Never smokers and lung cancer risk: A case-control study of epidemiological factors

Authors


Abstract

We performed an analysis of potential epidemiological risk factors for lung cancer using data from 280 cases and 242 hospital-based controls, all lifetime never smokers (those who had smoked <100 cigarettes in their lifetimes) and frequency matched on age, gender and ethnicity. The data on demographic characteristics, medical history of respiratory diseases (asthma, emphysema, pneumonia and hay fever), weight and height, family history, female characteristics and environmental tobacco smoke (ETS) and dust exposure were derived from personal interviews. We performed a logistic regression analysis of these variables adjusting for age, gender, ethnicity, income and years of education. Exposure to ETS (OR = 2.08, 95% CI [1.25–3.43]) and dusts (OR = 2.43, 95% CI [1.53–3.88]) were associated with significantly increased risk. In the analysis for joint effects, exposure to both ETS and dusts conferred a higher risk (OR = 3.25, 95% CI [1.58–6.70]) than exposure to either alone. Family history of any cancer with onset before age 50 in at least 1 first degree relative was a significant risk predictor (OR = 1.70, 95% CI [1.10–2.64]). Individuals with a self-reported physician-diagnosed history of hay fever, but not asthma, had a decreased lung cancer risk (OR = 0.57, 95% CI [0.35–0.92]). In the multivariate analysis, exposure to ETS and dusts, and family history of cancer with onset before age 50 were significant risk factors, while a history of hay fever (occurring without asthma) was significantly protective. © 2005 Wiley-Liss, Inc.

About 10% of all lung cancers occur in lifetime never smokers.1 Never smokers who develop lung cancer represent a unique, yet understudied, subset of all lung cancer patients. Environmental tobacco smoke (ETS) exposure is a frequently reported risk factor in never smokers.2, 3, 4, 5, 6 Other factors that have been implicated in lung cancer risk in never smokers include family history of any cancer and lung cancer,7 and select dietary variables.8, 9 The effect of BMI is contradictory.10, 11 Personal history of respiratory diseases was found to modulate risk in smokers (including former smokers),12, 13 but their effect in never smokers has been less studied. There are only a few studies examining the effects of occupational and environmental factors (other than ETS) on lung cancer risk in never smokers.14

To shed more light on the role of epidemiological risk factors in never smokers, we analyzed data from a case-control study designed to study genetic susceptibility to lung cancer.

Material and methods

Subject recruitment

From September 1995 through December 2003, patients with lung cancer and controls without a previous diagnosis of cancer were accrued for an ongoing and previously described molecular epidemiological study on susceptibility markers for lung cancer.15 There were 280 patients with histologically confirmed lung cancer recruited prior to initiation of radiotherapy or chemotherapy from The University of Texas M.D. Anderson Cancer Center, who reported themselves to be lifelong never smokers (defined as those who have smoked <100 cigarettes in their lifetime). There were no age, gender, ethnic or stage restrictions. The response rate for cases was about 80%. The reasons for refusal to participate include patient too ill, patient referred only for second opinion to M.D. Anderson Cancer Center or patient unwilling to donate blood for the study and complete the interview. Healthy controls who were also lifetime never smokers (n = 242) and were without a previous diagnosis of cancer (except for nonmelanoma skin cancer) were recruited from the Kelsey-Seybold Clinics, Houston's largest private multi-specialty physician group, that includes a network of 23 clinics and >300 physicians in the Houston metropolitan area. Patients newly arriving at the clinics were given a short survey form to determine their eligibility for the study. The completion of the form was strictly voluntary. On the basis of the survey forms, individuals most suitable for frequency matching to recruited cases were identified and contacted to schedule the interview and specimen collection. The matching criteria included age (±5 years), gender and ethnicity. The response rate for controls was about 83% when approached for an interview. The main reasons for declining participation included lack of time or difficulties related to transportation. All cases and controls were U.S. residents. This research was approved by the M.D. Anderson Cancer Center and Kelsey-Seybold Institutional Review Boards.

