Food allergy is an independent risk factor for decreased lung function in children with asthma

Introduction

Individuals with asthma and food allergy have increased asthma-related morbidity – daily cough/wheeze, lifetime hospitalizations, unplanned healthcare utilization – and mortality compared to individuals without food allergies.^1–4 In patients with both diseases, this higher morbidity and mortality may be related to direct and/or indirect effects of food allergy on a child or young adult’s bronchial hyperreactivity and lung function.^3,5–12 Indeed, the influence of food allergy on asthma appears to present early in life.⁷ Infants with sensitization to either cow’s milk or hen’s egg have significantly lower baseline forced expiratory flows and increased airway hyperreactivity before any clinical onset of wheeze, when compared with infants who had no sensitization.⁷

Yet, it is unclear if this association is due to effects on lung growth and development, is due to presumed airway inflammation related to food allergy,^11,12 or is simply a marker for worse asthma in more atopic individuals. If this association is because of airway inflammation or worse asthma, then we would not expect similar effects in individuals without asthma. Currently, any associations between food allergy and lung function in children and young adults with and without asthma remain unclear.

Therefore, we sought to determine if an association between food allergy and a child or young adult’s lung function exists by examining a larger cohort of participants with and without asthma from a family-based food allergy study. We hypothesized that regardless of asthma status, participants with food allergy would demonstrate decreased pulmonary function. We also hypothesized that among participants with asthma, food allergy would be associated with a further decrease in pulmonary function beyond that attributable to asthma status.

Methods:

We recruited participants from Chicago area general medical and allergy specialty clinics, community support groups, and media advertisements between August 2005 and June 2011. We recruited children and young adults, aged 0–21 years, with and without food allergy. After recruitment, subjects and/or subject representatives completed a standardized questionnaire interview by trained research staff to obtain information on each family member’s home environment, diet, lifestyle, history of food allergy (FA) and other atopic diseases. Participants then underwent allergy skin prick testing (SPT) to 9 foods and 6 aeroallergens (described below). We collected blood samples for food specific and total IgE measurements. For each participant, we obtained clinical data including type of symptoms on food exposure and timing of symptom onset through questionnaires and medical record review. All participating institutions’ Institutional Review Boards (IRBs) approved the study protocol prior to the trial onset. All participants or, if under 18 years old, parents/guardians provided verbal and written consent prior to participation in the study.

Food allergy definition

Participants were defined as having food allergy by physician diagnosis based on the following criteria: typical anaphylaxis symptoms (i.e. respiratory, cardiovascular, cutaneous, or gastrointestinal symptoms)¹³ within no more than 2 hours after food ingestion, as well as evidence of food specific IgE via skin prick test or Immunocap® (Phadia US Inc., Portage, MI, USA).¹⁴ We classified subjects by number of food allergies: no food allergy (Group 1), 1 food allergy (Group 2), or ≥2 food allergies (Group 3).

Asthma definition

We documented participants as having asthma if they reported a physician diagnosis. We classified participants as never having asthma, currently having asthma, or having outgrown their disease. We also stratified participants into two groups as either ever having asthma (current or outgrown) or never having asthma.

Skin prick test (SPT)

We performed SPT using Multi-Test II device (Lincoln Diagnostics) to 9 food allergens: cow milk, egg white, soybean, wheat, peanut, English walnut, sesame seed, fish mix (cod, flounder, halibut, mackerel, tuna), and shellfish mix (clam, crab, oyster, scallops, shrimp). We also performed SPT to 5 aeroallergens: Alternaria alternata, house dust mite mix (equal parts mixture of D. farinae and D. pteronyssinus), cat hair, dog epithelia, cockroach mix (American and German cockroach). All skin prick testing included negative (50% glycerinated saline) and positive (histamine, 1.0 mg/mL) controls. We obtained extracts and controls from Greer Laboratories (Lenoir, NC, USA). We measured SPT results 15 minutes after application. Positive tests included those results with a mean wheal diameter (MWD) of at least 3 mm or a MWD larger than the saline control. We excluded data if the saline control was ≥ 3 millimeters (mm), the histamine control was < 3 mm or if the difference of histamine minus saline was < 3 mm.

