Reproductive and hormonal factors and the risk of nonsmall cell lung cancer

Abstract

Although exposure to estrogen may directly influence or modify the association between cigarette smoking and lung cancer risk, results from epidemiologic studies examining the association between reproductive and hormonal factors and risk of lung cancer among women have been inconsistent. Between 1998 and 2008, 430 women diagnosed with nonsmall cell lung cancer, 316 hospital controls and 295 population controls were recruited into the multi-center Maryland Lung Cancer Study. Conditional logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) according to reproductive and hormonal exposures adjusting for age, smoking, passive smoking, education and household income. Results were similar for hospital and population based controls, so the control groups were combined. Reduced risks of lung cancer were observed among women with greater parity (≥5 vs. 1–2 births: OR = 0.50, 95% CI = 0.32, 0.78, p-trend = 0.002) and later ages at last birth (≥30 vs. <25 years old: OR = 0.68, 95% CI = 0.48, 0.98, p-trend = 0.04). After mutual adjustment parity, but not age at last birth, remained significantly inversely associated with risk (p-trend = 0.01). No associations were found for nonsmall cell lung cancer risk with age at menarche, age at first birth, menopausal status, oral contraceptive use or menopausal hormone use, including use of oral estrogens. Compatible with findings from recent epidemiologic studies, we observed a reduction in the risk of nonsmall cell lung cancer with increasing number of births. Other reproductive and hormonal exposures, including menopausal hormone therapy use, were not associated with risk.

Although male lung cancer incidence rates have been declining for the past two decades in the United States, the increasing incidence rates for females have only begun to stabilize in recent years.1 The gender-specific differences in lung cancer incidence and mortality are mostly thought to be attributable to secular trends in cigarette smoking.1 It has been hypothesized that women may have a greater susceptibility than men to lung cancer for the same amount of cigarette smoking2, 3; however, not all studies support this hypothesis.4, 5 Differences in lung cancer susceptibility by gender may depend, in part, on exposure to estrogen. Estrogen may play a direct role in lung tumorigenesis, as functional estrogen receptors are present in normal human lung tissue and lung tumors6 and could interact with cigarette smoking on the risk of lung cancer by accelerating the metabolism of cigarette smoking-derived carcinogens.7

Results from observational studies have not yielded consistent findings for reproductive and hormonal factors, such as number of live births,8–18 age at first birth,9, 11, 13–17 age at menarche,8, 9, 11–19 age at menopause8, 9, 12–15, 17–21, use of oral contraceptives (OCs),8, 10–12, 14, 15, 21 and use of menopausal hormone therapy8–15, 19, 22–29 and lung cancer in women. Some of these inconsistencies may reflect residual confounding by active or passive cigarette smoking or chance associations due to small sample sizes. Few studies had detailed reproductive and hormonal exposure information, such as combined estrogen plus progesterone versus estrogen-only menopausal hormone therapy.14

The purpose of our study was to examine whether reproductive or hormonal factors are associated with risk of nonsmall cell lung cancer in a case-control study conducted in Baltimore, MD, which includes a relatively large number of African-American and Caucasian female lung cancer cases (n = 430) and two different comparison groups (316 hospital-based and 295 population-based controls). Detailed information on lung cancer risk factors and reproductive history, including exogenous hormone use, has been collected in direct interviews with study participants.

Material and Methods

Study population

The Maryland Lung Cancer Study is a case-control study that was primarily designed to examine the associations between molecular and genetic factors and risk of nonsmall cell lung cancer.30, 31 Eligibility for our study was restricted to English-speaking, non-institutionalized, US-born African Americans and Caucasians residing in Baltimore City, adjacent counties and the Eastern Shore of Maryland. Patients diagnosed with primary nonsmall cell lung cancer were recruited from the following Baltimore hospitals: University of Maryland Medical Center, Veteran Affairs Medical Center- Baltimore, Johns Hopkins Hospital and Bayview Medical Center. Earlier cases were also recruited from Sinai Hospital, Bon Secours Hospital and Harbor Hospital. All nonsmall cell lung cancer cases were enrolled within 2 years of diagnosis, with approximately 80% of enrolled cases being interviewed within 180 days of diagnosis. Hospital controls were recruited from the same hospitals as cases, were frequency matched to cases on age, gender, race, smoking status (pack-years) and hospital and were excluded if they had a medical history of cancer other than nonmelanoma skin cancer or in situ cervical cancer. Population controls were identified from the Motor Vehicle Administration lists, contacted by telephone to request participation in the study and were frequency matched to cases on age, gender and race. Recruitment of cases and controls has been ongoing since 1998.