Collection of epidemiological data

After the study participants were briefed on the study and signed an informed consent, a 45-min structured personal interview was conducted by M. D. Anderson research interviewers, during which they obtained information on sociodemographic characteristics, smoking history, family history of cancer in first degree relatives, age and smoking status of affected relatives, medical history, occupation and select occupational exposures. In women, information was also collected on oral contraceptive (OC) use and hormone replacement therapy (HRT), parity, age at menopause and history of miscarriages. ETS exposure was defined as having been around someone else's cigarette smoke on a regular basis. The embedded probes within the questionnaire identified “regular” as daily or weekly exposure. ETS exposures at home and at work were reported separately. Exposed individuals also reported the number of years of exposure. Data on prior physician-diagnosed respiratory diseases, including asthma, emphysema, pneumonia and hay fever, were collected, but not validated by medical record review. The year or age at diagnosis was recorded for each positive response.

Statistical analysis

All analyses were performed using the SAS 8.0 statistical software package (The SAS system for Windows Release V8.2, SAS Institute Inc., 2001). To assess collinearity among potential predictors, nonparametric Kendall's tau-b was used as a measure of association between binary variables such as presence of environmental exposures or respiratory diseases. All statistical tests were two-sided. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated as estimates of the relative risk. Unconditional multivariate logistic regression analysis was performed to control for confounding by age, gender, ethnicity, income and years of education, where appropriate.

Family history was analyzed as a dichotomous variable for having first-degree relatives with any cancer, with any smoking-related cancer (including lung, head and neck, kidney, bladder and pancreatic cancers), with lung cancer specifically, with lung cancer in at least 1 nonsmoking relative and with any cancer with young age at onset (before 50) in at least 1 first-degree relative. We also created a set of dummy variables with respect to age at onset, to consider the following categories: (i) no cancer history reported in first-degree relatives (referent), (ii) cancer history present, but the age at diagnosis is 50 years or above for all reported cases and (iii) at least 1 first-degree relative with cancer diagnosed before age 50.

Of the exposure variables, we included only the summary measure of dust exposure in this analysis. This variable incorporated the following dust exposures (from work or hobby-related activities): metal, concrete, wood, cotton, textile fibers, fiberglass or sand.

Exposure to ETS was analyzed in several ways: as a dichotomous (presence/absence of the exposure), categorical (dummy variable for no exposure, exposure at home only, at work only and both) and as a quantitative variable (duration of exposure from home and workplace combined). Those who reported being exposed “every day” or “few times a week,” either at home or at work, were considered exposed, while those reporting no exposure or being exposed only “few times a month” or “rarely” were considered not exposed (only 4 individuals reported exposure “few times a month” or “rarely”). We used the maximum of the self-reported number of years of ETS exposure at work and at home as a cumulative measure of exposure for those who reported being exposed “every day” or “few times a week.” We did not ascribe any weight to exposure at work vs. exposure at home because we were not able to estimate the separate intensity of exposure from the questionnaire data. In the case of ETS exposure represented by the dichotomous or categorical variable, missing values were treated as a separate category, as data on ETS were missing for 84 (30%) controls and 55 (23%) cases.

Data on weight and height 5 years previous to enrollment were available only on a subset (67%) of participants, while current weight and height were available for almost all (94%) subjects. Thus, the results for weight and BMI 5 years prior to enrollment are reported only for the subset of the sample with complete responses.

For multivariate model selection, we used stepwise logistic regression. First, we assessed collinearity among potential predictors. The significance level for entry in the model was set at 0.05, and the significance level for retaining variables in the model was 0.10. The final model was based on the results of the stepwise procedure and on the clinical relevance of the variables. Age, gender, ethnicity, income (tertiles) and years of education were also included in the final multivariate model. The final multivariate model is also shown by gender, so that the gender-specific relevance of the variables could be evaluated.

Results

Data from 280 patients with lung cancer and 242 healthy controls, all never smokers, were available for this analysis (Table I). Women constituted the majority of the cases (68%) and, accordingly, the controls (73%). The majority of the cases (82%) and the controls (80%) were Caucasian (p = 0.73). There were no statistically significant differences between the cases and controls in terms of age, a matching variable (p = 0.095), or years of education (p = 0.11). For income, there was a significant case-control difference (p = 0.006): about 39% of the controls compared to only 26% of the cases reported an annual income within the $40,000–$74,000 range, while 35% of cases compared to only 26% of controls reported an annual income below $39,999.

Table I. Distribution of Host Characteristics by Case-Control Status
VariablesCasesControlsp-Value
  • 1

    Numbers do not add up to total because of the missing data.

