Reproductive Factors and Lung Cancer Risk among Never-Smoking Japanese Women with 21 Years of Follow-Up: A Cohort Study

Lung cancer is the third-most commonly diagnosed cancer and the second leading cause of cancer-related death among women worldwide (1). It is also one of the most lethal cancers, with 5-year net survival ranging from 10%–20% in most countries (2). Although lung cancer incidence has declined among men, most western countries have experienced an overall steady increase in incidence rates among women (3). In contrast, incidence rates in Japan are increasing in parallel for both sexes (4).

Compared with men, women have a lower proportion of lung cancer–related deaths attributable to cigarette smoking, are more likely to have adenocarcinomas (5), are more likely to have tumors with epidermal growth factor receptor mutations (6), which are usually associated with non-small cell lung cancer, and experience better survival (5). Moreover, among never smokers, the risk of lung cancer is higher in women than in men (7, 8). Gender differences in lung cancer occurrence and survival between men and women have led to speculation about a possible role of female sex hormones in lung carcinogenesis. This hypothesis appears to be supported by the presence of both estrogen and progesterone receptors and sex steroid-synthesizing enzymes in lung tissue (9), together with knowledge of the cell proliferative properties of estrogen (10) and the adverse effect of estrogen on the prognosis of patients with lung adenocarcinoma (11). In addition, estrogen has been shown to promote lung adenocarcinoma in animal models (12). Consequently, several studies have assessed the associations of reproductive and hormonal factors, as surrogate markers of exposure to estrogen, with lung cancer risk. However, findings have been inconsistent (13). Because smoking is the strongest risk factor for lung cancer, restricting analysis to never smokers may minimize residual confounding by smoking.

In our previous assessment of the association of lung cancer with reproductive factors and hormone use among 44,677 never-smoking Japanese women, with a follow-up time of 8 to 12 years and 153 incident lung cancer cases, none of the menstrual or reproductive variables analyzed showed a statistically significant association with lung cancer (14). However, that study was limited by a short follow-up period, small number of lung cancer cases and failure to examine associations by histologic subtype.

Here, we present updated results of the association of reproductive factors and hormone use with lung cancer incidence among never-smoking Japanese women after a longer period of follow-up and with the stratification of results by histologic subtype.

We used data from the Japan Public Health Center-based (JPHC) Prospective Study, whose details have been reported elsewhere (15). In brief, the JPHC Study recruited 140,420 subjects aged 40 to 69 years residing within 11 public health center (PHC) areas nationwide from 1990 to 1994. Comprehensive data regarding sociodemographic characteristics, diet, lifestyle, and reproductive events were collected using a self-administered questionnaire at recruitment. The institutional review board of the National Cancer Center Japan (Tokyo, Japan) approved the study protocol (2001–021, 2004–059) and the study was conducted following the ethical guidelines in Japan and the Declaration of Helsinki.

We limited our analysis to women who had never smoked, and excluded subjects from the Katsushika PHC area (Tokyo) due to lack of access to data on cancer incidence (n = 4,178). We also excluded those who did not respond to the baseline questionnaire (n = 11,613), were ineligible because of non-Japanese nationality, incorrect birthdate or duplicate registration (n = 21), or had moved residence before the start of follow-up (n = 115). We further excluded subjects with a cancer diagnosis before the baseline questionnaire (n = 11); history of any cancer (n = 1,494); missing data on smoking (n = 278); missing data on reproductive factors (n = 6,357); and ever smokers (n = 4,443). The final analysis sample included 42,615 women.

Subjects were followed from the date of recruitment until the date of diagnosis of lung cancer, death, migration from the study area, or end of follow-up (31 December 2013 for all PHC areas except Osaka, where follow-up ended on December 31, 2012), whichever came first.

Participants who died or relocated to other municipalities were identified annually through residential registers in each PHC area. Cause of death was confirmed using mortality data from the Ministry of Health, Labor and Welfare. Among study participants, 4,351 (10.2%) died, and 4,316 (10.1%) moved away from the study area during the study period.

Data on lung cancer incidence were obtained from local major hospitals in the respective PHC areas and data linkage with population-based cancer registries. These data were supplemented by information from death certificates. In our study sample, 4.3% of lung cancer cases during the study period were obtained through death certificates only (DCO). This low proportion of DCO cases is indicative of the high completeness of the cancer registry and is satisfactory for this study (16). During the 836,869 person-years of follow-up [median follow-up period: 21 y (IQR 20–23)], 400 subjects were diagnosed with lung cancer. Histologic type of lung cancer was coded using the International Classification of Diseases for Oncology, 3rd edition (C34.0–C34.9; ref. 17). Diagnosis of lung cancer was done by either cytology (18.3%), biopsy (22.1%), tissue diagnosis by surgical treatment (48.0%), or autopsy (0.3%). Histologic subtypes were coded according to the WHO histologic classification of lung tumors (18), with the most common subtypes being adenocarcinoma (76.3%), cancer/malignant neoplasm not otherwise specified (11.8%), and squamous cell carcinoma (3.5%).

