While the public health significance of asthma is well recognized,⁶ economic and social burden of allergic diseases is not limited to asthma. Many allergic conditions carry considerable direct and indirect cost implications and substantially affect quality of life.^{7, 8} In the United States, for example, health care and treatment expenditures for allergic rhinitis have almost doubled from the year 2000 ($6.1 billion) to the year 2005 ($11.2 billion).⁹ It has been estimated that allergic rhinitis results in 3.5 million lost work days and 2 million lost school days annually.⁷ Despite the increasing morbidity of allergic conditions, allergy-related outcomes other than asthma have been less well characterized in population-based studies.

Current rhinitis symptoms, including sneezing and/or nasal symptoms in the absence of a cold or the flu, were reported by approximately one-third of the population; 24.2% reported seasonal symptoms in the past 12 months, whereas 10.0% had symptoms year round (perennial symptoms). As expected, the prevalence of sneezing and nasal symptoms was significantly higher among those who reported current hay fever or who had current allergies than among those who did not report these outcomes (84.6% vs. 30.5%; 76.8% vs. 20.9%, respectively; P < .01 for difference). Almost two-thirds (61.2%) of those who reported current rhinitis symptoms had diagnosed allergies. Among hay fever sufferers, rhinitis symptoms were mainly seasonal in nature (i.e., occurring 1–3 seasons); 60.6% of those who had current hay fever reported seasonal symptoms of rhinitis.

Almost half of the total population (43.7%) had detectable levels of specific IgE to at least one of the tested allergens (Table I). The prevalence of allergic sensitization was significantly higher for those who reported current hay fever or current allergies (74.4% for current hay fever, 57.9% for current allergies). Correspondingly, at least half (52.7%) of the individuals who reported current rhinitis symptoms had detectable levels of specific IgEs to at least one allergen. The majority of the individuals who had detectable levels of specific IgEs were poly-sensitized; less than one-fourth of the sensitized subjects (21.6%) were mono-sensitized (Table I).

Figures 1 and ​and22 illustrate the prevalence of sensitization in the US population among those with and without current hay fever and current allergies. Hay fever and allergy sufferers were most often sensitized to grass and other plant-related allergens. For example, almost half (44.2%) of those who had current hay fever were sensitized to rye grass (Figure 1), and 60.1% had detectable levels of allergen-specific IgEs at least for one of the plant allergens (Figure 2). Sensitization to rodent allergens was least common; 3.1% of those who reported allergy-related outcomes were sensitized to rodent allergens. While the prevalence of sensitization tended to be higher in children, especially for pet- and mold-specific IgEs, the sensitization patterns between the age groups did not differ appreciably (Figure E1 in Online Repository).

To further characterize the associations between the primary outcomes and specific IgEs, we modeled specific IgE concentrations as continuous variables. Figures 3 and ​and44 show unadjusted and adjusted odds ratios for the associations between the allergy-related outcomes and allergen-specific IgEs and IgE clusters. The adjusted odds ratios for the outcomes correspond to a 10-fold increase in specific IgE concentrations. After adjusting for potential confounders (socio-demographic variables and other IgE clusters), increases in plant-, pet-, and mold-specific IgEs were most consistently associated with allergy-related outcomes (Figure 4). We did not find strong evidence for effect modification by age. The patterns of the associations were similar among children and adults, except for food-specific IgEs (data not shown). Detailed information on patterns of food allergy is published elsewhere.¹⁴ Current hay fever, as expected, was most strongly associated with increased levels of plant-specific IgEs (adjusted OR = 4.75 95% CI: 3.83–5.88).

Sensitization patterns for current rhinitis symptoms (Figure E2) were comparable to those with current hay fever and current allergies. The odds of having perennial symptoms of rhinitis increased most with elevated levels of pet-specific IgEs (adjusted OR = 2.56, 95% CI: 1.73–3.79). Seasonal symptoms were associated with elevated levels of plant-, pet- and mold-specific IgEs, most strongly with plant-specific IgEs (adjusted OR = 2.35, 95% CI: 1.71–3.22).

Regardless whether allergen-specific IgE was modeled as a continuous or as a dichotomous (data not shown) variable, the results did not change appreciably; presence of plant-, pet-, mold-specific IgEs remained most consistently associated with the studied outcomes.

Total IgE levels varied significantly across socio-demographic variables (Table I). A detailed description of the distribution of total IgE in the NHANES 2005–2006 can be found elsewhere.¹⁵ Table II shows unadjusted and adjusted effect estimates for the associations between the log-transformed total IgE levels and the primary outcomes, current hay fever and current allergies. Increased levels of total IgE were consistently associated with both primary outcomes. Irrespective of whether total IgE was modeled as a continuous or a dichotomized variable (cut offs: median and 100kU/L), the modeling results remained statistically significant (Table II, Table E1 in the Online Repository). The strongest associations were seen for current hay fever (adjusted OR for a ten-fold increase in total IgE = 1.86, 95% CI: 1.44–2.41). In Table II, we also present stratum-specific odds ratios for different age, gender, and race/ethnicity groups. The association between total IgE and current allergies was stronger among children than among adults, suggesting effect modification by age for current allergies. This may reflect differences in the prevalence of co-morbid conditions between age groups. Prevalence of asthma, for example, was significantly higher among children than adults.¹⁵

