IgE to common food allergens is associated with cardiovascular mortality in the National Health and Examination Survey and the Multi-Ethnic Study of Atherosclerosis

Abstract

Background:

In individuals without symptomatic food allergy, food-specific IgE is considered clinically irrelevant. However, recent studies have suggested that alpha-gal IgE is associated with cardiovascular (CV) disease.

Objective:

To determine whether sensitization to common food allergens is associated with CV mortality.

Methods:

The association between IgE sensitization to foods and CV mortality ascertained to 2019 was examined in the National Health and Examination Survey (NHANES) 2005–6 and the Wake Forest site of the Multi-Ethnic Study of Atherosclerosis (MESA) cohort; MESA enrolled adults without baseline clinical CV disease between 2000–2002. Total and specific IgE was measured to cow’s milk (CM), egg, peanut, shrimp, and a panel of aeroallergens (NHANES), and to CM, alpha-gal, peanut, dust mite and timothy grass (MESA). Cox-proportional hazard models were constructed, adjusting for sex, age, race/ethnicity, smoking, education, and asthma.

Results:

4414 adults from NHANES (229 CV deaths) and 960 from MESA (56 CV deaths) were included. In NHANES, sensitization to at least one food was associated with higher CV mortality (hazard ratio [HR] 1.7, 95%CI: 1.2–2.4, p=0.005). Milk sensitization was particularly associated (HR 2.0, 95 CI: 1.1–3.8, p=0.026), a finding replicated in MESA (HR 3.8, 95%CI: 1.6–9.1, p=0.003). Restricting analyses in NHANES to consumers of the relevant allergen strengthened food sensitization relationships, unmasking shrimp and peanut sensitization as additional risk factors for CV mortality.

Conclusions:

The finding that food sensitization is associated with increased risk of CV mortality challenges the current paradigm that sensitization without overt allergy is benign. Further research is needed to clarify mechanisms of this association.

Graphical Abstract

Capsule summary:

In two longitudinal studies, the presence of IgE to cow’s milk was associated with a significantly increased risk of cardiovascular mortality. When limited to those who routinely eat the allergen, shrimp and peanut sensitization were also associated with cardiovascular mortality.

Introduction:

The presence of IgE to food allergens is common, occurring in about 15% of unselected adults in the U.S.(1). While, by definition, food-specific IgE is necessary for IgE-mediated food allergy, most individuals who are sensitized to food are not clinically allergic and do not avoid the foods to which they are sensitized(2, 3). Whether these food-specific immunologic responses have health implications beyond traditional food allergy has been little explored. Recently, however, an association was reported between sensitization to an unusual food allergen, galactose-alpha-1,3-galactose (alpha-gal), a newly discovered cause of delayed meat allergy, and coronary artery disease (CAD) severity(4, 5).

The objective of this study was to understand whether sensitization to common food allergens was associated with cardiovascular mortality in the National Health and Nutrition Survey (NHANES) 2005–6. After a link between cow’s milk IgE and CVD mortality was identified in NHANES, the association between sensitization to cow’s milk and cardiovascular mortality was examined in a subset of the Multi-Ethnic Study of Atherosclerosis (MESA).

Methods:

Study Participants

NHANES is an in-person population-based cross-sectional survey conducted by the Centers for Disease Control and Prevention (CDC) that is designed to be representative of the non-institutionalized U.S. population(6, 7). It employs a complex study design to identify participants, with oversampling of certain populations to provide sufficient numbers for stratified analyses; weights are provided with the data to create estimates representative of the United States population. NHANES is conducted in 2 year blocks, with some questionnaires and laboratory studies varying by block. In NHANES 2005–6, allergen specific and total IgE were measured on all subjects aged 1 year and older. In the current analysis, participants 20 years and older at the baseline visit with information about food specific IgE, linked mortality data and non-missing smoking and education data were included. 557 of 4,971 participants were excluded because of missing data from the primary analysis (mostly food-specific IgE).

MESA is a National Heart Lung and Blood Institute sponsored cohort designed to study risk factors for subclinical and clinical heart disease. MESA enrolled participants 45 to 84 years of age from 6 sites in the United States who were initially free of CVD between July 2000 to July 2002(8, 9). Participants at the Wake Forest site were investigated as part of an ancillary study of stored sera designed to evaluate the prevalence of IgE to alpha-gal and the relationship between alpha-gal IgE and cardiovascular disease. Participants with available stored frozen sera and CT calcium coronary imaging were included in this analysis. Participants recruited at the Wake Forest Field Center were of White or Black race/ethnicity. Institutional review boards approved both studies and all participants provided written informed consent.

