Alcohol Consumption not Associated with Lung Cancer Mortality in Lifelong Nonsmokers

Alcohol consumption is known to increase the risk of cancers of the oral cavity, pharynx, larynx, esophagus, liver, colorectum, and breast (1). However, evidence linking alcohol consumption to risk of lung cancer is considered “limited” (1). As drinking is strongly associated with tobacco smoking, positive associations between alcohol consumption and lung cancer risk in studies that include smokers may reflect residual confounding by smoking, rather than an independent effect of alcohol (1). Seven studies have published data relating lung cancer risk to alcohol consumption separately in lifelong nonsmokers or former smokers who ceased smoking for at least 20 years (2-8). Five of these found no relationship between any level of daily alcohol consumption and lung cancer risk in this subgroup (2-6), whereas two studies reported increased risk associated with higher levels of alcohol consumption (7, 8). The most recent and larger of the positive studies (7) pooled data on lifelong nonsmokers from seven cohorts participating in the Pooling Project of Prospective Studies of Diet and Cancer (9, 10). Never-smoking men who reported drinking at least one drink per day (equivalent to ≥15 grams of alcohol per day) in this pooled analysis had six times higher lung cancer incidence [relative risk 6.38 (95% confidence interval (CI), 2.74-14.9)] than never-smokers who reported no drinking (7). The interpretation of this was unclear, however, because no association was observed in women, and the reference group in men included only 10 lung cancer cases among those who reported neither smoking nor drinking, increasing the potential for bias or chance.

We measured death rates from lung cancer in relation to self-reported alcohol consumption among 223,216 adult men and women who reported no history of regular smoking when enrolled in Cancer Prevention Study II (CPS-II), a large prospective cohort study begun by the American Cancer Society in the fall of 1982 and followed through 2006.

Participants were drawn from CPS-II, a nationwide prospective mortality study described in detail elsewhere (11). Briefly, 1.2 million participants age 30 y or older were enrolled by ACS volunteers in all 50 states, the District of Columbia, and Puerto Rico. The entire cohort was followed for 24 y, through December 31, 2006. Vital status was ascertained through personal inquiries by volunteers in September 1984, 1986, 1988, and thereafter through linkage with the National Death Index. Underlying cause of death was ascertained directly from death certificates through 1988 and thereafter from codes provided by National Death Index. As of December 31, 2006, 41.5% of participants were deceased, 58.3% alive, and 0.2% lost to follow-up in 1988 due to insufficient data for linkage with the National Death Index.

Information on tobacco use, alcohol consumption, and other exposures was obtained through confidential mailed questionnaires at the time of enrollment. We excluded from analyses individuals who reported ever having smoked cigarettes, cigars, or pipes regularly (n = 630,974, 53.3%), and participants who did not report their smoking habits or provided inconsistent information (n = 71,070; 6.0%). Updated information on smoking was not ascertained during the duration of the study; however, as participants were at least age 30 y when enrolled, they were consequently considered unlikely to initiate smoking during follow-up (12). We also excluded participants who, at baseline, reported a history of a cancer diagnosis other than nonmelanoma skin cancer (n = 33,852; 2.9%).

Baseline information on alcohol consumption derived from the question, “How many cups, glasses, or drinks of these beverages do you usually drink in a day, and for how many years?” Participants were instructed to report their current use separately for beer, wine, and “hard liquor,” and to report past consumption if their pattern had changed during the last 10 y. Consumption of less than one drink per day but at least thrice a week was to be indicated by writing “1/2.” The current analyses are based on total alcohol consumption, derived from responses on all three alcoholic beverages. As in a previous publication from this cohort (13), we defined five levels of consumption: nondrinkers; less than 1 drink per day, but at least thrice a week; 1 drink per day; 2 to 3 drinks per day; and 4 or more drinks per day. “Nondrinkers” were participants who explicitly recorded zero for current amount of consumption of all three alcoholic beverages and who either recorded zero for previous amount of alcohol consumption or failed to report previous drinking habits. Participants who did not report use of a certain alcoholic beverage were considered nonconsumers for that particular beverage if they indicated valid current consumption of at least one of the other alcoholic drinks.

Among the 448,566 people still available for analysis, 205,706 participants (45.6%) did not complete the alcoholic beverage section of the questionnaire and another 9,550 participants (2.1%) did not provide consistent data on alcohol consumption. We excluded these individuals from the primary analysis, but included them in additional analyses to examine possible biases caused by individuals with missing data on alcohol consumption. Of the 233,310 never-smokers whose report of alcohol consumption was complete and consistent, we excluded 4,316 individuals who were former drinkers but current abstainers, as well as 5,778 participants who reported seasonal or occasional drinking. After these exclusions, the primary analyses were based on 72,969 men and 150,247 women. Lung cancer (ICD 9, codes 162.0-162.9; ICD-10, codes C33-C34.9) was the primary cause of death for 406 men and 652 women in the analytic cohort during follow-up.

Cox proportional hazards modeling was used to calculate gender-specific hazard ratios (HR) and corresponding 95% CIs for the association between alcohol consumption and lung cancer mortality in lifelong nonsmokers, with time since enrollment as the underlying time metric. All models were stratified by single year of age at enrollment and adjusted for level of education (less than high school graduate, high school graduate or vocational school, some college, college graduate, graduate school, or missing), occupation (white collar, blue collar, housewife, or missing), and race (White, other), as ascertained on the 1982 baseline questionnaire. Tests of trend were obtained for increasing levels of alcohol consumption by assigning an ordinal value to each successive level and reporting the resulting two-tailed P value for the variable. Nondrinkers were excluded from the test for trend. The Cox proportional hazards assumption was evaluated by an examination of Kaplan-Meier curves, as well as by testing for an interaction by time in the model.

