Reproductive and hormonal factors and the risk of lung cancer: The EAGLE study
Evidence about the role for reproductive and hormonal factors in the etiology of lung cancer in women is conflicting. To clarify this question, we examined 407 female cases and 499 female controls from the Environment And Genetics in Lung cancer Etiology population-based case–control study. Subjects were interviewed in person using a computer-assisted personal interview to assess demographics, education, smoking history, medical history, occupational history, reproductive and hormonal factors. Associations of interest were investigated using logistic regression models, adjusted for catchment area and age (matching variables), cigarette smoking (status, pack-years and time since quitting). Additional confounding variables were investigated but did not substantially affect the results. We observed a reduced risk of lung cancer among women with later age at first live birth [≥31 years: odds ratio (OR) = 0.57, 95% confidence interval (CI) = 0.31–1.06, p-trend = 0.05], later age at menopause (≥51 years: OR = 0.49, 95%CI = 0.31–0.79, p-trend = 0.003) and longer reproductive periods (≥41 years: OR = 0.44, 95%CI = 0.25–0.79, p-trend = 0.01). A reduced risk was also observed for hormone replacement therapy (OR = 0.63, 95%CI = 0.42–0.95, p = 0.03) and oral contraceptive use (OR = 0.67, 95%CI = 0.45–1.00, p = 0.05) but no trend with duration of use was detected. Menopausal status (both natural and induced) was associated with an augmented risk. No additional associations were identified for other reproductive variables. This study suggests that women who continue to produce estrogens have a lower lung cancer risk. Large studies with great number of never smoking women, biomarkers of estrogen and molecular classification of lung cancer are needed for a more comprehensive view of the association between reproductive factors and lung cancer risk.
Lung cancer is the leading cause of cancer mortality worldwide. In the last three decades, its risk is dramatically increased in women, for whom is the fourth most common tumor (513,000 cases, 8.5% of all cancers) and the second in number of deaths (427,000 deaths, 12.8% of the total).1 In contrast, the risk has leveled off or decreased in men. Although different patterns of cigarette smoking over time between the two sexes is likely to account for most of the difference, the hypothesis of a higher female susceptibility to tobacco-related lung cancer has emerged, with conflicting epidemiological results.2 Still, there is consensus that important biological differences between the sexes exist in nonsmall-lung cancer risk. In particular, women are more likely to develop adenocarcinomas and experience younger age at onset.3 Moreover, a reduced DNA repair capacity has been showed in females tumors compared to men and the hypothesis that estrogens can play a role in lung cancer carcinogenesis has gained attention.4 Estrogen effects are likely to be mediated by the estrogen receptors (ER) α and β. ER α mRNA are expressed at low levels in the lung, whereas ERβ, the less active in inducing transcriptional activation, has been shown to be expressed in both normal and tumor pulmonary tissue.5–7 Estrogens could also interact with cigarette smoking by accelerating the metabolism of smoking-derived carcinogens.8
The first clue of a potential role for estrogens and reproductive factors came from a study of women in China that showed increased lung cancer risk with shorter menstrual cycle length.9 Afterwards, numerous observational studies examined the relationship between lung cancer and reproductive factors including: parity, age at first live birth, age at menarche and age at menopause.10–25 However, the findings across studies are far from being consistent. Most studies are small and examined only a few reproductive factors. The three most recent cohort studies, the Shanghai Women's Health Study,19 the Nurses' Health Study24 and the NIH-AARP Diet and Health Study,10 suggested a possible protective role of endogenous estrogens showing reduced lung cancer risk with later age at menopause10, 19, 24 and longer reproductive periods.19
In a small hospital-based case–control study, Taioli and Wynder,18 showed a positive association between hormone replacement therapy (HRT) and risk of lung cancer, but most subsequent case–control12, 14, 15, 26–30 and cohort studies10, 11, 13, 16, 19, 24, 31–35 did not support this hypothesis. Also investigations on oral contraceptive (OC) use and lung cancer risk failed to show a clear association.11, 12, 14–16, 18, 19, 24, 28, 36, 37
The purpose of our study was to examine whether reproductive or hormonal factors are associated with the risk of lung cancer in a population-based case–control study conducted in Italy, with a relatively large number of female cases and population controls.
