|
INDOOR NITROGEN DIOXIDE, GAS COOKING AND ASTHMA 1727
<br />Statistical methods
<br />We conducted meta-analyses to obtain summary risk
<br />estimates for the association between asthma, wheeze
<br />and household NO2 exposure and its surrogate, gas
<br />cooking. For every single exposure variable, to distin-
<br />guish the differing reporting times of symptoms
<br />between studies, we reported not only the overall
<br />meta -OR combining all the studies but also the sub-
<br />group meta-ORs in both 'current' and 'lifetime'
<br />asthma (or wheeze). When a study reported risk esti-
<br />mates For different strata of the population, e.g. for boys
<br />and girls ,3A,37-39 children with asthmatic mothers or
<br />non -asthmatic mothers40 and children living in single -
<br />or multi -family houses,41 we included these directly
<br />into the meta-analysis without combining them first.
<br />The risk estimates for the exposure ves non -exposure
<br />categories of gas cooking were summarized.
<br />For NO2 exposure, we calculated two types of pooled
<br />risk estimates: (i) for the comparison of asthma and
<br />wheeze risk at high vs low exposure independently of
<br />the exact definition of high and low exposure, and
<br />(ii) for asthma and wheeze risk per 15 -ppb increase
<br />in continuous NO, concentration. For inclusion in the
<br />meta-analysis, we converted all results in µg/m3 to
<br />l5 ppb using standard pressure and temperature. In
<br />the high vs low exposure meta-analysis, the included
<br />studies reported different specific ranges for NO2,
<br />which precludes a direct comparison of effect esti-
<br />mates from these studies. Some studies categorized
<br />NO2 levels into more than two categories; from
<br />these, we selected ORS for the highest compared
<br />with the lowest exposure category. We appreciate
<br />that this analysis is semi -quantitative.
<br />Heterogeneity
<br />We used standard chi-square tests to examine the
<br />heterogeneity among studies; results were defined as
<br />heterogeneous for P <0.10.2 The Iz statistic was used
<br />to quantify the extent of inconsistency among the
<br />studies. The I2 values <25% reflect low inconsistency,
<br />values of 25-75% reflect moderate inconsistency,
<br />whereas values >75% indicate high inconsistencies
<br />among studies." Due to the heterogeneity among stu-
<br />dies which were performed independently by different
<br />researchers in different populations, pooled risk esti-
<br />mates were calculated by random -effect models with
<br />inverse -variance weights 44 Summary estimates from
<br />fixed -effect models were also presented in the Forest
<br />plots for comparison.
<br />Influence analysis
<br />To evaluate the influence of individual studies on the
<br />summary effect estimate, we performed influence
<br />analysis. This method recalculated the summary esti-
<br />mate, omitting one study at a time.
<br />Sensitivity analysis
<br />Sensitivity analyses were employed to test whether
<br />the risk estimates varied by study region and age of
<br />the participants. Age was categorized into <6 years,
<br />6-10 years and >10 years. Further subdivision of the
<br />youngest category was not possible because of the
<br />number of studies performed within that age range.
<br />Study regions were divided into Europe, North
<br />America, and Asian and Pacific area.
<br />We noticed that the proportion of gas cooking varied
<br />considerably between studies. In order to examine
<br />whether observed exposure health relationships of a
<br />study were associated with the percentage exposed to
<br />gas cooking, stratified analyses were performed using
<br />30% of cooking with gas stoves as a cut-off.
<br />In our database, there were some studies which were
<br />conducted a long time ago. Since then, the actual use
<br />and the emissions of gas cookers as well as disease
<br />management strategies may have changed. To exam-
<br />ine the influence of older studies, we compared risk
<br />estimates between older and newer studies as part of a
<br />sensitivity analysis. For operational purposes, the pub-
<br />lication year 2000 was used as the cut-off.
<br />Subsequently, exploratory univariate meta -regres-
<br />sions were performed to assess whether heterogeneity
<br />in associations between gas cooking and asthma and
<br />wheeze between studies was related to age of the par-
<br />ticipants, study region, proportion of gas cooking and
<br />year of publication.
<br />Furthermore, random effects models were per-
<br />formed to determine the potential impact by asth-
<br />matic subjects. Asthmatic children may be more
<br />sensitive to the effects of indoor NO2. Therefore, we
<br />repeated analyses of the associations of gas cooking
<br />and indoor NO2 with wheeze by excluding two Stu-
<br />dies which focused on asthmatic children only at the
<br />initial recruitment.
<br />Assessing publication bias
<br />Publication bias was evaluated with funnel plots and
<br />the Egger's and Begg's tests. 45
<br />All statistical analyses were performed using STATA
<br />(version 10; StataCorp LP, College Station, TX, USA),
<br />employing the 'metan', 'metabias' and 'metainf com-
<br />mands for meta-analyses and bias evaluation.
<br />'Metareo was used to test differences in effect size
<br />between subgroups of studies_
<br />Results
<br />A now chart of the selection stages of the studies for
<br />analysis is shown in Figure 1. We extracted data from
<br />41 studies published since 1977 assessing the relation-
<br />ship between household NO2 or gas cooking and
<br />asthma and wheeze (Supplementary Table 1, available
<br />as Supplementary data at IJE online). 13,29,32-o,4!74
<br />Among those 41 studies, 19 studies were conducted
<br />in Europe (UK, Austria, Germany, Netherlands, Czech
<br />Republic, Spain and Russia), 14 in North America
<br />(USA and Canada), 3 in Asia (China and Japan), 4
<br />in Australia and 1 in New Zealand. Among them, four
<br />studies contributed information on infant s32,40,65,75
<br />0
<br />0
<br />a
<br />0
<br />11
<br />
|