We included the following reproductive factors based on previous studies: parity; age at menarche, menopause, first delivery, and last delivery; breastfeeding history; length of menstrual cycle; and use of exogenous hormones (14, 19). We categorized numerical variables based on their frequency distribution within the cohort as follows: number of births among parous women (1–2, 3, ≥4), age at first birth among parous women (≤23, 24–26, ≥27 years), and age at menarche (≤13, 14–15, ≥16 years), age at menopause (≤47, 48–50, ≥51 years), fertility span (≤32, 33–35, ≥36 years), and length of menstrual cycle (≤26, 27–29, ≥30 days). Fertility span referred to the difference between age at menarche and age at menopause. For premenopausal women at baseline (n = 13,710), we obtained age at menopause from the 5-year and 10-year follow-up surveys. Because these follow-up surveys collected data on age at menopause in categories (<40, 40–44, 45–49, 50–54, 55–59 or ≥60 years), we performed multiple imputation with 10 iterations using the predictive mean matching (PMM) method to impute the exact age at menopause. PMM is a semiparametric imputation method that ensures that imputed values are plausible and that the inference from missing data is robust to misspecification of the imputation model (20). Exogenous hormone use and breastfeeding history were dichotomous variables (yes, no) because further details on these variables were not collected. We categorized menstrual status as premenopause, natural menopause, and surgical menopause. Details about the type of surgery were not available.

We used Cox proportional hazards regression models to estimate the HRs with 95% confidence intervals (95% CI) for all lung cancer and the lung cancer subtypes adenocarcinoma and non-adenocarcinoma. The number of subtypes other than adenocarcinoma was small, which justified grouping them as nonadenocarcinomas. We used follow-up time as a time-scale and adjusted all models for PHC area (10 areas), age at entry in years (continuous), alcohol consumption (non-drinker, occasional drinker, regular drinker), and BMI (continuous). We also adjusted all models for the following binary variables: residence [urban (city), rural (town or village)], passive smoking at home and/or at the workplace (yes, no), sports activity (rarely, ≥1 day/week), family history of lung cancer (yes, no), and history of asthma (yes, no). Because premenopausal women were more likely to have a history of gynecologic disease than postmenopausal women, we further adjusted for menopausal status, age at menopause and total fertility span for history of gynecologic disease. Further adjustments varied depending on the exposure variable, as indicated in the footnotes of Tables 2 and 3. Variables adjusted for were identified on the basis of previous studies and the use of directed acyclic graphs. We checked for the proportional hazards assumption using Schoenfeld residual tests and found no evidence of its violation. We found no evidence to suggest multi-collinearity based on variance inflation factors and pairwise correlation coefficients. We tested for linear trends for ordinal variables by including them in the models as though they were continuous. We explored for interactions of reproductive factors with passive smoking (yes, no) and BMI (<25, ≥25) using likelihood ratio tests. We further performed sub-group analyses among postmenopausal women (including those with imputed age at menopause). Because we restricted age at menopause and fertility span to postmenopausal women in the main analyses, we did not consider these variables in sub-group analyses. Because imputation of age at menopause may introduce bias, we further restricted our analyses to postmenopausal women with complete data on age at menopause at the time of the baseline survey. The number of cases among premenopausal women was too small (16 for all lung cancer and 10 for adenocarcinoma) to warrant analyses restricted to this category. All analyses were conducted using Stata 16.0 and a two-tailed α of <0.05.

Among the study participants, mean age was 51.5 years, 87.5% were postmenopausal, and 11.5% were regular alcohol consumers (Table 1). Compared with subjects with natural menopause, subjects with surgical menopause were more likely to have a history of gynecologic disease (72.3% vs. 10.6%) or to have ever used exogenous hormones (19.4% vs. 11.5%). Most subjects were parous (93.4%) or had ever breastfed (81.9%). Mean ages at menarche, first birth, and menopause were 14.6, 24.9, and 49.1 years, respectively. The age-adjusted crude incidence rate of lung cancer in the study cohort was 48 (95% CI, 43–53) per 100,000 person-years.

