoma Cases | ||||||||
| All women (n=249) |
Self-reporting women (n = 210) |
Controls (n=650) |
OR (95% CI)‡ All women |
OR (95% CI)‡ Self-reporting women only |
||||
| Age at menarche | ||||||||
| ≤11 years old | 48 | (19) | 45 | (21) | 125 | (19) | 1.22 (0.81–1.84) | 1.28 (0.84–1.96) |
| 12–13 years old | 116 | (47) | 104 | (50) | 367 | (56) | 1.00 (ref) | 1.00 (ref) |
| ≥14 years old | 65 | (26) | 61 | (29) | 154 | (24) | 1.53 (1.05–2.23) | 1.64 (1.11–2.43) |
| Missing | 20 | 0 | 4 | |||||
Adjusted for age (restricted quadratic splines),ethnicity, and education
NOTE: Two controls were proxies and were accounted for in the adjusted OR for self-reporting respondents
Exogenous hormone use
Odds ratios for the association between exogenous hormone use and gliomas were consistently below 1.0 for all cases and self-reporting cases. An inverse association was observed for ever oral contraceptive use (ORindex = 0.72, 95% CI: 0.53–0.99, Table 4). Among ever users, there was no suggestive trend of decreasing odds ratios with longer duration of use, except when restricted to current users of oral contraceptives (β = −0.003 (95% CI: −0.005 – −0.001), p trend = 0.02, data not shown). Likewise, postmenopausal hormone use was inversely associated with gliomas (ORindex = 0.56, 95% CI: 0.37–0.85), particularly among those with surgical menopause (ORindex = 0.38, 95% CI: 0.13–1.07). The odds ratio for postmenopausal hormone use remained below 1.0 regardless of age at menopause. There was no evidence of linear trend with increasing duration of postmenopausal hormone use among ever users (p trend = 0.84). Among postmenopausal women, previous oral contraceptive use did not modify the inverse association between postmenopausal hormone use and gliomas (table 5). The combined effects of oral contraceptives and postmenopausal hormones were close to expectations for both additive (OR=0.38) and multiplicative (OR=0.46, p value > 0.10) models.
Table 4.
Frequencies and adjusted odds ratios (95% CI) of exogenous hormone use among glioma cases and controls, and self-reporting respondents only in the San Francisco Adult Glioma Study, 1991–2004
| Cases | ||||||||
|---|---|---|---|---|---|---|---|---|
| Exogenous hormone factors |
All women (n=619) |
Self-reporting women (n = 390) |
Controls (n=650) |
OR (95% CI)‡ All women |
OR (95% CI)‡ Self-reporting women only |
|||
| OC use | ||||||||
| Never | 284 | (48) | 133 | (34) | 244 | (38) | 1.00 (ref) | 1.00 (ref) |
| Ever | 312 | (52) | 257 | (66) | 405 | (62) | 0.62 (0.47–0.82) | 0.72 (0.53–0.99) |
| Missing | 23 | 0 | 1 | |||||
| OC use | ||||||||
| Never | 284 | (48) | 133 | (34) | 244 | (38) | 1.00 (ref) | 1.00 (ref) |
| Former | 290 | (49) | 238 | (61) | 352 | (54) | 0.65 (0.49–0.86) | 0.78 (0.57–1.07) |
| Current | 22 | (4) | 19 | (5) | 53 | (8) | 0.30 (0.16–0.60) | 0.26 (0.13–0.53) |
| Missing | 23 | 0 | 1 | |||||
| OC duration | ||||||||
| Never | 284 | (49) | 133 | (34) | 244 | (38) | 1.00 (ref) | 1.00 (ref) |
| <1 year | 38 | (7) | 37 | (10) | 87 | (13) | 0.35 (0.22–0.54) | 0.50 (0.31–0.80) |
| 1–4 years | 131 | (22) | 112 | (29) | 118 | (18) | 0.89 (0.63–1.25) | 1.08 (0.74–1.57) |
| 5–9 years | 67 | (12) | 56 | (15) | 86 | (13) | 0.62 (0.41–0.94) | 0.68 (0.44–1.06) |