Exposure definitions

In NHANES, total IgE and specific IgE to cow’s milk, hen’s egg, peanut, shrimp, and aeroallergens (dust mite [D. Farinae and D. Pteronyssinus], cat, dog, cockroach, Alternaria, ragweed, rye grass, Bermuda grass, oak tree, birch tree, aspergillus, thistle, mouse and rat) were measured as previously reported, with reported detection limit of 0.35 kU_A/L(10). Aeroallergen sensitization and food sensitization were defined as none or any.

In MESA, total IgE and specific IgE to cow’s milk, peanut, alpha-gal, Timothy grass and dust mite (D. Pteronyssinus) were measured by ImmunoCAP (Thermo-Fisher/Phadia) using commercially available assays. In MESA, IgE to alpha-gal was measured on undiluted serum, whereas total IgE and the remaining specific IgE were measured on serum diluted 2-fold and normalized using a 2x multiplication factor. The detection limit of the assays is 0.1 kU_A/L, but consistent with many studies the cut-off for sensitization was considered as 0.35 kU_A/L.

Covariates

Demographic factors and health behaviors:

Race/ethnicity was by self-report (Mexican, Other Hispanic, African American/Black, White, and “Other” in NHANES, White, Black, Hispanic and Chinese in MESA). Education was by self-report and categorized as greater than high school or less than equal to high school. Smoking status was defined as never, former and current.

Consumption of foods:

Dietary information in NHANES comes from a number of different questionnaire sources. Past 30 day liquid milk consumption frequency, quantity and type of milk was asked of all participants, as was past 30 day consumption of shrimp. A 12-month food frequency questionnaire (FFQ) quantified concentrated egg consumption (i.e. not baked goods) and gave additional details about milk consumption, including frequency of milk used in cereal. Peanut was also queried on the 12-month FFQ, but the question included seeds and nuts; thus it is possible to exclude those who do not eat peanut but not to definitively identify peanut consumers.(1, 11, 12) For the purposes of this analysis, milk consumption was classified as never if on the 30 day lookback and the 12 month FFQ the participant reported no liquid milk consumption. In both questionnaires, participants were asked to define the type of milk typically used, and only cow’s milk (i.e. nonfat/skim, 2% or whole milk) was considered.

Co-morbidities:

In NHANES, asthma was defined by self-reported doctor’s diagnosed asthma and “still” having asthma. Atopic dermatitis was defined as an affirmative answer to the questions “Have you ever had an itchy rash lasting for at least 6 months”, “has the rash occurred in the past 12 months” and “Has this itchy rash at any time affected any of the following places: the folds of the elbows, behind the knees, in front of the ankles, under the buttocks, or around the neck, ears, or eyes?”.

In MESA, asthma was defined by self-report of a history of doctor’s diagnosed asthma. Atopic dermatitis was not assessed in MESA.

Diabetes and hypertension were assessed in NHANES by self-reported history of doctor’s diagnosis of these conditions. In MESA, baseline diabetes mellitus was defined by 2003 ADA fasting criteria (impaired fasting glucose was not considered diabetes) or use of antidiabetic medication(13). Obesity was defined in both cohorts as body mass index of 30 kg/meter-squared or higher. Hypertension was defined by JNC V1(1997) criteria(14).

High cholesterol was defined as a total cholesterol of 200 mg/dL or higher(15). In NHANES, only part of the sample fasted prior to measuring cholesterol; in MESA, this was a fasting sample.

Outcome ascertainment

NHANES was matched to the National Death Index through December 31, 2019 using probability matching. For this analysis, CV mortality was defined by the category “Diseases of the Heart”, which includes ICD-10 codes I00-I09, I11, I13, and I20-I51 (but does not include cerebrovascular disease or other circulatory disease). Publicly available cause of death data is slightly perturbed in NHANES to maintain privacy(6, 7).

