Dr Bahi Takkouche Department of Preventive Medicine Faculty of Medicine University of Santiago de Compostela 15782 Santiago de Compostela Spain
Background: Exposure to pets has been implicated as a risk factor for asthma. However, this relation has been difficult to assess in individual studies because of the large potential of selection bias. We sought to examine the association between exposure to furry pets and asthma and allergic rhinitis by means of a meta-analysis.
Methods: We retrieved studies published in any language by searching systematically Medline (1966–March 2007), Embase, LILACS and ISI Proceedings computerized databases, and by examining manually the references of the original articles and reviews retrieved. We included cohort and case–control studies reporting relative risk estimates and confidence intervals of exposure to cats, dogs and unspecified furry animals and subsequent asthma and allergic rhinitis. We excluded cross-sectional studies and those studies that did not measure exposure but rather sensitization to pets.
Results: Thirty-two studies were included. For asthma, the pooled relative risk related to dog exposure was 1.14 (95% CI 1.01–1.29), that related to exposure to any furry pet was 1.39 (95% CI 1.00–1.95). Among cohort studies, exposure to cats yielded a relative risk of 0.72 (95% CI 0.55–0.93). For rhinitis, the pooled relative risk of exposure to any furry pet was 0.79 (95% CI 0.68–0.93).
Conclusions: Exposure to cats exerts a slight preventive effect on asthma, an effect that is more pronounced in cohort studies. On the contrary, exposure to dogs increases slightly the risk of asthma. Exposure to furry pets of undermined type is not conclusive. More studies with exact measurement of exposure are needed to elucidate the role of pet exposures in atopic diseases.
Asthma is a frequent disease, the prevalence of which varies more than 15-fold among countries worldwide (1). In Europe, the UK (20.7%) and Ireland (15.2%) have the highest rates, whereas Eastern European countries have the lowest (2).
In a similar fashion to the variation in prevalence of asthma, exposure to pets shows significant variations worldwide, being most frequent in Western countries. In the US, 39% of the households have a dog and 34% a cat (3). In the UK, about half of the households own a pet (4), whereas in Australia this proportion reaches 63% (5). On the contrary, pet keeping at home is rare to virtually nonexistent in the majority of Arab and African countries (6). Recently, Saudi Arabia has even prohibited pet possession (7).
Exposure to pet allergens has been implicated as a risk factor for asthma. However, this relation has been difficult to assess in individual studies because of the large potential of selection bias (8). In some studies, it has been shown that subjects with asthma refrain from having pets at home (8). This may lead to the erroneous conclusion that pet ownership provides a protective effect, mainly in cross-sectional studies but also in other studies where exposure timing is not determined with precision.
Recent reviews, essentially narrative in nature, have commented on the association between exposure to pets and asthma or atopy (8–10). The only quantitative review focused on studies published in the 1990s only, mainly included cross-sectional studies and did not consider pets separately (8). So far, no comprehensive meta-analysis is available. We, therefore, summarized the scientific evidence and carried out a meta-analysis on exposure to pets and risk of asthma, following the MOOSE guidelines for meta-analyses of observational studies (11).
Data sources and searches
We conducted a computerized Medline search from 1966 to March 2007 to identify potentially eligible studies. We applied the following algorithm both in Medical Subject Heading and in free text words: (ASTHMA OR RHINITIS OR RESPIR*) AND (CAT* OR DOG* OR PET* OR ANIMAL* OR FURRY) AND (CASE–CONTROL OR CASE-REFERENT OR RETROSPECTIVE OR COHORT OR FOLLOW-UP OR PROSPECTIVE). To check whether every article on the topic was retrieved, we performed a second search, introducing the words ‘asthma’, ‘rhinitis’, ‘cats’ and ‘pets’ in an unstructured fashion. We used similar strategies to search Embase (1980–2007) and LILACS databases (Latin America and Caribbean). We searched meeting abstracts using the ISI Proceedings database from its inception in 1990–2007. We also examined the references of every article retrieved and those of recent reviews of asthma and pets (8–10). We considered including any relevant article, independent of the language of the publication. Unpublished studies were not considered. All searches were carried out independently by an epidemiologist and a pneumologist (BT and FJGB) and results were merged.