| 10+ years | 62 | (11) | 48 | (12) | 114 | (18) | 0.44 (0.29–0.64) | 0.48 (0.31–0.74) |
| Missing | 37 | 4 | 1 | |||||
| PHT useω | ||||||||
| Never | 149 | (50) | 70 | (45) | 115 | (35) | 1.00 (ref) | 1.00 (ref) |
| Ever | 149 | (50) | 84 | (55) | 214 | (65) | 0.57 (0.41–0.79) | 0.56 (0.37–0.85) |
| Missing | 12 | 1 | 2 | |||||
| PHT useω | ||||||||
| Never | 149 | (50) | 70 | (45) | 115 | (34) | 1.00 (ref) | 1.00 (ref) |
| Former | 99 | (33) | 54 | (35) | 107 | (33) | 0.74 (0.51–1.08) | 0.78 (0.49–1.23) |
| Current | 50 | (17) | 30 | (19) | 107 | (33) | 0.38 (0.25–0.59) | 0.38 (0.22–0.65) |
| Missing | 12 | 1 | 2 | |||||
| PHT durationω | ||||||||
| Never | 149 | (53) | 70 | (46) | 115 | (35) | 1.00 (ref) | 1.00 (ref) |
| <1 year old | 11 | (4) | 6 | (4) | 29 | (9) | 0.27 (0.12–0.57) | 0.25 (0.09–0.70) |
| 1–4 years | 39 | (14) | 26 | (17) | 57 | (17) | 0.55 (0.33–0.89) | 0.57 (0.32–1.03) |
| 5–9 years | 23 | (8) | 15 | (10) | 31 | (9) | 0.61 (0.33–1.13) | 0.64 (0.31–1.30) |
| 10+ years old | 60 | (21) | 34 | (23) | 97 | (30) | 0.51 (0.34–0.77) | 0.57 (0.34–0.95) |
| Missing | 28 | 4 | 2 | |||||
adjusted for age (restricted quadratic splines), ethnicity and education
among postmenopausal women
Table 5.
Adjusted OR (95% CI) for interaction between oral contraceptive use and postmenopausal hormone use and gliomas among postmenopausal women, and self-reporting respondents only in the San Francisco Adult Glioma Study, 1991–2004
| Oral contraceptive use among self-reporting women |
Never postmenopausal hormone use among self-reporting women |
Ever postmenopausal hormone use among self-reporting women |
||
|---|---|---|---|---|
| Case/control | OR† (95% CI) | Case/control | OR† (95% CI) | |
| Never | 38/69 | 1.00 (ref) | 37/108 | 0.55 (0.30–0.98) |
| Ever | 32/45 | 0.83 (0.41–1.69) | 47/105 | 0.50 (0.25–0.98) |
| Oral contraceptive use among all women |
Never postmenopausal hormone use among all women |
Ever postmenopausal hormone use among all women |
||
| Case/control | OR† (95% CI) | Case/control | OR† (95% CI) | |
| Never | 102/70 | 1.00 (ref) | 84/108 | 0.57 (0.37–0.88) |
| Ever | 45/45 | 0.60 (0.33–1.10) | 61/106 | 0.36 (0.20–0.64) |
adjusted for frequency matched factors age (restricted quadratic splines), ethnicity and education
Exploratory analyses by smoking status and age at diagnosis did not uncover effect measure modification.
Discussion
In this large population based case-control study, we generally found no association between reproductive factors and gliomas. Possible exceptions include elevated odds ratios with older age at menarche and protective effects with exogenous hormone use (oral contraceptives and postmenopausal hormonal therapy). Previous studies exploring the relationship between older age at menarche and gliomas found similar results as ours (risk estimates ranged from 1.59 to 2.10).13,15,18–20 In contrast, we did not observe a difference in risk by menopausal status nor did we find support of a trend with increasing age at menarche as suggested in previous studies.15,20 Of the few studies stratifying by glioma subtype, ours is the first to demonstrate that older age at menarche is stronger among non-glioblastomas.