Specific IgE measurement

The Clinical Immunology Laboratory at Lurie Children’s, a CLIA-certified laboratory, measured specific IgE (sIgE) for nine food allergens (egg white, sesame, peanut, soy, milk, shrimp, walnut, cod fish and wheat) and six aeroallergens (Dermatophagoides pteronyssinus and Dermatophagoides farinae, cat dander, dog dander, Alternaria alternata, and German cockroach - Blattella germanica). The reported range for specific IgE was from 0.1 to 100 kU_A/L.

Pulmonary function measures

At site visits for individuals with and without asthma had the following lung function parameters measured: forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), the FEV1/FVC ratio, and Forced Expiratory Flow 25–75 %-predicted (FEF_25–75). We used FEF_25–75 %-predicted as the primary outcome measure given that small airway obstruction may be more important than decreases in FEV1 in asthmatic participants, particularly for long-term asthma prognosis.^15–18 In addition, %-predicted FEV1 may not accurately correlate with asthma severity in participants.^19,20 Secondary measures included %-predicted values of FEV₁, FVC, and FEV₁/FVC. We obtained lung function testing using a KoKo spirometer (nSpire Health, Longmont, CO) per American Thoracic Society/European Respiratory Society guidelines. Subjects required a minimum of three acceptable maneuvers to be included within analysis. We used the National Health and Nutrition Examination Survey III reference equations²¹ to determine percent predicted values for spirometry based on age, standing height (via calibrated stadiometer), sex, and ethnicity.

Statistical analysis

Demographic information included age in years, sex, maternal race, household income, history of hospitalization for RSV at any time in the past, length of asthma diagnosis, any smoke exposure, and parental asthma. We investigated descriptive statistics (mean ± SD, n (%)) for all variables by food allergy status (none, 1, or 2+). Associations between patient demographics and food allergy status were evaluated with ordinal logistic regression models. We included demographic factors associated with food allergy status (p-value < 0.2) in regression analyses. Mixed effect models assessed the effects of asthma classification (current, never, outgrown) and food allergy status on %-predicted lung function (FEV₁, FVC, FEV₁/FVC, and FEF_25–75). A random effect for family was included in models to account for correlation of subjects from the same family. All models assume complete cases analysis. Additional analyses considered the effects of food allergies in the subgroup with asthma. As a secondary analysis, we adjusted for the presence of aeroallergen sensitization. Due to a large amount of missing aeroallergen sensitization data we used multiple imputation methods to impute missing aeroallergen sensitization. Multiple are a rigorous approach to handling missing data, which allows for uncertainty imputation methods of the missing data by replacing missing values with a set of plausible values. Thus, multiple imputation methods treat imputed missing values differently than single-imputation methods. The multiple imputation methods as described by Rubin (1987) assume missing data are missing at random.²² While we can’t test this assumption explicitly, we have assessed patterns in the missing data, and have no reason to believe that missing aeroallergen sensitization is dependent on the actual missing value. We recognize a large portion of the variable were missing in this cohort (58%), and thus recommend interpreting the sensitivity analysis findings with caution. However, the primary analysis did not use imputed data and showed similar results as the sensitivity analysis. Specifically, we used five imputed datasets to estimate the average effect of food allergy status on lung function after adjustment for aeroallergen sensitization. We used SAS (version 9.4 (Cary, North Carolina, USA) for statistical analyses. All analyses assumed a two-sided type one error rate of 0.05.

Results:

Demographics

The study included 1068 participants enrolled in the Chicago Food Allergy Study. Of the 1068 participants, 403 (38%) had a diagnosis of at least one food allergy and 417 (39%) had a diagnosis of asthma at any point in their life (i.e. current or outgrown). We present demographics information for all study participants in Table 1. Compared to participants without any food allergies, participants with 1 or 2+ food allergies were more likely to be male, have a higher household income, have higher total IgE levels, have a diagnosis of asthma, have not been exposed to household tobacco smoke, and have at least one parent with asthma (Table 1). Interestingly, individuals with 2+ food allergies had a longer median asthma diagnosis duration compared with participants that had 0 or 1 food allergy (Table 1). Furthermore, most participants did not have aeroallergen sensitization data recorded (Table 1), which led to our decision to impute this missing data.

