Mite and pet allergen levels in homes of children born to allergic and nonallergic parents: the PIAMA study.

The Prevention and Incidence of Asthma and Mite Allergy (PIAMA) study is a birth cohort study that investigates the influence of allergen exposure on the development of allergy and asthma in the first several years of life. The objectives of this study were to investigate the relationship between a family history of allergy and/or asthma and exposure of newborn children to mite and pet allergen and to study the influence of different home and occupant characteristics on mite allergen exposure. Dust was sampled from the child's mattress and the parental mattress at 3 months after birth of the index child and analyzed for mite and pet allergens. Subjects were divided in groups according to history of asthma and allergy in their parents, and allergen exposure was studied in the different groups. Cat allergen exposure was significantly lower on parental mattresses in families with allergic mothers, but dog allergen exposure was not different. Mite allergen exposure was lower on parental mattresses in families with allergic mothers. Use of mite allergen-impermeable mattress covers reduced mite allergen exposure. Some other characteristics such as age of home and mattress were also found to influence mite allergen exposure. Parental mattresses in homes of allergic mothers had lower cat and mite (but not dog) allergen loadings than mattresses in homes of nonallergic parents. Paternal (as opposed to maternal) allergy seemed to have little influence.

A family history of asthma or allergy is associated with an increased risk of developing allergy and asthma (Horst and Halken 2000;Sears 1998;Strachan 1999). Allergic families are likely to take measures to decrease exposure to allergens (Brunekreef et al. 1992;van Strien et al. 1995). Previously, we showed that allergic parents of infants included in the Prevention and Incidence of Asthma and Mite Allergy (PIAMA) study more often reported taking allergen-avoidance measures and had a cat at home less often than did nonallergic parents (Wijga et al. 2001). Such selective allergen avoidance could complicate studies of the relationship between allergen exposure and development of allergic disease in children. Early allergen exposure is associated with sensitization in children Wahn et al. 1997), but it is still unclear whether the development of asthma is also affected Lau et al. 2000).
Allergen avoidance measures and several different housing and occupant characteristics might influence mite allergen concentrations, as has been shown in a number of studies (Chew et al. 1998(Chew et al. , 1999Simpson et al. 2001;van Strien et al. 1994;Wickens et al. 2001). Characteristics associated commonly with higher mite allergen concentrations were increasing numbers of occupants, older homes, older mattresses, and dampness of the home. A large part of the variation in mite allergen concentrations is usually not explained by housing and occupant characteristics; therefore these characteristics remain relatively poor predictors of mite allergen exposure levels (Chew et al. 1999). Furthermore, Dermatophagoides pteronyssinus (Der p 1) and D. farinae (Der f 1) concentrations in house dust are weakly correlated with each other, suggesting that different housing characteristics influence concentrations of these two mite allergens (Gross et al. 2000).
In this study we investigated whether parental allergy and asthma are associated with allergen concentrations in mattress dust and which housing characteristics influence Der p 1 and Der f 1 concentrations in mattress dust.