Our findings of an inverse association between gliomas and oral contraceptive use and postmenopausal hormone use need to be interpreted cautiously. Given that educational level has been inversely associated with gliomas, the protective effect we observed with use of exogenous hormones may be due to selection bias. If controls of higher education status had been more likely to participate, odds ratios may be overestimated since hormone use is more common among women earning higher education degrees. Nevertheless, our findings agree with most previous case-control studies.13,15,18,19,21,22 Only one prospective study reported hazard ratios slightly above the null for oral contraceptive use (HR = 1.04, 95% CI: 0.72–1.50),20 but it is difficult to directly compare their results to those of our case-control because self-reported hormone use was assessed only at baseline. Similar to other studies, there was generally no suggestive trend of decreasing odds ratios with increasing duration of exogenous hormone use.15,19–22 To our knowledge, this is the first study to examine potential interaction between oral contraceptives and postmenopausal hormone use. We found no evidence of synergism or antagonism between postmenopausal hormone use and previous oral contraceptive use and gliomas among self-reporting postmenopausal women. Whether these associations are causal or artifactual are unclear since women who use hormonal therapies may be generally healthier.30–31
Several studies have investigated the relation between parity and gliomas, but with conflicting results. Four of six studies reported parous women were at reduced risk of gliomas compared with nulliparous women. In the largest population based case control study to date (N=1657 glioma cases), Lambe et al. reported a 24% reduced risk in ever parous Swedish women compared to nulliparous women (OR = 0.76 95% CI: 0.66–0.87).12 A more recent but smaller hospital based case control study found similar, but non-significant, results (OR = 0.85 95% CI: 0.54 – 1.35),19 whereas odds ratio for parity were near the null in our study. These findings differ from a population-based study restricted to rural women in the upper Midwestern States (N=341 cases) which reported an elevated odds ratio (OR = 1.22 95% CI: 0.77–1.96) with ever giving birth and gliomas.15 Our results of no association with increasing number of children agree with most, but not all,12,19 previous studies.
No associations were detected for gravidity, age at first birth, or menopausal status. In addition, there was no clear pattern of association with age at menopause, even when excluding women with indeterminant or unknown menopausal status (that is, including only women reporting natural menopause or bilateral oopherectomy).
Given the evidence from animal studies of the protective effect of estrogen on gliomagenesis and survival, 32–36 the growing literature supporting the role of estrogen (and estrogen metabolites) as a neuro-protectant, particularly in estrogen receptor independent pathways,37–39 and observations of other female hormones known to restrict angiogenesis in vascular tumors40 like glioblastomas, it is reasonable to expect that greater hormonal exposure would be favorable. Findings from this study do not generally support our hypothesis that greater cumulative exposure to hormones decreases glioma risk. Earlier, rather than later, age at menopause was inversely associated with gliomas, albeit insignificantly. Earlier age at first birth among parous women was near the null and greater menstruation years did not decrease risk in this study of urban women. This is in contrast to findings of decreasing odds ratios with increasing number of menstruation months among rural women.15 Our definition of cumulative exposure (menstruation years) relied on varying definitions, all of which were assessed, but all without information about menstrual regularity. Odds ratios for menstruation years may be overestimated if cases reporting earlier age at menarche and later age at menopause had sporadic menstrual cycles compared to controls. The inconsistency between definitions of cumulative exposure suggests more robust measures are needed to tease out potential differences in effect between exogenous and endogenous hormones.
