The Journal of Nutrition Vol. 128 No. 4 April 1998, pp. 694-700

Alcoholic Beverage Consumption and Lung Cancer Risk among Residents of Los Angeles County^1,2,3,4,5,6

Catherine L. Carpenter^{*, 7}, Hal Morgenstern, and Stephanie J. London^{*, **, 8}

^* Department of Preventive Medicine, USC School of Medicine, Norris Comprehensive Cancer Center, Los Angeles, CA 90033;  Department of Epidemiology, UCLA School of Public Health, Los Angeles, CA 90095-1772; and ^** National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Although studies generally support a positive association between alcohol consumption and lung-cancer risk, the relationship between specific alcoholic beverages and lung-cancer risk has been inconsistent. We examined recent and past alcoholic beverage intake among 261 incident cases and 615 population controls enrolled in a lung-cancer case-control study of African Americans and Caucasians in Los Angeles County between 1991 and 1994. An in-person interview elicited information about past alcohol intake from ages 30 to 40 y, smoking, other lung-cancer risk factors, as well as recent intake of alcohol, and recent dietary intake. An association was observed between recent hard-liquor consumption and lung-cancer risk. The odds ratio (OR) for 1 or more drinks (1.5 oz or 0.051 mL) per day of hard liquor compared with infrequent liquor drinking (0-3 drinks per month), adjusted for smoking, the matching factors, saturated fat and other alcoholic beverages was 1.87 [95% confidence interval (CI) = 1.02-3.42]. No appreciable association was observed for total alcohol, whereas small inverse associations were observed for beer and wine, although confidence intervals were wide. An elevated lung-cancer risk was also observed for past liquor consumption (between ages 30 and 40 y). The adjusted OR for 1 or more drinks per day of liquor compared with infrequent drinkers was 1.83 (95% CI = 1.06-3.15). Confounding of the association between alcohol and lung cancer by smoking was apparent. Although we devoted considerable efforts to adjusting for smoking in our analyses, residual confounding is still possible because smoking and alcohol are closely associated. In addition, case-control studies including this study should be viewed with caution because of possible selection bias. An increased risk of lung cancer might occur with moderate drinking of hard liquor but confirmation is required in larger studies.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

Smoking is indisputably linked to lung cancer, yet only a small fraction of smokers develop lung cancer. Identification of genetic, environmental and nutritional factors that also affect lung-cancer risk might help to explain why some smokers are more likely to develop lung cancer than others.

Alcohol consumption is closely associated with smoking. It has been observed frequently that heavy drinkers are more likely to smoke than light drinkers or nondrinkers (Surgeon General 1988). As a consequence, measuring the association between alcohol and lung cancer has been problematic because of the difficulty in separating the effects of alcohol and smoking and the inability to fully account for the corresponding lifestyle risk factors associated with heavy alcohol consumption.

There are several ways in which alcohol consumption could be related to lung cancer. Alcohol consumption might indirectly affect lung-cancer risk through associated dietary changes (Hebert and Kabat 1991, Hillers and Massey 1985), nutrient deficiencies (Lieber et al. 1986, Morgan 1982, Vannucchi and Moreno 1989) or changes in plasma levels of carotenoids (Albanes et al. 1996, Forman et al. 1995, Rimm and Colditz 1993, Stryker et al. 1988). Alcohol could directly affect lung carcinogenesis by inducing cytochrome P-450 genes that activate procarcinogens in tobacco smoke (Seitz et al. 1981), by the production of lung surfactant phospholipids (Liau et al. 1981) or by serving as an immunosuppressant (Vitale and Gottlieb 1975).

Conflicting results were observed in early studies of alcohol intake and lung cancer. Several studies that found alcohol to be positively associated with lung cancer were limited by lack of a complete adjustment for smoking (Adami et al. 1992, Jensen 1979, Monson and Lyon 1975, Prior 1988, Schmidt and deLint 1972, Sigvardsson et al. 1996) or, when adjusted for smoking, the association went away (Restrepo et al. 1989), whereas other studies with null results (Mettlin 1989, Pierce et al. 1989) may have been affected by use of hospital controls. Several more recent studies that adjusted for smoking (Bandera et al. 1992, DeStefani et al. 1993, Hirayama 1990, Klatsky et al. 1981, Kvale et al. 1983, Omenn et al. 1996, Pollack et al. 1984, Potter et al. 1982 and 1992, Stockwell and Matanoski 1984) showed elevated risk estimates for alcohol consumption, but were inconsistent for which alcoholic beverages were associated.