Age and area-adjusted associations of reproductive factors with all lung cancer, adenocarcinoma, and nonadenocarcinoma are shown in Supplementary Table S1. Apart from menopausal status and breastfeeding, none of the factors evaluated showed a significant association with lung cancer or its subtypes. In multivariable-adjusted results (Table 2), there was an increased risk of lung cancer among women whose age of menarche was 14–15 years (vs. ≤ 13 years, HR, 1.33; 95% CI, 1.01–1.76), without an age at menarche-related linear trend. However, a likelihood ratio test showed an overall significant association between age at menarche and all lung cancer (P = 0.04). Late age at menopause (≥ 50 years) was associated with a higher adenocarcinoma risk (vs. ≤ 47 years, HR, 1.41; 95% CI, 1.01–1.96, P_trend = 0.04) but had no significant influence on all lung cancer or non-adenocarcinoma. The risk of adenocarcinoma was also elevated among women with a fertility span of ≥36 years (vs. ≤ 32 years, HR, 1.48; 95% CI, 1.07–2.06, P_trend = 0.01). Compared with premenopausal women, women with natural menopause (HR, 1.99; 95% CI, 1.02–3.88) or surgical menopause (HR, 2.75; 95% CI, 1.133–5.67) were at increased risk of adenocarcinoma. Breastfeeding was associated with a 51% reduced risk of nonadenocarcinoma (HR, 0.51; 95% CI, 0.28–0.92) but had no influence on all lung cancer or adenocarcinoma. Parity, age at first birth, exogenous hormone use and length of menstrual cycle showed no significant association with lung cancer or its subtypes. There was no evidence for interaction between reproductive factors and passive smoking or BMI.

In subgroup analyses of postmenopausal women (Table 3), the elevated risk of all lung cancer among women whose age at menarche was 14–15 years became nonsignificant but the association between breastfeeding and non-adenocarcinoma remained unchanged. In addition, there was an increased risk of lung cancer among women whose age of first birth was 24 to 26 years (vs. ≤ 23 years, HR, 1.29; 95% CI, 1.00–1.65), with no age-related linear trend and a non-significant likelihood ratio test (P = 0.14) for the overall association with lung cancer. Moreover, there was no evidence of an elevated risk of lung cancer among women with surgical menopause vs. those with natural menopause. The results of subgroup analyses of postmenopausal women with complete data on age at menopause at baseline were similar to those of the main analyses (Supplementary Table S2).

In this large-scale prospective cohort study of never-smoking Japanese women, longer fertility span and late menopause were associated with increased risk of adenocarcinoma but not with all lung cancer or nonadenocarcinoma. Breastfeeding was associated with a reduced risk of nonadenocarcinoma but had no association with all lung cancer or adenocarcinoma. The risk of adenocarcinoma was elevated among postmenopausal women compared with premenopausal women. We found no evidence for an association of lung cancer or its subtypes with parity, age at first birth, use of exogenous hormones or length of menstrual cycle.

The inconsistent results reported in previous epidemiologic studies on the association of lung cancer risk with hormonal, menstrual and reproductive factors may be attributable to differences in study design (hence design-related biases), varying adjustment for covariates, inadequate sample sizes, ethnic and lifestyle differences across study populations and residual confounding by tobacco smoking – the strongest risk factor for lung cancer. Supplementary Table S3 summarizes the characteristics, key findings and potential limitations of selected studies on the associations between reproductive factors and lung cancer.

Studies on the association between parity and lung cancer have yielded conflicting results. Whereas some cohort studies conducted in North America have shown that higher parity increases the risk of lung cancer (21, 22), two cohort studies from Asia have found higher parity to be protective of lung cancer (23, 24) while one, a nationwide cohort study from Korea, found no association between parity and lung cancer (25), which aligns with our current findings. In a recent meta-analysis, which did not include the study from Korea, higher parity was associated with reduced lung cancer risk among Asians but had no association with lung cancer among Caucasians (26).

Consistent with our study, most studies, including meta-analyses of case–control and cohort studies, reported no association between age at menarche and lung cancer (13, 22, 25–27). However, some studies found later age at menarche to be protective of lung cancer (28). Previous assessments of age at first birth yielded mixed findings, with studies reporting positive (29), negative (21, 30), or no association (23, 27) with lung cancer. In a recent meta-analysis, higher age at first birth was associated with a reduced lung cancer risk among Caucasians and a reduced adenocarcinoma risk among Asians (26), which is contrary to our results.