We clarified the role of specific IgEs in relation to total IgE by modeling them together (Online Repository Figures E3 and E4). Although the patterns of association between the specific IgEs and IgE clusters and the primary allergic outcomes remained similar after adjusting for total IgE levels, the odds ratios were apt to attenuate (compare Figure 3 to Figure E3, Figure 4 to Figure E4). Increases in total IgE had a small but independent effect on the primary allergic outcomes after adjusting for most of the specific IgEs and IgE clusters. However, the associations between current hay fever and total IgE did not remain statistically significant when adjusting for plant-specific IgEs, especially for rye grass-specific IgE (Figures E3 and E4).

Allergies affect a substantial proportion of the US population; more than one-third of the NHANES 2005–2006 participants reported diagnosed allergies. Of those who reported diagnosis of any type of allergic condition, over 50% had symptoms in the past 12 months. Although the prevalence of allergic conditions appears to be similar to that reported in other national surveys,^2–5 the prevalence rates may not be directly comparable. Case definitions of allergy-related outcomes tend to vary from one study to another, a fact which complicates comparisons of estimated prevalence rates.^{20, 21} Nonetheless, prevalence patterns across socio-demographic factors were largely similar to those reported in the literature.⁸

In the NHANES 2005–2006, prevalence of self-reported nasal symptoms was comparable to that reported in other national studies of rhinitis (30.2–31.5%).^{20, 22} Although the majority of the NHANES participants who reported current rhinitis symptoms also reported diagnosed allergies, a substantial proportion of the subjects, at least one-third, lacked any diagnosis of allergy. Several studies have shown that a large number of people who have rhinitis-related symptoms remain undiagnosed.^{20, 23, 24} On the other hand, it is well known that rhinitis symptoms can also be non-allergic in nature.²¹

Although allergic rhinitis is considered one of the most common chronic diseases in the US, the prevalence estimates have been highly variable, ranging from 9% to 42%.^{21, 25, 26} In the NHANES 2005–2006, more than half of the subjects with symptoms of rhinitis had detectable levels of specific IgE to at least one allergen, suggesting that approximately 18% of the US population may manifest symptoms of allergic rhinitis. Nathan and coworkers have reported similar rates (14.2%–22.0%) for allergic rhinitis, demonstrating increasing trends in the prevalence over the past decades.^{20, 22} A review focusing on diagnosis of allergy has recently suggested that about 50% respiratory symptoms may be allergic in origin.²⁷

Consistent with previous NHANES findings, sensitization to grass, dust mite, and ragweed allergens was most common in the US population.^{1, 28} Among those who reported any allergy-related symptoms in the past 12 months, sensitization rates were variable, ranging from 24.4% to 44.2% for grass-specific allergens; from 22.3% to 24.9% for dust mite allergens; and from 23.0% to 32.8% for ragweed allergen. It was not unexpected that a larger number of specific IgEs were associated with current allergies than with current hay fever; reported allergy symptoms may relate to a variety of allergic conditions.

Hay fever symptoms were predominantly associated with the presence of plant-specific IgEs, in agreement with published data.^{28, 29} Although hay fever is commonly referred to as seasonal allergic rhinitis, it can also be triggered by perennial allergens. In the NHANES 2005–2006, sensitization to pets was consistently associated with current hay fever and perennial symptoms of rhinitis. This is not an unforeseen finding because dog and cat allergens are not only found in virtually all US homes, but are also most often found in elevated levels.^{30, 31} Pet allergens tend to accumulate on many interior materials within the home, including carpeting, upholstery, and bedding, which can serve as continuous reservoirs for those allergens. Because of the aerodynamic characteristics of cat and dog allergen carrying particles, both allergens become aerosolized easily and remain airborne for long periods of time.^{32, 33}

While sensitization to molds has been associated with many atopic outcomes, studies suggest that sensitization is less frequent to molds than to pollen, animal or dust mite allergens.³⁴ Nonetheless, sensitization to molds is an important risk factor for manifestations of atopy. In NHANES 2005–2006, the magnitude of the effect of sensitization on reported symptoms was comparable between molds and pets.

