Characteristics of menstruation and pregnancy and the risk of lung cancer in women

Abstract

Lung tissue, both normal and cancerous, has been found to express estrogen receptors and patterns of expression have differed between men and women, suggesting a possible role for hormone-related factors in lung carcinogenesis in women. Few epidemiological studies have examined hormone-related variables and lung cancer risk and the findings have not been consistent. We investigated the association between characteristics of menstruation and pregnancy in relation to lung cancer risk in a population-based case–control study carried out in Montreal, Canada, including 422 women with lung cancer and 577 controls. For each variable, odds ratios (OR) and 95% confidence intervals (CI) were estimated using unconditional logistic regression modeling. Associations were also examined according to level of smoking and by lung cancer histology. All statistical tests were two-sided. Most characteristics of menstruation and pregnancy were not associated with lung cancer risk. However, an increased risk was observed for women who had had a non-natural menopause, which predominantly included women who had had a bilateral oophorectomy, compared with women who had had a natural menopause (OR = 1.92, 95% CI: 1.22–3.01). An inverse association with age at menopause was suggested. These results did not vary by level of smoking and they were similar for adenocarcinomas compared with other histological types. Our results suggest that hormonal factors, related to early menopause and/or ovary removal, may play a role in the risk of lung cancer. Further studies are needed to confirm these findings, and to assess the possible contribution of hormone replacement therapy. © 2009 UICC

Worldwide, lung cancer is the most frequently diagnosed cancer and the most common cause of cancer death.1 Between 80 and 90% of lung cancers are attributable to smoking in North American populations.2 However, only ∼15% of smokers eventually develop lung cancer,3 suggesting that other factors may modify the increased risk associated with tobacco carcinogens. Also, lung cancer occurs among individuals who have never smoked4, 5—the proportion of all lung cancers that occur among lifetime never smokers is approximately 15–25% among women, whereas just 10% among men.4 The distribution of lung cancers by histology also varies by sex, with the most frequent histological type being squamous cell carcinomas in men and adenocarcinomas in women.6 These differences between men and women suggest that the profile of etiologic factors for lung cancer may differ by sex.

Both normal and cancerous lung tissue have been found to express estrogen receptors,4, 5 and patterns of expression have differed between men and women,4 leading to speculation about a possible role for hormone-related factors in lung carcinogenesis in women. The first epidemiological evidence of a potential role of hormonal factors in lung cancer risk was from a study of women in China where it was observed that women with shorter menstrual cycle lengths were at an increased risk of adenocarcinoma.7 Other hormone-related variables that have been examined include age at menarche, menopausal status, use of hormone replacement therapy, oral contraceptive use, and parity.8–27 Overall, very few studies have been conducted on any one factor and the findings have not been consistent. In the context of a population-based case–control study of lung cancer carried out in Montreal, Canada, information was collected on a number of personal, lifestyle, and environmental factors, including characteristics of menstruation and pregnancy. The present analysis is intended to explore the possible role of these latter factors in the etiology of lung cancer in women.

Material and methods

Study population

This study, which has been described previously,28 included men and women aged 35 years and older who were Canadian citizens and resided in the greater Montreal area. Newly diagnosed lung cancer cases occurring between January 1996 and December 1997 were identified at 18 Montreal-area hospitals either through hospital tumor registries or active monitoring of pathology records. The 18 participating hospitals together diagnose 98% of the cases that occur in the greater Montreal area.28 A total of 1,429 men and women with histologically confirmed incident lung cancer were eligible and were approached for interview, of which 569 were women. The participation rate among the 569 eligible women was 81.7% (n = 465) and interviews with the woman herself, or her next of kin if she had died before the interview or was too ill (approximately 30%), were conducted an average of 12.1 months after diagnosis. Lung cancer histology was obtained from the pathology reports and coded according to the classification provided by the International Agency for Research on Cancer.29

Population controls were selected from electoral lists of the Province of Quebec, which provide a nearly complete listing of all Quebec residents who are Canadian citizens aged 18 or greater. Women residing in Montreal, the largest city in Quebec, were randomly selected, stratified to the expected age and sex distribution of cases. Among 885 eligible potential female controls that were approached for interview, the participation rate was 69.4% (n = 614). Approximately 4% of the interviews among female controls were conducted with proxy respondents due to illness, travel, or difficulties with communication. Written informed consent was obtained from all participants, and the study was approved by the Institutional Review Boards of the INRS-Institut Armand-Frappier, McGill University and the 18 participating hospitals.