Table 1:

Association of patient demographics and Food Allergy Status

	No Food Allergy (n=665)	1 Food (n=24)	2+ Foods (n=163)	P-value1.
Age (yrs) Mean ± SD	10.0 ± 3.3	9.8 ± 3.4	9.9 ± 3.2	0.5103
Sex n(%)
Male	316 (47.5)	144 (60.0)	93 (57.1)	0.0007
Female	348 (52.3)	96 (40.0)	69 (42.3)
Missing	1 (0.2)	0 (0.0)	1 (0.6)
Maternal race n(%)
White	524 (78.8)	208 (86.7)	135 (82.8)	0.3176
Nonwhite	98 (14.7)	25 (10.4)	25 (15.3)
Unknown/Missing	43 (6.5)	7 (2.9)	3 (1.8)
Household income n(%)
<$30,000	76 (11.4)	9 (3.8)	7 (4.3)	<0.0001
$30,000–$60,000	34 (5.1)	3 (1.3)	5 (3.1)
$60,000–$100,000	216 (32.5)	63 (26.3)	34 (20.9)
>$100,000	300 (45.1)	156 (65.0)	103 (63.2)
Missing	39 (5.9)	9 (3.8)	14 (8.6)
Total IgE Mean ± SD	153.6 ± 361.6	430.0 ± 473.0	845.1 ± 1416.4	<0.0001
Asthma n(%)
Yes	174 (26.2)	122 (50.8)	121 (74.2)	<0.0001
No	489 (73.5)	117 (48.9)	42 (25.8)
Missing	2 (0.3)	1 (0.4)	0 (0.0)
RSV hospitalization n(%)
Yes	26 (3.9)	11 (4.6)	13 (8.0)	0.0556
No	633 (95.2)	225 (93.8)	147 (90.2)
Missing	6 (0.9)	4 (1.7)	3 (1.8)
Length of asthma (months) median (IQR)	0.0 (0.0, 3.0)	0.0 (0.0, 30.0)	24.0 (0.0, 48.0)	<0.0001
Smoke exposure n(%)
Yes	134 (20.2)	29 (12.1)	19 (11.7)	0.0007
No	531 (79.9)	211 (87.9)	144 (88.3)
Parental asthma n(%)
Yes	177 (26.6)	72 (30.0)	58 (35.6)	0.0275
No	488 (73.4)	168 (70.0)	105 (64.4)
Aeroallergen
Sensitization	130 (19.6)	54 (22.5)	26 (16.0)	<0.0001
Yes	213 (32.0)	14 (5.8)	12 (7.4)
No	322 (48.4)	172 (71.7)	125 (76.7)
Missing

^1.P-value from ordinal logistic regression models

SD: Standard deviation; n: number

Comparison of lung function in participants by asthma status:

Mean lung function values for each asthma classification are presented in eTable 1. We found no statistically significant differences in FEV₁ %-predicted, FVC %-predicted or FEV₁/FVC between subjects with current asthma, subjects who had out outgrown their asthma, or those who never had an asthma diagnosis (Table 2). Participants with a current diagnosis of asthma showed an average decrease of 8.11% (SE: 3.21%) in FEF_25–75 %-predicted when compared with participants who never had asthma (p-value: 0.0123, Table 2). However, pairwise comparisons found no statistically significant differences in FEF_25–75 %-predicted between those who had outgrown asthma and those who never had asthma, or between those who had outgrown asthma and those with current asthma.

Table 2:

Comparison of FEV₁ percent predicted and FEF₂₅₇₅ percent predicted between subject asthma status (current, outgrown, and never smokers).