Gender, N (%)
 Male91 (32.5)65 (26.9)0.16
 Female189 (67.5)177 (73.1) 
Ethnicity, N (%)
 White231 (82.5)193 (79.8)0.73
 Hispanic25 (8.9)25 (10.3) 
 Black24 (8.6)24 (9.9) 
Age (years; mean ± SD)60.2 ± 12.761.9 ± 10.90.095
Education (years; mean ± SD)114.6 ± 3.514.1 ± 2.90.11
Income, N (%)1
 <$39,99989 (34.8)59 (26.1)0.006
 $40,000–$74,99966 (25.8)88 (38.9) 
 $75,000 or more101 (39.4)79 (35.0) 

The majority of lung cancer cases (63.1%) in this group of never smokers presented with adenocarcinoma (Table II). Women were more likely (66.7%) than men (55.8%) to present with a diagnosis of adenocarcinoma and bronchioalveolar carcinoma (BAC) and less likely to be diagnosed with small cell lung cancer. In particular, 20 of 23 patients with BAC were women (p = 0.013), while 6 of the 7 small cell lung cancer patients were men (p = 0.025). Small cell lung cancer cases were characterized by an earlier age at diagnosis (49 years on average) when compared to the other cell types (Table II). The overall mean age at diagnosis was 60.18 years. About a quarter of the patients were diagnosed before age 50. Men tended to be diagnosed at an earlier age than women (58.2 vs. 63.1 years; p = 0.07). The age at onset did not differ significantly between patients who reported and who did not report ETS exposure (61.3 vs. 58.2 years, respectively, p = 0.18).

Table II. Prevalence and Age at Diagnosis by Histology
Lung cancer cell typeTotalMenWomenAge at diagnosis
N (%)N (%)N (%)Mean ± SD
Adenocarcinoma166 (63.1)48 (55.8)118 (66.7)59.7 ± 12.9
BAC23 (8.8)3 (3.5)20 (11.3)61.5 ± 12.6
Small cell carcinoma7 (2.7)6 (7.0)1 (0.6)49.0 ± 13.8
Squamous cell carcinoma24 (9.1)11 (12.8)13 (7.3)62.8 ± 11.6
NSCLC43 (16.3)18 (20.9)25 (14.1)61.2 ± 12.3

Overall, ETS exposure was a significant risk predictor (OR = 2.08, 95% CI [1.25–3.43]) (Table III). However, only exposure at work was significantly associated with lung cancer risk, while exposure at home was not significant.

Table III. Prevalence and Main Effects of Select Risk Factors
VariablePrevalence by case-control statusMain effect
CasesControlsp–valueOR195% CI
  • 1

    Adjusted for age, gender, ethnicity (categorical variable), years of education and tertiles of income (categorical variable).

  • 2

    Numbers do not add up to total because of the missing data.

  • 3

    The average current BMI was 27.1 for cases and 28.3 for controls for which the information on weight 5 years prior to the enrollment was available.

