of the above. Airway inflammation, a cardinal feature of asthma, cannot practically be determined in population studies. Differences in diagnostic criteria may explain variable rates of disease. Another source of variation is diagnostic overlap with conditions such as bronchitis (Gregg, 1983; Lebowitz et al., 1990) which can occur in children or adults. Forty to 50 percent of children with asthma also have chronic bronchitis. In adults, 50 to 80 percent of smokers have chronic bronchitis, and only a minority of these have asthma (Snider, 1988).
Research indicates that of those persons with asthma, 90 percent of children, 70 percent of young adults, and 50 percent of older adults also have allergy (Lehrer et al., 1986; Peat et al., 1987; Platts-Mills et al., 1992). These data strongly suggest a role for allergy in the pathogenesis of asthma, particularly in childhood.
Public health officials are often asked to estimate the fraction of asthma in the population that could be attributed to allergy or to exposure to a specific allergen. Several conceptual issues must be understood before making these calculations. Last (1986) states that attributable fractions in the population cannot be estimated from prevalence data but must be calculated from incidence data obtained in cohort studies. This is particularly true for conditions such as cancer where the case-fatality rate is high, and the prevalence in a population grossly underestimates the cumulative incidence rate. It is also true for brief, self-limited illnesses such as influenza where the point prevalence would underestimate cumulative incidence. By contrast, asthma is a chronic disease with a low case-fatality rate, and therefore the prevalence and the cumulative incidence rates are similar. The prevalence of asthma is a reasonable measure of the burden of the disease in a society. Estimates of the fraction of asthma attributable to allergy or to a specific allergen may, therefore, utilize cumulative incidence or prevalence data.
Last (1986) also states that attributable fractions should not be obtained from the distribution of risk ratios obtained in case control studies. Approaches have been developed that use data from case control studies (Schlesselman, 1982), but these should be verified with population-based data. The essential limitation in using case control data to estimate population attributable risk is that it is difficult to know whether the control group is representative of the general population. This concern lessens when the outcome of interest is similar in multiple control groups.
Another problem with calculating attributable fractions is the issue of competing risks. Individuals may be exposed simultaneously to several risk factors, such as house dust mite, cockroach, and cat allergens. The assumption is generally made that exposure to competing risks is uniform in the