Preliminary results from this ongoing study demonstrate the feasibility of using an active user-intercept protocol to recruit smokers into a research trial. Linking a study recruitment protocol to highly utilized search engines through active user sampling results in very high initial reach (28,297) to a large sample of smokers looking for cessation assistance. There are significant portions of this initial pool of participants that drop out of the recruitment flow prior to enrollment. Using the broadest population denominator, preliminary results suggest that approximately 2.7% of Internet users looking for online cessation information will enroll in a research trial such as this one. However, this denominator may be overly conservative and not appropriate to estimate reach: It is analogous to counting all newspaper subscribers who read an advertisement for a cessation trial, many of whom will not even call the research center. Other potential denominators to estimate the reach of this recruitment approach to the population of interest
are those who indicated some preliminary interest (11,147) and/or those who were eligible (5,557) and recruited (3,614). Reach estimates using these denominators were 6.9%, 13.7%, and 21.1%, respectively. Given that more than 8 million smokers use the Internet to find cessation information each year (Fox, 2005
), active user sampling could potentially reach thousands of smokers with the opportunity to participate in a research trial.
The final sample differs from eligible participants on daily smoking rate, time to first cigarette, number of previous quit attempts, and age of first puff, suggesting a sample of heavier, more dependent smokers who may be more motivated to quit smoking. Education level did not appear to influence enrollment. Differences in race and gender were also observed, but may be attributed to investigator control of recruitment. Including a small number of additional questions in online eligibility screening allows for such comparisons and seems to be well accepted by participants. As in other studies of Internet cessation (Etter & Perneger, 2001
; Feil et al., 2003
; Koo & Skinner, 2005
; Stoddard et al., 2005
), the majority of our sample is comprised of younger, female, college-educated individuals with substantial Internet experience. The ability to adjust recruitment volume (oversampling or undersampling) based on race and gender has been critical to meeting goals for under-represented minority populations. To date, we have used this feature primarily to control enrollment of White females who comprised 62% of the participants enrolled in the first 3 months. Minority recruitment has proceeded more slowly, with Blacks enrolling at the highest rate among minority groups. Although online recruitment can be used to conduct preliminary eligibility screening and consent, eligibility and consent should be reconfirmed prior to randomization.
As reported in Cobb, Graham, Bock, Papandonatos, and Abrams (2005a)
, many individuals turn to the Internet for support within days of their quit. In the present study, individuals were excluded who had not smoked at all in the prior 24 hr. This stringent eligibility criterion was selected since this is one of the first large-scale randomized controlled trials of Internet smoking cessation. As a result, approximately 24% (181 of 764) of potential participants were determined to be ineligible. Since the majority of quit attempts in the United States are unassisted (Center for Disease Control and Prevention, 2002
), and since the Internet is a readily available resources for recent quitters, future studies should examine the efficacy of Internet-based cessation programs in preventing relapse.
The Internet raises unique reporting, methodological, and analysis issues. Traditional Web server log files may provide insufficient data to determine actual numbers of individuals reached. To understand barriers to studying enrollment encountered by smokers, and to determine the generalizability of the final sample, Internet smoking cessation trials should obtain and retain information about those not enrolled. This preliminary study must be interpreted cautiously and awaits independent replication. As with most other clinical trials, only a small portion of potential participants enrolled in this study. However, results suggest that Internet-based research using active user-intercept recruitment is feasible. Creative strategies can be used to sample participants, estimate a variety of parameters of reach (denominators), and measure aspects of potential selection bias that are implicit in all traditional clinical trials but often cannot readily be estimated (e.g., those who read a newspaper advertisement for a study but never respond).