The purpose of this study was to estimate the overall and beverage-specific effects of recent and past alcohol consumption on lung-cancer occurrence.

	MATERIALS AND METHODS

Abstract Introduction Methods Results Discussion References

Data from a case-control study designed to examine genetic markers for lung-cancer risk (London et al. 1995) were used to study the association between alcohol consumption and lung cancer. Between 1991 and 1994, incident cases of lung cancer were identified within 7 mo of diagnosis from 35 hospitals in Los Angeles County. These hospitals were chosen because they could provide the study with sufficient numbers of African Americans and Caucasians. Controls under age 65 y were randomly selected from licensed drivers who resided in Los Angeles County, whereas controls over 65 y were randomly selected from lists of MediCare Beneficiaries provided by the Health Care Finance Administration (HCFA).⁹ To try and achieve a balance in the distribution between cases and controls for age (within 10-y intervals), race and gender, controls were frequency matched for these factors to all lung cancer cases diagnosed at the provider hospitals in the previous 3 y. We attempted to enroll twice as many controls as cases (see London et al. 1995 for further details).

Eligibility criteria for the study were as follows: residence in Los Angeles County, age between 40 and 84 y, ability to complete a questionnaire in English, of Caucasian (non-Hispanic) or African American origin and having had no prior cancer other than nonmelanoma skin cancer.

A total of 1119 lung-cancer cases were identified from the 35 hospitals. After exclusions due to death (207), ineligibility (260) and physician refusal (36), 616 subjects remained. Of these 616 potentially eligible cases, 92 could not be located, 158 declined participation and the study ended before 10 could be enrolled; 356 eligible cases were enrolled in the study. The median interval from case diagnosis to enrollment was 3.7 mo and the maximum interval was 10 mo.

Letters inviting participation in the study were sent to 3193 potential controls. The address provided by the Department of Motor Vehicles (DMV) or HCFA tapes was incorrect for 752 subjects, and 157 persons had moved out of Los Angeles County, making these potential participants ineligible. A response was obtained from 1573 subjects of the 2284 subjects with valid addresses in Los Angeles County; of these, 94 were ineligible, 71 were deceased, 351 declined to participate and 1057 expressed willingness to participate. Among the 1057 willing subjects, 46 could not be reached before the study ended, 831 were eligible on screening, and 180 were ineligible. The study ended before we could enroll 100 of the 831 eligible. A total of 731 controls (473 Caucasians and 258 African Americans) were enrolled.

Enrollment consisted of completion of an in-person interview about known and possible risk factors for lung cancer, including smoking history, recent and past alcohol consumption between ages 30 and 40 y, dietary factors, occupational exposures, passive smoke exposure and family history of lung cancer.

The study was restricted to the 876 subjects (261 cases and 615 controls) who had complete information on smoking, past alcohol consumption, recent alcohol consumption and diet. A total of 95 cases and 116 controls were eliminated from the original sample of 1087 (356 cases and 731 controls), primarily because they did not complete the semiquantitative food-frequency questionnaire from which recent consumption was obtained.

The semiquantitative food frequency instrument (Willett et al. 1988) provided information on recent alcohol consumption and dietary intake. Cases reported their diets and alcohol intake for the year before lung-cancer diagnosis, and controls reported their diets and alcohol intake for the year before the interview date.

