We found strong statistical evidence that, in patients with HIV, current smoking was associated with an approximate 70% to 100% increased risk of bacterial pneumonia compared with non-smokers, and moderate evidence that stopping smoking decreased this by about 27%. There was some evidence that former smokers were at slightly increased risk compared with never smokers, which was not significant, however the estimate was insufficiently precise to exclude a substantial increased risk. There was no good evidence that the risk of smoking was modified either by the use of HAART or by CD4 count. There was mixed evidence that smoking increased the risk of PCP, but higher quality evidence indicated no substantial increase in risk of PCP from current smoking. The evidence from all studies was somewhat clouded by poor definition of exposure status (smoking) and failure to control for a full range of confounders. However, studies that reported the effects of adjustment suggested that adjustment for several confounders had an insubstantial effect on these estimates.
The strengths of this study relate to the comprehensive search of studies that had relevant outcomes even though the titles or abstracts included no mention of smoking. We found seven studies this way that did not have an investigation of smoking as a risk factor for pneumonia as their prime aim. It is plausible that several cohort studies investigating smoking as a risk factor failed to find a significant association, and therefore did not publish data. Incorporating studies where smoking was not the main focus offers reassurance that the association is not due to publication bias and funnel plots also suggested no evidence of bias.
There are some limitations with the review and the data we reviewed. We might have searched a larger number of databases and we did not search the grey literature. We included observational data, and there were inherent limitations. First, the measurement of smoking was suboptimal. Over time, people who were smokers at baseline would have stopped smoking, while some ex-smokers might have relapsed; this will mix exposure assignment and generally underestimate any true risk of smoking on pneumonia . Another limitation is that few studies adjusted for a full range of confounders. People who smoke are more likely than people who do not to have other factors associated with an increased risk of pneumonia, for example, intravenous drug use or failure to adhere to HAART. Our ability to assess the importance of this explanation was compromised by failure to report unadjusted and adjusted estimates, but those that were presented suggested a small effect of adjustment and this is unlikely to explain the associations observed.
Taken together, we believe this systematic review gives reasonable evidence that the association between smoking and bacterial pneumonia is causal in HIV patients. There are no strong reasons to doubt the validity of the association and there was a dose-response relationship between smoking and risk of pneumonia. There are supporting data that suggest biologically plausible mechanisms, such as reduced local defenses in the lung. Smokers with HIV have lower concentrations of CD4 and CD8 lymphocytes in lung tissue and lower concentrations of cytokines IL-1β and TNF-α . There is also evidence that smoking reduces the phagocytic function of alveolar macrophages in individuals infected with HIV . Another possible explanation is that free radicals in tobacco smoke increase oxidative stress within cells and therefore enhance the likelihood of infection .
This review produced inconclusive evidence on whether there is an association between current smoking and risk of PCP in patients with HIV. There was no evidence of a dose-response relationship and, given the weight of evidence, we conclude that smoking is not a risk factor for development of PCP. HAART prevents the occurrence of PCP and antibiotic prophylaxis against PCP can be stopped in people with a CD4 count greater than 100 cells/μL . It is clear then that CD4 count is critical and this reinforces our conclusion that smoking is not an important risk factor for this pneumonia.
In this meta-analysis, the relative risk of bacterial pneumonia in current smokers compared to non-smokers was 26% higher than the relative risk of current smokers compared to non-smokers. In western countries, where nearly all of these studies were completed, most non-smokers aged around 40 years (the average age of participants in included studies) are never smokers rather than former smokers . The risk of former smokers in comparison to never smokers was imprecisely estimated and we could not therefore exclude a continued risk in former smokers. These data are clouded further by the definitions of smoking status used in these studies. We cannot therefore be clear that the risk of pneumonia for former smokers is the same as never smokers, but it is clear that it is lower than among current smokers.
To our knowledge, no other systematic reviews have been published on this topic, but narrative reviews have addressed it [32–34]. This is the first review to explicitly compare current smokers with former smokers. All but one study published data comparing the risk of current smoking to never smokers, but a person who smokes cannot become a never smoker. Using indirect comparisons allowed us to estimate the association that is most clinically relevant. We were, however, unable to estimate how long the risk of current smoking remains elevated after stopping before the benefits of cessation are manifest. There are numerous cohort studies of people with HIV and they are likely to have recorded data on smoking repeatedly during the follow-up of the cohorts. It would therefore be useful for these cohorts to examine the difference in risk by time since cessation and including these data would allay concerns that smoking status is likely to have changed over time. There is evidence that smoking predisposes to tuberculosis, but rather less data on the effect of smoking cessation on tuberculosis . Such associations could also be investigated in these studies.
These results have direct clinical implications. The prevalence of smoking was high in these cohorts, much higher than the population in general, in keeping with the higher prevalence in gay men and intravenous drug users [36, 37]. Two studies reported incidence rates for bacterial pneumonia, the weighted mean of which was 7.8 per 100 person-years in smokers [14, 15]. This means that 25% of smokers would develop pneumonia over 10 years due to their smoking that would have been prevented by smoking cessation. Of cases of pneumonia in people with HIV, 22% are due to current smoking if the median prevalence of smoking in these studies (60%) applies generally. Thus it seems imperative that physicians caring for people with HIV become proficient in smoking cessation treatment, regularly offer treatment, and support their patients to stop smoking. Brief advice to stop smoking is insufficient. The large majority of people will fail to stop with advice alone . Many smokers who are gay or current or former intravenous drug users want to stop smoking but find this difficult due to nicotine addiction; effective treatment programs exist that can ameliorate this . The decision on whether such programs become routine practice in HIV medicine depends upon showing that they are cost-effective in this population. Smoking cessation programs have been called 'among the most cost-effective of all healthcare interventions' (p7 of ) in the general population, so it seems likely that they would be similarly cost-effective in people with HIV. There are several small scale trials published of treatment for smoking cessation in people with HIV [40–43], though there is no reason to imagine that treatment regimens in this subgroup need be very different from standard smoking cessation treatment. However, it was outside the scope of this review to examine evidence of efficacy in supporting cessation or its cost-effectiveness.