In our large representative sample of the US population, over 50% of participants reported a history of allergic conditions and 44% were specific IgE-positive. We had information on 5 allergic conditions and 19 specific IgEs to food and inhalant allergens. Unlike the case in many clinical studies, we had an unselected sample for allergy history, and because the study was population-based, we had the ability to calculate predictive values. Individual allergic conditions had high specificity for atopy but low sensitivity. Measures that combined all allergic conditions had higher sensitivity but reduced specificity. Most of this discordance appeared to be driven by the 40% of persons who were specific IgE-positive but reported no allergic conditions.
The prevalence of atopy as defined by specific IgE varies by population. In the European Community Respiratory Health Study, a study of adults aged 20–44 years conducted in the 1980s, the prevalence of at least 1 positive specific IgE ranged from 16% at a center in Spain to 45% at a center in New Zealand (4
). The prevalence of specific IgE positivity at a given center was unrelated to the geometric mean total IgE level among all participants at that center. Our prevalence of specific IgE positivity of 44% is comparable to the high end of specific IgE positivity in the European Community Respiratory Health Study (4
In adult populations, specific IgE positivity appears to be relatively stable over time (12
), although a person's IgE status may change over time (13
). Among 695 older adults followed from age 40 years to age 60 years, 41% lost IgE sensitization, while 7% were newly sensitized (13
). Persons who were most likely to remain or become sensitized were those with a history of allergic illness (13
), which suggests that questionnaires may be able to capture a more consistent chronic picture.
The prevalence of atopy and the predictive nature of allergy questions have been assessed in both clinical and population-based samples. Atopy has been characterized according to skin-prick testing, total IgE, or specific IgE. Our study was among the largest carried out to date, and it included persons from multiple ethnic groups who were unselected for allergy and ranged in age from less than 1 year to more than 70 years. While we observed differences in allergic conditions and atopy by age, ethnic group, and gender, we did not observe differences in the predictive ability of the questionnaire to assess atopic status.
The results of our analysis are similar to those of other studies conducted in different populations using different measures of atopy, even though the prevalence and definition of atopy varied by population. In a study of the reliability of allergy questions among 118 health-care workers in Texas, responses were internally consistent, and the test-retest reliability for allergy-related questions was high (14
). The investigators observed that the questions on allergy had a wide range of sensitivity (19%–74%) but high specificity in comparison with specific IgE results (14
). Among 2,120 children in the International Study of Asthma and Allergies in Childhood (ISAAC), skin-test positivity was strongly associated with hay fever, and this hay-fever question was very specific but not very sensitive; the positive predictive value of hay fever was 70% (15
). In a study of 290 Finnish young adults—half with asthma and half without—hay fever or nasal allergy diagnosed by a doctor had a positive predictive value of 76%, with a specificity of 93% and a sensitivity of 52% for atopy defined as either specific IgE positivity or a positive skin-prick test (16
). While this sample was enriched for asthmatics, the positive predictive value and specificity were similar to what we observed; the sensitivity was much higher, however, perhaps because of the enrichment of the sample or the use of 2 measures of atopy. The prevalence of atopy in the Finnish nonasthmatic group was 48% (16
), similar to our estimate of 44% for the US population as a whole.
Designing questionnaires for large-scale studies is challenging in that there is always tension between questionnaire length and the amount of information collected to reduce misclassification of the outcomes. Question wording is critical and will influence the predictive value. In NHANES 2005–2006, the survey questions were standard questions regarding allergic conditions. Self-report of hay fever, allergy, and eczema was based on a doctor's diagnosis of these conditions. While people may accurately report what their health professional told them, the diagnosis may not meet standard diagnostic criteria. In a study of 152 children seen by board-certified allergists, allergy was overdiagnosed by approximately 20% in comparison with skin-prick and IgE test results (17
In the NHANES questionnaire, current symptoms could only be reported if a respondent had received a health professional's diagnosis; thus, persons with allergic symptoms who managed their symptoms with over-the-counter medication would not have been captured, and these undiagnosed but symptomatic persons would have led to an underestimate of affected persons and thus misclassification of allergic status. For rhinitis, respondents were queried about having a sneezy, runny, or blocked nose when they did not have a cold; this broad wording would have identified people who did not seek medical care, but may also have included people who did not meet the clinical definition of rhinitis. Since rhinitis can have both atopic and nonatopic forms, with approximately 70% of rhinitis in adults being atopic (18
), inclusion of rhinitis as an allergic condition will result in some misclassification of atopic status. Consistent with that, current rhinitis had some of the lowest positive predictive values (50%); only adult-onset allergy had a lower positive predictive value. Additionally, there is some possibility that people may have allergic rhinitis without being serum IgE-positive; in some cases, specific IgE is present only in the nasal fluids and not the serum (19
The survey queried about ever being diagnosed with specific diseases, regardless of whether the person was currently symptomatic. Over time, people who are no longer bothered by allergic conditions may tend to forget a previous diagnosis, which would lead to a lack of predictive value of the questions if the sensitization remained. With regard to allergies, respondents were only asked whether they had allergy, not about agents that triggered their allergies. Other investigators have asked specific questions about allergic agents to help refine their questionnaire and its predictiveness (20
). Questionnaires have been shown to be predictive of change in sensitization over time in occupational settings (22
); these questionnaires incorporate baseline characteristics and changes in allergic symptoms over time to predict sensitization. Repeated measurement of symptom information may help to improve the positive predictive values of questionnaires.