Asthma Status	Compariso n Group	FEV1 %-predicted		FVC %-predicted		FEV1/FVC		FEF25–75 %- predicted
		Estimat e (SE)	P-value	Estimate (SE)	P-value	Estimat e (SE)	P- valu e	Estimate (SE)	P-value
Current	Never	−2.2467 (2.2599 )	0.3212	−2.6503 (2.2631)	0.2428	−2.2360 (2.2529 )	0.32 20	−8.1069 (3.2135)	0.0123
Current	Outgrown	1.5881 (4.2407 )	0.7084	−3.7467 (4.2231)	0.3759	1.5216 (4.2280 )	0.71 93	−1.7130 (6.0155)	0.7761
Never	Outgrown	3.8348 (4.3988 )	0.3842	−1.0964 (4.3896)	0.8030	3.7576 (4.3855 )	0.39 24	6.3939 (6.2459)	0.3070

Adjusted for parental asthma, prior RSV hospitalization, household income, smoke exposure, total IgE, and length of asthma diagnosis.

FEV₁: Forced expiratory volume in 1 second; FVC: Forced vital capacity; FEF_25–75: Forced expiratory flow 25% – 75%; SE: standard error

Comparison of food allergy classifications with lung function:

In the entire (unstratified) cohort, we observed no differences in any lung function parameter by food allergy status (eTable 2). This remained true after including aeroallergen sensitization in the model (eTable 3). Among participants with asthma (n=417), those with allergies to 2 or more foods had a significant decrease in FEF_25–75%-predicted compared to those with no food allergies (β: −7.46%, SE: 3.67%, p-value = 0.0416; Table 3). However, when comparing participants with only one food allergy with those who had none we observed no statistically significant difference in %-predicted FEF25–75. We also found no significant differences in %-predicted FEV₁, FVC, or FEV₁/FVC among individuals with asthma and either one or two or more food allergies when compared to those participants with no food allergies. After adjusting for aeroallergen sensitization, no statistically significant effects of food allergy status remained (eTable 4). However, most individuals (~60%) did not have aeroallergen sensitization data recorded (Table 1), which may affect this result. Despite this lack of information, we chose to explore the potential confounding of aeroallergen sensitization in subjects with multiple food allergies and asthma by using multiple imputation methods to account for the missing data. After imputing the missing data, we performed a sensitivity analysis investigating the relationship between food allergy status and %-predicted FEF_25–75. After inclusion of aeroallergen status in models for the subgroup of participants with asthma, FEF_25–75 %-predicted remained significantly lower on average in those with 2+ food allergies compared to those with none (β: −6.83%, SE: 3.26%, p-value = 0.0360).

Table 3:

Association of lung function with food allergy category in children with asthma without including aeroallergen sensitization.

Food Allergy (number)	FEV1 %-predicted		FVC %-predicted		FEV1/FVC		FEF25–75 %-predicted
	Estimate (SE)	P-value	Estimate (SE)	P-value	Estimate (SE)	P-value	Estimate (SE)	P-value
None	Ref		Ref		Ref		Ref
1	−0.4383 (2.0799)	0.8339	−0.4938 (2.0360)	0.8093	−0.3952 (2.0705)	0.8494	−1.3284 (3.4482)	0.7016
2+	−0.5623 (2.1540)	0.7951	−3.4295 (2.1094)	0.1099	−0.5544 (2.1443)	0.7970	−7.4551 (3.5707)	0.0416

Adjusted for parental asthma, prior RSV hospitalization, household income, smoke exposure, and total IgE.

Discussion:

While we found no overall association between food allergy status and lung function among the entire unstratified cohort of participants, when investigating the asthma subgroup, we found that on average subjects with 2 or more food allergies had a lower FEF₂₅₋₇₅ %-predicted compared to individuals with no food allergies. Participants with asthma and only 1 food allergy did not show any significant differences in FEF_25–75 %-predicted compared to those with no food allergies. We found no significant associations with the other %-predicted lung function measures (FEV1, FVC, or FEV₁/FVC). After inclusion of aeroallergen status (including imputed data) in models for the subgroup of participants with asthma, FEF_25–75 %-predicted remained significantly lower on average in those with 2+ food allergies compared to those with none (β: −6.83%, SE: 3.26%, p-value = 0.0360). With these findings, our investigation suggests that the effect of multiple food allergies on lung function exists only when a child or young adult also has asthma.