Among the relatively few studies that have assessed the link between breastfeeding and lung cancer, most have reported no association (24, 30, 31), which is consistent with our overall findings with respect to all lung cancer and adenocarcinoma. However, a pooled analysis of data from eight case–control studies participating in the International Lung Cancer Consortium found a reduced lung cancer risk among women who had ever breastfed (27), which is consistent with our results for nonadenocarcinoma.

Case–control and cohort studies on the role of age at menopause in lung carcinogenesis have reported either no association (24, 27, 32, 33) or a negative association (29, 30). In a meta-analysis of 11 case–control and five cohort studies (13), Zhang and colleagues found no significant association between age at menopause and lung cancer risk, which is in line with our overall results. However, we observed increased risk of adenocarcinoma among women with a higher age at menopause, contrary to case–control and cohort studies that reported either no association (24, 29) or a negative association (30). Several studies have reported increased risk of lung cancer among women who were postmenopausal at recruitment, which is consistent with our results, particularly for adenocarcinoma. In a meta-analysis of five case-control and three cohort studies, Min and colleagues reported a statistically significant increased risk of lung cancer among postmenopausal women; however, this increase was restricted to cohort studies and studies conducted outside Asia (34).

Compared with other reproductive and hormonal factors, the role of fertility span in lung carcinogenesis has received less attention. Contrary to our findings, a cohort study in China and a case–control study in Italy found longer fertility span to confer protection against lung cancer (23, 30). In contrast, cohort studies in Singapore and Korea found no association between fertility span and lung cancer (24, 25), which is consistent with our results on all lung cancer. However, in our results stratified by subtype, we found a significant positive association between fertility span and adenocarcinoma, which is contrary to a cohort study that reported no association (24) and a case a case–control study that reported a negative association (30). Our study confirms the previously observed no significant association between menstrual cycle length and lung cancer risk (30, 32, 33). However, a recent meta-analysis found a longer menstrual length to be protective of lung cancer among Asian women (26).

Similar to our findings, most individual-level studies (21, 23–25, 35) and two meta-analyses (13, 26) have reported no association between oral contraceptive use and lung cancer. However, one cohort study conducted in the United States found an increased risk of lung cancer among subjects who used oral contraceptives for >5 years (29), whereas a case–control study in German reported a reduced risk (32). In both studies, the association was restricted to ever smokers, which may be an indication of residual confounding by smoking. The German study also reported increased risk of lung cancer among long-term users of hormone replacement therapy (HRT), although the association became nonsignificant when data were stratified by smoking status (32), which again suggests residual confounding by smoking. An analysis of data from the Women's Health Initiative trial found no increased risk of lung cancer incidence among women treated with estrogen plus progestin for 5.6 years and followed for 2.4 years (36). Two meta-analyses of observational studies found no evidence to support a role of HRT in lung carcinogenesis (13, 37). Nonetheless, one meta-analysis reported a reduced risk of lung cancer among users of HRT (38). When the results were stratified by study design, however, this association was only significant in case-control studies (38), suggesting that the observed association is spurious.

In our study, the risk of lung cancer, particularly adenocarcinoma, was elevated among postmenopausal women compared with premenopausal women, which agrees with evidence from a case–control study (30) but not from a cohort study (31). However, as in a previous cohort study (24), the risk of lung cancer did not vary by type of menopause. Premenopausal women were younger, with lower BMI, healthier, and more likely to reside in cities (Table 1). Thus, the increased cancer risk in this group may be due to confounding by other lifestyle and health-related factors.

The expression of estrogen receptors in the normal lung and in lung tumor cell lines (39) led researchers to hypothesize a link between estrogen and lung carcinogenesis, and lead to studies on the relationship between reproductive factors and lung cancer. Reproductive factors are proxy measures of the activity of endogenous estrogen, and associations with such factors have been interpreted to imply a role of endogenous estrogen in lung carcinogenesis. Among the factors we investigated, longer fertility span and late menopause are perhaps the best proxy indicators of sustained high levels of endogenous estrogen exposure and therefore fit this hypothesis well. This hypothesis is further supported by the significant linear trends observed in the associations between these variables and adenocarcinoma, suggesting that both variables are indicators of the same phenomena, although we found no evidence to suggest multi-collinearity in the assessed variables. Although the precise mechanisms through which estrogen may induce lung cancer are unclear, suggestions include promotion of cellular proliferation, formation of polycyclic aromatic hydrocarbons-related deoxyribonucleic acid adducts, activation of growth factors, and influence on the expression of genes involved in metabolizing components of cigarette smoke (40–42). Estradiol activates estrogen receptor-β (ER-β), the predominant estrogen receptor in the lungs, causing proliferation of lung cancer cells (41, 43, 44). ER-β has been shown to function as a transcriptional activator at increased estradiol levels (45). ER-β can be found in the columnar epithelium, intermediate cells, and basal cells of the lung (46). The effects of estrogen on cell proliferation and gene expression appear to be tissue and cell type-specific, and the distribution of estrogen receptors in the lung tissue may determine the cellular responses to estrogen (39), and hence lung carcinogenesis. This could partly explain why age at menopause and longer fertility were associated with adenocarcinoma but not with non-adenocarcinoma, albeit that more research is required to elucidate this.

