Meta-analysis of the Association Between Second-hand Smoke Exposure and Ischaemic Heart Diseases, COPD and Stroke

Florian Fischer; Alexander Kraemer


BMC Public Health. 2015;15(1202) 

In This Article


In this study, the effect sizes for IHD, COPD and stroke attributable to SHS exposure were estimated. For all three outcomes, the effect sizes were larger for women than for men. In men, statistically significant results were revealed only for the association between SHS exposure and stroke. According to the calculated effect sizes for all three disease entities, the risk factor of SHS exposure seems to be particularly important for COPD. A 66 % excess risk of COPD was calculated for people exposed to SHS for both sexes combined. For stroke (RR = 1.35, 95 % CI: 1.22–1.50) and IHD (RR = 1.27, 95 % CI: 1.10–1.48), the RR was considerably lower.


The calculated association between SHS exposure and IHD is consistent with several meta-analyses calculating the overall RR of coronary heart diseases associated with SHS exposure among non-smokers. In a meta-analysis including 18 studies (10 prospective cohort studies and eight case–control studies), the estimated RR was 1.25 (95 % CI: 1.17–1.32).[20] A meta-analysis by Wells[12] focused on the association between IHD mortality and SHS exposure. According to this study, a RR of 1.23 (95 % CI: 1.12–1.35) was calculated for both sexes combined (men: RR = 1.25, 95 % CI: 1.03–1.51; women: RR = 1.23, 95 % CI: 1.11–1.36).[12] These estimations are comparable to the calculation of the effect size for both sexes combined. Nevertheless, the study by Wells[12] provided effect sizes which are almost equal for both sexes. In our study the results for the association between SHS exposure and IHD indicated much higher effect sizes for women. Wells[12] also calculated the effect size associating IHD morbidity with SHS exposure. Here, the RR for women was 1.51 (95 % CI: 1.16–1.97), which is comparable to the estimation of the results of our study. Therefore, it seems that the associations for IHD morbidity and mortality differ substantially, and this leads to differences in the effect sizes estimated in this study compared to previous ones.


The estimation of the effect size for the association between SHS exposure and COPD cannot be compared to other meta-analyses, because this is the first attempt to calculate a synthesis for the primary studies dealing with this association. Up to now, the number of studies on SHS exposure as a risk factor for adult onset COPD is small compared with the number on the adverse health effects of SHS exposure on childhood respiratory symptoms and diseases.[22] The estimation for both sexes combined led to a RR of 1.66 (95 % CI: 1.38–2.00), which is higher than the estimation for the association between SHS exposure and IHD. This also applies to the gender stratified estimations: in women a RR of 2.17 was calculated with a fairly broad confidence interval (95 % CI: 1.48–3.18). This can be explained by the fact that three of the total of five studies dealt with the association in women. The studies by Wu et al.[75] (RR = 3.12, 95 % CI: 1.56–6.50) and McGhee et al.[24] (RR = 2.59, 95 % CI: 1.30–5.27) in particular contributed to the broad confidence interval. Therefore, the few existing studies on SHS exposure and COPD differ considerably, although the results indicate a positive association. No judgement on the consistency of the results of primary studies on the association between SHS exposure and COPD for men is possible, because only the study by McGhee et al.[24] provided results for the male subgroup (RR = 1.50, 95 % CI: 0.96–2.28).


The estimations for the association between SHS exposure and stroke (RR = 1.35, 95 % CI: 1.22–1.50) are comparable with previous meta-analyses. In our study, the effect sizes showed a significantly increased risk for people exposed to SHS in both sexes, with RRs that are almost equal between men (RR = 1.40, 95 % CI: 1.09–1.81) and women (RR–1.43, 95 % CI: 1.28–1.61). Lee and Forey[76] provided a comprehensive review of epidemiological evidence relating stroke to SHS exposure in lifelong non-smokers. Overall, including 16 studies (seven prospective cohort studies, six case–control studies and three cross-sectional studies) which used current spousal smoking (or nearest equivalent) as the exposure index led to an overall estimate of 1.25 (95 % CI: 1.16–1.36), which is slightly lower than our calculations. The study results also indicated no significant heterogeneity and no differences between men and women,[76] which is consistent with our study results. Eight studies in the meta-analysis provided information regarding a possible dose–response relationship between SHS exposure and stroke. According to this, the synthesis for the highest level of exposure led to a RR of 1.56 (95 % CI: 1.34–1.82).

Another meta-analysis,[49] included 20 studies (10 cohort studies, six case–control studies and four cross-sectional studies) published between 1984 and 2010. All of these reported results for non-smokers, who were mainly defined as never-smokers, although some studies also included ex-smokers or infrequent current smokers. Eleven studies in the meta-analysis by Oono et al.[49] measured the dose of SHS exposure, which was either defined as the number of smokers, cigarettes per day, hours per week, pack years, or cotinine concentration and score. Our calculations for the effect size of the increased risk of stroke attributable to SHS exposure (RR = 1.35, 95 % CI: 1.22–1.50) are in line with the results of SHS exposure of either 10 cigarettes per day (RR = 1.31, 95 % CI: 1.12–1.54) or 15 cigarettes per day (RR = 1.45, 95 % CI: 1.19–1.78).[49]

Dose–Response Relationship

The results of the primary studies that were included in the meta-analysis on the associations between SHS exposure and IHD as well as stroke indicate a distinct dose–response relationship. Even low levels of SHS exposure increase the risk of adverse health effects, indicating that there is no safe level of exposure.[42,49] The effects of SHS exposure are lower than those of active smoking, but it has been consistently shown that the effects of SHS exposure on the cardiovascular system are much larger than might be expected from a comparison of the doses of toxins delivered to active and passive smokers. Therefore the effects of SHS are estimated to be on average 80–90 % as harmful as those of active smoking.[10] The effects of a dose–response relationship between SHS exposure and adverse health outcomes were not depicted in this study, because it focused on regular exposure to SHS. Although the dose–response function might supply important additional information, Sauerbrei et al.[77] argued that aggregated data are too limited to perform a meta-analysis including a dose–response analysis. Nevertheless, regular SHS exposure, irrespective of the dose is still an important risk factor, because it may lead to both acute and chronic diseases.