Materials and Methods
The PIAMA study is a prospective birth cohort study, consisting of two parts: a natural history part, in which no intervention takes place, and a double-blind placebo-controlled intervention part in which the effect of miteimpermeable mattress encasings is studied. Pregnant women were recruited into the study during the first trimester of pregnancy and were selected on the basis of self-reported allergies and/or asthma, using a validated questionnaire (Lakwijk et al. 1998) that was distributed by 52 midwife practices and obstetric clinics in the Netherlands. In total, 3,291 pregnant women were included in the natural history study (NHS), of which 472 reported an inhalant allergy and/or asthma (NHS-a, response rate 62%) and 2,819 did not report an allergy, or asthma (NHS-na, response rate 55%). Wijga et al. (2001) described recruitment and inclusion of the natural history subjects in more detail. In the intervention study, 855 pregnant women reporting allergies and/or asthma were included (response rate 42%) and randomly assigned to a placebo and an active group.
Selection criteria for the group of allergic mothers in the natural history study and in the intervention study were the same. In 810 subjects, the intervention measures were applied-416 in the active group (IS-act) and 394 in the placebo group (IS-pl). The intervention consisted of mite-impermeable mattress covers for the parents' and the child's bed in the active group and cotton mattress covers in the placebo group, applied in the third trimester of pregnancy. All subjects in the intervention study were eligible for a home visit at 3 months after the birth, and all subjects with an allergic mother and a random sample (n = 655) of the subjects with a nonallergic mother were eligible in the natural history study. At 3 months after the birth in 716 (367 in the active group and 349 in the placebo group) homes in the intervention study, dust samples were collected. In the natural history study, dust samples were collected in 427 homes (91%) of mothers reporting an allergy and/or asthma, and 610 homes (93%) of mothers not reporting an allergy or asthma. Allergy and asthma in the father were assessed using the same questions that were used for the mother; these questions were asked in the third trimester of pregnancy and were not used to select subjects for the different subgroups of the study. We obtained written informed consent from all parents. Der p 1, Der f 1, Fel d 1, and Can f 1, respectively. The lower limit of detection was 8 ng/mL for Der p 1, 6 ng/mL for Der f 1, 0.4 mU/mL for Fel d 1, and 20 ng/mL for Can f 1 for 5-fold diluted samples. Samples with undetectable amounts of allergen were given a value of two-thirds of the detection limit because the distributions were log-normal. The undetectable values were not distributed equally over the range below the limit of detection, but there will be more close to the detection limit than close to zero. To reflect this, we arbitrarily chose a value between 0.5 and 1 at two-thirds. This was done for the results of 36%, 27%, 1%, and 60% of the samples taken from the parental mattress, and 65%, 52%, 13%, and 68% of the samples taken from the child's mattress, respectively, for Der p 1, Der f 1, Fel d 1, and Can f 1. Of dust samples with an undetectable amount of dust (< 11 mg), the amount of allergen per gram dust was not calculated and is therefore missing from all analyses. This was true for two samples taken from the parental mattress and 93 (5%) samples taken from the child's mattress. About 20% of the samples were analyzed in duplicate from a second Statistical analysis. Differences in median dust and allergen levels between groups of parents with and without allergies and/or asthma in the natural history study were tested using Wilcoxon's two-sample test. We also assessed the influence of season of sampling on both mite allergens separately, using ln-transformed allergen concentrations and 3-month moving averages, to diminish random variation (Chatfield 1996). Because sampling in the PIAMA study was done throughout the year, we had to account for season. We did so using multiple linear regression analysis, with dummy variables for the different seasons (January-March, April-June, July-September, and October-December). The months October, November, and December (autumn) were used as the reference season in all analyses. All housing and occupant characteristics were categorized and entered in multiple linear regression models, using ln-transformed dust and mite allergen concentrations as dependent variables. We used the regression coefficient for each variable to calculate the relative difference the value of this variable made for the dependent variable (i.e., the dust or allergen concentration). The reference concentration is the concentration a subject would have had when all variables were in the reference category. Dust samples with undetectable allergen levels were included in all analyses. Because the results for allergen loads expressed in unit per square meter were similar to those expressed per gram dust, the results are presented for allergen loads expressed in units per square meter only-a more direct measure of exposure to the total amount of allergen per surface unit. To illustrate the similarity, a graphic presentation of mite allergen levels by season is presented for both expressions. p-Values < 0.05 were regarded as statistically significant. Table 1 shows percentages of subjects with different housing and other characteristics, possibly associated with mite allergen concentrations. Tables 2 and 3 show dust and allergen quantities per square meter for the child's bed and the parents' bed, respectively. Both tables show data from the NHS only. On the mattresses of the children, there were no differences in mite and pet allergen loadings in relation to parental allergy. The mattresses of the parents showed a slightly different picture: Cat allergen loadings were lower when the mother or both parents were allergic. Mite allergen levels may vary by season. Samples in this study were taken throughout the year. The geometric mean amounts and concentrations of Der p 1 and Der f 1 for both beds per month of sampling are shown in Figure 1, which shows 3-month moving averages. The graphs show that both mite allergens follow a clear seasonal pattern. Levels of both allergens were higher in autumn and reached their lowest level at the end of winter. Seasonal influence was stronger for Der f 1 (factor 2 change) than for Der p 1 (factor 1.5 change).

Results
In Tables 4 and 5 various home and occupant characteristics are shown with regard to their influence on the amount of dust and Der p 1 + Der f 1 (Der 1, ng/m 2 ). Table 4 shows that the amount of dust, as well as the amount of mite allergens per square meter from the child's bed, was significantly lower for children in the active group of the intervention study, compared to the natural history group with allergic mothers. Child's mattresses in the intervention study placebo group had comparable amounts of dust and mite allergen concentrations as mattresses in the natural history group with allergic mothers. Other characteristics associated with higher concentrations of Der 1 in dust   Der p 1 Der f 1 from the child's bed were age of the mattress, absence of double-glazed windows in the child's bedroom, season, and the child having older siblings. Table 5 shows that the amount of dust as well as both mite allergen concentrations in the parents' bed, were significantly lower in the intervention study active group, compared to the natural history group with allergic mothers. Placebo mattress covers lowered Der 1 loadings as well, and dust and Der 1 loadings were higher on mattresses of nonallergic than allergic mothers. In addition, Der 1 levels varied with season and region, were higher in older mattresses and houses, and lower when the child was firstborn. Despite the fact that many determinants were investigated, only a small fraction of the variance in Der 1 loadings could be explained by our models. Table 6 shows that the overwhelming determinant of cat and dog allergen in the child's mattress was presence of a cat or dog in the home. In addition, there was a modest influence of the presence of mattress covers and mechanical ventilation (with dog allergen), and age of mattress (with cat allergen). Table 7 shows the results for the parents' mattresses. Again, the influence of the presence of a cat or dog in the home is overwhelming. Presence of mattress covers were more important on parents' than on babies' mattresses; few other factors reached significance.