Material and Methods
Environment And Genetics in Lung cancer Etiology (EAGLE) is a population based case–control study primarily designed to investigate the genetic and environmental determinants of lung cancer. The study protocol has been extensively described elsewhere.38 Incident lung cancer cases (2,100) and 2,120 population controls were enrolled between 2002 and 2005 from a catchment area including 216 municipalities in the Lombardy region of Italy. Lung cancer cases were enrolled in 13 hospitals and population controls were randomly sampled from the catchment area to frequency match the cases by residence, gender and age. Subjects were 35–79 years at diagnosis or at sampling. Participation rates were 86.6% for cases and 72.4% for controls. Extensive clinical data were collected for lung cancer cases including morphology coded according to the International Classification of Diseases for Oncology, Third Edition.39 The study was approved by the institutional review boards and all participants signed a written informed consent. Our study population includes the EAGLE females study subjects (448 cases and 500 controls).
All cases and controls underwent a computer-assisted personal interview and completed a self-administered questionnaire (both questionnaires are available on the EAGLE website (http://eagle.cancer.gov/). The interview included questions on demographics, education, current and past smoking behavior, environmental tobacco smoke (ETS) at home (during childhood and adulthood) and at the workplace, alcohol consumption, medical history, family history of cancer and occupational history. For women, the questionnaire also included a detailed section on reproductive history. Information on age at menarche, age at menopause, average length of menstrual cycle, age at first pregnancy, number of full-term pregnancies, breast-feeding, and history of ovariectomy and use of OC was recorded. Menopausal women were asked on the type of menopause (natural, or induced by surgery, radiation or chemotherapy), on age at start and duration, and on use of HRT. No information was collected on type of hormones or dosage. The self-administered questionnaire investigated behavioral aspects related to smoking habits and diet.
Odds ratios (ORs), 95% confidence intervals (CI) and tests for trend were calculated using unconditional logistic regression. All models included the matching variables: area (five categories including five large cities and their surrounding municipalities) and age (5-year categories). Additional potential confounders or effect modifiers examined in the models were: smoking status (ever/never); pack-years (continuous, mean-centered: linear, quadratic, and cubic terms); time since quitting (0 for never/current smokers, 0.5, 1, 2, 5, 10, 20, ≥30 years),40 education, body mass index (BMI) and ETS at home (during childhood and/or adulthood) and at work. Risk estimates were little affected by the inclusion of other covariates such as alcohol consumption, occupational exposure to lung carcinogens, personal history of other cancers and family history of lung cancer, so they were not included in the final model.
Several reproductive characteristics were investigated as independent variables: age at menarche (<13, 13–14, ≥15 years), cycle duration (<27, 27–28, ≥29 days), parity (0, 1, 2, ≥3 live-births), age at first live-birth (<22, 22–25, 26–30, ≥31 years), breastfeeding (ever/never), menopausal status (no, natural menopause, induced menopause), age at menopause (<46, 46–50, ≥51 years), ovariectomy (both ovaries removed yes/no), OC and HRT (ever, never and duration of use in years). We also calculated duration of the reproductive period—time from age at menarche to age at menopause (if in menopause) or from age at menarche to age at diagnosis/enrollment (if not in menopause)—and duration of menopause—time between age at menopause and age at diagnosis/enrollment.
When adequate, we also ran all models including the independent variables as continuous.
To evaluate interaction between reproductive variables and smoking or BMI, we compared the log-likelihood of models with and without product terms.
Multinomial logistic regression models were performed to test the homogeneity of the association between reproductive variables and lung cancer risk across histological types.
All tests were two-sided. Statistical analyses were performed using Stata 11 (Stata Corporation, College Station, TX).
Out of 448 female cases and 500 female controls enrolled in the study, interviews or questionnaire information were available for 407 cases (90.8%) and 499 controls (99.8%; Table 1). Cases and controls did not differ for residential area, and age (the effect of matching), education, ETS or alcohol consumption, and occupational exposure to known (list A) or suspected (list B) lung carcinogens according to the International Agency for Research on Cancer evaluations.40 Almost half of the cases (46%) were current smokers compared with 21% of the controls; cases had smoked a greater number of cigarettes. The proportion of subjects underweight was higher among cases and that of obese higher in controls. Cases had a higher frequency of primary cancer other than lung cancer and reported more frequently a positive family history of lung cancer. As expected, the majority of lung cancers were adenocarcinomas.