Past alcohol consumption was measured by asking subjects to recall how many cans or bottles of beer, glasses of wine, and drinks or shots of liquor, sherry or brandy, were consumed on the average between the ages of 30 and 40 y using the following discrete categories: never or <1 time per month; 1-3 times per month; 1 time per week; 2-4 times per week; 5-6 times per week; 1 time per day; 2-3 times per day; 4-5 times per day; or 6+ times per day. Responses for each beverage type were assigned the midpoint of the frequency category and transformed to number of drinks per month. Subjects who responded in the uppermost category, 6+ times per day, had their values truncated to 6.0 times per day and were transformed to 180 drinks per month. Total alcohol consumption was represented by summing drinks per month of beer (12 oz or 0.41 mL), wine (4 oz or 0.14 mL) and hard liquor (1.5 oz or 0.05 mL). The same categories were used to elicit recent alcohol consumption on the food-frequency questionnaire. Exposure categories for beer, wine and hard liquor were created by dividing the subjects into the same four intake categories: 3 drinks per month (infrequent drinkers or nondrinkers); 1-6 drinks per week (occasional drinkers); 1-2 drinks per day (regular drinkers) and 3+ drinks per day (heavy drinkers). Because of small numbers in the uppermost (3+ drinks per day) beverage consumption categories (for instance, 5 cases and 4 controls drank 3+ glasses of wine per day), we collapsed the regular and heavy consumption categories for beer, wine and liquor.

Smoking status was established using a reference date as follows: 1 y before lung cancer diagnosis for the cases and 1 y plus the median time in which cases were diagnosed before the interview date for the controls (15.7 mo). Never smokers answered no to the question, "Have you smoked 100 cigarettes or more over your lifetime?" Subjects who smoked were asked the age at which they started smoking regularly, as well as usual amount smoked. In addition, smokers were asked to recall up to three times they started and stopped smoking for at least 6 mo, and one time where they had smoked more heavily than their usual amount.

Pack years were computed by multiplying the usual amount smoked by the total duration of time subjects smoked at that level and then summed across periods of differing consumption. The number of years since quitting was computed by subtracting the date of quitting smoking from the reference date.

Unconditional logistic regression models were fit to the data (Hosmer and Lemeshow 1989). Odds ratios (OR) and their 95% confidence intervals (CI) were estimated using Epilog Plus, version 3.0 (Epicenter Software 1993). Trend tests were computed by fitting a logistic model to mean values of alcohol for each intake category.

The strong association between alcohol consumption and smoking prompted careful examination of confounding. Initially, efforts were made to best characterize the relationship between smoking and lung cancer (see London et al. 1995 for further details). We considered smoking variables that produced the largest log-likelihood score with the fewest degrees of freedom when added to a model with alcohol. The model that was best suited for confounding adjustment included indicator variables for approximate quartiles of pack-years (0, >0-15, 16-31, 32-53, 54+), and indicator variables for approximate quartiles of years since quitting smoking (<5, 5-10, 11-14, 15+ and never smokers).

In addition to smoking, other known or possible risk factors for lung cancer were included as covariates in the model if their inclusion changed the estimated effect of any alcohol variable by 10% or more, or if their two-sided P value, adjusting for other predictors, was <0.05. The final model used to estimate the association between alcohol consumption and lung-cancer risk included indicator terms for ethnicity, gender, pack-years, years since quitting and saturated fat [divided into categories based on USDA guidelines for daily consumption (U.S. Department of Agriculture 1995)], as well as a continuous term for age. Terms for fruits, vegetables, foods high in saturated fat, occupational exposures, family history and passive smoking had no appreciable influence on the association between alcohol and lung cancer or on model precision and therefore were not included. Terms for alcoholic beverage effects were estimated in individual models and estimated with all three beverages in the same model. Alcoholic beverage effects were estimated separately from total alcohol consumption effects.

	RESULTS

Abstract Introduction Methods Results Discussion References

The distribution of gender, ethnicity, smoking status, average age and histologic type of lung cancer are presented for cases and controls (Table 1). Mean ages for cases (63.6) and controls (62.5) did not differ (P = 0.10). Distributions for gender and ethnicity differed by 10 and 11%, respectively, for cases and controls. Adenocarcinoma and squamous cell carcinoma were the most common histologic type of lung cancer.

The small number (8) of never-smoking cases, (see Table 1) did not allow for assessment of the alcohol association among never smokers. As expected, smoking patterns were very different for cases and controls.