ETS exposure2
 Overall ETS exposure, N (%)156 (80.8)126 (67.4)0.0072.081.25–3.43
 ETS from home, N (%)128 (66.3)114 (61.0)0.3801.300.82–2.06
 ETS from work, N (%)82 (42.5)56 (30.0)0.0111.981.23–3.18
 Categories of ETS exposure
  Neither at home nor at work37 (19.2)61 (32.6) 1.0 
  At home only74 (38.3)70 (37.4) 1.640.94–2.88
  At work only28 (14.5)12 (6.4) 4.261.76–10.30
  Both54 (28.0)44 (24.0) 2.241.22–3.80
 Duration of exposure (years; mean ± SD)24.4 ± 16.124.5 ± 13.60.950  
Dusts (summary measure), N (%)283 (30.3)43 (17.8)0.0012.431.53–3.88
Combined ETS and dust exposure (by categories)
 Neither (reference)31 (16.1)54 (28.9) 1.0 
 Dust only6 (3.1)7 (3.7) 1.250.34–4.53
 ETS only112 (58.0)97 (51.9) 1.941.11–3.39
 Both44 (22.8)29 (15.5) 3.251.58–6.70
Respiratory diseases, N (%)2
 Pneumonia
  Current112 (41.2)66 (27.3)<0.0012.031.38–3.00
  >3 yr58 (21.5)52 (21.3)0.9640.920.58–1.45
 Asthma
  Current49 (18.0)25 (10.3)0.0131.821.05–3.15
  >3 yr44 (16.5)19 (8.1)0.0042.121.16–3.88
  >5 yr40 (15.2)18 (7.7)0.0092.031.09–3.80
  >10 yr35 (13.6)15 (6.5)0.0102.201.12–4.32
  >20 yr33 (12.9)14 (6.1)0.0112.261.12–4.54
 Hay fever
  Current70 (25.7)71 (29.3)0.3000.830.55–1.25
  >3 yr66 (24.6)67 (28.2)0.3690.830.55–1.25
  >5 yr65 (24.3)65 (27.5)0.4140.830.54–1.27
  >10 yr62 (23.5)61 (26.3)0.4700.870.56–1.35
  >20 yr58 (22.3)57 (25.0)0.4840.870.55–1.37
 Duration of asthma (tertiles)
  No disease (reference group)223 (82.0)217 (89.7) 1.0 
  <9 yr10 (3.7)8 (3.3) 1.300.49–3.52
  9–45 yr19 (7.0)9 (3.7) 1.470.62–3.50
  46+ yr20 (7.3)8 (3.3) 2.931.17–7.32
 Duration of hay fever (tertiles)
  No disease (reference group)202 (74.5)171 (70.7) 1.0 
  <32 yr19 (7.0)23 (9.5) 0.590.30–1.16
  32–47 yr25 (9.2)25 (10.3) 0.840.45–1.57
  48+ yr25 (9.2)23 (9.5) 0.880.41–1.88
 Joint effects of asthma and hay fever
  Neither181 (66.5)157 (64.9) 1.0 
  Asthma only21 (7.7)14 (5.8) 1.050.50–2.21
  Hay fever only42 (15.4)60 (24.8) 0.570.35–0.92
  Both28 (10.3)11 (4.5) 2.431.11–5.35
 Anthropometric factors
  BMI, mean (SD)226.53 (5.1)28.53 (5.9)<0.001  
  BMI 5 years ago, mean (SD)226.6 (4.8)27.3 (6.0)0.2220.960.92–1.01
  Weight change over 5 years (kg), mean (SD)21.27 (8.7)2.91 (11.9)0.143  
Cancer family history (first degree relatives), N (%)2
 Any cancer188 (68.9)153 (64.0)0.1981.410.94–2.10
 Smoking-related cancer257 (20.8)48 (19.9)0.791.210.75–1.95
 Lung cancer36 (13.2)37 (15.4)0.4831.030.54–1.03
 Lung cancer in at least one nonsmoker26 (2.3)10 (4.2)0.2160.940.31–2.89
Family history by age at onset
 None (reference)85 (31.1)88 (36.7) 1.0 
 Onset at age 50+102 (37.4)101 (42.1) 1.190.76–1.84
 Onset at age <5086 (31.5)51 (21.3) 1.871.13–3.10
Female characteristics
 Parity (nulliparous vs. parous), N (%)17 (9.0)17 (9.6)0.8411.200.56–2.55
 Use of oral contraceptives, N (%)24 (2.1)4 (2.3)0.9310.650.14–3.01
 Use of hormone replacement therapy, N (%)275 (41.0)73 (41.7)0.8881.060.67–1.68
 Menopausal status, N (%)2157 (83.1)156 (88.6)0.1610.680.31–1.47
 Age at menopause, mean (SD)46.3 (7.7)45.7 (7.1)0.4801.010.98–1.05
 History of miscarriages, N (%)241 (22.5)44 (25.0)0.5830.900.54–1.51

When the participants were categorized into not exposed (reference), exposed at home only, exposed at work only and exposed at both home and work, those exposed at work only were, unexpectedly, at the highest risk, followed by those exposed at both venues. Exposure at home only did not confer a significant risk (Table III). Exposed cases and controls did not differ significantly by duration of ETS exposure (mean duration 24.5 (13.6) years for the controls, 24.4 (16.11) for the cases, p = 0.95). There also was no significant case-control difference in the duration of home or work ETS exposure (data not shown).

A positive response to the summary measure of dusts was associated with a 2.43-fold increased risk (Table III). In the analysis of joint effects, exposure to both ETS and dusts conferred a higher risk (OR = 3.25, 95% CI [1.58–6.70]) than to either exposure alone (ETS, OR = 1.94, 95% CI [1.11–3.39]; dust, OR = 1.25, 95% CI [0.34–4.53]).