To explore possible biases introduced into the results by excluding the large number of lifelong nonsmokers who were missing data on alcohol consumption, we conducted additional analyses. An indicator variable designating missing data was added to the previously described multivariate models, and the HR for lung cancer associated with this group was compared with that of nondrinkers. We also determined the HR for this group compared with nondrinkers for death from other cancers known to be causally related to alcohol consumption (esophagus, larynx, pharynx, oral cavity, and liver).

Table 1 shows the numbers of lifelong nonsmokers, deaths from lung cancer, and selected demographic characteristics by level of alcohol consumption. Approximately 45% of women and 36% of men in the analytic cohort reported no alcohol consumption. Men who reported abstaining from alcohol or consuming 4+ drinks per day were somewhat less educated and more likely to have worked in a blue collar occupation than men who reported drinking <1 to 3 drinks per day. Women who reported no alcohol consumption had the lowest educational attainment of any group.

Table 2 indicates that the HR estimates associated with each level of alcohol consumption were almost identical whether adjusted for age only or for age, education, occupation, and race. Lung cancer risk was not positively associated with any level of alcohol consumption in men or women, even among those who reported four or more alcoholic drinks per day. Lung cancer mortality was inversely associated with the level of alcohol consumption in women, although the trend was not statistically significant (P_trend = 0.058). None of the analyses violated the Cox proportional hazards assumption.

We assessed the validity of the self-reported information on alcohol consumption by testing the association between heavy consumption (≥4 drinks per day) and mortality from conditions that are causally related to excessive drinking. Compared with death rates in nondrinkers, men and women, respectively, who reported ≥4 drinks per day had increased death rates from liver cirrhosis [HR, 5.85 (95% CI, 3.59-9.52); HR, 3.05 (95% CI, 2.05-4.54) and from five cancers related to alcohol consumption [oral cavity, pharynx, larynx, esophagus, and liver, HR, 2.00 (1.35-2.97); HR, 1.52 (0.83-2.47)].

To assess the stability of alcohol consumption during follow-up, we compared information provided at two time points in a subset (n = 32,869) of the never-smokers who reenrolled in the CPS-II Nutrition subcohort in 1992 (14). Consumption remained relatively stable in this subgroup over the 10-year interval. Eighty-two percent of those who reported abstaining from alcohol in 1982 again reported abstinence in 1992, whereas another 15% reported consuming alcoholic beverages only one to four times each month. Among current drinkers in 1982, 75.8% of participants continued to consume alcohol at the same level or at adjacent levels of frequency according to the classification used here.

Finally, we assessed whether the exclusion of almost half of the never-smokers due to missing information on alcohol consumption could have biased our results by removing very heavy drinkers from the analysis. The death rates from liver cirrhosis, alcohol-related cancers, and lung cancer were not significantly different among people missing all alcohol data than in nondrinkers. The multivariate-adjusted HRs associated with missing alcohol data were as follows in men and women, respectively: liver cirrhosis [HR, 1.48 (95% CI, 0.96-2.27); HR, 0.74 (0.59-0.91)], alcohol-related cancers [HR, 1.19 (95% CI, 0.91-1.55); HR, 0.92 (95% CI, 0.76-1.11)], and lung cancer [HR, 1.07 (95%CI, 0.88-1.30); HR, 1.08 (95% CI, 0.94-1.23)].

Results from this large prospective study provide no support for and substantial evidence against the hypothesis that alcohol consumption independently influences lung cancer risk. The death rate from lung cancer was not associated with any level of alcohol consumption among the never-smoking men and women in our study, even in those who reported drinking four or more alcoholic beverages per day. No positive trends in lung cancer risk were observed with the level of alcohol consumption in men or women. Indeed, lung cancer risk seemed to be inversely related to the level of alcohol consumption among the never-smoking women, although the trend was not statistically significant.

Strengths of our study are its size, length of follow-up, and the availability of repeated questionnaire information with which to assess the stability of alcohol consumption. Very large studies with long-term follow-up are essential when measuring lung cancer risk prospectively in lifelong nonsmokers. We based our analyses on 1,058 lung cancer deaths among lifelong nonsmokers, over thrice the number of incident cases among never-smokers in the pooled analysis by Freudenheim et al. (7), and more than five times larger than in other cohort studies in analyses restricted to lifelong nonsmokers (2-6, 8). Because smoking is strongly associated with both alcohol consumption and lung cancer, restricting analyses to men and women who report no history of smoking provides the most effective way to minimize potential confounding by active smoking.

Limitations of our study were that information on alcohol consumption was self-reported, that repeated questionnaire information was available on only a subset of participants after the first 10 years of follow-up, and that almost half of the never-smokers had to be excluded from the analyses because of missing information on alcohol consumption. Despite this, the heaviest level of alcohol consumption (≥4 drinks daily) was strongly associated with mortality from cirrhosis and cancers related to alcohol consumption during the 24-year follow-up of this cohort, as was reported previously after 9 (13) and 14 years of follow-up (15). There was also no evidence that the exclusion of people with missing data might obscure a relationship between alcohol consumption and lung cancer because alcohol was not associated with lung cancer among subjects with missing data.

In summary, the findings from this large prospective study provide the strongest evidence to date that alcohol consumption is not related to lung cancer risk, independent of its association with smoking.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part bythe 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.