Because of the strong association between alcohol consumption and smoking observed in previous studies, it was important to adjust for confounding by smoking. Tables 2 and 3 present odds ratios unadjusted for smoking and odds ratios adjusted for smoking. Marked changes in the magnitude of odds ratios were noted after adjustment for the smoking variables. To assess patterns of smoking in relationship to alcohol consumption, the mean number of pack-years of smoking was computed for categories of total alcohol and beverage-specific consumption in the past. Among the controls, average pack-years increased as alcohol consumption increased. When this association was examined by type of alcoholic beverage, the mean number of pack-years was highest among heavy beer-drinkers (46.5), followed by heavy hard-liquor drinkers (41.6) and heavy wine drinkers (41.3). Pearson correlation coefficients between pack-years and alcoholic beverages (drinks per month) among ever-smoking controls were slight to moderate, with correlation between pack-years and past beer consumption equal to 0.20, followed by hard liquor (0.06) and wine (0.01).

Associations between recent alcohol consumption and lung-cancer risk are presented in Table 2. Models that estimated effects for each beverage unadjusted for other beverages generally produced odds ratios that were close in magnitude to odds ratios estimated from models with all three beverages in same model. In both cases, a positive association was observed between liquor consumption and lung-cancer risk. When adjusting for wine and beer consumption, the odds ratio for 1 or more drinks per day of hard liquor, compared with 3 drinks per month, was 1.87 (95% CI = 1.02-3.42; P for linear trend = 0.06). The observed effect was a little weaker and a little more precisely estimated without adjustment for the other beverage types. Small inverse associations were observed for the upper intake categories of wine (OR = 0.79; 95% CI = 0.34-1.86) and beer (OR = 0.86; CI = 0.44-1.75) with no clear trend evident (P = 0.66 for wine; P = 0.45 for beer). No association was observed for total alcohol consumption (OR for 3+ drinks per day = 1.07; CI = 0.46-2.47).

Presented in Table 3 are results for past consumption of alcohol for all subjects between the ages of 30 and 40 y. For models that estimated effects of each beverage adjusting for the other two, an 83% increased risk of lung cancer was observed for drinkers of 1 or more drinks per day of hard liquor, compared with the reference group (OR = 1.83; 95% CI = 1.06-3.15; P for linear trend = 0.06). These results observed for the effect of past liquor consumption are similar to the results for recent liquor consumption in Table 2.

To assess whether the hard-liquor association occurred predominantly in heavy smokers, we stratified alcoholic beverage consumption by light (<35 pack-year smoking history) vs. heavy smoking (35+ pack-year smoking). The estimated odds ratio for 1+ drink per day of hard-liquor in the past was 1.48 (95% CI = 0.58-3.80) in light smokers, and 2.11 (95% CI = 1.03-4.31) in heavy smokers. A test for heterogeneity indicated that differences were compatible with chance (P = 0.76).

Associations observed for the other alcoholic beverages were less precise. The odds ratio for 1 or more drinks of beer in the past, relative to infrequent drinking was 0.72 (95% CI = 0.44-1.19), whereas the odds ratio of 1 or more drinks of wine in the past was 0.60 (95% CI = 0.29-1.25). When we evaluated each beverage effect unadjusted for other beverages, associations were close in magnitude to associations observed in mutually adjusted models.

When odds ratios for the alcoholic beverages were stratified by histology (adenocarcinoma, squamous cell and small cell, and all other histologic types combined), the liquor association was positive for each histologic type (see Table 4). The strongest association was observed among the other histologic types (OR = 2.06, 95% CI = 0.93-4.54), and the weakest association among cases with adenocarcinoma (OR = 1.41, 95% CI = 0.63-3.15), although the differences may be due to chance given the small numbers within each histologic category.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

The results of this study suggest that the consumption of one or more drinks of liquor per day increases the risk of lung cancer. This association was observed for both recent and past consumption, and it does not appear to be due to confounding by smoking or other risk factors for lung cancer. As shown in Tables 2 and 3, the estimated effect of liquor on lung-cancer risk was a little stronger when adjusting for the effects of wine and beer consumption than when not adjusting for the other beverages. We believe that the larger adjusted estimate is probably less biased because the effects of different beverage types may confound each other, as pointed out by Kimball and Friedman (1992). Although small inverse effects were observed for the consumption of wine and beer, these associations did not exhibit a dose-response gradient and may have been chance findings.