Pneumonia showed a significant association with lung cancer risk (Table III). However, since pneumonia can be a symptom of preexisting lung cancer rather than a risk factor, we excluded cases and controls diagnosed with pneumonia within 3 years of the interview, and the apparent effect of pneumonia disappeared.

Self-reported physician-diagnosed asthma was significantly associated with risk of lung cancer. After exclusion of subjects diagnosed within 3, 5, 10 or 20 years from the time of interview, the association remained significant and relatively stable (Table III). There was a dose–response pattern for duration (p = 0.007 for trend) (Table III), adjusted for age. The risk for those with the longest reported duration was almost 3-fold as compared with asthma-free individuals: OR = 2.93, 95% CI [1.17–7.32].

The OR associated with a history of hay fever was 0.83, but it did not achieve statistical significance (Table III). Since we have previously shown in this case-control study that hay fever was significantly protective against lung cancer for all smoking categories combined,16 we performed a more detailed analysis of this condition. When we excluded subjects diagnosed within 3, 5, 10 or 20 years from the time of interview, the point estimates were essentially unchanged (Table III). Since asthma and hay fever were significantly associated (tau-b = 0.23, p < 0.001), we evaluated their joint effects. Individuals with hay fever but not asthma were at significantly lower risk (OR = 0.57) when compared to those with no condition, while those with both asthma and hay fever were at a significantly elevated risk (OR = 2.63) (Table III). Asthma alone did not confer increased risk. Predictably, only 1 subject in this group of lifetime never smokers, a lung cancer patient, reported having emphysema.

Cases at the time of diagnosis appeared to be significantly leaner (BMI = 26.5) than controls at the time of interview (BMI = 28.5; p < 0.001). Cases also tended to have been leaner 5 years prior to enrollment when compared to controls, and to have reported less gain in weight over the 5 years prior to diagnosis (1.27 kg) than controls over a similar time period (2.91 kg; p = 0.143) (Table III). Since disease status could affect BMI, we analyzed only BMI reported as of 5 years before the time of diagnosis (cases) or the interview (controls). BMI 5 years prior to enrollment analyzed as a quantitative variable tended to be protective, but the association did not reach statistical significance.

There was no significant difference between cases and controls with respect to family history of any cancer, smoking-related cancers, lung cancer or lung cancer in nonsmoking relatives. However, a history of early onset (before age 50) cancer among first-degree relatives was associated with a significantly increased risk (OR = 1.87, 95% CI [1.13–3.10]) (Table III). The top 5 cancers driving the association were breast (15.6%), skin (8.5%), colon (8.1%), lung (6.2%) and melanoma (5.2%).

Parity, use of OCs or hormone-replacement therapy, menopausal status, age at menopause and history of miscarriages did not demonstrate case-control differences and did not confer risk (Table III).

Before performing the multivariate analysis, we tested all the potential predictors for collinearity. We noted that exposure to dusts was associated with overall ETS exposure (tau-b = 0.132, p = 0.01), and, particularly, with ETS exposure at the workplace (tau-b = 0.259, p < 0.001). As mentioned earlier, a self-reported diagnosis of asthma was associated with diagnosis of hay fever. These respiratory conditions did not show any association with overall ETS exposure; however, asthma was associated with ETS exposure at home (tau-b = 0.143, p = 0.005) and with exposure to dusts (tau-b = 0.116, p = 0.009). Men were more likely to be exposed to dusts (tau-b = 0.205, p < 0.001) and ETS exposure at work (tau-b = 0.182, p < 0.001). Women were more likely to have ETS exposure at home (tau-b = 0.182, p < 0.001). Although some of the associations were statistically significant, their magnitude was modest, suggesting that collinearity would not constitute a major problem in the multivariate analysis. Because BMI 5 years prior to enrollment did not show a statistically significant association with risk in the univariate analysis and about one third of the values were missing, we did not include this variable in the multivariate model.

