Validity of self-reported smoking habits

Editor Smoking is one of the most significant negative health factors in the civilized world. A comprehensive follow-up study by British doctors has shown that on average smokers live 7 years less than non-smokers (1). Estimation of smoking habits is consequently an important clinical and scientific problem. Haug et al. have recently studied 146 Norwegian pregnant women and 102 nonpregnant women’s self-reported smoking habits and changes in the bio-chemical marker serum thiocyanide (2). We have validated information on smoking habits among 905 Danes between 30 and 50 years of age; they are part of the Ebeltoft Health Promotion Study, which is a prospective randomized study of the value of prophylactic measures (health checks, health talks) in general practice (3). We believe that the results have a broad interest, which is why we summarize them here. In a postal questionnaire the participants gave information as to whether they were non-smokers, ex-smokers or occasional smokers, and they stated their average daily tobacco consumption. Number of cigarettes, cheroots, cigars, and grams of pipe tobacco per day were converted to corresponding numbers of cigarettes (Q-cigarettes). The questions concerning self-reported smoking habits were repeated at the health check 2-6 months later at an interview conducted by a nurse. The total tobacco consumption (I-cigarettes) was converted in the same way and registered as follows: 0, 0-10, 11-20, > 20 cigarettes per day. Without the participants’ knowledge, the CO content in the expiratory air was subsequently measured. A concentration of 8 ppm has previously proved a useful value for discriminating between smokers and non-smokers (43). Seven participants were excluded because of incomplete data; categorization of the others can be seen in the table. The correlation between CO in expiratory air and daily tobacco consumption was assessed with Spearman’s test, and the degree of correlation was measured with Spearman’s p. The association between CO and self-reported tobacco consumption in the questionnaire showed a significantly strong correlation (pd.82). The correlation was even stronger (pS.85) when the participants were interviewed face to face. The design of the study, with an interval between completion of the questionnaire and the interview/CO measurement, hinders a direct comparison of the validity of the two data collection methods since several participants might have changed their smoking habits. At the interview, 16 of the previous 43 occasional smokers, 6 of the ex-smokers, but only 1 of the non-smokers, had become daily smokers since completion of the questionnaire. Moreover 13 of these had a CO concentration > 8 ppm, which causes a stronger correlation. Consequently our analysis, like bio-chemical studies, conf‘iis that self-reported tobacco consumption is a reliable way to measure smoking habits in a population. However, an American study has shown that the reported consumption was approximately 70% of what it ought to be according to consumption data from cigarette excise taxes (6). This could indicate some underreporting. On the other hand a Danish study on health status found a near-correlation between the two measures of tobacco consumption (7). Differentiated mis-classification is a threat to the inter-