This study attempted to circumvent the effect of disease symptoms on recent alcohol consumption by asking subjects to recall alcohol intake between ages 30 and 40 y. Even though recent alcohol intake was assessed 1 y before diagnosis for the cases, lung-cancer symptoms may have nonetheless affected consumption patterns. We examined the correlation between recent and past alcohol consumption and found that recent consumption was not correlated with past consumption among the controls (Pearson correlation coefficient = 0.12). Although recent and past alcohol consumption were not strongly correlated, the estimated effects on lung-cancer risk were similar in magnitude.

Because smoking is such a strong risk factor for lung cancer, it is important to carefully control for cigarette smoking, particularly when estimating the effects of smoking-related characteristics (Friedman 1996). Although some residual confounding by smoking might exist in our data, we took extra measures to adjust for smoking. In addition, we demonstrated that liquor drinkers did not smoke any more on average than did beer drinkers or wine drinkers.

Selection bias may have also contributed to the observed hard-liquor association. That is, study controls may have consumed less hard liquor than did the population at risk from which study cases arose. To address possible selection bias, we collected smoking information for potential controls who did not participate in the full study. A random sample of 201 persons was drawn from the 868 nonrespondents with a known address who were either not home when visited or who could not be reached by telephone. A mini-questionnaire was mailed out to these 201 individuals, and it was returned by 66 (33%). An additional 55 subjects who elected not to participate when visited at their homes also answered the mini-questionnaire. The proportion who ever smoked among the 121 nonrespondents who answered the questionnaire (61%) resembled the proportion of ever smokers in the enrolled control group (66%). The small difference in smoking between the nonrespondents and controls suggests that the selective sampling of eligible controls may not have influenced the results, at least with respect to smoking history. Although we have no information to determine whether control participation was associated with hard liquor consumption, the data on smoking provide some reassurance.

The exclusion of subjects because of missing alcohol information could have also resulted in selection bias. Nevertheless, the distributions of smoking history for cases and controls were similar for subjects with alcohol data and subjects with missing alcohol data. Eliminating subjects from the study with missing alcohol information did not therefore appreciably alter the distribution of smoking history in the overall sample.

Several recent studies have found positive associations for alcohol consumption. Two case-control studies (Bandera et al. 1992, DeStefani et al. 1993) and one nested study (Potter et al. 1992) that reported an alcohol-lung cancer association found that the alcohol effect was due to beer consumption. Both case-control studies were restricted to men, and both studies elicited alcohol consumption 1 y before the interview date. The nested case-control study was conducted in a cohort of 41,837 Iowa postmenopausal women. Measurement of alcohol consumption in the Iowa cohort was fairly recent, because the follow-up period was 3 y. It is possible that disease symptoms may have affected alcohol consumption in these studies. However, without measurement of alcohol in the past, it is difficult to conjecture how disease symptoms might have influenced recent alcohol consumption in these studies.

One prospective cohort study (Pollack et al. 1984) and a short report from a case-control study (Stockwell and Matanoski 1984) found positive effects for alcoholic beverages including hard liquor. The cohort study was conducted among 8006 Japanese men in Hawaii. Measurement of alcohol occurred 20 y before the end of follow-up. Positive effects were found for hard liquor [relative risk (RR) = 2.62] and wine (RR = 2.19) but not beer. Estimated effects from the case-control study of health hazards in the painting trades were positive for beer (OR = 2.27) and whiskey (OR = 2.26), but not wine.

Results from this study and the Hawaiian cohort study, both of which were positive for hard liquor, measured alcohol intake in the past. Despite a greater percentage of cases drinking more alcohol in the past than recently for this study, the estimates for both periods were relatively consistent, particularly for liquor consumption. In this study, the adjusted odds ratio for past hard liquor consumption was nearly unchanged when a continuous term for recent consumption was added to the model (OR = 1.84; 95% CI = 1.06-3.15 changed to OR = 1.80; 95% CI = 1.04-3.09). Thus in this study we were not able to separate the effects of recent and past alcohol consumption.