The following variables were included in the multivariate analysis: summary measure of dust exposure; ETS exposure at home, work or both; respiratory conditions (asthma, hay fever or both) and family history of cancer in first-degree relatives (history of cancer occurring at age 50 or later and history of cancer occurring at an age below 50). Age, gender, ethnicity, education and income were controlled for in this analysis (Table IV). We found that exposure to dusts (OR = 2.13), ETS exposure at work (OR = 3.94) and at both home and work (OR = 2.02) and family history of young onset cancer (OR = 1.82) were associated with risk, and history of hay fever occurring without asthma was significantly protective (OR = 0.50).

Table IV. Multivariate Lung Cancer Risk Models for Never Smokers
VariableOR195% CI
  • 1

    Adjusted for age, gender, ethnicity (categorical variable), years of education and tertiles of income (categorical variable).

Overall (247 cases, 224 controls)
 ETS exposure by category
  At home1.590.89–2.86
  At work3.941.54–10.06
  Both2.021.06–3.85
 Dusts (summary measure)2.131.29–3.53
 Joint Effects of asthma and hay fever
  Asthma only0.920.41–2.06
  Hay fever only0.500.30–0.83
  Both2.100.93–4.73
 Family history
  Onset after age 501.110.70–1.76
  Onset age <501.821.07–3.08
Adenocarcinoma (149 cases, 227 controls)
 ETS exposure by category
  At home1.400.72–2.71
  At work4.671.70–12.85
  Both2.020.98–4.16
Bronchioalveolar carcinoma (21 cases, 226 controls)
 Dusts (summary measure)4.651.65–13.07
Squamous carcinoma (23 cases, 226 controls)
 Dusts (summary measure)4.531.59–12.90

Exploratory multivariate cell type-specific analyses were also performed. Because of the small sample size, they were limited to adenocarcinoma, squamous cell carcinoma and BAC (Table IV). For adenocarcinoma, only ETS exposure at work was significant, while only exposure to dusts was significant for both BAC and squamous cell carcinoma (Table IV).

We also performed an exploratory gender-specific analysis. Only the frequency of ETS exposure at home (50.8% of men vs. 69.6% of women, p < 0.001) and at work (49.2% of men vs. 30.4% of women, p < 0.001), but not overall (74.2% of men and 74.2% of women, p = 0.99), and frequency of dust exposure (37.8% men vs. 18.6% of women, p < 0.001) showed significant gender differences (Table V).

Table V. Multivariate Lung Cancer Risk Models for Never Smokers, by Gender
VariableWomen (164 cases, 165 controls)Men (83 cases, 59 controls)
No. of exposed/affectedOR195% CINo. of exposed/affectedOR195% CI
CasesControlsCasesControls
  • 1

    Adjusted for age, gender, ethnicity (categorical variable), years of education and tertiles of income (categorical variable).

ETS exposure by category
 At home55591.290.65–2.5719113.240.93–11.25
 At work9311.661.26–107.601993.841.04–14.17
 Both37301.880.87–4.0717142.560.69–9.47
Dusts (summary measure)45222.151.11–4.1638212.921.16–7.35
Joint Effects of asthma and hay fever
 Asthma only1691.390.51–3.75550.270.06–1.23
 Hay fever only27420.510.27–0.9615180.440.16–1.20
 Both2291.980.81–4.83624.440.45–43.52
Family history
 Onset age 50+ yr76771.380.78–2.4426240.480.18–1.25
 Onset <5063322.581.35–4.9423190.780.29–2.09

The multivariate model for women identified a set of significant risk predictors very similar to the overall model (Table V): exposure to dusts (OR = 2.15), ETS exposure at work only (OR = 11.66) and family history of young onset cancer (OR = 2.58) conferred risk, while history of hay fever without asthma was protective (OR = 0.51). In men, represented by much fewer participants, only exposure to dusts (OR = 2.92) and to ETS at work (OR = 3.84) was significantly associated with risk. The point estimate of the effect of dusts was similar in both genders, while the ORs for ETS exposure at work differed substantially, probably because of a very low number of women exposed only at work. The effects of family history or history of respiratory conditions did not reach statistical significance in men (Table V).

Discussion

In this article, we used data on epidemiologic and exposure risk factors to assess their impact on lung cancer risk in never smokers. Overall, the most important risk factors for never smokers were exposure to ETS at work, exposure to dusts and family history of early onset cancer, while a history of hay fever was protective.

