Many investigators have studied the relationship between living in damp or mouldy homes and respiratory health.1 Although an ongoing Swedish study reports infrequent occurrence of damp,2 several European studies have reported dampness or mould occurring in roughly 10–50% of homes.3^–6 These conditions may adversely affect health in a number of ways.7^–9 Viable and non-viable mould fragments may induce respiratory disorders by a variety of potential mechanisms, such as immunoglobulin E-mediated hypersensitivity reactions proteins, volatile organic compounds or cell wall components such as ergosterol and 1,3-β-d-glucan, inducing non-specific inflammatory reactions and mycotoxins, which reduce the defence system of the lung by interfering with macrophage functioning.10^–13 Dampness in homes may also indicate increased presence of dust mites, which may explain some of the association.14 15

Several studies have found associations between health and direct measures of mould, such as spore counts, ergosterol and glucan levels, but others failed to show clear association between total fungal counts and respiratory disorders in children.16^–19 However, directly measured mould exposures such as spore counts in indoor air samples, and even in indoor dust reservoirs, have been shown to have only low to moderate reproducibility.20 21 Indirect measures such as visible moulds or mildew on surfaces have been commonly used instead for assessment of mould exposure, and may better represent long-term exposure to moulds than direct measurements during a short sampling time.22

The Pollution and the Young (PATY) project was a collaborative enterprise. Researchers from several comparable cross-sectional studies pooled their original data on exposures and health status for 58 561 children. The studies were carried out during 1988–1999 in 10 countries in Eastern and Western Europe and in Russia and North America. PATY was established, with EU funding, primarily to investigate associations between air quality and respiratory health. Here, we make use of the subset of the study population (comprising 57 161 children aged 6–12 years) with self-reported information on household mould. Pooling the original data allows the harmonisation of data analysis, the pursuit of research questions that were not addressed in the original studies, and the inclusion of unpublished studies, giving a powerful opportunity to explore the effects of mould on children’s health. Relationships between home dampness or mould and respiratory health have been reported previously for only two of the countries included in the PATY project.23 24

METHODS

Studies and participants

Studies were identified from the published literature and others (unpublished at the time of planning PATY) of which we were aware. The primary aim of the PATY study was to assess associations between ambient air pollution and children’s respiratory health, and recent cross-sectional studies with this focus were sought. Studies were considered which had (a) assessed respiratory symptoms and individual risk factors by questionnaire, (b) included cough and wheeze as primary health outcomes and (since a primary aim of the PATY study was to investigate effects of outdoor air pollution) (c) allowed the calculation of annual mean particulate matter measures within a number of study areas.

Table 1 provides a summary overview of the studies which contributed to this paper. More details may be found in published reports.23^–30

Health endpoints

Eight outcomes were identified, on the basis of the compatibility of their wordings, their scientific interest and their availability across the studies. Where possible, symptoms for the last 12 months were preferred to symptoms “ever”, to reduce recall bias. The outcomes were: wheeze in the last 12 months, asthma ever, bronchitis in the last 12 months, nocturnal dry cough in the last 12 months, morning cough, “sensitivity to inhaled allergens”, hay fever ever, and “woken by wheeze in the last 12 months”. The exact wordings have been published elsewhere.31

Definition of exposure

It was not possible to construct an exposure variable which was identical in all studies. We constructed a binary variable “indoor mould” to be as consistent as possible, but there were some between-study variations. For 10 of the studies, it referred to any visible presence of mould in the household. For the study carried out in The Netherlands, it referred only to mould in the last 12 months, and, for the Italian study, only to mould in the child’s bedroom. Exact wordings of the questions are provided in Appendix A. We assessed the sensitivity of the results to differences in these definitions by comparing, when data were available, effects of mould in the child’s bedroom with mould in other rooms, and of recent mould with mould in previous years.

Potential confounders

We controlled for age, sex, maternal education, paternal education, nationality, household crowding, use of gas for cooking, presence of unvented gas/oil/kerosene heaters, maternal smoking during pregnancy, presence of a current smoker in the household, birth order and “ever had a pet”. “Year” was controlled for in the Austrian analyses, since questionnaires were distributed in three separate years in all areas. Confounding at higher levels and possible geographic clustering of outcomes were accounted for by including a “study area” term in the model. Not all potential confounding variables were available for all studies (though most were); the impact of this was tested in a sensitivity analysis.

Potential sources of heterogeneity

We assessed the role of several potential sources of differences between country-specific estimates of associations between indoor mould and respiratory health:

• season of questionnaire (defined as two-thirds or more of study questionnaires completed in the Spring, February–April), since the reported prevalence of symptoms and also exposure to indoor mould may vary by season

• age, by comparing effect estimates between younger (6–8 years) and older (9–12 years) children

• earlier studies versus later studies (using an arbitrary cut-off point of 1995)

• a rough East/West categorisation to compare results from the more “affluent Western” countries with those in the former “Eastern block”

• studies with higher response rates (⩾80%) versus lower response rates (<80%).