Exposure assessment

Data was collected with an in-person interviewer-administered questionnaire that assessed various factors including detailed smoking history, sociodemographic characteristics, residential history, occupational exposures, and medical history. For women, the questionnaire included items on hormone-related factors including age at first and last menstrual period, reason for cessation of menstrual periods, ovary removal (oophorectomy) status, age at oophorectomies, age at first pregnancy and first live birth, number of pregnancies and live births, and lactation history.

We determined menopausal status according to whether or not the woman reported that she was still menstruating at the time of diagnosis or time of interview for controls, along with information on her oophorectomy status. We also classified women according to type of menopause (natural or non-natural, i.e., surgical, or due to radiation or chemotherapy). Women who had at least one remaining ovary and who still had their menstrual periods at the time of diagnosis/interview were classified as premenopausal. Women whose menstrual periods had stopped naturally, surgically (by hysterectomy with bilateral oophorectomy), or because of radiation or chemotherapy before the date of diagnosis/interview were classified as postmenopausal. Because women who had had a hysterectomy but had at least one intact ovary remaining after their surgery may still produce estrogen, they are not postmenopausal with respect to their hormone levels.30 Because they were not still menstruating when they were interviewed, since their uterus was removed, and we did not ask any further questions regarding menopause or symptoms, their menopausal status at interview was unknown. However, many of these women were interviewed at an age when they likely would have experienced natural menopause. Thus, for these women (n = 71 cases, 102 controls), we imputed their menopausal status. Based on the distribution of age at natural menopause among the base population, i.e., the controls, we observed that the median age at natural menopause was 50 years and the 90th percentile for age at menopause was 55 years. Thus, for women who had had a hysterectomy but had at least one intact ovary after surgery, if their age at interview was greater than 55 years (53 cases, 84 controls), we classified them as postmenopausal (natural) and if their age at interview was less than 50 (10 cases, 7 controls), we classified them as premenopausal. For women whose age at interview was between 50 and 55 years, we classified their menopausal status and type as missing (8 cases, 11 controls). For the 53 cases and 84 controls now classified as naturally postmenopausal, their age at menopause was missing.

Statistical analysis

To evaluate the association between each of the hormone-related variables and lung cancer risk, unconditional logistic regression modeling was used to estimate odds ratios (ORs) and 95% confidence intervals (CI). The p-value for trend across categories was calculated by assigning women the ordinal value of that category, entering the variable as a continuous term in the regression model, and assessing the Wald chi-square test statistic for that variable. Variables were categorized in a manner consistent with previous literature. We used the distribution among controls that reported their age at menopause (any type) and selected cutpoints at the 25th percentile (age 45 years) and the median (age 50 years) to categorize age at menopause into 3 groups (<45 years, 45–<51 years, 51+ years).

A separate model was run for each hormone-related variable, including the same set of covariates, namely, age at diagnosis (or age at interview for controls), respondent status (self or proxy), ethnic origin (French Canadian or other), and socioeconomic characteristics including number of years of schooling and mean census tract family income. In addition, lifetime smoking history was included and was represented by the comprehensive smoking index (CSI), which has previously been adapted for use in this case–control study.31 The CSI incorporates measures of smoking status, duration of smoking, time since cessation of smoking, and smoking intensity into one aggregate measure. While the parameter estimate for the CSI is not readily interpretable, the value of the CSI, particularly in studies where smoking is of interest as a covariate rather than as a main effect, is that the variable provides both a good fit to the data while maintaining a parsimonious representation of several smoking-related variables. We additionally conducted analyses where smoking was adjusted for using a more conventional approach, which we have used in previous studies,28, 32 with smoking parameterized by smoking status, cigarette-years (cigarettes per day × years smoked), and time since quitting,33 to compare the results to that obtained using the CSI. Because of the importance of smoking on lung cancer risk, women with missing smoking information (n = 14) were excluded from all analyses. Similarly, we excluded women that were missing data on all of the hormone-related variables of interest (n = 66). While there were small differences between these 66 participants with the rest of the study population, these differences were not appreciable and the overall exclusion fraction was small (6.2%) such that the exclusion of these 66 women was unlikely to have caused any measurable distortion in the reported ORs.