There are only a few studies analyzing epidemiological risk factors for lung cancer in never smokers.6, 7, 8, 9, 12, 14 ETS is one of the factors consistently shown to confer increased lung cancer risk in never smokers2, 3, 4, 5: almost one fourth of lung cancer cases among never smokers have been attributed to exposure to passive smoking.17 However, the magnitude of the effect of ETS varies from study to study and most studies have focused on women.18 A cohort study in Japan19 reported that nonsmoking women whose husbands smoked cigarettes were at a higher risk (RR = 1.4) for lung cancer when compared to nonsmoking women whose husbands were nonsmokers. Similarly, Trichopoulos et al.20 found an elevated lung cancer relative risk in wives of smokers (2.4 if the husband smoked <1 pack and 3.4 if the husband smoked >1 pack of cigarettes per day). In the present analysis, ETS exposure at work and both at work and home was significantly associated with risk, while exposure at home only was not. Very few women were exposed to ETS at work only, which resulted in an unstable estimate. Unfortunately, the questionnaire did not collect information on the number of smokers around the exposed individual. In a recent cohort study, Vineis et al.6 noted, as did we, that ETS exposure at work was a risk factor for lung cancer (OR = 2.17, 95% CI [1.16–4.08]), while exposure at home was not (OR = 0.82, 95% CI [0.37–1.82]).

Previous studies, in both smokers and nonsmokers, have reported that occupational and environmental exposure to particles are associated with an increased risk of lung cancer.21, 22 Borm et al.23 believe that insoluble particles cause pulmonary inflammation, which leads to genotoxic, proliferative and tissue remodeling processes including fibrosis and malignant transformation. Particles classified as human carcinogens are respirable crystalline silica (quartz and cristobalite), asbestos fibers and some hardwood dusts.23 In the present study, we found that exposure to at least 1 of several types of dusts was a significant risk factor in never smokers. Neuberger and Field14 noted that most studies on occupational lung cancer are based on data from male smokers, and residual confounding of smoking with lung cancer risk remains a problem even with adjustment for smoking. Our study of only never smokers suggested that the effect of occupational and environmental exposures such as dusts and ETS was real. Previously, Wu et al.24 using data on African-Americans and Mexican-Americans from this same case-control study provided evidence of a joint effect between wood dust exposure and smoking in lung cancer risk. In parallel with that finding, we also observed that individuals exposed to both ETS and dusts had a higher risk than individuals exposed to either of these alone.

Physician-diagnosed asthma was a risk factor in our study (OR = 1.82, 95% CI [1.05–3.15]); the risk remained relatively constant if individuals diagnosed within 3, 5, 10 or 20 years were excluded, and there was an increasing risk with increasing duration of the disease. Likewise, Santillan et al.25 in their meta-analysis found asthma to be a significant risk factor for lung cancer in never smokers. The combined estimate was 1.8, 95% CI [1.2–2.3], if the original studies reported estimates not necessarily adjusted for ETS exposure, or 1.9, 95% CI [1.4–2.5], if the original estimates were ETS-adjusted.

The overall effect of hay fever (OR = 0.83) was not statistically significant. However, we observed an association between diagnoses of hay fever and asthma, as have others.26 In a more detailed analysis looking at the joint effects of asthma and hay fever, we noted that reporting both conditions conferred risk (which did not reach strict statistical significance in the multivariate analysis), while hay fever alone (without asthma) was significantly protective and asthma alone (without hay fever) did not confer risk. We believe that more studies are needed to clarify this complex association, because the data on respiratory diseases in our study were self-reported and not validated. For this reason, the effects, while interesting, should not be viewed as conclusive. Recently, a prospective analysis of CPS-II data focusing on cancer mortality among US men and women with asthma and hay fever was conducted.27 This analysis was also based on self-reported history of physician-diagnosed conditions. Asthma was significantly and positively associated with lung cancer mortality (RR = 1.11, 95% CI [1.02–1.20], while hay fever alone and a history of both hay fever and asthma showed significant inverse associations (RR = 0.85, 95% CI [0.80–0.90] and RR = 0.73, 95% CI [0.65–0.83], respectively). However, when never smokers were analyzed separately, none of these associations remained significant.