Although this is one of the largest case-control studies examining reproductive factors and hormonal use in gliomas, there were still important limitations that may affect interpretation of our results. First, delays in interviewing cases may have resulted in a greater amount of exposure misclassification among cases since gliomas can rapidly affect memory and mental functioning. Rapid case ascertainment strategies used in this study likely minimized this potential bias. In addition, non-differential misclassification may have occurred with age at menopause since the menopausal transition makes it difficult to precisely recall age at menopause. Excluding proxies as a strategy to avoid information bias is not optimal since exclusion may create selection bias.41 Second, controls selected using random digit dialing (RDD) may have created selection bias if responders differed systematically from non-responders.42 Since we do not know more about non-responders, it is difficult to assess the magnitude and direction of this bias. If non-response was driven by telecommunication advances (more cell phone use during series 3 than series 1), we might have expected the odds ratios to differ by series, which we did not. Third, our study lacked detailed information on the hormonal composition of exogenous hormones used and time since last use in order to account for appropriate latency periods. Finally, reproductive factors may be too broad to capture the ‘windows’ of relevant exposure when one hormone has more impact than another. Exploring such a hypothesis would benefit from the additional consideration of testosterone, for which there is currently no good surrogate. Understanding the independent function of testosterone, estrogen, progesterone in the nervous system is complex due to their apparent dual functions. The same hormone may create a different response in males than females and this difference may vary from one brain region to another 43–44 adding to the complexity of which hormone (or more likely which group of hormones) would be largely responsible for lower incidence of gliomas in females.
Strengths of this present study were the relatively large number of histologically confirmed glioma cases and the availability of more detailed information to detect subtle differences in the effect of potential cumulative hormonal exposure. However, we cannot rule out the possibility that the few notable findings were due to other explanations, such as exposure misclassification and unknown or uncontrolled confounders including demographic or social correlates of the reproductive experience.
This study is an important addition to existing literature regarding reproductive factors and gliomas. We provided some of the first results comparing reproductive factors with glioma histologic subtype. Although we observed no overall association with reproductive factors in this study, there was a suggestion of increased risk with older age at menarche, consistent with several previous studies. Our study provided evidence that this positive association is stronger in non-glioblastomas. We acknowledge that non-glioblastomas are still a heterogenous group of tumors, but these results may give insight into differences in risk factors by glioma subtype and impress the need for larger studies (or combining existing studies) in order to garner adequate power for subtype-specific analysis. Furthermore, this study used cumulative exposure as the important next step toward moving beyond assessing reproductive events that are relevant to one point in time. Cumulative exposure, as defined multiple ways in our study, does not appear to be an important in glioma risk, contrary to an observed protective effect among rural women. Given the consistent protective effect with oral contraceptive and postmenopausal hormone use across previous studies, this study was the first to assess whether a potential cumulative or synergistic effect may be relegated to exogenous hormone use only. Future epidemiologic studies could benefit from a more detailed account of exogenous hormone, but without a greater understanding of hormone function in the nervous system interpretation will remain limited. Potentially as informative would be developing a grouping classification system of reproductive factors that could serve as meaningful indices of hormone-specific aggregates to more closely resemble their physiologic function in the body. Future epidemiologic studies could benefit from a more detailed account of exogenous hormone use to confirm findings, but without a greater understanding of hormone function in the nervous system in relation to mechanisms of gliomagenesis interpretation will remain limited. Potentially as informative would be developing a grouping classification system of reproductive factors that could serve as meaningful indices of hormone-specific aggregates (which more closely resembles their physiologic function in the body).
Acknowledgements
This research was supported in part by a grant from the National Institutes of Health: CA52689, CA097257, CA89032, ES06717, EX04705, and P30ES10126. Additional funding was provided by a gift from the Robert J. and Helen H. Glaser Family Foundation. The authors would like to thank Lucie McCoy for her technical assistance with the data.