Modification of the ORs by level of smoking was evaluated by including product terms for the hormone-related variable and the potential effect modifier. The p-value for multiplicative interaction was based on the likelihood ratio test comparing models with and without the product term. Level of smoking was categorized as lifetime nonsmokers, lighter smokers, and heavier smokers. The categories of lighter versus heavier smokers were based on dichotomizing at the median CSI value among all smokers in the study. We also examined associations separately for adenocarcinomas and other histological types grouped together. All statistical tests were two-sided and a p-value of 0.05 was considered as statistically significant.

Results

Overall, there were 422 cases and 577 controls that were eligible, had complete smoking information, and responded to at least one of the hormone-related questionnaire items. The mean age was 61.5 years (standard deviation (SD) = 9.3) for cases and 61.5 (SD = 9.4) for controls. Compared with controls, cases were more likely to have had a proxy respondent, to be of French Canadian origin, to have had a lower family income, to have had fewer years of schooling, and to have ever smoked cigarettes (Table I). Among ever smokers, the mean number of cigarette-years was 1082.0 (SD = 545.3) for cases and 595.6 (SD = 485.0) for controls. The most common histological type of lung cancer among cases was adenocarcinoma (Table I).

Table I. Characteristics of the Study Population

	Cases (n = 422)	Controls
1 Current smokers include individuals that quit <2 years ago.
Age
<45 years	23 (5.5)	34 (5.9)
45–<55years	76 (18.0)	113 (19.6)
55–<65 years	147 (34.8)	184 (31.9)
65+ years	176 (41.7)	246 (42.6)
Respondent status
Self	286 (67.8)	552 (95.7)
Proxy	136 (32.2)	25 (4.3)
Ethnic origin
French ancestry	332 (78.7)	393 (68.1)
English ancestry	38 (9.0)	31 (5.4)
Other	52 (12.3)	153 (26.5)
Mean census tract family income
Low	202 (47.9)	190 (32.9)
Middle	141 (33.4)	194 (33.6)
High	79 (18.7)	193 (33.4)
Years of schooling
<7	90 (21.3)	96 (16.6)
7–<12	205 (48.6)	203 (35.2)
12+	127 (30.1)	278 (48.2)
Cigarette Smoking
Never	32 (7.6)	297 (51.5)
Former (quit 10+ years ago)	43 (10.2)	109 (18.9)
Former (quit 2–<10 years ago)	38 (9.0)	45 (7.8)
Current1	309 (73.2)	126 (21.8)
Histological type
Adenocarcinoma	201 (47.6)
Squamous cell carcinoma	83 (19.7)
Small cell carcinoma	73 (17.3)
Large cell carcinoma	37 (8.8)
Other histology	28 (6.6)

Table II shows associations between lung cancer and each measured characteristic of menstruation. The proportion of missing values was generally less than 10% for each of the variables, except for age at menarche among cases (22.5%) for which most of the missing values were from proxy respondents. Also, age at menopause was missing for all women with an imputed menopausal status, as previously explained. Age at menarche was not associated with lung cancer risk nor was menopausal status (Table II). However, among women who were postmenopausal (n = 338 cases, 466 controls), an elevated risk of lung cancer was observed among women who had experienced a non-natural menopause type compared with a natural menopause (Table II). When restricting the same analysis to those women whose menopause status was not imputed (n = 285 cases, 382 controls), the observed association was similar (adjusted OR = 2.06, 95% CI: 1.29–3.27 for non-natural versus natural menopause type).

Table II. Multivariate1 Odds Ratios (95% Confidence Intervals) of Lung Cancer Associated With Characteristics of Menstruation

