A systematic review and network meta-analysis of population-level interventions to tackle smoking behaviour

Abstract

This preregistered systematic review and meta-analysis (PROSPERO: CRD 42022311392) aimed to synthesize the effectiveness of all available population-level tobacco policies on smoking behaviour. Our search across 5 databases and leading organizational websites resulted in 9,925 records, with 476 studies meeting our inclusion criteria. In our narrative summary and both pairwise and network meta-analyses, we identified anti-smoking campaigns, health warnings and tax increases as the most effective tobacco policies for promoting smoking cessation. Flavour bans and free/discounted nicotine replacement therapy also showed statistically significant positive effects on quit rates. The network meta-analysis results further indicated that smoking bans, anti-tobacco campaigns and tax increases effectively reduced smoking prevalence. In addition, flavour bans significantly reduced e-cigarette consumption. Both the narrative summary and the meta-analyses revealed that smoking bans, tax increases and anti-tobacco campaigns were associated with reductions in tobacco consumption and sales. On the basis of the available evidence, anti-tobacco campaigns, smoking bans, health warnings and tax increases are probably the most effective policies for curbing smoking behaviour.

Fig. 1. Study selection.

PRIMSA flow chart detailing the identification and screening of identified records for the systematic review and meta-analysis.

Fig. 2. Network meta-analysis of the effects of tobacco policies on smoking cessation, by outcome type and policy category, ranked by effect magnitude.

a–d, Network forest plots are shown on the left, and rank plots are shown on the right. The axis ranges differ across panels. Two or more studies available for each policy and outcome are included in the network meta-analysis as mentioned in Table 1. In the left panels, the squares indicate pooled estimates of effect size, and the whiskers indicate the 95% CIs. The white lines within the grey squares indicate pooled ORs with narrow CIs, while the black lines indicate the other values. The p-score was used to rank interventions on the basis of their effectiveness, and the right panels visually represent this ranking score by showing higher probabilities of effectiveness in dark green and lower probabilities in orange.

Fig. 3. Network meta-analysis of the effects of tobacco policies on smoking prevalence and e-cigarette use, by outcome type and policy category, ranked by effect magnitude.

a,b, Network forest plots. c,d, Rank plots. The axis ranges differ across panels. Two or more studies for each policy and outcome are included in the network meta-analysis as mentioned in Table 2. In a and b, the squares indicate pooled effect estimates, and the whiskers indicate 95% CIs. The p-score was used to rank interventions on the basis of their effectiveness, and c and d visually represent this ranking score by showing higher probabilities of effectiveness in dark green and lower probabilities in orange.

Fig. 4. Narrative summary of tobacco behaviour change following the implementation of different tobacco policies (n = 238 studies).

The number of studies is indicated in each circle. ‘Multicomponent tobacco control programme’ includes different combinations of multiple tobacco control policies; miscellaneous policies include NRT, Quitline, point-of-sale tobacco display ban, ban on advertising, youth access laws and plain packaging mandates. Green indicates a positive effect of the tobacco policy—that is, a statistically significant increase or decrease in the targeted outcome favouring the policy. Orange indicates a negative effect of the tobacco policy—that is, a statistically significant increase or decrease in the targeted outcome favouring the control. Grey indicates no effect of the tobacco policy—that is, a statistically insignificant increase or decrease in the targeted outcome.