Abbreviations used are as follows
- GBM
glioblastoma
- OR
odds ratio
References
- 1.Martinez-Cerdeno V, Noctor S, Kriegstein A. Estradiol stimulates progenitor cell division in the ventricular and subventricular zones of the embryonic neocortex. Eur J Neurosci. 2006;24:3475–3488. doi: 10.1111/j.1460-9568.2006.05239.x. [DOI] [PubMed] [Google Scholar]
- 2.Kuppers E, Ivanova T, Karolczak M, Lazarov N, Fohr K, Beyer C. Classical and nonclassical estrogen action in the developing midbrain. Horm Behav. 2001;40:196–202. doi: 10.1006/hbeh.2001.1671. [DOI] [PubMed] [Google Scholar]
- 3.Marin R, Guerra B, Alonso R, Ramirez CM, Diaz M. Estrogen activates classical and alternative mechanisms to orchestrate neuroprotection. Curr Neurovasc Res. 2005;2:287–301. doi: 10.2174/156720205774322629. [DOI] [PubMed] [Google Scholar]
- 4.Wright DW, Kellermann AL, Hertzberg VS, Clark PL, Frankel M. ProTECT: A Randomized Clinical Trial of Progesterone for Acute Traumatic Brain Injury. Ann Emerg Med. 2007;49:391–402. doi: 10.1016/j.annemergmed.2006.07.932. [DOI] [PubMed] [Google Scholar]
- 5.Chamaon K, Stojek J, Kanakis D, et al. Micromolar concentrations of 2-methozyestradiol kill glioma cells by an apoptotic mechanism, without destroying their microtubule cytoskeleton. J Neurooncol. 2005;72:11–16. doi: 10.1007/s11060-004-2158-4. [DOI] [PubMed] [Google Scholar]
- 6.Lis A, Ciesielski MJ, Barone TA, Scott BE, Fenstermaker RA, Plunkett RJ. 2-methoxyestradiol inhibits proliferation of normal and neoplastic glial cells, and induces cell death, in vitro. Cancer Lett. 2004;213:57–65. doi: 10.1016/j.canlet.2004.04.021. [DOI] [PubMed] [Google Scholar]
- 7.Kang SH, Cho HT, Devi S, et al. Antitumor effect of 2-methoxyestradiol in a rat orthotopic brain tumor model. Cancer Res. 2006;66:11991–11997. doi: 10.1158/0008-5472.CAN-06-1320. [DOI] [PubMed] [Google Scholar]
- 8.Dhandapani KM, Wade FM, Mahesh VB, Brann DW. Astrocyte-derived transforming growth factor- β mediates the neuroprotective effects of 17 β-estradiol: involvement of nonclassical genomic signaling pathways. Endocrinology. 2005;146:2749–2759. doi: 10.1210/en.2005-0014. [DOI] [PubMed] [Google Scholar]
- 9.Behl C. Oestrogen as a nueroprotective hormone. Nat Rev Neruosci. 2002;3:433–442. doi: 10.1038/nrn846. [DOI] [PubMed] [Google Scholar]
- 10.Yague JG, Lavaque E, Carretero J, Azcoitia I, Garcia-Segura LM. Aromatase, the enzyme responsible for estrogen biosynthesis, is expressed by human and rat glioblastomas. Neurosci Lett. 2004;368:279–284. doi: 10.1016/j.neulet.2004.07.010. [DOI] [PubMed] [Google Scholar]
- 11.Gonzales A, Martinez-Campa C, Mediavilla MD, et al. Inhibitory effects of pharmacological doses of melatonin on aromatase activity and expression in rat glioma cells. Br J Cancer. 2007;97:755–760. doi: 10.1038/sj.bjc.6603935. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lambe M, Coogan P, Baron J. Reproductive factors and the risk of brain tumors: a population based study in Sweden. Int J Cancer. 1997;72:389–393. doi: 10.1002/(sici)1097-0215(19970729)72:3<389::aid-ijc2>3.0.co;2-l. [DOI] [PubMed] [Google Scholar]
- 13.Schlehofer B, Blettner M, Preston-Martin S, et al. Results from the International Adult Brain Tumor Study. Role of Medical History in Brain Tumor Development. Int J Cancer. 1999;82:155–160. doi: 10.1002/(sici)1097-0215(19990719)82:2<155::aid-ijc1>3.0.co;2-p. [DOI] [PubMed] [Google Scholar]