The study was approved by the institutional review boards of all participating institutions and the National Institutes of Health, and all participants provided written informed consent.

Exposure assessment

All case and controls were interviewed directly using a standardized questionnaire including comprehensive ascertainment of tobacco use and medical, family, and reproductive history, alcohol use, physical activity, diet and socio-demographics. A detailed section on reproductive history included items regarding pregnancy outcomes, breastfeeding, menarche and menstrual cycles, OC use, menopause, menopausal hormone therapy and gynecological surgeries. First, female participants were asked if they had ever been pregnant and, if so, the ages that they became pregnant and the outcomes of each pregnancy, in chronological order up to the 12th pregnancy. Participants were also asked about use of various birth control options, including birth control pills and age starting and stopping use. Other questions included “At what age did your menstrual periods become regular?” and “At what age was your last menstrual period” to inquire about age at menarche and menopause, respectively. If they responded negatively to the question “Are you still menstruating?” participants chose from a list of reasons that their menstrual periods stopped including change of life/natural menopause, hysterectomy (still has ovaries, ovaries removed, unknown), currently pregnant or they could specify another reason. Finally, participants were asked “Have you ever used hormonal medications just before, during or after menopause, such as pills, vaginal creams, shots, suppositories or skin patches?” A list of various medications were listed, including estrogen pills, progesterone pills, estrogen and progesterone pills, estrogen and testosterone, estrogen vaginal creams, shots, skin patches, estrogen patch and progesterone pills, suppository or “other.” For each medication, participants specified the ages at which they started using them and total number of years they were used. All questions about age in these sections were open-ended.

Statistical analysis

Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using conditional logistic regression. Regression models comparing lung cancer cases to hospital and population controls were initially performed separately. Population-based models were conditioned on the factors used in the frequency matching, including age (<60, 60–69 or ≥70 years) and race (Caucasian or African American), and adjusted for age (continuous), smoking (never smoker, former smoker <15 cigarettes/day for <40 years, former smoker ≥15 cigarettes/day for <40 years, former smoker <15 cigarettes/day for ≥40 years, former smoker ≥15 cigarettes/day for ≥40 years, current smoker <15 cigarettes/day, current smoker 15–25 cigarettes/day or current smoker ≥25 cigarettes/day), education (did not complete high school or high school graduate), number of smoking adults in household (0, 1 or ≥2) and current household income (<$10,000 or ≥$10,000). Models using hospital controls were also conditioned on factors which were used in the matching process including age (<65 or ≥65 years), hospital (Johns Hopkins or other institution), pack-years of smoking (<20 or ≥20) and race (Caucasian or African American) and included the same adjustment variables as population-control models though pack-years of smoking (continuous) were substituted for the eight-level smoking variable (described above) because hospital controls were matched to cases by pack-years of smoking. The categories used in conditioning were deliberately broad to avoid losing cases without a matched control. The purpose of the additional adjustment for age and pack-years was to account for any potential residual confounding that remained after conditioning on these variables. We defined former cigarette smokers as having quit 3 or more years before the interview to account for changes in smoking habits after a diagnosis of lung cancer. Hormone therapy used was defined as any use of hormone medications, including pills, creams, patches, shots or suppositories, just before, during or after menopause. Categories of oral estrogen and estrogen plus progestin use were created after excluding women who reported ever using more than one form of menopausal hormones; these categories were compared to women who reported never using menopausal hormone therapy. Furthermore, because clinical guidelines since the early 1990s have recommended against prescribing estrogen-only therapy for menopausal symptoms among women with intact uteri,32 we examined oral estrogen use only among women who had ever had a hysterectomy.