In a previous study, Friedlander et al investigated the association between food allergy and lung function in inner city participants.³ This study showed that individuals with asthma who had multiple food allergies (allergies in ≥2 distinct food groups) had significantly decreased FEV₁ %-predicted and FEV₁/FVC compared to individuals with asthma but without any food allergies.³ This study differed from our current investigation in a several ways. First, the prior study only included asthmatic subjects and could not address the question of whether similar lung function changes would be seen in participants both with and without asthma. Second, Friedlander et al included a smaller number of participants (300 subjects), a younger recruited patient population (5–13 years old), and had higher proportion of black and Latino participants and participants from lower income households.³ Our study showed no significant lung function associations in the cohort as a whole, however we found an association with lung function (FEF_25–75) when looking at participants with asthma. While our study did not find similar effects in %-predicted FEV₁ and FEV₁/FVC in a more affluent and largely Caucasian population, we did show similar effects on small airway obstruction in subjects with asthma. In addition, other pediatric studies suggest that %-predicted FEF_25–75 alterations have greater long term clinical significance than other lung function parameters.^15–18 Also, %-predicted FEV1 may not be as closely associated with asthma disease severity in participants compared to adults.^19,20 Furthermore, the greater associations from aeroallergen exposure and sensitization found in Friedlander et al may result from the high perennial allergen exposure in inner city populations, potentially resulting in residual confounding.^23–27

In addition, other studies have shown that having a food allergy is associated with asthma.^4,28–30 Participants with food allergies have an increased risk of developing asthma at any age^28–30 and are more likely to develop asthma at an earlier age.⁴ Moreover, sensitization to certain foods appear to play a key role in this association, with food allergies to hen’s egg, cow’s milk, and tree nuts serving as the most frequently implicated in wheezing and/or asthma development.^{10,29,31–37} One prospective study found that a positive skin prick test to hen’s egg or cow’s milk during the first year of life had a ten-fold increase in the odds of having asthma as an adult.³⁸ In addition, the more food allergies a child or young adult has, increases the odds of asthma development.³⁴ Furthermore, food allergy is not only a risk factor for asthma, but it is also a risk factor for asthma morbidity and mortality.^1–3,28,39 In a case control study by Roberts et al, participants with food allergy had 8.58 times the odds of developing life-threatening asthma requiring ventilation compared with controls who required ICU admission without intubation.¹ Prior studies also show that subjects with any food allergy have more unsatisfactory levels of daytime symptom control, to have more asthma-related hospitalizations, to need a daily controller medication, and to need more rescue bronchodilator therapy.^3,28,39 These effects appear to increase with the presence of multiple food allergies.^3,28,39 For example, participants with more than one food allergy had 3 times the risk of daytime symptoms, over 5 times the risk of having a lifetime hospitalization, over 3.5 times the risk of having a hospitalization within a year, and 4.5 times the risk of unplanned healthcare utilizations with their primary healthcare provider.³ One potential explanation for our findings is that participants with food allergies represent a population who had both earlier asthma and worse exacerbations, with associated airway remodeling.