Studies were included if they met the following criteria: (i) presented original data from case–control or cohort studies, (ii) the outcome of interest was asthma or rhinitis. Other atopic manifestations were not considered, (iii) the exposure of interest was ownership of cats, dogs or other furry pets, (iv) provided relative risk estimates and their confidence intervals or provided enough data to calculate them (raw data, P-value or variance estimate).
Because they are beyond the scope of this meta-analysis, we excluded studies that used single asthma symptoms as an outcome, such as wheezing, and did not establish a diagnosis of asthma. For the same reason, we did not consider those studies that dealt with sensitization to pets. In some instances, sensitization is used by authors either as a proxy for exposure or as an outcome and is generally measured through IgE or skin tests (9).
If data were duplicated in more than one study, the most recent one was included in the analysis. When relative risks of different pets were available in the same publication, we considered each pet separately.
We developed a questionnaire and recorded study name, year of publication, study design, sample size (cases and controls or cohort size), type of controls for case–control studies (hospital or population controls), variables used for adjustment or matching, and association measures that compared exposed subjects with unexposed. Because of their inadequacy to assess causation, cross-sectional studies were discarded from this meta-analysis.
As no universal scale is available for measuring quality of observational studies, we followed the recommendations of the MOOSE guidelines and assessed the quality of key components of design separately rather than generating a single aggregate score (11). Following this recommendation, we assessed study quality on the basis of the following five criteria labeled as ‘yes’ or ‘no’: (i) whether the target population was clearly defined (yes) or on the contrary, based on convenient sampling of subjects (no), (ii) whether asthma diagnosis included clinical features and flow rate and/or reversal after treatment (yes) or was based on clinical examination only (no), (iii) whether exposure assessment included measurement of pet allergens at home (yes) or was based on questionnaire only (no), (iv) (yes) or else (no), and, finally, (v) whether results were adjusted for age, sex and other potential confounders (yes), or else (no). Throughout this assessment, when the information on a specific item was not provided by the authors, we graded this item as ‘no’.
Within each item, we calculated two pooled odds ratios: one for those studies that were labeled ‘yes’ and another for those labeled ‘no’. As a secondary analysis, we carried out a pooled analysis on those studies that fulfilled more than three criteria and compared with those that scored 3 or less. The complete protocol for quality scoring is available upon request.
Data synthesis and analysis
We weighted the study-specific adjusted log odds ratios for case control studies and log relative risks for cohort studies by the inverse of their variance to compute a pooled relative risk and its 95% confidence interval. We presented both fixed and random effects pooled estimates, but used preferentially the latter when heterogeneity was present. The fixed effects model assumes that there is no between-study variance, i.e. that the results of the studies used in the meta-analysis are homogeneous and that their variation is because of sampling only. The random effects model, on the contrary, assumes that study results are heterogeneous. The random effects model yields pooled results that are less precise in nature (have wider confidence intervals) but are closer to the true value in case heterogeneity exists.
We used a version adapted to small samples of the DerSimonian and Laird Q test to check for heterogeneity (12). The null hypothesis of this test is the absence of heterogeneity. To quantify this heterogeneity, we calculated the proportion of the total variance as a result of between-study variance (Ri statistic) (12). To further explore the origin of heterogeneity, we restricted the analysis to subgroups of studies defined by study characteristics such as case–control/cohort design and type of controls (hospital-based or population-based).
We used funnel plots to assess publication bias visually. As funnel plots have several limitations and represent only an informal approach to detect publication bias (13). we further carried out formal testing using the test proposed by Egger (14). All analyses were performed with the software hepima® version 2.1.3 (15) and stata version 8.0 (StataCorp LP, College Station, TX, USA).
Our search strategy retrieved 3311 references, but only 32 different articles met our inclusion criteria (16–47). The large majority of these references were excluded because they did not provide data on exposure to pets, they presented cross-sectional data, failed to present any measure of uncertainty of the relative risk, or because they presented data on wheezing, bronchitis, sensitization or atopy but not on asthma or allergic rhinitis and thus, did not fulfill our inclusion criteria (48–53).
We finally included 19 case–control studies and 13 cohort studies carried out in 19 countries. One case–control study provided data on only allergic rhinitis and not on asthma (47).
Asthma and exposure to cats
The 18 case–control studies (including 5887 cases and 4222 controls) and 13 cohort studies were published between 1985 and 2006 (Tables 1 and 2). Fourteen case–control studies used population controls and four used hospital controls.
