Annex I
Executive Summary
Systematic Quantitative Review of the Effect of Environmental Tobacco Smoke Exposure on Respiratory Health in Children
Derek G Cook
David P Strachan
H Ross Anderson
Department of Public Health Science, St George's Hospital Medical School, London SW17 ORE
Background
During the last two decades, many epidemiological studies have reported on the association between parental smoking and respiratory diseases in childhood. These were considered in both the US Surgeon General's 1986 report and the Environmental Protection Agency Report (1993). However these and most other reviews have been neither systematic nor meta-analytic in their approach. Moreover, a large number of publications have occurred since the completion of the EPA report.
Our aim was to systematically review the health effects of Environmental Tobacco Smoke (ETS) in children's respiratory health and where possible to produce meta-analytic estimates of the relative risks. In carrying out the review we were particularly concerned to consider the importance of residual confounding from other environmental factors as a possible explanation for any differences found and to assess the relative importance of exposure at different ages. We also distinguished wherever possible between the effects of smoking by different household members and between pre- and post-natal exposure.
Report Structure
Summary of the review process, findings and conclusions (this document)
Separate chapters on:
Sudden Infant Death Syndrome
Lower Respiratory Tract Illness in pre-school children
Prevalence of asthma & respiratory symptoms in schoolchildren
Incidence, severity and prognosis of asthma
Bronchial Hyper-Reactivity
Allergic sensitisation
Ear disease and adenotonsillectomy
Review Process
Published papers, letters and review articles were selected by an electronic search of the Embase and Medline databases, using the search strategy described in the Appendix. briefly all passive smoking references were selected by the MESH heading Tobacco smoke pollution and/or relevant text-words in the title, keywords or abstract. Papers were then restricted to children by selecting all papers classified as containing data on neonates to 18 years and/or by relevant text-words in the title or abstract. Embase searches were entirely based on text-word searches. This search, completed in March 1996, yielded 3365 references which were downloaded into Reference Manager. After further electronic text-word searching (see appendix) and review of the on-line abstracts, 605 articles were identified as relevant to the broader overview. 385 (64%) of these had been published during 1990-96, the remainder during 1972-89. The number of studies included under each heading are summarised in table 1.
Main Findings (See Table 2 for overall summary of odds ratios for different outcomes)
Sudden Infant Death Syndrome
Maternal smoking during pregnancy was clearly associated with a risk of sudden infant death, with all 35 studies showing a relative risk greater than 1 which was statistically significant in all but 2 studies. The pooled odds ratio for studies not adjusting for confounding variables was 2.49
(95% CI 2.28-2.72). In studies which adjusted for potential confounding variables the pooled odds ratio was 2.08 (95% CI 1.96-2.21). While it is difficult to distinguish the independent effects of pre- and post-natal smoking by the mother, those studies which examined the issue found evidence of a post-natal effect. Such a conclusion is strengthened by evidence from 3 studies reporting on risk of paternal smoking where the mother was a non-smoker. Two reported significant effects, 1 no effect with a pooled odds ratio of 1.63 (95% CI 1.26-2.11).
Lower Respiratory Illness in infancy and early childhood
The pooled odds ratio for either parent smoking across all studies was 1.48 (95% CI 1.40 to 1.57) and was consistent across different study types: 1.45 (95% CI 1.34 to 1.57) for community based studies of lower respiratory illness, bronchitis and/or pneumonia; 1.54 (95% CI 1.30-1.81) for community studies of wheezing illness; 1.45 (95% CI 1.27-1.66) for studies of hospitalisation for lower respiratory illness, bronchitis, bronchiolitis or pneumonia. The associations were robust to adjustment for confounding factors, and showed evidence of dose response where this was investigated. Importantly, there was a significantly elevated risk of early chest illness associated with smoking by other household members in families where the mother did not smoke (relative odds 1.29, 95% CI 1.19-1.41). There was insufficient evidence to evaluate the independent contribution of pre- and post-natal maternal smoking.