Statistical analysis

A two-stage approach was used. In stage 1, study-specific estimates of effects of mould were estimated using logistic regression with robust standard errors, in STATA version 8. In stage 2, study-specific effect estimates and their standard errors were entered into a meta-analysis, from which we obtained forest plots of the estimates, a combined (mean) estimate, and a measure and Cochran χ² test of between-study heterogeneity. The study-specific estimates were assumed to follow a random distribution about a mean, and the estimation of this mean and its confidence intervals takes into account the variation in study-specific estimates and the uncertainty (due to sampling variability) related to each study-specific estimate.32 The approach taken allows covariates to have different effects in different studies; this is convenient when their definitions differ slightly between studies, and also reflects the diverse study populations.

Lastly, meta-regressions were conducted on the study-specific estimates, including study-level potential sources of heterogeneity as covariates, to (a) assess the association between the study factor and effect estimate, (b) estimate a combined effect stratified by study factor and (c) estimate the degree of residual heterogeneity between study-specific effect estimates after taking the study factor into account.

Sensitivity analyses

We tested four individual factors (sex, age, household density and passive smoking) as potential effect modifiers by comparing results in boys with those in girls; younger (6–8 years) versus older (9–12 years) children; children from crowded versus non-crowded households; and children from homes with current smokers versus households without current smokers.

We also tested parental allergy (either parent ever had eczema, skin or nasal allergy, or hay fever) as a potential effect modifier. Seeing a stronger association when parents had allergy might suggest recall bias. Alternatively, parental symptoms of allergy may also have been caused by indoor mould, which could lead to overadjustment; the stratified analysis is a way to address that.

RESULTS

Table 2 shows the categorisation of studies with respect to potential sources of heterogeneity, and the prevalences of respiratory symptoms and of mould. Bronchitis was the most commonly reported symptom, with a prevalence exceeding 30% in 4 out of 10 studies. “Woken up by wheeze in the last 12 months” was generally the most infrequently reported, with prevalences between around 2% and 16%.

The percentage of children exposed to mould ever ranged from 13.9% in Russia to 39.1% in North America (table 2). Mould in bedrooms was less prevalent.

With only two exceptions (the associations in The Netherlands and in Germany between hay fever and mould), confounder-adjusted odds ratios (ORs) for exposure to indoor mould were greater than 1 in all studies for all outcomes (fig 1). Estimates from The Netherlands were usually among the lowest, and those from Russia among the highest.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Forest plots of study-specific ORs and combined ORs for the effects of mould exposure. The vertical line indicates the null position (OR 1). The horizontal line represents the 95% CI of the estimate. The diamond shape at the bottom indicates the position and CI of the mean of the estimates. Extreme CIs are truncated at 0.5 and 4.0. ORs are adjusted for area and individual risk factors: age, sex, parental education, nationality, household crowding, gas for cooking, unvented heater, post- and prenatal exposure to tobacco smoke, birth order, and “ever had a pet”.

Combined ORs were similar for all outcomes, ranging from 1.30 (95% CI 1.22 to 1.39) for nocturnal cough to 1.50 (1.31 to 1.73) for morning cough (table 3). Study-specific results were heterogeneous for four of the eight outcomes (asthma, bronchitis, morning cough and woken by wheeze), within this clear pattern of positive associations.

Analyses controlling only for age, sex and area gave similar results to those from the fully adjusted model (results not shown). The largest change was seen for morning cough: the combined OR from the age–sex adjusted model was 1.56 (1.36 to 1.79), whereas that from the fully adjusted model was 1.50 (1.31 to 1.73).

Results within groups of studies, stratified by potential source of heterogeneity

Few of the potential sources of heterogeneity showed any consistent association with the magnitude of study-specific results. Studies with response rates above 80% had, on average, higher ORs than studies with lower response rates for all outcomes except wheeze (significantly so for only bronchitis, sensitivity to inhaled allergens, and morning and nocturnal coughs).

Sensitivity analyses

No consistent impact of variations in the wording of mould questions was found in these analyses. Table 4 shows a comparison of combined ORs among the studies which had information on mould ever and on recent (in the last 1 or 2 years) mould. Mean ORs are lower for recent mould than for mould ever for some outcomes, but the direction of change was not consistent across studies, and differences were generally small (study-specific results not shown). We note, however, that The Netherlands, which showed generally low effect estimates in the main analyses, had “recent mould” as its main exposure variable. No consistent interactions were seen between mould and current smoking, age or sex.

ORs for asthma were higher in more crowded households in 10 out of 11 countries, with a combined OR of 1.25 (1.10 to 1.41) in less crowded households and 1.60 (1.35 to 1.89) in more crowded ones, which is a significant difference between groups (p = 0.005). For sensitivity to inhaled allergens also, ORs were higher in more crowded households for 8 out of 10 households, with a combined OR of 1.28 (1.16 to 1.41) in less crowded households and 1.57 (1.35 to 1.81) in more crowded ones (p<0.001). For other outcomes, ORs were not consistently or significantly higher in more crowded households.