Table 1. Relative risks of asthma according to exposure to furry pets in case–control studies
Age, sex, family atopy, number of sibs, smoking, dust exposure, wheeze
The fixed effect pooled relative risk for cat exposure was 3.52 (95% CI: 3.46–3.59) but the large amount of heterogeneity between studies explains that the random effects estimate was of opposite sign (RR = 0.87, 95% CI, 0.52–1.44). The amount of heterogeneity decreased considerably when we restricted our analysis to cohort studies, group in which exposure to cats shows a protective effect against subsequent asthma (RR = 0.72, 95% CI: 0.55–0.93). On the contrary, heterogeneity did not subside in the case–control group of studies, including after stratification by type of controls (hospital vs population) and the random effects pooled relative risk is inconclusive: 1.23 with 95% CI: 0.65–2.53.
Restricting the analysis to those 12 studies that were carried out on children only did not alter the results (21, 22, 24, 26–28, 32, 34, 39, 42, 44, 46). We did not observe any meaningful changes in the results when we carried out the quality analysis and stratified the results consequently. Globally, there was no heterogeneity between studies of high quality compared with between low-quality studies.
The funnel plot was distorted by one influential point: the study by Al-Mousawi et al. (32) which showed an exceptionally high RR with a narrow confidence interval. When this study was excluded, there was no evidence of publication bias (P-value of Egger’s symmetry test: 0.42).
Interestingly, when we excluded the studies carried out among Arab and African populations that provided data on exposure to cats (26, 32), among which pet keeping at home is very infrequent, the fixed-effects RR turned protective and similar to that observed for cohort studies (RR = 0.88, 95%CI 0.79–0.98). These two studies yielded a pooled relative risk of 3.67 with 95% CI: 3.60 to 3.74.
Asthma and exposure to dogs
Nine studies were available on dog ownership and asthma (21, 22, 28, 30, 39, 41, 42, 44, 45), four case–control and five cohort studies. There was no heterogeneity between study-specific effects, and thus, fixed effects pooled estimates were similar if not identical to random effects estimates.
Globally, dog exposure increases slightly the risk of asthma with RR’s between 1.10 and 1.20 and confidence intervals that reach statistical significance (Table 3). We did not observe any substantial change in the results when we excluded the only available study carried out in a population with a low pet-keeping frequency (30), nor when we restricted our analysis to pediatric studies. The funnel plot (Figure 1) and Egger’s regression test (P-value: 0.27) did not provide any evidence of publication bias.
Table 3. Pooled relative risks (RR) and 95% confidence intervals (CI) of asthma and pet exposure
Number of studies
RR (95% CI) fixed effects
RR (95% CI) Random effects
Q test P-value
*Birth cohorts and adult population excluded.
†Proportion of total variance as a result of between-study variance
We retrieved 18 articles that presented data on exposure to any furry pet (generally, cats and dogs) and asthma (16, 18–23, 28–31, 33, 35–37, 41, 43, 44), which included 12 case–control and six cohort studies. Globally, there was a considerable amount of heterogeneity between studies. This heterogeneity decreased dramatically among cohort studies when we stratified by design and when we excluded studies carried out among populations with low pet-keeping frequency (Ri = 0.53 and 0.69, respectively) (20, 30). The results of this analysis were inconclusive. The pooled RR among cohort studies is compatible with an absence of effect (RR = 0.88, 95% CI: 0.64–1.19) in a similar fashion to the pooled result of studies in which populations with low pet-keeping frequency were excluded (RR: 1.09, 95% CI: 0.91–1.32). On the contrary, when all studies were taken into account, or when the analysis was limited to case–control studies or pediatric studies, there is an indication of an increase in the risk of asthma, although with a large amount of heterogeneity (Table 3). The results did not display any meaningful variation when we stratified the analysis by quality items nor when excluded the large case–control study by Anyo et al.(29). Further, we did not find any evidence of publication bias (P-value of Egger’s symmetry test: 0.13).
The two studies carried out among low pet-keeping populations (20, 30) yielded a pooled relative risk of 26.89 with 95% CI: 1.29 to 561.37.
