J Clin Epidemiol Vol. 52, No. 1, pp. 83–89, 1999 Copyright © 1999 Elsevier Science Inc. All rights reserved. 0895-4356/99/ $–see front matter PII S0895-4356(98)00132-2 Comparing Participants and Nonparticipants in a Smoking Cessation Trial: Selection Factors Associated With General Practitioner Recruitment Activity Carlo Senore, 1,* Renaldo N. Battista, 3 Antonio Ponti, 1 Nereo Segnan, 1 Stanley H. Shapiro, 2 Stefano Rosso, 1 and Daniela Aimar 1 1 Unita’ di Epidemiologia, Dipartimento di Oncologia, CPO-Piemonte, Torino, Italy; 2 Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, Canada; and 3 Department of Epidemiology and Biostatistics, McGill University and the Division of Clinical Epidemiology, the Montreal General Hospital, Montreal, Quebec, Canada ABSTRACT. We studied the relationship between smokers’ sociodemographic characteristics, their smoking habit, health status, and the probability of their having been approached for recruitment in the smoking cessation trial performed in Turin, Italy, with the aim of gathering information on the role of selection criteria adopted by general practitioners (GPs) in offering anti-smoking counseling. The 965 smokers who were offered participation in the trial were matched to a sample of eligible smokers (n = 277), selected from the rosters of the 42 GPs collaborating in the trial, who had not been invited to participate. The probability of being offered enrollment, estimated through a multiple conditional logistic regression model, assuming the GP as the matching variable, was significantly increased for intermediate (10–19 cigarettes per day: odds ratio [OR] = 4.13; 95% confidence interval [CI]: 2.63–6.47) and heavy (20 cigarettes per day or more: OR = 10.12; 95% CI: 6.51–15.75) smokers, for smokers diagnosed with chronic cardiovascular (OR = 2.06; 95% CI: 1.19–3.58), or respiratory (OR = 2.50; 95% CI: 1.40–4.48) diseases, and for smokers mentioning an intermediate number (2–4) of past quit attempts (OR = 3.70; 95% CI: 2.18–6.28). General Practitioners focused their recruitment activity on higer-risk smokers or smokers who had tried to quit, to offer more clues for intervention. Assessing the potential public health benefit of preventive interventions requires a more systematic evaluation of the generalizability of the reported findings. j clin epidemiol 52;1:83–89, 1999. © 1999 Elsevier Science Inc. KEY WORDS. RCT, smoking cessation, selection factors, general practitioner, counseling, external validity INTRODUCTION Various reports [1–4] suggested that smokers’ characteristics influence their likelihood of being offered smoking cessa- tion advice in the context of routine clinical activity. Physicians apparently focus on the smokers they perceive to be at increased risk of ill health effects of cigarettes. Pre- vious diagnoses of smoking related diseases [1–4] higher levels of tobacco consumption [1,3], and age [3] emerged as important predictors of counseling. Smokers reporting sev- eral attempts to quit in the past [4] or a longer duration of smoking habit [1,4], as well as those reporting a higher level of attendance at a doctor’s office [1–3], were more likely to have been told to stop. It has been suggested [5] that this kind of selective coun- seling practice be maintained when recruiting smokers for intervention studies. The differential decline of the effec- tiveness rate, observed across physicians participating in a cessation trial [5], was ascribed to an increased selectivity in delivering counseling. Indeed, some physicians not only counseled patients less often but also adopted more restric- tive criteria than indicated in the study protocol, recruiting subjects with symptoms of pulmonary disease, or heavy smokers. Although the low number (n = 4) of physicians involved in the study [5] limits the strength of this conclu- sion, the reported findings suggest that the generalizability of the results of smoking cessation trials in general practice might be hampered by the persistent tendency among gen- eral practitioners (GPs) to target high-risk smokers. The introduction of unplanned selection criteria could affect the external validity of the observed results, thus im- pairing their application to clinical practice. Indeed, im- pact assessments of such interventions are based on the as- sumption that GPs can take the opportunity of office visits to offer their support for quit attempts to all smokers, re- gardless of their health status or level of addiction. The role of selective criteria adopted by some GPs in of- fering antismoking advice was explored by analyzing the re- *Address correspondence to: Dr. Carlo Senore, Unita’ di Epidemiologia, CPO–Piemonte, ASL 1 Torino, Via S. Francesco da Paola 31, 10123 Torino, Italy. Accepted for publication on 6 July 1998.