Our analysis was based on NHANES 2005–2006 and was designed to reflect the US population. While the NHANES sampling strategy is designed to be nationally representative, participation in the study and the blood draw may influence the results to a limited extent. The population for this analysis comprised persons aged 1 year or more who had a blood sample analyzed for IgE. Participation in NHANES was higher among younger persons (>80% in those under age 18 years) and lower among the elderly (approximately 65% in those aged ≥60 years). Both the lowest and highest age groups were least likely to meet the definition of atopy; it is possible that persons who did not participate had a differential association between atopy and self-reported allergic disease. Persons who had rashes on both arms were excluded from the blood draw; comparisons of allergic outcome history for people with and without blood measurements did show that persons with eczema were less likely to give a blood sample but that persons with allergy or hay fever were more likely to have blood drawn. Thus, it is possible that there were some small differences between the characteristics of persons included and the whole sample; however, unless the relation between self-reported allergic outcomes and specific IgE prevalence was very different among the people without a blood draw, this difference is unlikely to have influenced the results or to explain the lack of concordance between these measures.
We included all persons who had specific IgE measured, even if they did not have a complete panel of allergens. Of those with incomplete panels (n = 213), 35% were considered atopic as compared with 44% of the entire sample. While it is possible that we underestimated the prevalence of atopy by including these persons in the analysis, it is highly unlikely, because few people were sensitized only to the missing allergens. Sensitization to only these allergens was rare (1.5%), so at most, only 1–2 more of the 136 participants classified as nonatopic among those with an incomplete panel would be considered atopic. Therefore, while differences existed between persons with and without IgE information, these differences were unlikely to have influenced the results in any meaningful way.
While differences in questionnaire phrasing and the finite number of specific IgEs tested may contribute to the observed discordance between questionnaire measures, the high percentages of asymptomatic persons (37%) and of allergic persons without atopy (23%) are the primary challenges to the lack of predictive value of the allergic conditions. Among 2,612 children in a birth cohort study, 17% of those who were specific IgE-positive at age 4 years did not have any allergic disease (23
); Wickman et al. (23
) estimated that it would take 7 or more positive specific IgEs or a sum of specific IgEs greater than 400 kU/L (much greater than the 0.35 kU/L needed for detection) to predict allergic disease at 90%. In our sample, the median sum of specific IgEs was 8 kU/L (interquartile range, 2–32) for persons with symptoms and 4 kU/L (interquartile range, 1–17) for persons without symptoms, suggesting that few would meet the cutpoint of 400 kU/L for allergic disease prediction. In a population-based study of more than 8,000 Swiss adults, 14% of those reporting hay fever were not IgE- or skin-prick test-positive (21
), which is comparable to our findings. The factors that contribute to being asymptomatic have been reviewed by Bousquet et al. (24
). Asymptomatic persons tend to be younger and monosensitized and are less likely to have a family history of atopy (24
). In our nationally representative sample, asymptomatic persons were more likely to be younger and monosensitized (13%) than symptomatic persons (9%). However, over 75% of both asymptomatic and symptomatic participants were sensitized to 2 or more antigens, suggesting that monosensitization is rare in the general population.
The lack of agreement is not unique to questionnaire measures. Measures of atopy themselves are discordant, with the 2 most common measures, specific IgE positivity and skin-prick test positivity, having varying agreement across populations (2
). Among children aged 8–12 years at the 18 sites of the multinational ISAAC study, 8.3%–71% were IgE-positive but skin-prick test-negative. Among 8,340 adult participants in the Swiss Study on Air Pollution and Lung Diseases in Adults, 9.3% of the serum IgE-positive persons were not skin-prick test-positive (21
). Even with these discordances, these measures, along with questionnaires, are useful in the evaluation of clinical allergy. Given these inconsistencies among the measures of atopy and allergy, it is clear that there are no gold standards for atopy and allergy diagnosis and that clinical judgment remains a piece of the puzzle.
Questionnaires remain the best and sometimes the only way to collect medical information, particularly historical information, in a large study sample. Increasing the precision of questionnaires will improve their utility; however, not all aspects of the allergy-atopy spectrum are well characterized using questionnaires. Researchers need to be aware of the limitations of all potential atopy measures, given the lack of concordance in many large population-based studies.