Missing values were imputed using the median (for continuous variables) or the mode (for categorical variables). Less than 1% of values were missing for all variables except current income (14.7%), education (6.0%) and age at menopause (1.8%).

We tested for interaction between reproductive and hormonal factors by smoking status (never, former, current smoking) using the likelihood ratio test by comparing models with and without cross-product terms.

All statistical analyses were performed using Stata statistical software package version 9 (Stata Corporation, College Station, TX). All statistical tests were two-sided, and p ≤ 0.05 was considered statistically significant.

Results

Hospital controls were well-matched to lung cancer cases according to age at interview (median age for cases = 66; controls = 64) and smoking habits (median pack-years for cases = 35.1; controls = 31.5) (Table 1). Population controls were also well-matched to cases by age at interview (median age for controls = 67). Because smoking was not a matching factor for population controls they were less similar to lung cancer cases in terms of tobacco exposure and socioeconomic status; population controls had the highest proportion of never smokers (48.5%), the lowest median pack-years of smoking (13.9 pack-years), the highest proportion of high school graduates (91.2%), and the lowest proportion of women with household income below $10,000 (4.1%) and adult smokers in the same household (60.3%).

Table 1. Distribution of select characteristics for nonsmall cell lung cancer cases, hospital control, and population controls, Maryland Lung Cancer Study (1998–2008), n = 1,041

Among hospital controls, current compared to never smokers included a higher proportion of women who were menopausal, parous and reported ever having used menopausal hormone therapy and a lower proportion of women reporting having ever used birth control pills (Table 2). Among population controls, current compared to never smokers had slightly fewer live births, a higher proportion of women reporting having ever used birth control pills and a lower proportion of women reporting having ever used menopausal hormone therapy. Among both hospital and population controls, current compared to never smoking was associated with a slightly younger age at first and last birth.

Table 2. Distribution of select characteristics for never, former, and current smokers among hospital (n = 316) and population controls (n = 295)

After adjusting for smoking, number of adult smokers in household, education and household income, the results were generally similar, both quantitatively and qualitatively, when using either the hospital- or the population-based controls. After performing separate analyses, we combined the control groups to increase power and conditioned on age and race; additional conditioning on pack-years of smoking yielded very similar results but with less precision. After combining both control groups, we observed that a greater number of live births was associated with a significantly reduced risk of lung cancer (≥5 vs. 1–2 births: OR = 0.50, 95% CI = 0.32, 0.78, p-trend = 0.002) (Table 3). Later age at last birth was also associated with a reduced risk (≥30 vs. <25 years: OR = 0.68, 95% CI = 0.48, 0.98, p-trend = 0.04), but there was no association with age at first birth. After mutual adjustment for parity and age at last birth, we observed a similar result for parity (≥5 vs. 1–2 births: OR = 0.53, 95% CI = 0.33, 0.86, p-trend = 0.01), but the estimates for age at last birth were attenuated and no longer significant (≥30 vs. <25 years: OR = 0.82, 95% CI = 0.56, 1.20, p-trend = 0.26). Older age at menarche was associated with a slightly reduced risk of lung cancer but it was not significant (≥14 vs. <12 years: OR = 0.75, 95% CI = 0.50, 1.12, p-trend = 0.17). Menopausal status was not clearly associated with lung cancer risk.

Table 3. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for nonsmall cell lung cancer by reproductive factors

In models adjusted only for age, we observed that use of menopausal hormone therapy was associated with a significantly reduced risk of nonsmall cell lung cancer; however, this inverse association was observed only in comparisons with population controls, who were not matched to cases by smoking and was attenuated after adjusting for smoking and other covariates (data not shown). After adjusting for smoking, number of adult smokers in household, education and household income, we did not observe significant associations for OC or menopausal hormone use, including long-term use (Table 4). We observed no association according to time since stopping OC or menopausal hormone use.