The inconsistency in results for alcoholic beverages observed across studies may have been due to different distributions of histologic cell types. The one study that reported results by histology (DeStefani et al. 1993) found high risk associated with beer consumption among all histologic types (squamous cell, small cell and adenocarcinoma), with squamous cell the most frequent type. In this study, adenocarcinoma was the most frequent histologic type. Because the power to detect differences in association for the histologic cell types was limited, we could not assess whether the liquor association was due to the high proportion of adenocarcinoma cases in our study, although the positive association we observed for each cell type leads us to believe that histologic differences were probably not responsible.

The inconsistent beverage associations may also have been due to different patterns of alcoholic beverage usage in different populations. That is, an association between alcohol and lung cancer observed for a particular beverage may be due to the predominant usage of that beverage in the population. Examining mean beverage intake among three population-based studies (Bandera et al. 1992, Pollack et al. 1984, Potter et al. 1992), and this study shows that in most cases, the beverage most often used was not the one associated with lung-cancer risk. Beer, the most predominant beverage consumed in the Hawaiian cohort (mean = 21 drinks per month), the Bandera study with New York male controls (mean = 18.2 drinks per month) and controls from the present study (mean = 12.5 drinks per month), was positively associated with lung-cancer risk only in the Bandera study. The most frequent beverage consumed among Iowa postmenopausal women was hard liquor (mean = 0.8 drinks per week), whereas lung-cancer risk was positively associated with beer.

Because the effect of liquor was observed primarily for the highest (open-ended) category of consumption (1+ drinks per day), the contrasting results for other alcoholic beverages might be due to differences in consumption distributions for the highest categories; it is possible, for example, that very heavy drinkers preferred liquor to wine or beer. To address this possibility, we also fit logistic models in which the amount of each beverage type was expressed as a continuous variable. The results were consistent with the categorical approach. The estimated odds ratio corresponding to one drink per month was 1.006 for liquor (95% CI = 1.00-1.009; P = 0.06), 0.999 for wine (95% CI = 0.990-1.009; P = 0.88) and 1.002 for beer (95% CI = 0.997-1.008; P = 0.43).

The liquor effect may have been a function of concentration, that is, ethanol is more highly concentrated in hard liquor than in beer or wine. The association between hard liquor and esophageal cancer reported in the literature has been partially attributed to ethanol concentration (Craddock and Henderson 1991), but whether the same explanation is applicable to lung cancer is unclear.

In conclusion, our results provide some evidence for an increased lung-cancer risk among subjects who consumed one or more drinks per day of liquor both recently and in the past. Hard-liquor drinkers in our study were not any more likely to smoke than were beer drinkers or wine drinkers. However, we were not able to account for selection bias that would result if participation by the controls was associated with hard-liquor consumption. We demonstrated that at least with respect to smoking, controls who refused were similar to controls who participated.

	ACKNOWLEDGMENTS

The authors recognize the following people for their assistance in conducting the study: Jan Lowery and Lena Masri for project coordination and Corinne Singer, Regina Olivas-Ho, Kisha Barnes, Steve Grossman and Leanna Wolfe for enrollment of subjects. We thank the physicians and staff at participating hospitals for their cooperation. The authors acknowledge Sander Greenland for advice about the analysis and Diana Petitti for review of the manuscript.

	FOOTNOTES

Manuscript received 10 April 1997. Initial reviews completed 10 July 1997. Revision accepted 8 December 1997.