In MESA, participants were contacted every 9–12 months, with 6 additional in-person clinic visits occurring to date and mortality ascertained through December 31, 2019. Eligible cardiovascular events and mortality were captured by linkage to the NDI, questionnaire, medical records search, obituaries, and next of kin interviews, and then verified by autopsy report, hospitalization records and/or interview as appropriate, with adjudication by an expert committee. In order to match the NHANES definition of CV death, cerebrovascular disease was excluded from the definition of CV death in MESA for these analyses.

Statistical Analysis

Descriptive data are presented as percentages or means and 95% confidence intervals (CI) as appropriate, and account for the survey weighting and complex study design for NHANES.

To study the association between food sensitization and cardiovascular mortality in NHANES, a Cox proportional hazard model was created of any food sensitization (defined as any positive food specific IgE), adjusting for race/ethnicity, sex, age, education, smoking, and asthma (base model/Model 1). The assumption of proportional hazards was evaluated by testing of pooled scaled Schoenfeld residuals using the “estat” command in STATA; because this does not accommodate survey set data, this was done without survey setting. Because smoking status and race violated the assumption of proportional hazards, analyses incorporated stratification by these variables for NHANES. The relationship between individual food sensitizations, sensitization to any aeroallergen and total IgE and cardiovascular mortality were similarly examined.

To replicate the finding of an association between cow’s milk sensitization and cardiovascular mortality in NHANES, a Cox proportional hazard model was created in MESA, adjusting for the same variables. Using the same methods as above, the assumption of proportional hazards was tested; there were not violations except as detailed below. The relationship between sensitization to peanut, total IgE and aeroallergen sensitization was similarly examined. Alpha-gal sensitization was too infrequent to analyze relationships with outcomes.

In both studies, the base model was additionally adjusted for hypertension, obesity, high cholesterol, and diabetes mellitus (Model 2). In NHANES, a model additionally adjusting for atopic dermatitis was created (Model 3). In NHANES, models stratified by allergen consumption status were created to explore whether allergen consumption modified the observed associations. Sensitivity analyses in NHANES were done to assess whether higher levels of IgE sensitization (defined as ≥2 kU_A/L) conferred more risk of cardiovascular mortality. In NHANES, all analyses used the survey weights, population sampling units (PSUs) and strata provided in the survey except for the Kaplan Meier curves.

All statistical analyses were done in STATA SE version 17 (College Station, TX). A p-value of less than 0.05 was considered significant.

Results:

Participant characteristics

In NHANES, 4,414 individuals were included, of whom 874 died by the time of follow-up (229 of cardiovascular disease). Median follow-up time was 13.8 years (interquartile range [IQR]: 13.2–14.3). Of 4,414, 731 (weighted percentage: 15%) were sensitized to at least one food, 166 (4%) to milk, 358 (7%) to peanut, 127 (3%) to egg, 311 (6%) to shrimp, and 1896 (42%) to aeroallergens, while 172 had higher level IgE sensitization (≥2 kU_A/L) to at least one food (3.5%), 23 to milk (0.6%), 86 to peanut (1.6%), 10 to egg (0.2%) and 73 to shrimp (1.4%). Mean total IgE was 133 kU/L (95% CI: 118–148).

In MESA-Wake Forest, 960 individuals were included, of whom 290 died in follow-up (56 classified as non-stroke cardiovascular causes). Median follow-up time was 17.7 years (IQR: 13.5–18.4). Thirty seven (4%) were sensitized to milk, 65 to peanut (7%), 8 to alpha-gal (0.8%), and 273 (28%) to grass or dust mite. Mean total IgE was 117 kU/L (95% CI: 99–135). Supplemental table 1 shows event rates stratified by exposure.

Demographic characteristics of the cohorts and univariate associations with food sensitization are shown in Table 1/Supplemental Table 2. In univariate analyses, there were no significant differences in age, sex, race/ethnicity, education level or smoking between those sensitized and not sensitized to milk in either cohort (Table 1). In contrast, those sensitized to peanut in both cohorts, and shrimp in NHANES, were younger, less likely to be female, and less likely to be of White race/ethnicity (Supplemental Table 2). In both cohorts, those sensitized to milk were more likely to have diabetes than those unsensitized in univariate analyses, while relationships between diabetes and sensitization to other foods were not significant and/or were inconsistent between cohorts. Obesity was not consistently associated with sensitization across the cohorts. Neither hypertension nor high cholesterol showed a statistically significant association with sensitization in either cohort.