Table 1. Distribution of selected characteristics for cases and controls, EAGLE study (2002–2005)1
Table 2 shows associations between lung cancer and each exposure variable. After adjusting for area, age and smoking, reduced risk of lung cancer was observed among women with later age at first live birth (OR = 0.57; 95% CI 0.31–1.06 for ≥31 vs. <22 years, p-trend = 0.05), later age at menopause (OR = 0.49; 95% CI 0.31–0.79 for ≥51 vs. <46 years, p-trend = 0.003) and longer reproductive duration (OR = 0.44; 95% CI 0.25–0.79 for ≥41 vs. <33 years, p-trend = 0.01). Menopausal status (both natural and induced) was associated with an augmented risk which increased with menopause duration (p-trend = 0.002); exclusion of women who declared to be in surgical menopause but not to have undergone bilateral ovariectomy (thus, potentially, maintaining their ovarian function unaltered) did not substantially modify the effect estimates (i.e., about 5%, data not shown).
Table 2. Adjusted odds ratios (OR) and 95% confidence intervals (95%CI) for lung cancer by reproductive and hormonal factors1
HRT and the use of OCs were associated with a reduced lung cancer risk (OR = 0.63; 95% CI: 0.42–0.95 and OR = 0.67; 95 CI%: 0.45–1.00, respectively). The risk did not substantially change among women with longer duration of HRT or OC use.
Lung cancer risk was not associated with age at menarche, cycle duration, breastfeeding and parity.
Further adjustment for passive smoking, education and BMI did not substantially change the risk estimates of each exposure variable.
We also examined the results separately for never and ever smokers (current and former) and we found no evidence of interaction between the reproductive variables and smoking status. Although based on small numbers, the direction of the associations was similar in the two groups for most of the examined variables (Supporting Information, Table S1).
Because the residual sources of estrogen production in menopausal women are related to the production of estrone from androstenedione in adipose tissue, the risks for menopause status and menopause duration were separately examined for women with BMI < 25 and ≥ 25. OR estimates did not differ (p for interaction = 0.68 and 0.93, respectively; Supporting Information, Table S2)
We also examined the increasing risk of lung cancer with menopause duration separately in never and ever users of HRT: the pattern of increasing risk did not differ (p for interaction = 0.33; Supporting Information, Table S3). In the same way, when we considered lung cancer risk for reproductive duration stratified by OC use (ever, never) a similar decreasing risk was observed in the two groups (p for interaction = 0.79).
Similar OR patterns were observed across histological types (adenocarcinoma vs. other morphologies, Table 3).
Table 3. Adjusted odds ratios (OR) and 95% confidence intervals (95%CI) for lung cancer by histological type1
In a population-based case–control study conducted in Italy (EAGLE), endogenous hormones during premenopausal years appear to have a protective role in lung cancer development. Specifically, we found a reduced risk of lung cancer with later age at menopause (and, correspondingly, increasing risk with menopause), later age at first live birth and longer reproductive period. These findings were consistent across the major lung cancer histology groups. In contrast, lung cancer risk was not associated with age at menarche, cycle duration, breastfeeding or parity. These findings suggest a complex pattern of hormone-related factors in association with lung cancer risk.
Our results on the protective effect of later menopause or longer reproductive periods, support previous findings from large prospective studies (the Nurses' Health study and the NIH-AARP Diet and Health Study, respectively),24, 10 which found increased risk of lung cancer with younger age at menopause. Similarly, a prospective study in lifetime nonsmokers Chinese women showed an inverse association between age at menopause and the risk of lung cancer.19 Similar results, although not statistically significant, were observed in three case–control studies from Canada,25 Czechoslovakia,21 and China.41 However, other studies showed positive associations with lung cancer risk9, 18, 20, 22 or null results.12–16
Age at first live birth was not associated with lung cancer risk in most studies.13–15, 17–19, 25 A few22, 24 reported an increased risk with older age at first live birth, whereas others11, 12 reported an inverse association.
The only study which examined the length of the reproductive period19 showed a pattern consistent with ours.
