The paper by West and Zhou published in this issue of Thorax (see page 998) is an important contribution to the literature because it addresses and refutes the questions raised, on the basis of retrospective case-control studies, about the effectiveness of nicotine replacement therapy (NRT) in “real world” settings.1 2 Contrary to those retrospective analyses, West and Zhou’s prospective population study found that NRT helps smokers to quit, roughly doubling the odds of successful quitting. The odds ratio of 2.2 reported by West and Zhou is quite consistent with the odds ratios reported in randomised clinical trials.3

In this sense, West and Zhou’s finding that NRT is effective should come as no surprise. The efficacy of NRT has been demonstrated in over 100 randomised controlled trials encompassing over 35 000 smokers.3 Moreover, the effectiveness of NRT has also been demonstrated in multiple trials that tested NRT under over-the-counter conditions.4 These studies not only showed that over-the-counter NRT was more effective than placebo, but also that it was just as effective as using NRT under the care of a doctor. Although critics have argued that “effectiveness” (in contrast to “efficacy”) has not been shown, in fact these trials were effectiveness trials: there was little or no screening or exclusion of participants, no extra support, instruction or therapy offered, and minimal data collection.5 The benefit of NRT has therefore been proven in both efficacy and effectiveness studies. West and Zhou’s data extend this to a prospective study of population samples not enrolled in a treatment trial, and thus help to address the concern that subjects enrolled in treatment trials may not be representative of the population.

Why are the analyses from retrospective surveys, which appear to show little or no efficacy either for NRT or for any behavioural treatment,2 6 7 so discrepant with the findings of West and Zhou and those of dozens of other studies? One important reason, alluded to by West and Zhou, is that brief failed quit attempts are easily forgotten unless something makes them memorable.8 Spending money on an NRT product (or being stuck with acupuncture needles, for that matter) is unusual and therefore memorable, with the likely result that failures on NRT are recalled and reported while failures in unaided quitting are forgotten or perhaps dismissed as not being “real” quit attempts. This can create substantial bias in retrospective reporting of quit success, making treatments of all kinds appear ineffective and making retrospective surveys unsuitable for evaluating effectiveness.

Another important bias in uncontrolled population studies of cessation methods is that smokers self-select which method they use for quitting. West and Zhou’s data confirm the findings from other studies that more dependent smokers—who have a lower probability of success in the first place—gravitate towards treatment.9 This artificially depresses success rates in the treatment group compared with the self-selected non-treaters. West and Zhou try to control for self-selection by including in their models a single measure of nicotine dependence. It is unlikely, however, that this completely controls for this “indication bias”;10 the realisation by smokers that they need help to quit smoking is based on a host of additional factors that were not controlled for in these analyses such as low self-efficacy and lack of social support.9 This suggests that the treatment effect reported by West and Zhou probably understates the actual benefit of treatment and again highlights the value of randomised trials for obtaining unbiased estimates of treatment effects.

While West and Zhou’s data show that NRT is effective even when used without support or behavioural treatment in an over-the-counter setting, this is not the optimal use of NRT. There is ample evidence that behavioural intervention can further additively improve success rates for smokers using NRT,3 11 so the optimal approach for any smoker is to use NRT and behavioural treatment. Unfortunately, very few smokers use behavioural treatments,6 12 even though behavioural support is increasingly available free of charge through convenient channels such as telephone quitlines13 and internet programs.11 All smokers should be encouraged to get behavioural treatment.

The way NRT itself is used is also often not optimal. Smokers who use NRT too often fail to follow the prescribed regimen: they use too little medication14 and use it for too short a time,15 both of which reduce the effectiveness of the medication.16 17 One reason smokers under-dose is that too many of them mistakenly believe that nicotine medications are unsafe—in fact, many believe NRT is as dangerous as cigarettes.18 We need to do more to educate smokers and to encourage them to use the dose and duration recommended on the label. But the biggest source of NRT “under-dosing” is those smokers who don’t use NRT at all. West and Zhou report that about two-thirds of their sample did not use NRT or other treatment in their attempts to quit, which is consistent with recent US data on NRT use.6 Those two-thirds fail to take advantage of an accessible, safe and effective treatment that, even when used as an over-the-counter product, can double their odds of quitting smoking and possibly save their lives.