Another potential explanation for our findings may be that participants with multiple food allergies have altered nutritional status, which ultimately affects their underlying airway disease.^40–46 Studies show that participants with food allergy who are placed on preventative food elimination/restriction diets have an increased risk for nutrient deficiency including that of vitamin D and essential fatty acids.^40,41 This is relevant as nutrient deficiency is associated with increased presence of respiratory symptoms.⁴² Low intake of vitamin E, magnesium, sodium, potassium, and calcium have all been implicated as risk factors for increased wheeze.⁴² Vitamin E has been shown to cause a three-fold increase in wheezing.⁴² Furthermore, a diet low in omega-3 fatty acids of the diet may increase the risk of developing a TH2 phenotype after birth and subsequent increased predisposition to atopy.⁴³ In addition, inclusion of oily fish into an individual’s diet may serve as a protective factor against the development of childhood asthma.⁴⁴ There is also evidence that decreases in dietary antioxidants consumption may lead to increased airway hyperreactivity and wheeze in adults.^45,46 Therefore, altered nutrition with associated decreases in antioxidants may be one possible link of food allergy with worse asthma morbidity, airway inflammation and remodeling, and altered airway function. Future longitudinal studies should look at the mediation of dietary intake by food frequency questionnaires on the effects of food allergy on lung function.

The first limitation of this study is that only 36% of the cohort had aeroallergen sensitization status documented. We chose to include aeroallergen sensitization in our model since it appears to modulate lung function and airway inflammation.^47–51 Yet, this lack of data reduced our sample size in multivariable models, consequently reducing power to detect statistically significant associations after inclusion of non-imputed data. While we further investigated our findings by including an imputation of missing aeroallergen sensitization, these results should be cautiously interpreted due to the amount of missing data. However, the results after including imputed aeroallergen data showed similar effect size and direction as the results prior to aeroallergen inclusion in the model. This suggests that more complete aeroallergen sensitization data might also reveal significant changes in FEF_25–75 %-predicted. Another limitation is that we did not have a sufficient sample size to evaluate specific foods in relation to the altered pulmonary function. As previously stated, in other studies certain food allergies which present earlier in life appear to have a greater influence on asthma. However, our results suggest that regardless of specific food type, the presence of multiple food allergies is associated with decreased pulmonary function, specifically FEF_25–75 %-predicted. Additionally, due to the cross-sectional design, we could not investigate the temporal effect of food allergy diagnosis and alterations in lung function, nor did our study capture participants’ food allergy diagnosis duration. Identifying potential temporal effects of having food allergies, such as food allergy duration, on lung function may be useful for targeted preventative interventions. Future studies plan to investigate potential temporal associations. Also, the number of participants with asthma is greater in the group with two or more food allergies. As a result, this may partially account for the lower FEF_25–75 %-predicted observed in this patient population. However, we included asthma within the analytic models, which would account for this effect but there may be some residual confounding. Lastly our study only showed changes in FEF_25–75 %-predicted, and not FEV₁ %-predicted. Several studies indicate the importance of FEF_25–75 %-predicted as a significant marker of pediatric asthma severity and prognosis.^15–18 Also, FEV1 %-predicted may not closely associate with asthma severity when compared to other measures.^19–21 However, while FEF_25–75 has long-term prognostic value, this measure has potential limitations such as high variability, and a large normal reference range.^{15–18,52–54} In addition, the observed decrease within our study was less than those in the studies that showed a clinically significant relationship between FEF_25–75 and long-term prognosis,^15–18 and was also above the established cutoff values for normal values for FEF_25–75.⁵³ Therefore, the clinical ramifications of our findings remain unclear until additional studies determine the association of food allergy with longitudinal clinical outcomes.

In summary, we found that children and young adults with asthma have lower small airways lung function on average when they have at least two food allergies. We speculate this may be due to the severity of illness, or due to nutritional effects of food avoidance on lung growth or airway inflammation. We suggest that further studies should investigate the temporal relationship of multiple food allergy diagnoses with pediatric lung function trajectory. Future studies should further confirm the relationship between aeroallergen sensitization, multiple food allergies and asthma. Finally, additional investigation research should investigate mechanisms driving the association between multiple food allergies and lung function, including nutrient deficiencies.

Abbreviations:

FA

Food allergy

SPT

Skin prick test

IRB

institutional review board

MWD

Mean wheal diameter

sIgE

specific IgE

FEV₁

Forced expiratory volume in 1 second

FVC

Forced vital capacity

FEV₁/FVC

Forced expiratory volume in 1 second/Forced vital capacity

FEF_25–75

Forced expiratory flow, 25 – 75