	Cases	Controls	Adjusted OR (95% CI)
1 ORs adjusted for age (in years), respondent status, ethnic group, number of years of schooling, mean census tract family income, and smoking (represented by the CSI). 2 Total number of cases and controls for each variable vary due to missing values. 3 Refers to oophorectomy regardless of menopausal status; 5 cases and 15 controls had a bilateral oophorectomy after having experienced natural menopause. 4 Women with imputed menopausal status are missing from the analysis of age at menopause.
All women	n = 4222	n = 5772
Age at menarche
<13 years	153 (46.8)	235 (41.8)	1.00 (reference)
13 years	73 (22.3)	136 (24.2)	0.89 (0.56–1.40)
14 years	52 (15.9)	98 (17.4)	0.93 (0.56–1.54)
15+ years	49 (15.0)	93 (16.5)	0.73 (0.43–1.23)
			ptrend = 0.29
Menopausal status
Premenopausal	42 (11.1)	92 (16.5)	1.00 (reference)
Postmenopausal	338 (88.9)	466 (83.5)	1.14 (0.61–2.12)
Oophorectomy3
Never	267 (69.7)	425 (75.8)	1.00 (reference)
Unilateral	31 (8.1)	45 (8.0)	1.29 (0.70–2.37)
Bilateral	85 (22.2)	91 (16.2)	1.77 (1.11–2.83)
			ptrend = 0.02
Postmenopausal women	n = 3382	n = 4662
Menopause type
Natural	242 (71.6)	378 (81.1)	1.00 (reference)
Non-natural	96 (28.4)	88 (18.9)	1.92 (1.22–3.01)
Surgical, with bilateral oophorectomy	80 (23.7)	76 (16.3)	1.90 (1.18–3.06)
Radiation, chemotherapy or other menopause	16 (4.7)	12 (2.6)	2.07 (0.70–6.10)
Age at menopause4
<45 years	82 (32.8)	81 (21.5)	1.00 (reference)
45–<51 years	101 (40.4)	166 (44.0)	0.69 (0.42–1.15)
51+ years	67 (26.8)	130 (34.5)	0.66 (0.38–1.16)
			ptrend = 0.14
Age at natural menopause
<45 years	26 (16.9)	31 (10.7)	1.00 (reference)
45–<51 years	76 (49.4)	138 (47.8)	0.78 (0.38–1.57)
51+ years	52 (33.8)	120 (41.5)	0.69 (0.32–1.48)
			ptrend = 0.36
Age at non-natural menopause
<45 years	56 (58.3)	50 (56.8)	1.00 (reference)
45–<51 years	25 (26.0)	28 (31.8)	1.46 (0.53–4.03)
51+ years	15 (15.6)	10 (11.4)	1.86 (0.48–7.16)
			ptrend = 0.30

For the analysis of age at menopause, a nonstatistically significant inverse association with lung cancer risk was suggested (Table II). When examined by menopause type, a similar inverse association was suggested for natural menopause but not for non-natural menopause where an increase in risk was suggested (Table II), though these analyses were based on small numbers. When looking at oophorectomy status regardless of menopause type and status, an increased risk of lung cancer was observed among women who had previously had a bilateral oophorectomy, but not among women with a unilateral oophorectomy, compared with women with both ovaries intact (Table II).

Ever being pregnant was not associated with the risk of lung cancer (adjusted OR = 1.04, 95% CI: 0.64–1.70 for gravid versus nulligravid) nor was ever having a live birth (adjusted OR = 0.87, 95% CI 0.56–1.38 for parous versus nulliparous). Similarly, other pregnancy-related variables, including number of pregnancies, age at first pregnancy, number of live births, age at first live birth, and lactation duration were not associated with lung cancer risk (Table III). When the analyses in Tables II and III were further adjusted for the potential confounding affects of occupational lung carcinogens and previous medical conditions, the observed ORs did not change appreciably (not shown).

Table III. Multivariate1 Odds Ratios (95% Confidence Intervals) of Lung Cancer Associated With Characteristics of Pregnancy

	Cases	Controls	Adjusted OR (95% CI)
1 ORs adjusted for age (in years), respondent status, ethnic group, number of years of schooling, mean census tract family income, and smoking (represented by the CSI). 2 Total number of cases and controls for each variable vary due to missing values.
All women	n = 4222	n = 5772
Number of pregnancies
Nulligravid	58 (14.0)	89 (15.7)	1.00 (reference)
1	52 (12.5)	73 (12.9)	0.82 (0.43–1.55)
2	96 (23.1)	134 (23.6)	1.16 (0.65–2.05)
3	75 (18.1)	125 (22.0)	0.93 (0.52–1.67)
4+	134 (32.3)	147 (25.9)	1.20 (0.68–2.10)
			ptrend = 0.42
Number of live births
Nulliparous	70 (16.9)	98 (17.3)	1.00 (reference)
1	66 (16.0)	106 (18.7)	0.62 (0.35–1.10)
2	118 (28.6)	151 (26.6)	1.05 (0.62–1.79)
3	65 (15.7)	112 (19.7)	0.85 (0.47–1.51)
4+	94 (22.8)	101 (17.8)	1.00 (0.56–1.79)
			ptrend = 0.62
Gravid women	n = 3572	n = 4792
Age at first pregnancy
<22	148 (41.7)	135 (28.4)	1.00 (reference)
22–<25	87 (24.5)	125 (26.3)	0.93 (0.57–1.51)
25–<29	72 (20.3)	121 (25.4)	1.02 (0.61–1.72)
29+	48 (13.5)	95 (20.0)	0.69 (0.39–1.20)
			ptrend = 0.31
Parous women	n = 3432	n = 4702
Age at first live birth
<23	159 (48.5)	151 (32.4)	1.00 (reference)
23–<26	66 (20.1)	112 (24.0)	0.87 (0.53–1.45)
26–<30	61 (18.6)	124 (26.6)	0.77 (0.46–1.30)
30+	42 (12.8)	79 (17.0)	0.93 (0.52–1.65)
			ptrend = 0.55
Lactation duration
None	277 (83.2)	311 (66.5)	1.00 (reference)
1–6 months	29 (8.7)	97 (20.7)	0.62 (0.34–1.11)
>6 months	27 (8.1)	60 (12.8)	1.07 (0.53–2.15)
			ptrend = 0.62