Table 4. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for nonsmall cell lung cancer by hormonal factors

We also examined the results for all reproductive and hormonal factors by smoking status (never, former, current; Table 5). The reduced risk associated with five or more births compared to one to two births was consistent among never, former and current smokers (OR for never smokers = 0.26, 95% CI = 0.06, 1.07; OR for former smokers = 0.48, 95% CI = 0.24, 0.99; OR for current smokers = 0.54, 95% CI = 0.28, 1.05; p-interaction = 0.58). We did not observe any substantial differences in lung cancer risk with greater age at menarche or age at last birth by smoking status. In general, none of the risks were stronger in former or current smokers.

Table 5. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for nonsmall cell lung cancer stratified by smoking status

We also examined the associations stratified by race (Caucasian vs. African American) and found that the results were not statistically significantly different between these two groups (Supporting Information Table 1). Additionally, when we restricted the analysis to cases interviewed within 6 months of diagnosis (n = 317) the results did not substantially change. For instance, when we restricted the cases to those interviewed less than 6 months after diagnosis, we saw only slight differences in the strength of the associations for parity (OR for 1–2, 3–4, ≥5 births: 1.00 (reference), 0.66 (0.46–0.93), 0.57 (0.35–0.91), p-trend = 0.01). When we further restrict to cases interviewed within 30 days of diagnosis the inverse association remained (OR for 1–2, 3–4, ≥5 births: 1.00 (reference), 0.62 (0.35–1.10), 0.32 (0.13–0.83), p-trend = 0.02). However, the association for younger age at natural menopause (<50 vs. 50–54) became stronger and statistically significant (OR = 1.64, 95% CI = 1.02, 2.64) after restricting to cases interviewed within 6 months of diagnosis.

Discussion

This case-control study is among only a few observational studies to have comprehensively examined the relationship between reproductive and hormonal factors and lung cancer in women in the United States. After controlling for exposure to active and passive exposure to cigarette smoke, education and income, the results using hospital-based and population-based controls did not differ substantially. We found that greater number of live births was associated with a reduced risk of nonsmall cell lung cancer. Other reproductive and hormonal factors, including long-term use of OC and menopausal hormone therapy, were not associated with risk. Although we hypothesized that greater exposure to endogenous or exogenous estrogens would be associated with an increased risk of lung cancer, especially among current smokers, the results overall were not stronger among current compared to former or never smokers.

We observed that, compared to women who had one or two births, women with five or more births had a significantly lower risk of lung cancer, regardless of smoking status. Our results are in agreement with two recent prospective studies, one among nonsmoking women in the Singapore Chinese Health Study12 and another among nonsmoking women in Shanghai,13 both having found significant inverse relationships between parity and risk of lung cancer. Two hospital-based case-control studies also observed inverse associations with parity, including one conducted in Singapore, particularly among lifetime never-smokers17 and another conducted among nonsmoking and smoking women in the Czech Republic.18 However, other epidemiologic studies found either null8, 9, 14–16 or positive associations.10, 11 We did not identify any potential common biases arising from these studies that could clearly explain the range of inconsistent findings. Although confounding by cigarette smoking may be the most obvious explanation, analyses restricted to never-smoking women have yielded positive,11 null,9, 15 and, including the current study, inverse associations12, 13, 17 between parity and risk of lung cancer. These contradictory findings may have arisen due to chance alone, as the number of lung cancer cases was typically low. It is plausible that a greater number of births may reduce the risk of lung cancer through a decline in circulating estrogens after pregnancy. For instance, among 216 healthy postmenopausal women in the Nurses' Health Study, higher parity was associated with reduced levels of estrone sulfate, and older ages at first birth were associated with lower percent bioavailable estradiol.33

No clear patterns emerge from the scant evidence on lung cancer risk in relation to other reproductive factors, such as menstrual cycle regularity, type of menopause and ages at menarche, first and last birth.8, 9, 11–21 Some studies indicate that women who reach menopause naturally at relatively young ages are at slightly increased risk of lung cancer,9, 13, 15, 16 which may partly reflect smoking behaviors since smokers tend to reach menopause at earlier ages.34 Nonetheless, two of these studies were conducted among never-smoking women.9, 13 The present study shows that women who reached natural menopause before age 50 were at slightly increased risk (vs. 50–54, OR = 1.50, 95% CI = 0.97, 2.31), though the results were weaker among former and never smokers. Future studies examining serum estrogens and lung cancer may help to more fully understand the role of reproduction, endogenous hormones and lung cancer risk.