- 14.Ryan P, Lee M, North B, McMichael A. Risk Factors for Tumors of the brain and meninges: Results from the Adelaide adult brain tumor study. Int J Cancer. 1992;51:20–27. doi: 10.1002/ijc.2910510105. [DOI] [PubMed] [Google Scholar]
- 15.Huang K, Whelan EA, Ruder AM, et al. Reproductive factors and risk of glioma in women. Cancer Epidemiol Biomarkers Prev. 2004;13:1583–1588. [PubMed] [Google Scholar]
- 16.Cantor KP, Lynch CF, Johnson D. Reproductive factors and risk of brain, colon, and other malagnancies in Iowa. Cancer Causes Control. 1993;4:505–511. doi: 10.1007/BF00052425. [DOI] [PubMed] [Google Scholar]
- 17.Cicuttini FM, Hurley SF, Forbes A, et al. Association of Adult glioma with medical conditions, family and reproductive history. Int J Cancer. 1997;71:203–207. doi: 10.1002/(sici)1097-0215(19970410)71:2<203::aid-ijc13>3.0.co;2-i. [DOI] [PubMed] [Google Scholar]
- 18.Hochberg F, Toniolo P, Cole P. Nonoccupational risk indicators of glioblastoma in adults. J Neurooncol. 1990;8:55–60. doi: 10.1007/BF00182087. [DOI] [PubMed] [Google Scholar]
- 19.Hatch EE, Linet MS, Zhang J, et al. Reproductive and hormonal factors and risk of brain tumors in adult females. Int J Cancer. 2005;114:797–805. doi: 10.1002/ijc.20776. [DOI] [PubMed] [Google Scholar]
- 20.Silvera SA, Miller AB, Rohan TE. Hormonal and reproductive factors and risk of glioma: a prospective cohort study. Int J Cancer. 2006;118:1321–1324. doi: 10.1002/ijc.21467. [DOI] [PubMed] [Google Scholar]
- 21.Wigertz A, Lonn S, Mathiesen T, et al. Risk of brain tumors associated with exposure to exogenous female sex hormones. Am J Epidemiol. 2006;164:629–636. doi: 10.1093/aje/kwj254. [DOI] [PubMed] [Google Scholar]
- 22.Benson VS, Pirie K, Green J, Casabonne D, Beral V. Lifestyle factors and primary glioma and meningioma tumours in the Million Women Study cohort. British J Cancer. 2008;99:185–190. doi: 10.1038/sj.bjc.6604445. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Wrensch M, Lee M, Miike R, et al. Familial and personal medical history of cancer and nervous system conditions among adults with gliomas and controls. Am J Epidemiol. 1997;145:581–593. doi: 10.1093/oxfordjournals.aje.a009154. [DOI] [PubMed] [Google Scholar]
- 24.Wiemels JL, Wiencke JK, Sison JD, Miike R, McMillan A, Wrensch M. History of allergies among adults with glioma and controls. Int J Cancer. 2002;98:609–615. doi: 10.1002/ijc.10239. [DOI] [PubMed] [Google Scholar]
- 25.Krishnan G, Felini M, Carozza SE, Miike R, Chew T, Wrensch M. Occupation and adult gliomas in the San Francisco Bay Area. J Occ Environ Med. 2003;45:639–647. doi: 10.1097/01.jom.0000069245.06498.48. [DOI] [PubMed] [Google Scholar]
- 26.Geerlings MI, Muitenberg A, Witteman JC, et al. Reproductive period and risk of dementia in postmenopausal women. JAMA. 2001;285:1475–1481. doi: 10.1001/jama.285.11.1475. [DOI] [PubMed] [Google Scholar]
- 27.de Kleijn MJ, van der Schouw YT, Verbeek AL, Peeter PH, Banga JD, van der Graaf Y. Endogenous estrogen exposure and cardiovascular mortality risk in postmenopausal women. Am J Epidemiol. 2002;155:339–345. doi: 10.1093/aje/155.4.339. [DOI] [PubMed] [Google Scholar]
- 28.Jansen SC, Temme E, Scouten EG. Lifetime estrogen exposure versus age at menopause as a mortality predictor. Maturitas. 2002;43:105–112. doi: 10.1016/s0378-5122(02)00183-4. [DOI] [PubMed] [Google Scholar]
- 29.Robins JM. Data, design, and background knowledge in etiologic inference. Epidemiology. 2001;12:313–320. doi: 10.1097/00001648-200105000-00011. [DOI] [PubMed] [Google Scholar]