The analyses summarized in Tables II and III were repeated twice, once by using smoking status, cigarette-years, and time since quitting as a means of adjusting for smoking instead of the CSI, and once restricting the analysis to subjects who were self-respondents. The results adjusted for smoking status, cigarette-years, and time since quitting were not appreciably different from those in Tables II and III (not shown). When the analysis was restricted to self-respondents, the ORs in Tables II and III did not change by more than 10% for any variable except for age at natural menopause. Compared with age at natural menopause of <45 years, the adjusted ORs (95% CI) in the self-respondent-restricted analysis were 0.57 (0.27–1.20) for 45–<51 years 0.57 (0.26–1.26) for 51+ years.

Table IV shows the associations for each characteristic of menstruation and pregnancy applied to adenocarcinomas. The pattern of results observed was similar to that seen for lung cancers overall in Tables II and III. The results for other histological types were similar (not shown). When analyzed according to smoking level (not shown), associations among lifetime never smokers (n = 32 cases, 297 controls), lighter smokers (n = 128 cases, 206 controls), and heavier smokers (n = 262 cases, 74 controls) did not greatly differ (not shown).

Table IV. Multivariate1 Odds Ratios (95% Confidence Intervals) of Adenocarcinoma of the Lung2 Associated With Characteristics of Menstruation and Pregnancy

	Adjusted OR (95% CI)
1 ORs adjusted for age (in years), respondent status, ethnic group, number of years of schooling, mean census tract family income, and smoking (represented by the CSI). 2 n = 201 cases of adenocarcinoma.
Age at menarche
<13 years	1.00 (reference)
13 years	0.95 (0.55–1.65)
14+ years	0.83 (0.51–1.36)
Menopausal status
Premenopausal	1.00 (reference)
Postmenopausal	1.15 (0.54–2.45)
Menopause type
Natural	1.00 (reference)
Non-natural	1.83 (1.08–3.10)
Age at menopause
<45 years	1.00 (reference)
45+ years	0.73 (0.46–1.16)
Age at natural menopause
<45 years	1.00 (reference)
45+ years	0.95 (0.52–1.73)
Age at non-natural menopause
<45 years	1.00 (reference)
45+ years	1.41 (0.46–4.29)
Oophorectomy
Never	1.00 (reference)
Unilateral	1.23 (0.59–2.56)
Bilateral	1.79 (1.03–3.10)
Number of pregnancies
Nulligravid	1.00 (reference)
1–2	1.05 (0.57–1.96)
3+	0.92 (0.49–1.71)
Age at first pregnancy
<22	1.00 (reference)
22–<25	1.03 (0.57–1.86)
25+	1.12 (0.66–1.88)
Number of live births
Nulliparous	1.00 (reference)
1	0.66 (0.33–1.32)
2	1.14 (0.61–2.13)
3+	0.85 (0.45–1.58)
Age at first live birth
<23	1.00 (reference)
23–<26	0.93 (0.51–1.71)
26+	1.03 (0.61–1.73)
Lactation duration
None	1.00 (reference)
1+ months	0.82 (0.46–1.47)

Discussion

Using data from a population-based case–control study, we explored whether characteristics of menstruation and pregnancy may be associated with the risk of lung cancer among women. Most hormone-related variables were not associated with lung cancer risk. However, our results suggested that non-natural menopause, most of which resulted from having a bilateral oophorectomy, may be associated with an increased lung cancer risk. Consistent with this finding, an inverse association with age at menopause was suggested. Associations did not differ for adenocarcinomas versus other histological types, suggesting that hormonal factors do not likely account for the predominance of adenocarcinoma among women with lung cancer. The observed associations also did not vary by level of smoking.