The use of OC and HRT in our population was associated with a reduced lung cancer risk; however; no trend in risk was found with increasing duration of use. Observational studies that investigated the association between HRT and lung cancer risk have shown mixed results. After the increased risk reported by Taioli and Wynder,18 only one study found a doubled risk limited to induced menopausal women treated with HRT.13 No effect,11, 14, 16, 19, 24, 28, 33 or reduced risk were found in other studies.12, 15, 29, 30, 34, 42 A recent review of 18 studies (15 observational studies and three randomized clinical trials) found mixed results;43 however, the authors cautioned the interpretation of these findings because most studies did not report smoking status and HRT type. The findings of another meta-analysis of 11 studies published between 1968 and April 2008 (eight case–control studies and three cohort studies)44 do not support the hypothesis of an association between HRT and lung cancer risk. Slatore et al.35 examined a large cohort of menopausal women and found that the use of combined (estrogen + progestin) hormone therapy was associated with an increased risk of lung cancer in a duration-dependent manner with a 50% increase in women using HRT for a period longer than 10 years. In the Women's Health Initiative randomized controlled trial study, the use of estrogen alone was not associated with increased incidence or mortality from lung cancer,32 whereas estrogen plus progestin therapy was associated with an increased mortality, but not incidence, from lung cancer.31 In our study population, we did not collect information on HRT type and therefore cannot address this issue.
With regard to OC use, most observational studies found no association with lung cancer risk.11, 14–16, 18, 19, 24, 28 Kreuzer et al.12 observed a reduction in lung cancer risk, but no trend in risk with increasing duration as in our study. On the contrary, in the Nurses' Health Study24 duration of OC use longer than 5 years was associated with a slightly increased risk. Two other large cohort studies in UK, the Royal College of General Practitioners cohort study and the Oxford FPA cohort study, compared long-term mortality in ever versus never-users of OC and did not find major differences across the two groups.36, 37
We found no association with cycle duration, breastfeeding or parity. These findings appear in contrast with the quite consistent pattern of an inverse association of estrogen-related variables with lung cancer risk in our study. A possible explanation is that reproductive factors may not be consistently associated with estrogen levels. A recent study in premenopausal women reported no association between 15 different estrogens measured in the urine and age at menarche, OC use, parity or breastfeeding,45 but showed a significant association with menstrual cycle length and regularity and age at first birth. This suggests a complex association between estrogen levels and reproductive factors. Moreover, the effects of these hormones are likely to be tissue-specific, as the different patterns of association between breast and lung cancer risk seem to indicate.46
Studies on the role of ER expression in lung cancer have been inconclusive with high levels of ER expression in some studies, but very low levels in others.47, 48 Schwartz et al.15 found that postmenopausal hormone use was associated with reduced lung cancer risk only in nonsmall cell lung cancers characterized as ER α and/or β positive. In addition, Chlebowski et al.32 hypothesized that the stimulation of ER alone is not sufficient to increase lung cancer growth but stimulation of progesterone receptor is also needed. Additional evidences suggest that estrogens can interact with growth factors, in particular epidermal growth factor (EGF). Although EGF is known to be involved in cell growth, protection from apoptosis and angiogenesis, it is possible that ER and EGF pathways are alternatively activated, because EGFR protein expression was shown to be down-regulated in response to estrogens.49 It is also possible that specific subgroups of lung cancer may be differentially associated with estrogens, as it appears for ER positive versus ER negative or basal-like breast cancer types.50 Our results were consistent across histology groups of lung cancer, but more subtle, molecularly related subgroups may be present and account for some of these differences.
Moreover, different study designs, modest sample sizes and populations with different ethnicities, life styles, diet or proportion of smokers could have contributed to the disparate findings across the studies. Our study is a population-based case–control investigation conducted in Caucasians, with life style and diet typical of a Western population. Confounding by smoking is often raised as a possible explanation of the inconsistent findings. Although the analyses in never smoking women in our study were limited by the small sample size, EAGLE benefited from comprehensive and detailed data on smoking, which allowed for tight adjustment by smoking status, intensity, duration and time since quitting. Although residual confounding by smoking can never be completely ruled out, it is unlikely that smoking can substantially explain our results.
Additional strengths of our study were a quite large sample size as a whole, the high participation rates, which reduces the likelihood of selection bias, and the detailed and comprehensive assessment of reproductive factors and major confounding variables. Study limitations include lack of data on hormone therapy type. In addition, analyses by smoking status were limited by the small number of cases in each category.
In conclusion, in a population-based study of lung cancer, we found that women who continue to produce endogenous hormones have a lower lung cancer risk. Larger studies (or pooled datasets) with rigorously defined reproductive variables, large number of never smoking women, biomarkers of estrogen and molecular classification of lung cancer are needed for a more comprehensive view of the association between reproductive factors and lung cancer risk.
The authors express their gratitude to all the EAGLE study participants and the collaborators (listed on the EAGLE website at http://eagle.cancer.gov/), whose commitment made this study possible.