INDIVIDUALIZED VERSUS POPULATION‐BASED APPROACHES TO TOBACCO CONTROL – A FALSE DICHOTOMY

We agree in principle with many of the broad concerns spelled out by Lawrence et al. [1], but also disagree with the general tone and some of the particular complaints levelled at population-based intervention studies. We agree that there is a risk that population-wide interventions can lead potentially to disparities in smoking rates, although the evidence accumulated thus far on this question does not support this concern. In fact, the evidence available suggests that population-level interventions such as advertising bans, mass media campaigns, smoke-free policies and even tax hikes actually work at least as well and, in the case of taxes, better for more disadvantaged segments of society [2–5]. Also, we would point out that rates of smoking in Australia and most other developed countries have been declining at similar rates in all socio-demographic strata (that are represented in surveys). That said, there remain important disparities in access to treatments for smoking cessation services favouring those in wealthy compared to poor countries and in countries without universal access to health services favouring those who can afford to purchase treatments [6,7]. Certainly, governments need to make sure that access to evidence-based cessation services is something that is accessible to all segments of society, not just those who can afford them, as this has the potential to unnecessarily amplify existing disparities in smoking rates and mortality. We disagree with Lawrence et al.'s central theme [1], which suggests that we should be looking towards randomized clinical trials as the basis for judging the efficacy of population-level strategies. We think this suggestion is unrealistic and unnecessary for several reasons. First, randomized experimental designs are not practical for evaluating the effects of policy interventions because governments, not researchers, control policy implementation. Secondly, the effects of population interventions are weak per person, although they have a large net benefit because they reach so many people. This means that huge sample sizes are needed to observe effects, compounded by the fact that the impact of many interventions is presumed to be influenced, in part, by the degree to which others respond similarly, creating huge complications for trials. While Lawrence et al. [1] seem to recognize some of these limitations, in that they recommend cluster randomization, it is not clear what the clusters could be. Tobacco control researchers have led the advance in thinking about more appropriate ways to evaluate the impact of population-level interventions, where the use of naturalistic experiments and meditational models are spelled out [8–10], including the use of the so-called ‘natural experiment’ and the use of a longitudinal cohort design in which individuals are measured on the same key outcome variables at multiple times, preferably before and after a policy is implemented. Combining these two strategies in a single study yields a two (or multi-) group, pre–post design, which offers a higher degree of internal validity than either feature alone. This methodology has been employed by the International Tobacco Control Policy Evaluation consortium (the ITC Project) to evaluate the impact of a wide range of population-based interventions to control tobacco [11–18]. We applaud the authors on addressing some of the more dubious arguments in the recent paper by Chapman & McKenzie [19]. The imbalance they identify between trials of individual-orientated and population-level interventions (that they curiously call unaided) are real, but has a more prosaic explanation than the conspiracy they posit. As part of a group that has published a fair proportion of the world's studies on the impact of population-level tobacco control interventions, we know how difficult and costly these studies can be. Further, the per-person effects are typically quite small and the size of studies to detect direct effects on cessation outcomes, especially given the range of uncontrolled other ongoing activity, needs to be enormous, well beyond the budgets of all but the richest countries. Fortunately, for the most part, we do not need this level of evidence to have confidence that these interventions work. The important question to answer for population-based interventions is not whether they work [the question for which randomized controlled trials (RCTs) are the strongest measure of efficacy], but whether they work in the broader context of changing outcomes on a population-wide basis? Unlike pharmacotherapies, where RCTs provide ample evidence of efficacy, the real question is whether any intervention, regardless of whether it is individually focused or more broadly focused, works in particular contexts [10]. Thus, we do not believe that a lower standard is set for population-level interventions. On the contrary, the criterion for success is stronger as the outcome is a population-wide impact. By contrast, this is only ever inferred for individually focused behavioural and pharmacological studies which rarely, if ever, test external validity and effectiveness on a population-wide basis. Unfortunately, Lawrence et al. [1] seem to be stuck within the same narrow view of evidence that the drug industry has to use for their products, and the narrow thinking that sees this as a gold standard for all forms of intervention research, something which gives ontological priority to some questions over others. This is not a productive way to identify evidence-based interventions that can truly alter population health. There is a place for both individualized and population-based interventions to control tobacco use. The relevant question we should be focusing on is which combination of interventions is likely to yield the greatest impact at the lowest cost. In this regard, we agree with Chapman & McKenzie [19] that population-based interventions are likely to continue to be the main approach in tobacco control, because these approaches have been demonstrated to have the greatest overall impact at the lowest possible cost. K.M.C. has research funding from NABI Biopharmaceutical for research investigating a nicotine vaccine for smoking cessation; K.M.C. has also served as a paid expert witness in litigation against the tobacco industry. R.B. has nothing to declare in terms of industry funding. The authors have received salary support from the following sources: the National Cancer Institute, US P50 CA111326) (P01 CA138389) (RO1 CA 100362) (R01 CA125116); the Canadian Institutes for Health Research (79551); and the National Health and Medical Research Council of Australia (450110).

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.