A role for hormonal factors was first suggested based on an observed increased risk of lung adenocarcinoma associated with shorter menstrual cycle lengths,7 which in subsequent reports12, 13, 15, 18 was corroborated in just one study.15 Other hormonal factors, including age at menarche,7, 9, 11–13, 15, 17, 18, 20, 22 age at first pregnancy,7, 12, 17 age at first live birth,9, 15, 18, 22 parity,7–9, 11, 12, 15, 17, 18, 20–22 and lactation,12, 20 have been inconsistently associated with lung cancer risk, though most studies did not observe evidence of an association, consistent with our findings for these variables.

Only few previous studies have examined menopausal status and type9, 12, 17, 20, 23 and among those that examined natural postmenopausal status compared with premenopausal status,9, 17, 20, 23 an association was observed in just one study, where an increase in risk was reported.17 However, in two population-based case–control studies9, 17 and one prospective study20 that additionally examined surgical menopause, women who had had a surgical menopause had statistically significant and nonsignificant 1.6–2.0-fold increased risks of lung cancer compared with premenopausal women, consistent with our results. Also consistent with our findings, among two population-based case–controls studies that examined history of hysterectomy,8, 11 a marginally significant increased lung cancer risk was observed in one study.8 In contrast, surgical menopause compared with natural menopause was not associated with lung cancer risk in another study,12 though this analysis was based on only 180 cases and 303 hospital controls. Only one previous US population-based case–control study of 336 adenocarcinomas and their 336 matched controls had information on oophorectomy status,9 and in contrast to our findings, the risk associated with surgical menopause was greatest among women whose ovaries were not removed.

An increased risk with non-natural menopause, particularly surgical menopause, may represent an increased risk with younger age at menopause given that surgery is usually done before natural menopause occurs. Indeed, in our study, the median age at menopause (and interquartile range) was 50 (48–53) years for natural menopause and 43 (37–49) years for non-natural menopause and our results were suggestive that age at menopause may be inversely associated with lung cancer risk. Only few previous studies have examined age at menopause not accounting for menopause type.12, 13, 15, 20, 23, 24 Among four case–control studies,12, 13, 15, 24 observed associations were null except in one hospital-based study, which opposite to our findings, reported a reduced risk with lower age at menopause.12 Among two prospective cohort studies of lifetime never smokers conducted in Japan20 and China,23 an inverse association was reported in one study,23 whereas the association was null in the other. Similarly, only few studies have examined age at natural menopause,7, 9, 11, 17, 18, 20 with a greater age at natural menopause associated with reduced risks of 33–61% in some studies,9, 17, 18 which was statistically significant in one.18 We also observed some suggestion of an inverse association with age at natural menopause but not with non-natural menopause though most of these women had non-natural menopause before the age of 45.

Because very few studies have examined menopause type and age at menopause in relation to lung cancer risk, we had no prior expectation on these associations. Although bilateral oophorectomy is associated with lower risks of ovarian and breast cancers, it is possible that the change in hormone levels following a bilateral oophorectomy has a different relationship with lung cancer risk from that of breast and ovarian cancers, such that risk is increased when estrogen levels are reduced at earlier than normal ages. Furthermore, unlike natural menopause where circulating estrogen levels decrease gradually over several years, bilateral oophorectomy is associated with a sudden decrease in estrogen within days.34, 35 Thus, the observed increased risk associated with non-natural menopause in our study may reflect the effect of an early and sudden decrease in estrogen levels. Because of the rapid decrease in circulating hormones following bilateral oophorectomy, symptoms of menopause, particularly vasomotor symptoms, are more severe compared with natural menopause, thus hormone replacement therapy is indicated for many women.35, 36

It is therefore also possible that our findings, if real, may reflect early and potentially long-term35 use of hormone replacement therapy. Previous studies examining hormone replacement therapy use in relation to lung cancer risk9, 10, 12, 14, 16, 19–27, 37 have been inconsistent; risk associated with ever versus never use has been higher in one study,12 lower in 4 more recent studies,19, 25–27 and null in others. In two studies that reported no association with ever use, duration of hormone replacement therapy use was inversely associated with lung cancer risk in one study,17 whereas in another study a longer duration increased risk.22 Interestingly, in one study, a statistically significant increased risk of lung cancer was observed among women who reported induced menopause and hormone replacement therapy use.20 For most of these studies, information on type of hormone replacement therapy and timing of use (with respect to age at initiation and duration) was not reported. Unfortunately, we did not have data on the use of hormone replacement therapy, thus we were unable to examine this hypothesis.