Table 1:

Characteristics of the cohorts

	NHANES				MESA
	Overall (n=4414)	Milk Sensitized (n=166)	Not Milk Sensitized (n=4248)	P*	Overall (n=960)	Milk Sensitized (n=37)	Not Milk Sensitized (n=923)	P*
Age, mean. y	46 (45–48)	49 (46–51)	47 (45–48)	0.06	62 (62–63)	60 (56–63)	63 (62–63)	0.12
Gender, % female	52 (51–53)	44 (32–58)	52 (51–54)	0.21	53 (50–56)	41 (26–57)	54 (50–57)	0.12
Race/ethnicity, %				0.23				0.92
Black	11 (8–16)	12 (7–21)	11 (8–16)		43 (40–46)	41 (26–57)	43 (40–47)
Hispanic/Latino (Mexican)	8 (6–10)	3 (1–6)	8 (6–11)		<1	0	<1
Hispanic/Latino (Non-Mexican)	3 (2–5)	6 (3–12)	3 (2–5)
White	72 (66–78)	76 (65–84)	72 (66–78)		57 (54–60)	57 (53–60)	57 (53–60)
Other than the above	5 (4–7)	4 (1–15)	5 (4–7)		0	0	0
Education level HS or greater, %	58 (54–62)	61 (49–73)	58 (54–62)	0.49	71 (68–74)	84 (68–93)	71 (68–74)	0.09
Smoking Status				0.40				0.37
Never	51 (48–54)	55 (46–63)	51 (48–54)		45 (42–48)	54 (38–69)	45 (41–48)
Former	25 (23–27)	23 (18–31)	25 (23–27)		41 (38–44)	30 (17–46)	41 (38–45)
Current	24 (22–27)	22 (18–26)	24 (22–27)		14 (12–17)	16 (7–32)	14 (12–17)
Diabetes, %	8 (7–9)	14 (8–23)	8 (7–9)	0.046	11 (9–13)	22 (11–38)	10 (9–13)	0.032
Obese, %	34 (31–38)	42 (37–51)	34 (31–37)	0.052	35 (32–38)	46 (31–62)	34 (31–38)	0.15
Hypertension, %	30 (28–32)	36 (29–43)	30 (27–32)	0.08	54 (51–57)	65 (48–78)	54 (50–57)	0.18
High cholesterol, %	45 (43–48)	53 (43–63)	45 (43–47)	0.11	40 (33–39)	41 (26–57)	36 (33–39)	0.56
HbA1c, %	5.5 (5.4–5.5)	5.7 (5.5–5.9)	5.4 (5.4–5.5)	0.027	Not assessed.
Asthma, %	8 (7–10)	14 (7–25)	8 (7–9)	0.08	10 (8–12)	22 (11–38)	9 (7–11)	0.01
Atopic Dermatitis, %	3 (3–4)	2 (1–4)	4 (3–4)	0.10	Not assessed.
Milk Consumption, %				0.19	Sample size too small.
Never	7 (6–8)	3 (1–10)	7 (6–8)
Some	93 (92–94)	97 (90–99)	93 (92–94)

Expressed as means or percentages with 95% CI in parenthesis (medians and standard deviations are not calculable for NHANES survey methodology).

^*Chi-squared or T Test (MESA) or univariate regression (NHANES) as appropriate. NHANES estimates are weighted, analyses account for survey strata and sampling unit.

Association between sensitization to foods and cardiovascular mortality (Table 2/Figure 1)

Table 2:

Association between cardiovascular death and sensitization.