- 30.Cauley JA, Cummings SR, Black DM, Mascioli SR, Seeley DG. Prevalence and determinants of estrogen replacement therapy in elderly women. Am J Obstet Gynecol. 1990;163:1438–1444. doi: 10.1016/0002-9378(90)90602-4. [DOI] [PubMed] [Google Scholar]
- 31.Derby CA, Hume AL, McFarland M, McPhillips JB, Lasater TM, Carleton RA. Correlates of postmenopausal estrogen use and trends through the 1980s in two southeastern New England communities. Am J Epidemiol. 1993;137:1125–1135. doi: 10.1093/oxfordjournals.aje.a116616. [DOI] [PubMed] [Google Scholar]
- 32.Jung-Testas I, Baulieu EE. Steroid hormone receptors and steroid action in rat glial cells of the central and peripheral nervous system. J Steroid Biochem Mol Biol. 1998;65:243–251. doi: 10.1016/s0960-0760(97)00191-x. [DOI] [PubMed] [Google Scholar]
- 33.Hopewell JW. The effects of castration on the induction of experimental gliomas in male rats. Br J Cancer. 1970;24:187–190. doi: 10.1038/bjc.1970.21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Verzat C, Delisle M, Courrierre P, Hollande E. Influence of host sex on growth of a human glioblastoma line in athymic mice. Neuropathol Appl Neurobiol. 1990;16:141–151. doi: 10.1111/j.1365-2990.1990.tb00943.x. [DOI] [PubMed] [Google Scholar]
- 35.Plunkett RJ, Lis A, Barone TA, Fronckowiak MD, Greenberg SJ. Hormonal effects on glioblastoma multiforme in the nude rat model. J Neurosurg. 1999;90:1072–1077. doi: 10.3171/jns.1999.90.6.1072. [DOI] [PubMed] [Google Scholar]
- 36.Avtsyn AP, Yablonovskaya LY. Effects of disturbances in the hormonal status on experimental brain tumors. Acta Unio Int Contra Cancrum. 1964;20:1519–1522. [PubMed] [Google Scholar]
- 37.Klauber N, Parangi S, Flynn E, Hamel E, D’Amato RJ. Inhibition of angiogenesis and breast cancer in mice by the microtubule inhibitors 2-methoxyestradiol and taxol. Cancer Res. 1997;57:81–86. [PubMed] [Google Scholar]
- 38.Fotsis T, Zhang Y, Pepper MS, et al. The endogenous oestrogen metabolite 2-methoxyestradiol inhibits angiogenesis and suppresses tumor growth. Nature. 1994;368:237–239. doi: 10.1038/368237a0. [DOI] [PubMed] [Google Scholar]
- 39.LaVallee TM, Zhan XH, Herbstritt CJ, Kough EC, Green SJ, Pribluda VS. 2-methoxydestradiol inhibits proliferation and induces apoptosis independently of estrogen receptors alpha and beta. Cancer Res. 2002;62:3691–3697. [PubMed] [Google Scholar]
- 40.Bengtson NW, Linzer DI. Inhibition of tumor growth by the antiangiogenic placental hormone proliferin-related protein. Mol Endocrinol. 2000;14(12):1934–1943. doi: 10.1210/mend.14.12.0573. [DOI] [PubMed] [Google Scholar]
- 41.Greenland S, Rothman K. Fundamentals of Epidemiologic Data Analysis; Philadelphia, PA: Lippincott-Raven Publishers; 1998. pp. 253–279. [Google Scholar]
- 42.Bunin G, Spector L, Olshan A, et al. Secular Trends in Response Rates for Controls Selected by Random Digit Dialing in Childhood Cancer Studies: A Report from the Children’s Oncology Group. Am J Epidemiology. 2007;166(1):109–116. doi: 10.1093/aje/kwm050. [DOI] [PubMed] [Google Scholar]
- 43.Dhandapani KM, Brann D. Estrogen-astrocyte interactions: implications for neuroprotection. BMC Neuroscience. 2002;3:6–10. doi: 10.1186/1471-2202-3-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.McEwen BS, Alves SE, Bulloch K, Weiland NG. Clinically relevant basic science studies of gender differences and sex hormone effects. Psychopharmacol Bull. 1998;34:251–259. [PubMed] [Google Scholar]