Discussion
This study showed that in the PIAMA birth cohort, children of mothers with inhalant allergy and/or asthma were born into an environment with lower cat allergen concentrations in mattress dust of parents, which was completely explained by a lower prevalence of cat ownership. Der 1 loadings were also lower on parents' mattresses in families with allergic mothers. Several home characteristics such as age of home and mattress were shown to increase mite allergen concentrations. Families with allergic fathers but not allergic mothers did not have lower mattress loadings of any allergen compared to families in which no parent was allergic. Wijga et al. (2001) showed that parents with an allergy reported taking allergen-avoidance measures while furnishing their homes. This study shows that mite allergen levels on parents' mattresses are indeed lower when the mother is allergic. Once maternal allergy was taken into account, however, reported allergen-avoidance measures had no further influence. In previous studies, the Der p 1 concentration in mattress dust from homes of allergic subjects was also found to be lower than in homes of nonallergic subjects (van Strien et al. 1995;Verhoeff et al. 1994).
Parents' but not children's mattresses in families with allergic mothers had less cat allergen, as was expected, since Wijga et al. (2001) showed that families in which both parents are allergic were less likely to own a cat than families without parental allergies. The difference between allergic and nonallergic mothers disappeared after taking the difference in cat ownership into account.
For dog allergen there was no difference in the allergen concentration of mattress dust between families with or without maternal allergies and/or asthma. This corresponds to the fact that owning a dog was not influenced by allergies and/or asthma reported by the parents (Wijga et al. 2001  1,135 1,081 Adjusted R 2 0.14 0.17 a Concentration ratio between presence and absence of characteristic described in table. b Concentration is the reference concentration a subject would have had when all variables in the model assume the reference value (0). *p < 0.1; **p < 0.01; # p < 0.05.
Mite allergen concentrations were low in this study compared to earlier studies we conducted in the Netherlands (van Strien et al. , 1995. Concentrations of mite allergen in children's mattresses were comparable, however, to concentrations found by Custovic et al. (2000) on babies' mattresses, and also comparable to the levels found by Wahn et al. (1997). A direct comparison between samples taken in previous studies and samples taken in the PIAMA study is fairly difficult because the subjects as well as dust sampling and analysis were different. More study of the changes in mite allergen concentrations and the possible reasons for these changes ) is justified.
Different home and subject characteristics were studied with regard to their influence on mite allergen concentrations. This revealed that even at low concentrations some characteristics were still associated with mite allergen levels. Firstborn children had, on average, lower mite allergen concentrations in dust from their own beds, and the parents also had lower mite allergen concentrations in dust from their beds. This effect was not caused by the use of new mattresses because the age of the mattress was included in all models, but it could have been caused by the use of new bedding for the child's bed and the parents' bed. Chew et al. (1999) showed that dust from mattresses in apartments had significantly lower mite allergen concentrations. In this study, parents' mattress Der p 1 concentrations were lower, but Der f 1 concentrations were significantly higher (data not shown), with a net result that Der 1 levels were not significantly higher in apartments. Der p 1 concentrations seemed to be influenced more by housing characteristics associated with dampness and the age of the house, whereas Der f 1 was more influenced by seasonal changes and age of the mattress (data not shown). Overall, the characteristics studied here explained only about 10% of the variation in mite allergen concentrations, whereas in a previous study 25% of the variation in the Der p 1 concentration of dust from the beds of schoolchildren was explained by the same type of characteristics . This is likely related to the much lower allergen levels in this study. Comparison of children using mite-impermeable mattress covers with similarly selected children without mattress covers showed that mite allergen loadings were lower in dust from the parents' mattress as well as children's mattresses. For children's mattresses, this seemed primarily an effect of smaller dust loadings, but for parents' mattresses, the effect on mite allergen loadings was much larger than for dust loadings.
In conclusion, cat allergen exposure was lower in parents' mattresses in families of children born to allergic parents, and dog allergen was not affected. Mite allergen exposure was lower on the parents' mattress in families with allergic mothers compared to nonallergic mothers. Only a low percentage of the variation in mite allergen concentrations could be explained by housing and occupant characteristics. The major determinant of pet allergen loadings was having a pet in the home.
1,031 1,021 Adjusted R 2 0.35 0.40 a Concentration ratio between presence and absence of characteristic described in table. b Concentration is the reference concentration a subject would have had when all variables in the model assume the reference value (0). *p < 0.1; **p < 0.01; # p < 0.05.