Combined effects were greater for all outcomes except hay fever and sensitivity to inhaled allergens among children with healthy parents than among children with reported parental illness. However, mean interaction terms were not statistically significant for any outcome, differences were generally small, and their directions were not highly consistent across individual studies.

DISCUSSION

Relationships between mould in the household and a spectrum of respiratory and allergic disorders among children aged 6–12 years were demonstrated in our study. Estimates were highly consistent for wheeze, nocturnal cough, sensitivity to inhaled allergens and hay fever. For the four other outcomes there was some evidence of heterogeneity, but within a pattern of almost entirely positive results. Effect estimates were not high but exposures were common—over 20% in most countries.

Mould growth depends critically on moisture availability and relative humidity. Moisture in homes can result from a number of conditions, including lack of ventilation, leakage, high humidity and crowding. Crowding may be associated with more serious mould problems because of more moisture-generating activities in the home. There was some evidence in our analyses of effect modification by crowding in the home, although this was statistically significant for only asthma and sensitivity to inhaled allergens, and may have been a chance finding.

Our analysis is based on self-reported questionnaire data for exposure and for outcomes. Ambient outdoor air pollution was known to be the focus of all studies, yet, even so, parents who themselves have allergies, for example, may be more alert to their child’s health and also to the presence of allergens in the home. Similarly, parents of sick children may be more likely to report mould. These could lead to systematic reporting bias. However, sensitivity analyses showed that estimated effects of mould were no higher among children with parents reporting respiratory symptoms in themselves than among children with healthy parents. This also encourages us to think that reporting bias from parents with sick children was probably not substantial, although we cannot rule out the possibility.

A limitation of these analyses is that, since the PATY study originally focused on recent air pollution, studies specifically examining mould are not represented. In addition to the loss of these studies, our questions on mould were not highly tailored to the purpose, and may underestimate true effects. There are compensations. Using these databases to investigate mould effects may reduce bias: recall bias, as mentioned above, and biases such as publication bias when selecting studies.

A concern with meta-analyses in general is the comparability of studies. Since these studies were initiated independently, they sometimes differed in details of design and of wording of symptoms and exposures. We assessed symptom and confounder questionnaires to extract the symptoms assumed to be most comparable across studies. In all studies, self-reported symptoms, which are non-specific as predictors of mould-related respiratory disorders, were used as the outcome variables. Most of these non-specific symptoms can result from allergic and non-allergic responses to different mould components. Wordings of outcomes and of risk factors (mould/moisture) were generally similar across the study questionnaires, and, when there were differences, sensitivity analyses showed no systematic effect of these differences on results. Because the same statistical model and the same confounder model were used for all studies, the current analyses will be more comparable than traditional meta-analyses when only published effect estimates are compared across studies.

Comparison with original analyses of the same data

Two PATY studies had previously analysed associations between self-reported mould and symptoms, and original results were close to those from the PATY analyses.23 24 In addition, the German researchers, on a subset of 272 children, had measured allergic sensitisation and household spore counts, finding a positive association between the two, and also that sensitised children exposed to high spore counts were more likely to have rhino-conjunctivitis.3

Comparison with other previous studies

The 2001 NORDDAMP review reported ORs ranging between 1.4 and 2.2 for associations between dampness and health effects such as cough, wheeze and asthma.1 A number of subsequent studies (on adults and/or children) have reported similar associations between respiratory symptoms and indoor damp or mould.2 33^–35 Some used self-reported exposures. Others, with access to homes, were able to measure actual current exposure. Evidence has accumulated although not all previous results have been statistically significant, nor have they all been positive, or consistent across symptoms, or across groups of children. For example, Andriessen et al,36 in one of the few studies to include Eastern Europe, found no overall associations between mould/damp and prevalence of respiratory symptoms. Among symptomatic children, however, frequency of reporting cough or phlegm was higher in the presence of mould/damp (perhaps suggesting exacerbation), although further analysis showed this association to be restricted to atopic children.36

Effect estimates from the PATY study are of a similar magnitude to those from previous studies, with the bulk of individual (study outcome-specific) estimates ranging from 1.2 to 1.6, and occasional estimates below 1.0 or exceeding 2.5. About one-third of these estimates were not individually statistically significant, but in combination the evidence was overwhelming of the harmful effects of mould across the whole range of symptoms measured. Yet despite these and previous results, it may be that these associations have not come to the public’s attention. Parents in the five CESAR component studies, as part of the health questionnaire, were given a tick list of perceived causes of children’s breathing troubles. Their answers showed mould and dampness not to be widely considered hazardous to health. While 79% of parents ticked “industrial pollution” as a cause of children’s breathing troubles, and 53% and 49% ticked traffic and tobacco smoke, respectively, only 13% indicated that they thought household dampness a cause. Yet our results confirm mould to be hazardous among the general population of schoolchildren across this diverse group of countries. Respiratory illnesses during childhood play an important role in modifying children’s health, with long-lasting effects for their future. This highlights the important role of mould and the indoor environment in relation to public health.