In contrast to previous findings,²⁸ sensitization to dust mites did not remain significantly associated with any of the allergic outcomes in completely adjusted models. This may reflect differences in age distributions and/or analysis methods. The previous NHANES data covers a narrower age range than the NHANES 2005–2006, and because of smaller number of tested allergens, clustering effects of positive test responses have not been accounted for in the previous surveys. Moreover, the allergenic composition of the house dust extract that was used for skin prick testing in NHANES II is unknown, which complicates interpretation of the results. It is also likely that patients with allergy diagnosis, particularly those who see an allergist, have greater awareness and compliance with environmental control measures for dust mites than those who have not consulted a specialist.³⁵

The level of total IgE is influenced by genetic and environmental factors.²⁷ Because levels of total IgE may overlap between non-atopic and atopic individuals, as well as between different allergic diseases, the level of total IgE is generally considered less clinically relevant than specific IgE data.^{27, 36} In the NHANES 2005–2006, increase in total IgE appeared to have a small, but independent effect on allergy-related outcomes. After controlling for total IgE, the precision of the effect estimates in the models improved, though the patterns of the associations remained similar. In studies where accounting for clustering of sIgEs is not feasible (e.g., studies with limited number of allergens in the test panel), adjusting for total IgE may help to control confounding by other sIgEs. In fact, the effect of total IgE was less pronounced for grass- and ragweed-specific IgEs among individuals with current hay fever. Rye grass and ragweed were the most common sensitizers among mono-sensitized hay fever sufferers (data not shown), which may partially explain this finding. Furthermore, total IgE levels were significantly lower among individuals who did not report asthma than among asthmatic individuals (Figure E5), in agreement with published data.^{15, 37} Among those participants who reported allergy-related outcomes and who had asthma, a wide spectrum of allergen-specific IgEs were significantly elevated.

Previous studies have shown that not only asthma, but also other co-morbid conditions are associated with hay fever and other forms of allergic rhinitis.^{7, 8, 26, 38, 39} In the NHANES 2005–2006, both asthma and sinus infections were significantly more prevalent among those who reported current hay fever or current allergies than among individuals without allergy-related outcomes.

Because sensitization is a dynamic process and affected by age,⁴⁰ we examined total and specific IgEs in relation to allergy-related outcomes across different age groups. The association between total IgE and current allergies was stronger among children than among adults, perhaps reflecting a higher asthma prevalence in children. Although the specific IgE results did not suggest strong evidence for effect modification by age, the odds ratios for some of the specific IgE clusters tended to be higher among children than among adults. In particular, sensitization to pets and molds seemed to increase the odds of having allergy-related outcomes in children. Sensitization to food allergens was also more common in children than in adults, as previously shown.¹⁴

The cross-sectional nature of the NHANES 2005–2006 is an important limitation of the study. Because temporal relationships are often difficult to determine in cross-sectional studies, we primarily focused on outcomes that required the presence of symptoms in the past 12 months. We recognize that misclassification of outcomes or exposures can introduce bias. In the NHANES, outcome measures were based on questionnaire responses alone, a practice which is common in most large-scale surveys. Allergic symptoms can be difficult to distinguish from symptoms which are non-allergic in origin, because many allergic and non-allergic conditions have similar symptoms.²⁷ Some conditions can even have both non-allergic and allergic etiologies. For example, a large proportion of rhinitis suffers may have mixed rhinitis, a combination of allergic and non-allergic rhinitis.^{8, 41} A considerable number of individuals with allergies may also remain undiagnosed, because many people tend to use over-the-counter products, instead of seeking medical attention.²⁰ However, we included an allergy-related outcome that was not based on physician’s diagnosis (i.e., symptoms of current rhinitis) in our analysis. Although serological measurements provide objective evidence of atopy, the diagnosis of clinically relevant allergy also depends on symptom history. It is well known that the presence of allergen-specific antibodies does not necessarily mean that a person has clinically relevant symptoms when exposed to an allergen.^{19, 36} Because patterns of sensitization tend to vary with climatic and geographic factors, it is possible that the test panel may not have covered all relevant allergens. For example, a recently published list of the major clinically important outdoor aeroallergens in North America includes a number of allergens that were not included in the test panel.²³ On the other hand, in national population-based studies, the test panel cannot be optimized for any specific region. Despite the limitations of the study, we believe that serious differential misclassification is unlikely; both outcomes and serum IgE levels were assessed independently, without prior knowledge of the atopic status.

One of the major strengths of the study is that the sample for the survey was selected to represent the entire US population. Indeed, the NHANES 2005–2006 provides the largest nationally representative dataset of serum IgE levels that has ever been collected on the US population. Although skin testing is often used to determine IgE-mediated sensitivity, ⁸ serum-specific IgE immunoassays enabled the expansion of our test panel to include a larger number of allergens. Similar sensitivities have been reported when serum-specific IgEs and skin-prick tests have been compared with respect to the presence of symptoms, although the performance characteristics of these two immunoassays are known to vary.^{8, 42} None of the previous population-based studies in the US have examined both qualitative and quantitative relationships between sensitization and allergy-related symptoms. The NHANES 2005–2006 was also the first study to account for clustering of specific IgEs. Studies have shown that many allergens share structural similarities and can be cross-reactive.^{43, 44} Because IgE-mediated allergy tends to occur to clusters of allergens,^{45, 46} problems of collinearity may arise during the statistical analysis. To discover patterns and relationships in the allergen-specific data, we used cluster analysis, a widely-used method to analyze correlated data. As we have previously reported in abstract form,¹³ the NHANES 2005–2006 specific IgE data group into seven clusters, which not only have optimal statistical properties, but also are biologically relevant.