The protective effect from hay fever confirms similar findings in all smoking categories combined from the same study population16 (OR = 0.58, 95% CI [0.48–0.70]). Similarly, Osann28 observed that never smokers with a history of asthma or hay fever experienced a significantly lower risk for lung cancer. However, her sample of never smokers was small (28 cases and 91 controls provided information on respiratory conditions) and limited to women. Moreover, separate effects of hay fever and asthma were not evaluated. In contrast, Mayne et al.12 found a more than 4-fold increased risk (4.33, 95% CI [1.23–15.21]) from asthma in former smokers, although only a nonsignificant 10% excess risk was found in never smokers. Cockcroft et al.29 have suggested that patients with respiratory atopy appeared to have some degree of protection against developing malignancies of endodermal origin. Other studies report no association between lung cancer and allergic disorders or atopy.30, 31 There are 2 different contradictory hypotheses on the role of allergies in cancer risk. The protective effects could be attributed to enhanced immune surveillance resulting in a stimulated immune system.28, 29 Anti-inflammatory agents to treat hay fever could also contribute to this protection. The other hypothesis maintains that chronic immune stimulation leads to random pro-oncogenic mutations in actively dividing stem cells and to an increased risk of cancer.32

Higher BMI was observed in controls than in cases at the time of enrollment, while BMI 5 years prior to enrollment did not show a significant case-control difference. Several previous studies including both smokers and nonsmokers10, 33, 34, 35, 36 reported higher BMI as protective against lung cancer. Henley et al.,37 on the other hand, found in analyses restricted to lifelong nonsmokers from a prospective study that among those who did not report preexisting disease, leanness was not substantially associated with lung cancer mortality. They concluded that the association, often observed between leanness and lung cancer in studies that include smokers and attempt to control for smoking, is likely an artifact of the effects of smoking and preexisting disease.

Overall, a family history of lung cancer and any cancer in first degree relatives did not affect the risk for lung cancer in this group of never smokers, unlike the data from Mayne et al.'s7 analysis, but in agreement with the finding of Hu et al.9 We observed, however, that early onset cancer was more often reported in relatives of cases than in relatives of controls, both for the whole group and among women, suggesting a role for genetic susceptibility. Previously, it was noted that family members of younger (both smoking and nonsmoking) probands had an increased lung cancer risk.38, 39, 40, 41 In particular, Schwartz et al.38 noted that family members of nonsmokers with early onset lung cancer had a 6-fold elevated risk of lung cancer (adjusting for the smoking history of the family members).

The sample size limited reliability of the multivariate analysis in men. We noted that the cell type distribution was somewhat different in males and females. Unfortunately, the sample size did not permit gender-specific analysis of risk factors separately by histology. OC use, HRT, menopausal status, parity and number of miscarriages did not affect lung cancer risk in this study. Yet, in a study by Kreuzer et al.42 as well as in female former and current smokers drawn from the same case-control study,43 HRT has been demonstrated to be protective.

The strength of this study includes its size as one of the largest reported for never smokers, and the availability of detailed risk factor data elicited by personal interview. By using weight 5 years prior to the diagnosis/interview, we reduced any influence of preclinical disease on weight. The information on ETS separately at home and at work was collected, reducing the probability of misclassification of individuals not exposed from their spouses as not exposed to ETS at all. Among the study limitations are absence of quantitative information on ETS exposure, the considerable amount of missing ETS exposure data, lack of specific information on intensity of ETS exposure, childhood ETS exposure (which was found to modify the effect of subsequent ETS exposure in adult life3, 5) and on ETS exposure at public places. Another limitation of the study is that the dust exposure data used lacked any information on the length of exposure. Misclassification and recall bias of exposure are of particular concern. Respiratory disease diagnoses were self-reported, and we were not able to validate the self-reported prior respiratory disease history. A higher prevalence of respiratory diseases in our hospital-based study population was reported (hay fever, 29% in controls, 26% in cases; asthma, 10.3% in controls, 18% in cases) when compared to the general US population (hay fever, 14–20%44, 45, 46; asthma, 4.5–6%26, 46, 47) or to the CPS-II data (hay fever, 11.7%; asthma, 4.5%),27 which may partly be due to overreporting. However, this is unlikely to be a bias attributed to our control selection, especially since asthma was reported more often in cases than in controls.

Further study on never smokers incorporating information on dietary factors and polymorphisms in genes involved in pathways related to DNA repair, carcinogen metabolism and inflammatory response is clearly warranted to throw more light on the etiology of lung cancer in lifetime never smokers.

Ancillary