Results

Table 1 presents a description of the study subjects (the matching was broken in the table). Among controls, 19% of mothers and 33% of fathers smoked, and in 10% of controls both parents smoked. The proportion of smoking parents was substantially higher among cases, and in families with at least one smoker, the average daily number of cigarettes smoked at home was higher among cases than among controls. Parental education and family material conditions were also more favourable among controls.

Among controls, parental smoking was associated with most indicators of socioeconomic status, most strongly with education. All socioeconomic factors were mutually associated: the correlation coefficient between mother's and father's education was 0.56; other coefficients were between 0.2 and 0.3 (not shown in table).

Table 2 shows the associations between invasive meningococcal disease and parental smoking and socioeconomic circumstances. Maternal smoking was related to an almost fivefold increase in the risk of the disease; smoking of the father increased the risk almost four times. Children of parents who both smoked had almost nine times higher risk than children of both non-smoking parents. An increase in the average number of cigarettes smoked at home by 20 was associated with a 3.5-fold increase in risk of meningococcal disease. When the number of cigarettes smoked at home was categorised, the relative risk (95% confidence intervals) compared to non-smoking homes were 1.86 (0.56–6.24) for 1–9 cigarettes, 2.74 (1.03–7.27) for 10–19 cigarettes, and 4.16 (1.89–9.16) for 20 cigarettes or more (not shown in the table). Although adjustment for parental education and material circumstances somewhat reduced these estimates, they remained strong and highly statistically significant. For example, each 20 cigarettes smoked at home were associated with a rate ratio of 2.65 (1.33–5.35). Using the adjusted effects, between 34% and 42% (depending on the indicator of parental smoking) of invasive meningococcal disease in children could be attributable to parental smoking.

Among socioeconomic factors, maternal education was the strongest predictor of meningococcal disease but education of the father, not crowded housing, ownership of a car, ownership of a weekend house/cottage, and spending weekends outside of town were all significantly associated with a reduced risk of the disease in bivariate analyses (table 2, second column). In multivariate analyses, none of these variables remained significantly related to the disease (table 2, third column). However, the loss of statistical significance of the effects of car and weekend house ownership was primarily a result of loss of statistical power, rather than to the removal of a confounding factor.

Finally, we examined in more detail to what extent were the effects of socioeconomic factors mediated by parental smoking. The effect of crowding was eliminated after adjustment for parental smoking (fully adjusted rate ratio 1.24, 95% confidence interval 0.54–2.84). The effect of maternal education was also reduced after controlling for parental smoking; the adjusted rate ratios were 0.36 (0.08–1.59), 0.38 (0.06–2.36), and 0.14 (0.01–1.57) for children of mothers with apprenticeship, secondary, and university education, respectively, compared to children of primary educated mothers (p value for trend 0.232, not shown in table). The rate ratios for car and weekend house ownership did not change substantially after controlling for parental smoking, but the confidence intervals became wider owing to loss of power (not shown).

Discussion

This study suggests that parental smoking and the socioeconomic environment strongly influence the risk of invasive meningococcal disease in children; among socioeconomic factors, maternal education seems to play a prominent role in this population.

There are two main potential methodological problems with case-control studies: selection bias (related to recruitment of cases and selection of appropriate controls), and reporting bias (related to the retrospective nature of exposure ascertainment). In this study, selection bias is not likely. The study was population based, and subjects were recruited prospectively, as cases occurred in the study population. Reporting of invasive meningococcal disease is compulsory in the Czech Republic, and the coverage has been good.12In addition, an active surveillance was introduced after the outbreaks in 1993, in order to improve the accuracy and completeness of data collection and to speed up the reporting.9 ,10 Local epidemiologists searched for new cases in collaboration with local clinicians in district hospitals. It is extremely unlikely that a substantial number of cases were missed. Diagnostic bias is also unlikely; excluding the five cases who were diagnosed on the basis of clinical signs only did not change the results. The controls were selected from the same population from which the cases came, and their characteristics were similar to census data or to distribution found in previous studies in the Czech parents.13 ,14

Reporting bias is also improbable. The socioeconomic factors examined in this study are relatively stable and objective and, as mentioned above, their distribution among controls was similar to other recent population samples. It is not likely that parents of cases would systematically overestimate their smoking or underestimate their socioeconomic status and education. One would rather expect that parents of cases, if they suspected that smoking may be related to the disease, would tend to underreport their smoking habits; this would underestimate the effects of parental smoking. We think that it is unlikely that selection or reporting bias have affected the validity of this study.