Prevalence of Asthma and Respiratory Symptoms in School Children
The pooled odds ratios for either parent smoking were 1.17 (95% CI 1.10-1.25) for asthma; 1.24 (95% CI 1.19-1.30) for wheeze; 1.33 (95% CI 1.27-1.39) for cough; 1.33 (95% CI 1.14-1.55) for phlegm; and 1.31 (95% CI 1.14-1.50) for breathlessness. Adjustment for confounding had little effect on these estimates. Evidence of heterogeneity between studies appeared largely explicable in terms of publication bias with a superfluity of small studies with large odd ratios. However, excluding these had little effect on the pooled odds ratios. There was clear evidence that maternal smoking had a greater effect than that of paternal smoking for all conditions, though there was a significantly increased risk of each symptom associated with smoking by the father only. For all symptoms, children exposed to two parents smoking were at greater risk, the pooled odds ratios compared to children on non-smoking parents being 1.52 (95% CI 1.34-1.72) for asthma, 1.40 (95% CI 1.29-1.51) for wheeze and 1.61 (95% CI 1.50-1.73) for cough.
Incidence, severity and prognosis of asthma
Case control studies looking at ETS and asthma prevalence provided a slightly greater odds ratio for either parent smoking than the cross-sectional surveys: 1.35 (95% CI 1.19-1.54). In longitudinal studies, maternal smoking was associated with an increased incidence of wheezing illness up to age 6 (pooled odds ratio 1.31, 95% CI 1.22-1.41) but less strongly thereafter (1.13, 95% CI 1.04-1.22). The long term prognosis of early wheezing illness was better if the mother smoked, reflecting the fact that children of smoking parents are more likely to develop mild wheezing illness at younger ages. Such an interpretation is supported by 3 studies that suggest that parental smoking is more strongly associated with wheezing amongst non-atopic children. The effect of ETS exposure on asthmatics is not however benign; indicators of disease severity, attack frequency, medication use and life-threatening bronchospasm were in general positively related to household smoking, but could not be combined in a quantitative meta-analysis.
We thus face a contradiction. Longitudinal studies demonstrate that maternal smoking is associated with an increased incidence of wheezing illness, particularly at younger ages. This excess incidence of early wheezing illness appears to be largely non-atopic “wheezy bronchitis” and to run a relatively benign course. However, amongst children with established asthma, parental smoking is associated with more severe disease. We believe that this paradox is explained by viewing ETS as a trigger of wheezing attacks (probably acting in conjunction with infection), rather than as a cause of the underlying asthmatic tendency. Our interpretation is supported by the lack of a positive association between ETS and atopic sensitisation (see below). However, the true test of the hypothesis lies in whether long term measures of asthmatic tendency such as bronchial hyper-responsiveness are associated with ETS exposure.
Bronchial hyper-reactivity (BHR)
We were able to extract effect measures from 8 studies in the form of relative odds of measured bronchial hyper-reactivity for ETS exposed children compared to non-exposed children. There was no evidence of heterogeneity between studies and no single study dominates. The pooled estimate of the relative odds was 1.28 (95% CI, 1.08 to 1.52). For 6 studies not providing odds ratios none found statistically significant effects. A further 4 studies were identified as having collected data but not published. The studies included in the meta-analysis covered 4976 children, those reporting non-significant results but not odds ratios covered 3714 children and the unpublished studies covered 4793 children.
We conclude that a clear effect of ETS exposure on BHR in the general population has not been established. While the meta-analysis suggests a small, but real, increase in BHR in school aged children, it seems likely that this estimate is biased upwards due to publication bias. In contrast limited evidence from 4 studies suggests greater variation in peak flow in children of smoking parents. Such a finding would be in keeping with acute effects of ETS exposure on airflow rather than chronic effects on BHR.
Ear Disease and Adentonsillectomy
Evidence for middle ear disease was remarkably consistent, with pooled odds ratios if either parent smoked of 1.41 (95% CI 1.19-1.65) for recurrent otitis media, 1.38 (95% CI 1.23-1.55) for middle ear effusion, and 1.21 (95% CI 0.95-1.53) for out-patient or inpatient referral for glue ear. These associations were robust to adjustment for confounding factors and are likely to be causal. Few studies have assessed dose response. Large French and British Studies were inconsistent regarding the association between parental smoking and tonsillectomy.