Few studies have examined the association between OC use and risk of lung cancer, but these have yielded mostly null results.8, 10–12, 14, 21 A case-control study in Germany showed a reduced risk with use of OCs, particularly among smokers15; however, no relationship was observed with increasing duration of use. Our findings also do not suggest that OC use influences the risk of lung cancer in women, even among current smokers.

The evidence as a whole for hormone therapy is equivocal, with some studies providing evidence of a positive association,8 some showing inverse associations,14, 15, 23–29 or no overall association.9–13, 19, 22, 23 Combined estrogen and progestin was not associated with lung cancer in a randomized controlled trial, the Women's Health Initiative, though follow-up for our study was only 5.2 years.35 Differences in the results across studies may be attributable to a variety of factors, including recall or selection biases, differences in outcome (i.e., total lung cancer or a particular histologic type), or differences in definition or categorization of menopausal hormone therapy. However, our study provides evidence that uncontrolled confounding by cigarette smoking may be partially responsible for some of the reported inverse associations between menopausal hormone therapy use and lung cancer in women. For instance, we observed a significantly reduced risk for the use of menopausal hormone therapy when comparing cases to population controls, which was attenuated after the adjustment for other factors, specifically cigarette smoking status, duration and intensity. In contrast, age-adjusted comparisons of lung cancer cases to hospital controls, who were matched to cases by pack-years of smoking, showed no association with risk.

Recent clinical and laboratory evidence suggests that estrogen may directly or indirectly influence the risk of lung cancer, though the potential mechanisms in humans are unclear. Estrogen receptors, including estrogen receptor α (ERα) and estrogen receptor β (ERβ), are expressed in normal and cancerous human lung tissue.36 Activation of ERα induces lung differentiation and maturation, while stimulation of ERβ by estradiol leads to proliferation of lung cancer cells.37 ERα appears to play a direct role in the upregulation of cytochrome p450 (CYP) enzymes in the liver, which oxidize components of cigarette smoke, i.e., polycyclic aromatic hydrocarbons (PAHs) such as benzo[a]pyrene (BP) into carcinogenic derivatives.38 The metabolism of nicotine and cotinine by CYP enzymes is accelerated in women compared to men, particularly during pregnancy or when taking estrogen-only birth control pills.7 Estrogen may also aid in the formation of PAH-related DNA adducts.39 Laboratory studies in rodents support a role of estrogen in the promotion of nonsmall cell lung cancer.40 Although this evidence suggests that greater exposure to endogenous and exogenous estrogen would, if anything, lead to an increased risk of lung cancer, in general, this hypothesis has not been strongly supported by results from epidemiologic studies.

The ability to compare results using both population and hospital controls was a major strength of our study. In general, case-control studies are prone to several biases, including differential recall of past exposures between cases and controls, greater likelihood of population controls comprising healthy volunteers and the inclusion of hospital controls who are more or less likely to be hospitalized due to the exposure of interest (Berkson's bias).41 These types of biases were unlikely to have strongly influenced findings from the present study given that hospital- and population-based controls were not recruited based on their reproductive history or hormone use, and despite the greater similarities of the cases to the hospital compared to population controls in terms of smoking, education and household income, the results did not differ substantially between the two groups after adjustment for smoking and other covariates. A detailed questionnaire regarding lung cancer risk factors and reproductive history, which allowed us to control for active and passive smoking, income and education, and to examine a variety of reproductive and hormonal exposures, was an additional strength of our study. Given the wide range of smoking behaviors among the women in our study, we were able to directly address the hypothesis that cigarette smoking modifies the association between endogenous and exogenous estrogen exposure and lung cancer risk.

Our study suggests that, in general, endogenous and exogenous hormone exposure was not associated with lung cancer risk after accounting for active and passive cigarette smoking and indicators of socioeconomic status. However, we observed that a higher parity was associated with lower lung cancer risk among never, former and current smokers. The question of whether the inverse association between number of births and lung cancer risk may be explained by hormonal changes during or after pregnancy warrants further investigation.