Only five studies that measured risk of rhinitis were available (Table 4). The two studies that measured the risk among subjects exposed to cats yielded a pooled relative risk of 0.69 (95%CI: 0.41–1.16). Among subjects exposed to any furry pet (four studies), the RR was 0.79 (95% CI: 0.68–0.93) with no heterogeneity.
Table 4. Relative risks of rhinitis according to exposure to furry pets
*This study was carried out on a population of children, and used 124 cases and 245 controls. The analysis was adjusted for age, sex, family history of atopy and siblings. Details on the rest of the studies are available in Tables 1 and 2.
Our results indicate that, globally, exposure to cats exerts a preventive effect against the development of asthma. The results may be deceptive at first sight if one focuses on fixed effects results, as these results are in apparent opposition to those of the random effects. In fact, heterogeneity between effects exists essentially among case–control studies but to a much lesser extent among cohort studies. As case–control studies are much more prone to bias than cohort studies, results referring to the latter reflect probably better the real effect.
On the contrary to cat ownership, exposure to dogs seems to increase the risk of asthma. Exposure to undetermined furry pets is inconclusive, with relative risk estimates slightly higher than one, but with large heterogeneity, except among cohort studies. As this group includes mainly studies with exposure to both cats and dogs, the effect observed is probably the result of a mixture of the effects exerted by each pet category.
Globally, it seems that frequency of pet keeping is an important effect modifier. Exposure to pets is a strong risk factor for asthma in those countries where pets are not frequently kept at home for cultural or religious reasons. It has been hypothesized that high exposure to cat allergen may induce tolerance in individuals (54). It is then plausible that a low frequency of pet-keeping leads to an absence of this tolerance. However, these results should be interpreted with caution because of the small number of studies carried out in countries with a low prevalence of pet ownership. In addition, exposure to pets at home may be different from exposure outside of the home. Furthermore, it is difficult to exclude the possibility of the existence of a genetic or other environmental risk factor among these populations. Family inheritance may also be a confounder of the relation between pets and asthma. Only a few studies were adjusted for this factor in the analysis.
The differences in the effects between dog and cat exposure may have several explanations. It is likely that sensitization to dog allergens requires a shorter exposure time and lower dose than sensitization to cats (55). Moreover, cat allergens are ubiquitous and may be found in environments where cats are absent (56, 57), as they can be carried by clothes (58). Another reason is that pet avoidance subsequent to asthma symptoms was observed for cats during childhood and adulthood but not for dogs (59). Further, Fel d 1 the major cat allergen belongs to the group of secretoglobins, whereas Can f 1, the major dog allergen, belongs to the family of lipocalins that presents different biochemical and pathogenic characteristics (60, 61). Last, some pets produce both allergen and endotoxin or bacterial exposures (62), while others may not.
As for allergic rhinitis, although the limited number of studies included in our meta-analysis does not allow for firm conclusions, one can observe a protective pattern.
Our meta-analysis may be subject to several limitations. Recall bias may have distorted the results in case–control studies as shown by the large difference in pooled results between case–control and cohort studies. In the large majority of the studies, measurement of exposure to pets throughout studies was made by questionnaire, which allows for a qualitative assessment of the exposure only. However, this shortcoming was not confirmed by our quality analysis that did not find evidence of any large difference in effects between groups of studies with different exposure assessments.
Further, residual confounding (confounding from unknown variables that is not eliminated by adjustment), as in any meta-analysis of observational studies, may have introduced considerable bias. Genetic factors may be included among these variables. This meta-analysis is limited to those studies with a diagnosis of asthma and allergic rhinitis. The fact that we excluded studies the outcome of which was wheezing or unspecified atopy was likely to influence our results. However, we consider that asthma, wheezing and atopy are distinct entities and that they should not be mixed in the same meta-analysis.
In conclusion, our results show that exposure to cats may have a protective role in asthma occurrence. Exposure to dogs increases the risk slightly, while ownership of undetermined furry pets provides inconclusive results so far. Further studies in which exposure is measured in a more detailed fashion are needed.
Dr Takkouche’s work in meta-analysis is funded by “CIBER en Epidemiología y Salud Pública”(CIBER-ESP), Spain. ME is funded by a Canadian Institutes of Health Research postdoctoral fellowship award. Dr. Fitzgerald is a Michael Smith Foundation for Health Research Distinguished Scholar and a BC Lung/Canadian Institute of Health Research Scientist. There is no specific funding for this study.