	LITERATURE CITED

Abstract Introduction Methods Results Discussion References

• Adami H. O., McLaughlin J. K., Hsing A. W., Wolk A., Ekbom A., Holmberg L., Persson I. Alcoholism and cancer risk: a population based cohort study. Cancer Causes Control 1992; 3:419-425 [Medline]
• Albanes D., Heinonen O. P., Taylor P. R., Virtamo J., Edwards B. K., Rautalahti M., Hartman A. M., Palmgren J., Freedman L. S., Haapakoski J., Barrett M. J., Pietinen P., Malila N., Tala E., Liippo K., Salomaa E. R., Tangrea J. A., Teppo L., Askin F. B., Taskinen E., Eorzan Y., Greenwald P., Huttunen J. K. Alpha-tocopherol and beta-carotene supplements and lung cancer incidence in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study: effects of base-line characteristics and study compliance. J. Natl. Cancer Inst. 1996; 88:1560-1570 [Abstract/Free Full Text]
• Bandera E. V., Freundenheim J. L., Graham S., Marshall J. R., Haughey B. P., Swanson M., Brasure J., Wilkinson G. Alcohol consumption and lung cancer in White males. Cancer Causes Control 1992; 3:361-369 [Medline]
• Craddock, V. M. & Henderson, A. R. (1991) Potent inhibition of oesophageal metabolism of N-nitrosomethylbenzylamine, an oesophageal carcinogen, by higher alcohols present in alcoholic beverages. In: Relevance to Human Cancer of N-Nitroso Compounds, Tobacco Smoke and Mycotoxins. IARC Scientific Publications, No. 105, International Agency for Research on Cancer. Lyon, France.
• De Stefani E., Correa P., Fierro L., Fontham E.T.H., Chem V., Zavala D. The effect of alcohol on the risk of lung cancer in Uruguay. Cancer Epidemiol. Biomarkers Prev. 1993; 2:21-26 [Abstract]
• Epicenter Software (1993) Epilog Plus: statistical Package for Epidemiology and Clinical Trials. Epicenter Software, Pasadena, CA.
• Forman M. R., Beecher G. R., Lanza E., Reichman M. E., Graubard B. I., Campbell W. S., Marr T., Yong L. C., Judd J. T., Taylor P. R. Effect of alcohol consumption on plasma carotenoid concentrations in premenopausal women: a controlled dietary study. Am. J. Clin. Nutr. 1995; 62:131-135 [Abstract/Free Full Text]
• Friedman G. D. Invited commentary: depression, smoking, and lung cancer. Am. J. Epidemiol. 1996; 144:1104-1106 [Free Full Text]
• Hebert J. R., Kabat G. C. Implications for cancer epidemiology of differences in dietary intake associated with alcohol consumption. Nutr. Cancer 1991; 15:107-119 [Medline]
• Hillers V. N., Massey L. K. Interrelationships of moderate and high alcohol consumption with diet and health status. Am. J. Clin. Nutr. 1985; 41:356-362 [Abstract/Free Full Text]
• Hirayama, T. (1990) Contribution of a long-term prospective cohort study to the issue of nutrition and cancer with special reference to the role of alcohol drinking. Recent Prog. Res. Nutr. Cancer (Proc.): 179-187.
• Hosmer, D. W. & Lemeshow, S. (1989) Applied Logistic Regression. Wiley Series in Probability and Mathematical Statistics, John Wiley & Sons, New York, NY.
• Jensen O. M. Cancer morbidity and causes of death among Danish brewery workers. Int. J. Cancer 1979; 23:454-463 [Medline]
• Kimball A. W., Friedman L. A. Alcohol consumption regression models for distinguishing between beverage type effects and beverage preference effects. Am. J. Epidemiol. 1992; 135:1279-1286 [Abstract/Free Full Text]
• Klatsky A. L., Friedman G. D., Siegelaub A. B. Alcohol and mortality: a ten-year Kaiser-Permanente experience. Ann. Intern. Med. 1981; 95:139-145
• Kvale G., Bjelke E., Gart J. J. Dietary habits and lung cancer risk. Int. J. Cancer 1983; 31:397-405 [Medline]
• Liau D. F., Hashim S. A., Pierson R. N., Ryan S. F. Alcohol-induced lipid change in the lung. J. Lipid Res. 1981; 22:680-686 [Abstract]
• Lieber C. S., Garro A., Leo M. A., Mak K. M., Worner T. Alcohol and cancer. Hepatology 1986; 6:1005-1019 [Medline]