The strengths of this study include its large sample with a long follow-up period, stratification of lung cancer by sub-type, homogenous study population with respect to ethnicity and smoking status, and comprehensive assessment of reproductive factors. Several limitations of this study also warrant mention. Information on some of the exposure variables may have been subject to recall bias because it was based on individual recollection. Moreover, the study lacked data on exogenous hormone type and did not distinguish between HRT and contraceptives. Nevertheless, use of exogenous hormones in Japan at the time of study initiation was low, with 2.5% and 6.3% of women aged 45–64 years being current and ever users of hormone replacement therapy, respectively (47), and about 1% of women aged 16–49 being users of oral contraceptives (48). These statistics suggest that the proportion of hormone users was higher in our study sample than in the general population. This may be due to differences in the age of the included women and the way the questions were framed: women in the JPHC study were asked whether they had ever used any exogenous hormone (without specifying the type or indication for use). Although the study is based on a large sample of women followed for a long period, the number of incident lung cancer cases was not large enough allow detailed assessment of subtypes of lung cancer. Thus, the results of non-adenocarcinoma should be interpreted cautiously. This study did not collect data on estrogen receptor expression in lung tissue, yet there is evidence suggesting that the effect of estrogens may depend on the receptor status of lung cancer cells (49), and specific subgroups of lung cancer may be differentially associated with estrogens. Moreover, we did not have information on EGFR status to allow us to stratify the results by this variable. Finally, although we adjusted for a large number of confounders, information on other potential confounders such as air pollution, radiation exposure, and income was not available. In a subgroup of participants with information on education (n = 17,582), the results were similar to those of all subjects and did not change after adjusting for education, suggesting that lack of adjustment for socio-economic status may not have substantially biased our results.

In conclusion, we found that longer fertility span and late age at menopause were associated with increased risk of adenocarcinoma, whereas breastfeeding was associated reduced risk of nonadenocarcinoma. Given that reproductive factors are only proxy measures of exposure to estrogen, future studies that include serum measures of estrogen and progesterone concentrations and the status of receptors may help clarify the role of endogenous hormones in lung carcinogenesis. If our findings of increased risk of adenocarcinoma among women with longer fertility span or late menopause are true, then fertility span and age at menopause may be useful variables in developing risk prediction models for lung adenocarcinoma among non-smoking women. Further, the question of whether the association of fertility span and age at menopause with adenocarcinoma is explained by duration of exposure to estrogen warrants further investigation.

T. Yamaji reports grants from Ministry of Health, Labour and Welfare of Japan during the conduct of the study. S. Tsugane reports grants from National Cancer Center Research and Development Fund and grants from Ministry of Health, Labour and Welfare of Japan during the conduct of the study. No disclosures were reported by the other authors.

C. Wilunda: Conceptualization, formal analysis, writing–original draft, writing–review and editing. N. Sawada: Conceptualization, data curation, supervision, methodology, writing–review and editing. T. Yamaji: Conceptualization, data curation, methodology, writing–review and editing. M. Iwasaki: Conceptualization, data curation, supervision, methodology, writing–review and editing. M. Inoue: Resources, data curation, supervision, methodology, writing–review and editing. S. Tsugane: Conceptualization, resources, data curation, supervision, funding acquisition, methodology, writing–review and editing.

We are indebted to the Aomori, Iwate, Ibaraki, Niigata, Osaka, Kochi, Nagasaki and Okinawa Cancer Registries for providing their incidence data. Members of Japan Public Health Center-based Prospective Study are listed at the following site (as of April 2017: http://epi.ncc.go.jp/en/jphc/781/7951.html). This study was supported by the National Cancer Center Research and Development Fund (23-A-31[toku], 26-A-2 and 29-A-4, since 2011) and a Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (19shi-2, from 1989 to 2010) awarded to (S. Tsugane). The study funders played no role in study design; the collection, analysis and interpretation of data; in the writing of the article; and in the decision to submit the article for publication.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.