Allergic Sensitisation
No consistent association were found in neonates or older children between parental smoking and total serum IgE concentration, allergic rhinitis or eczema. Some evidence of a weak inverse assocation was found with skin prick sensitivity with a pooled odds ratio of 0.86 (95% CI 0.77-0.97) for current passive smoke exposure and 0.94 (95% CI 0.79-1.13) for perinatal exposure. The combined odds ratio was 0.88 (95% CI 0.88-0.98. However, significant and unexplained heterogeneity of odds ratios between studies suggest the need for cautious interpretation, particularly of the confidence intervals.
Conclusions
- The relationships between parental smoking and sudden infant death and acute lower respiratory illness in infancy are almost certainly causal.
The elevated risks associated with smoking by other household members provide good evidence that postnatal exposure from both mother and father are important.
Because pre-natal smoking is almost invariably associated with post-natal smoking, the role of pre-natal maternal smoking will be difficult to resolve using epidemiological studies.
- There is convincing evidence that parental smoking is associated with increased prevalence of asthma and respiratory symptoms in school children.
Among children with established asthma, parental smoking is associated with more severe disease.
Parental smoking probably acts, alone or in combination with infection, as a trigger of wheezing attacks rather than as a cause of the underlying asthmatic tendency. ETS exposure is not consistently related to allergic sensitisation and the case for a relationship with BHR has not been established.
- It seems likely that parental smoking causes both acute and chronic middle ear disease in children. The evidence regarding tonsillectomy is inconsistent.
- Reducing parental smoking would result in important reductions in respiratory morbidity and mortality in infants and children.
Output
Thorax have agreed to publish our systematic reviews as a peer reviewed series. A series editor has been appointed and it planned that the first papers will appear in the latter half of 1997. Reviews will be updated with 1996 references prior to publication. This may results in some changes to the pooled odds ratios presented in table 2.
Table 1 Number of Papers selected for the review process
|
| Outcome |
Potentially relevant after
brief review
of abstracts |
Included in review |
Additional
references
identified |
Total in
review |
| SIDS |
67 |
30 |
10 |
40 |
| Ears & Tonsils |
51 |
37 |
5 |
42 |
| Allergy |
172 |
30 |
2 |
32 |
| Spirometry |
199 |
30–40 |
|
30–40 |
| Bronchial hyper–reactivity |
73 |
27 |
1 |
28 |
| LRI in infancy |
78 |
47 |
0 |
47 |
| Respiratory symptoms in schoolchildren |
85 |
43 |
3 |
46 |
| Asthma incidence, severity & prognosis |
62 |
52 |
2 |
54 |
| Total |
605* |
|
| *some papers appear under several headings |
Appendix
Medline Search Strategy
To identify all passive smoking references ($=wildcard):
- MESH heading “Tobacco smoke pollution”
- {passive OR second-hand OR second hand OR involuntary OR parent$ OR maternal OR mother$ OR paternal OR father$ OR household$} AND {smok$ OR tobacco$ OR cigarette$}
- Combine (a) OR (b)
To restrict to children:
- Restrict (c) to all relevant age groups
- Search within (c) for:
{Paediatric$ OR pedatric$ OR infan$ OR child$ OR adolescen$
- Combine (1) OR (2)
EMBASE Search Strategy
Textword searches of titles, keywords and abstracts were carried out as above. That is (b) AND (2).
Electronic search strategies for indentifying specific endpoints in Reference Manager
Among references downloaded from Medline or Embase as above, search for any of the following text strings in title, abstract or keyword fields:
| Disease | Strings searched for |
| SIDS | ‘infant death' OR ‘SIDS' |
| Ear | ‘tympanom' OR ‘otitis' OR ‘middle ear' OR ‘glue ear' |
| Tonsils | ‘tonsil' |
| Allergy' | globulin E' OR ‘IgE' OR ‘atopic' OR ‘atopy' OR ‘allergy' OR ‘skin prick' |
| Spirometry | 'lung' OR ‘fev' OR ‘pefr' OR ‘fvc' OR ‘pulmonary' OR ‘flow rate' OR ‘spirometr' |
| Asthma | ‘asthma' |
| Symptoms | ‘cough' OR ‘wheeze' OR ‘breathless' OR ‘phlegm' OR ‘mucous' |
| Others | ‘respirat' OR ‘bronch' OR ‘pneumon' |
|