Gender Differences

The stratification for sex performed in this study is highly relevant, because the effect sizes as well as the prevalence of diseases and the prevalence of SHS exposure differ between the sexes. Until now, it has been largely men who have been considered in many studies dealing with IHD, because of their higher prevalence of coronary diseases. In most parts of the world women are at least 50 % more likely to be exposed to SHS than men.[78] Until now, only a few studies have investigated possible mechanisms underlying sex differences in adverse health outcomes such as IHD related to SHS exposure. It is assumed that the anti-oestrogenic effect of cigarette smoking–and therefore also the exposure to SHS–may be at least partly related to the increased risk of IHD in young females smokers.[79] Furthermore, a study by Geisler et al.[80] indicated that in smoking women undergoing oestrogen replacement therapy, plasma levels of oestrogen were 40–70 % lower than in non-smoking women. Additionally, a decrease in both oestradiol and testosterone concentrations in smoking men has been reported.[81] Therefore, hormonal factors seem to considerably influence vulnerability due to SHS exposure. This might be one explanation for gender differences in the effects of SHS exposure.[82]


There are methodological restrictions in data quality of primary studies, which have to be considered when interpreting the results. Among these, particularly the differences in study designs and misclassification bias due to different definitions and measurements of SHS exposure have to be mentioned. Another limitation of major importance in the context of a systematic literature review is a possible publication bias, although a review of published and unpublished studies on the health effects of SHS exposure showed no evidence of publication bias against statistically non-significant results in the peer-reviewed literature.[83]

Another limitation in the identification of primary studies on the association between SHS exposure and the three selected diseases leads back to the decision to perform the systematic literature search only in one literature database, PubMed. Therefore, some studies might have been missed, although an additional manual search in the reference lists of publications was performed, which led to only eight further articles. A broader search strategy with another search algorithm may have led to further articles eligible for the meta-analysis.

The quality assessment led to the exclusion of nine studies. Although the development of criteria for the quality assessment was based on established instruments, different criteria may have led to the exclusion of more or fewer articles, depending on their strictness. The quality checklist was used as a scale, although the criticism has been made that these scales do not provide a transparent estimation of the degree of bias.[84] Furthermore, quality scores neglect information about individual items and no empirical basis for the different weights that are implicitly given to each item exists.[85] Nevertheless, this approach was chosen, to allow for the exclusion of studies with low methodological quality.

Since only cohort studies and case–control studies were selected, a large number of studies had to be excluded either during the screening of titles and abstracts or during the assessment of full-texts. Also, comparatively small studies with low effect sizes or rather broad confidence intervals were included in the meta-analysis. These studies carried a smaller weight in the synthesis of results. To make the results of the primary studies comparable, all the OR provided in case–control studies were re-calculated into RR using quite a conservative approach, which is more likely to underestimate the true association. Therefore, overall, the effect sizes calculated in the meta-analysis represent a conservative estimate.

Besides the identification and data quality of primary studies, the combination of research results from multiple studies performed in the meta-analysis faces several limitations and uncertainties. Although this meta-analysis indicates only low heterogeneity, diversity between studies, for example due to different populations (e.g., countries, age groups), inclusion and exclusion criteria (e.g., more severe patients), study designs (e.g., inadequate follow-up of lost patients), statistical methods used, and various sources of bias, is still an important issue. Formal heterogeneity tests face low statistical power. In this study, the Q-statistic and I2-test were used. A shortcoming of the Q-statistic is that it has low power to detect true heterogeneity among studies when the meta-analysis includes only a small number of studies.[86] The Q-statistic is useful to test for the existence of heterogeneity, but not to assess the magnitude of heterogeneity. For that we used the I2.

For the calculation of the effect sizes for COPD, it has to be kept in mind that the estimation for men is based on only one study. Particularly for COPD, the synthesis is based on very few studies, which limits its reliability. The combination of studies will often result in small confidence intervals, suggesting a false precision.[58] In this context, it is relevant to point out that random effects models, as used in this study, are not sufficient to explain the heterogeneity between studies, since the random effect merely quantifies an unexplained variation by estimating it.[59]

Conclusion and Implications

Up to now, the effects of SHS exposure on population health are still controversial, although several studies and meta-analyses have revealed comparable results on the association between regular SHS exposure and adverse health outcomes. However, further studies with sound methodological approaches due to large prospective epidemiological studies using biomarkers for exposure assessment are still required to determine the risks associated with SHS exposure.[16,76] Furthermore, there is only a little evidence for the effects of SHS on health-related quality of life, which is a very important parameter for well-being besides objective parameters such as morbidity and mortality.

To address this research need, this study was conducted. It is the first study to have calculated effect sizes for the association between SHS exposure and the disease outcomes IHD, COPD, and stroke, stratified by sex. The effect sizes calculated in the meta-analysis are overall comparable with previous findings in meta-analyses for IHD and stroke. This suggests that the results are reliable. Although no previous meta-analysis for the association between SHS exposure and COPD is available, the results are assumed to be reliable as well, because the methodological approach in this study was the same for all three disease entities. Nevertheless, further research is needed, to provide more adequate primary studies which account for confounding and other biases.