Several biological mechanisms can provide the link between tobacco smoke and meningococcal disease. Smoking is negatively associated with cell mediated and humoral immunity,15 causes a fall in salivary IgA and a rise in IgM concentrations,16 affects respiratory epithelial functions, and increases bacterial adherence and the risk of inflammation and other infections.17-19 The observation that smokers are more likely to be meningococcal carriers is consistent with the increased risk of invasive meningococcal disease.20 Exposure to smoke causes direct damage to the nasopharyngeal mucosa; passive smoking is associated with an increased risk of respiratory disease in young children.21-23 As children have more delicate mucous membranes, they may be more likely to acquire meningococcal disease if they are chronically exposed to passive smoking.20

The epidemiological literature, although not extensive, is also consistent. Two British studies3 ,4 and two US studies5 ,6 found an increased risk in children of smoking parents, and a South African study found that parental smoking increased the risk of meningococcal disease in the presence of a recent respiratory infection.7 The adjusted relative risks for maternal smoking in the US studies were 3.8 (1.6–8.9)5and 2.9 (1.5–5.7)6; in the more recent British study, passive smoking in the home was associated with an odds ratio of 7.5 (1.5–38.7).4 Our results are consistent with these studies.

We have used several markers for parental smoking. The largest relative risk (more than eightfold) was observed in the children of parents who both smoked. This may be an overestimate, as families where both parents smoke are likely to have multiple social disadvantage which would not be taken into account by adjusting for the socioeconomic factors in the questionnaire. A rate ratio of 2.6 per 20 cigarettes smoked at home daily seems a conservative estimate of the real effect of passive smoking in children, but it indicates a relatively strong effect.

Smoking is a major source of indoor air pollution. We have included two other indirect measures of indoor pollution previously linked with respiratory symptoms and disease in children: type of heating (particles are a byproduct of coal burning) and cooking (burning gas generates nitrogen dioxide). Spending weekends out of town may also be related to lower indoor exposures, in addition to being a socioeconomic indicator. However, the numbers of children in individual categories were small and the estimates were imprecise; the results do not allow firm conclusions on indoor environment and meningococcal disease.

Existing data suggest that socioeconomic disadvantage increases the risk of meningococcal disease. In Brazil, the incidence of meningococcal disease was about two times higher in deprived areas than in more affluent areas.8 In both US studies, low maternal education and other social characteristics were associated with increased risk of the disease.5 ,6 In the UK, crowding and several other social indicators were related to meningococcal disease.3 ,4 Our results are consistent with these reports, and confirm our previous findings that maternal education is a powerful predictor of children's health.13 ,24 Some of the association between maternal education and meningococcal disease was (statistically) mediated by maternal smoking, but the fully adjusted rate ratios, despite the loss of statistical significance, suggest that a part of the effect may be independent from parental smoking. The effects of the father's education were largely eliminated in multivariate analysis. The observation that the effects of ownership of a car and a weekend house remained relatively unchanged in multivariate models indicates that material conditions may also contribute to the risk of the disease.

The socioeconomic factors measured in our study represent different aspects of living conditions relevant to the Czech situation. Education appears the best predictor of individual health behaviours of adults in Czech Republic,25 and available evidence suggests that maternal education is the best predictor of child health in the Czech population.13 ,24 ,26 The distribution of income in the Czech Republic is still relatively equitable, despite the increase in income inequality since 1992.27 Crowding, ownership of a car, and a weekend house or a cottage may be better indicators of the material circumstances than official income in post-communist societies. In particular, car and weekend house (or cottage) are both sufficiently desirable and common to be used as proxies for material circumstances. Because of the limited housing market, crowding is probably a poor measure of financial situation; it rather measures the housing conditions per se. Weekends spent outside of the town are mostly those spent in the weekend house but they also include trips, holidays, and visits to relatives and friends. The number of siblings were included as a possible indicator of both socioeconomic circumstances and, possibly, the risk of contacts with the infectious agent.

The findings on meningococcal disease are similar to other infections. For example, social deprivation and crowding was found to be associated with bacterial meningitis.28 Maternal smoking during pregnancy increased risk for sudden infant death syndrome and childhood infections.23 Passive smoking and low socioeconomic status now appear to be established as risk factors for invasive meningococcal disease. Although the exact mechanisms are not known, improving social conditions and reducing parental smoking seem appropriate interventions, not only with respect to meningococcal disease.