• London S. J., Daly A. K., Cooper J., Navidi W. C., Carpenter C. L., Idle J. R. Polymorphism of glutathione S-transferase M1 and lung cancer risk among African Americans and Caucasians in Los Angeles County, California. J. Natl. Cancer Inst. 1995; 87:1246-1253 [Abstract/Free Full Text]
• Mettlin C. Milk drinking, other beverage habits, and lung cancer risk. Int. J. Cancer 1989; 43:608-612 [Medline]
• Monson R. R., Lyon J. L. Proportional mortality among alcoholics. Cancer 1975; 36:1077-1079 [Medline]
• Morgan M. Y. Alcohol and nutrition. Br. Med. Bull. 1982; 38:21-29 [Free Full Text]
• Omenn G. S., Goodman G. E., Thornquist M. D., Balmes J., Cullen M. R., Glass A., Keogh J. P., Meyskens F. L. Jr., Valanis B., Williams J. H., Barnhart S., Cherniack M. G., Brodkin C. A., Hammar S. Risk factors for lung cancer and for intervention effects in CARET, the Beta-Carotene and Retinal Efficacy Trial. J. Natl. Cancer. Inst. 1996; 88:1550-1559 [Abstract/Free Full Text]
• Pierce R. J., Kune G. A., Watson L. F., Field B., Merenstein D., Hayes A., Irving L. B. Dietary and alcohol intake, smoking pattern, occupational risk, and family history in lung cancer patients: results of a case-control study in males. Nutr. Cancer 1989; 12:237-248 [Medline]
• Pollack E. S., Nomura A.M.Y., Heilbrun L. K., Stemmermann G. N., Green S. B. Prospective study of alcohol consumption and cancer. N. Engl. J. Med. 1984; 310:617-621 [Abstract]
• Potter J. D., McMichael A. J., Hartshorne J. M. Alcohol and beer consumption in relation to cancers of bowel and lung: an extended correlation analysis. J. Chron. Dis. 1982; 35:833-842 [Medline]
• Potter J. D., Sellers T. A., Folsom A. R., McGovern P. G. Alcohol, beer, and lung cancer in postmenopausal women: The Iowa Womens Health Study. Ann. Epidemiol. 1992; 2:587-595 [Medline]
• Prior P. Long-term cancer risk in alcoholism. Alcohol and Alcoholism 1988; 23:163-171 [Abstract/Free Full Text]
• Restrepo H. E., Correa P., Haenszel W., Brinton L. A., Franco A. A case-control study of tobacco-related cancers in Colombia. Bull. PAHO 1989; 23:405-413
• Rimm E., Colditz G. Smoking, alcohol, and plasma levels of carotenes and vitamin E. Ann. N. Y. Acad. Sci. 1993; 686:323-334 [Medline]
• Schmidt W., de Lint J. Causes of death of alcoholics. Q. J. Stud. Alcohol 1972; 33:171-185 [Medline]
• Seitz H. K., Garro A. J., Lieber C. S. Enhanced pulmonary and intestinal activation of procarcinogens and mutagens after chronic ethanol consumption in the rat. Eur. J. Clin. Invest. 1981; 11:33-38 [Medline]
• Sigvardsson S., Hardell L., Przybeck T. R., Cloninger R. Increased cancer risk among Swedish female alcoholics. Epidemiology 1996; 7:140-143 [Medline]
• Stockwell, H. G. & Matanoski, G. M. (1984) Letters to the editor. N. Engl. J. Med. 311: 256.
• Stryker W. S., Kaplan L. A., Stein E. A., Stampfer M. J., Sober A., Willett W. C. The relation of diet, cigarette smoking, and alcohol consumption to plasma beta-carotene and alpha-tocopherol levels. Am. J. Epidemiol. 1988; 172:283-296
• U.S. Department of Agriculture (1995) Nutrition and Your Health: Dietary Guidelines for Americans, 4th ed. U.S. Government Printing Office, Washington, DC.
• U.S. Department of Health and Human Services (1988) Report of the Surgeon General. The Health Consequences of Smoking: Nicotine Addiction. U.S. Government Printing Office, Washington, DC.
• Vannucchi H., Moreno F. S. Interaction of niacin and zinc metabolism in patients with alcoholic pellagra. Am. J. Clin. Nutr. 1989; 50:364-369 [Abstract/Free Full Text]
• Vitale J. J., Gottlieb L. S. Alcohol and alcohol-related deficiencies as carcinogens. Cancer Res. 1975; 35:3336-3338 [Abstract/Free Full Text]
• Willett W. C., Sampson L., Browne M. L., Stampfer J. J., Rosner B., Hennekens C. H., Speizer F. E. The use of a self-administered questionnaire to assess diet four years in the past. Am. J. Epidemiol. 1988; 127:188-199 [Abstract/Free Full Text]

0022-3166/98 $3.00 ©1998 American Society for Nutritional Sciences