It should also be considered that the factors that lead to the indications for bilateral oophorectomy at a young age may also predispose to lung cancer. Current indications for bilateral oophorectomy include some reproductive cancers (i.e., ovarian, endometrial, and fallopian tube), increased ovarian cancer risk, carriers of mutations in BRCA1/2, severe endometriosis, severe premenstrual syndrome, and tubo-ovarian abscess.35 Given the recency of BRCA 1/2 testing, it is unlikely that this would account for the observations in our study; however, other conditions may themselves predispose to lung cancer. We did not have data on indications for bilateral oophorectomy. It is also possible that surgical menopause and the ensuing reduction in estrogen levels may interact with smoking in ways that were not captured in our assessment of smoking.

This study had a large sample size, with 422 cases and 577 controls. This allowed us to conduct separate analyses for adenocarcinoma, the most common lung cancer histological type in women, though we were unable to examine other histological types in detail. Our examination of potential effect modification by smoking was also limited as only 32 cases had never smoked throughout their lifetime. Cases and controls were selected from the same underlying study base, i.e., Canadian citizens residing in the greater Montreal area, and the participation rates were relatively high, thus minimizing the potential for selection bias. Because a potential role for hormonal factors in lung cancer risk is not a well-known hypothesis, bias due to differential reporting of these factors by cases and controls was not likely.

The use of proxy respondents for some participants, of which the majority were the spouse (25%) or offspring (48%) of the women, may have resulted in exposure misclassification, although our findings for most variables did not appreciably differ when restricted to self-respondents. Unfortunately, we were unable to reliably analyze age at menarche because of the relatively high proportion of missing values from proxy respondents. As adipose tissue is the primary source of circulating estrogen following menopause,38 the analysis of obesity (i.e., body mass index) would have also been of interest. However, reverse causality (i.e., lung cancer leading to a lower body mass) is difficult to rule out in a case–control study. As previously mentioned, we were unable to analyze use of hormone replacement therapy since this data was not collected. Similarly, we could not examine oral contraceptive use.

A major strength of this study was the detailed assessment of history of smoking, which is important in an analysis of hormonal factors and lung cancer as smokers generally have lower estrogen levels.39 We used various measures of smoking, including smoking status, duration of smoking, time since cessation of smoking, and smoking intensity, which were incorporated into one parsimonious measure (i.e., the CSI). Nonetheless, residual confounding by smoking is still possible. However, our results for adenocarcinoma, the histological type that has been most weakly associated with smoking,4 did not greatly differ from the results for other histological types, thus minimizing the likelihood of residual confounding by smoking.

In summary, our results suggest that non-natural menopause, particularly menopause involving removal of both ovaries, may increase the risk of lung cancer in women, and this finding is consistent with the results from most of the few other previous studies that have examined menopause type or hysterectomy.8, 9, 11, 17, 20 Our results also suggest that this finding may be due to an earlier age at menopause. This would result in a shorter duration of exposure to the premenopausal hormone profile and a longer duration of exposure to the postmenopausal hormone profile. Nonetheless, further studies are needed to confirm these findings, particularly from studies that have information on menopause type (including oophorectomy status) and menopause age along with use of hormone replacement therapy, so that whether the mechanism involved relates to the early and sudden drop of endogenous estrogen or from early and/or long-term use of hormone replacement therapy can be investigated. While smoking remains the most important modifiable factor associated with lung cancer risk, understanding the role of hormones in lung carcinogenesis may provide further insight into the etiology of the disease, particularly for women.

Acknowledgements

Dr. A. Koushik holds a Career Investigator Award from the Cancer Research Society-Fonds de la recherche en santé du Québec-Université de Montréal. Dr. M-É Parent holds a Career Investigator Award from the Fonds de la recherche en santé du Québec. Dr. J. Siemiatycki holds a Canada Research Chair in Environmental Epidemiology and Population Health and the Guzzo Chair in Environment and Cancer. The sponsors had no role in the design, data collection, data analysis, interpretation of the results, the preparation of the manuscript, or the decision to submit the manuscript for publication.