	NHANES (HR, 95% CI, p value)			MESA (HR, 95% CI, p value)
	Model 1 (n=4414)	Model 2 (n=4338)	Model 3 (n=4410)	Model 1 (n=960)	Model 2 (n=944)
Milk Sensitization	2.0 (1.1–3.8), 0.026	1.9 (1.1–3.4), 0.035	2.1 (1.1–3.8), 0.027	3.8 (1.6–9.1), 0.003	3.5 (1.4–8.8), 0.007
Peanut Sensitization	1.8 (0.8–3.9), 0.14	1.4 (0.7–2.8), 0.30	1.9 (0.9–4.0), 0.11	1.4 (0.5–3.5), 0.53	1.3 (0.5–3.4), 0.59
Egg Sensitization	1.3 (0.7–2.2), 0.36	1.2 (0.6–2.2), 0.57	1.3 (0.7–2.2), 0.33	Not measured
Shrimp Sensitization	1.6 (0.9–2.7), 0.08	0.8 (0.4–1.4), 0.38	1.2 (0.8–2.0), 0.36	Not measured
Food sensitization overall	1.7 (1.2–2.4), 0.005	1.4 (1.1–1.8), 0.01	1.6 (1.1–2.2), 0.008	Not applicable
Aeroallergen sensitization	1.1 (0.8–1.5), 0.49	1.0 (0.7–1.4), >0.99	1.1 (0.8–1.4), 0.70	1.4 (0.8–2.5), 0.25	1.4 (0.8–2.4), 0.30
Total IgE (per log10 increase)	1.1 (1.0–1.3), 0.02	1.1 (1.0–1.2), 0.09	1.1 (1.0–1.2), 0.032	1.3 (1.9–1.6), 0.006	1.3 (1.1–1.6), 0.007

Hazard Ratio (HR), 95% CI, p value. Adjusted for age, race/ethnicity, sex, smoking history, asthma and education. Model 2 additionally adjusted for diabetes, obesity, high cholesterol and hypertension. Model 3 is model 1 additionally adjusted for atopic dermatitis. HR: Hazard rate.

Figure 1.

Kaplan-Meier curves of cumulative cardiovascular death in NHANES (left) and MESA (right), stratified by the presence of IgE to cow’s milk (milk sensitization). Curves are unadjusted and unweighted.

Sensitization to at least one food was significantly associated with cardiovascular mortality (Hazard Ratio [HR]: 1.7 [95% CI: 1.2–2.4], p=0.005) in the base adjusted NHANES model. Of the individual food allergens, sensitization to cow’s milk was associated with increased cardiovascular mortality (HR 2.0 [95% CI: 1.1–3.8], p=0.026) in NHANES. The association between milk sensitization and cardiovascular mortality was replicated in MESA-Wake Forest (HR 3.8 [95% CI: 1.6–9.1], p=0.003).

Adjustment for diabetes, obesity, high cholesterol, and hypertension only minimally attenuated the relationship between milk sensitization and CV death in both cohorts. In NHANES, where atopic dermatitis was ascertained, adjustment for atopic dermatitis did not attenuate the relationship between milk sensitization and CV mortality. Notably, atopic dermatitis itself was not a risk factor for cardiovascular mortality in NHANES (HR: 1.1, [95% CI: 0.5–2.2], p=0.89) in the base model.

In a sensitivity analysis of higher level sensitization and cardiovascular mortality in NHANES, the relationship between any food sensitization and cardiovascular mortality was strengthened (HR 2.7 [95% CI: 1.2–6.0], p=0.02), and egg and peanut sensitization emerged as a risk factor for CV mortality (HR: 5.9, [95% CI: 1.1–32.0], p=0.040 and HR: 3.5 [95% CI: 1.1–11.8], p=0.049, respectively). The relationship with milk sensitization was essentially unchanged (Supplemental Table 3). The sample size in MESA was too small to conduct this sensitivity analysis.

Total IgE was associated with cardiovascular mortality in both NHANES and MESA (Table 2). This association was robust to adjustment for diabetes, obesity, high cholesterol, and hypertension in MESA, but not NHANES. Aeroallergen sensitization was not associated with CV death in either cohort.

Modification of associations by food allergen consumption status (Table 3)

Table 3.

Association between food sensitization and CV mortality in NHANES stratified by consumption of allergen.

	Consumers overall	Non-consumers
Milk sensitization	2.3 (1.3–4.0), 0.006 (n=3,885)	*(n=271)
Peanut sensitization	2.5 (1.1–5.6), 0.029 (n=2,919)	*(n=156)
Egg sensitization	1.3 (0.7–2.8), 0.40 (n=2,851)	*(n=198)
Shrimp sensitization	3.7 (1.3–10.1), 0.016 (n=2,031)	1.0 (0.3–3.3), 0.96 (n=2,228)

Hazard ratio (HR), 95% CI, p value. Adjusted for age, sex, race/ethnicity, smoking status, education, and asthma..

^*Too few participants in these categories to calculate HRs.

To explore whether food allergen consumption modified identified relationships between sensitization and mortality, participants were stratified by whether they reported consuming the relevant allergen in NHANES. Although sensitization to egg, peanut, and shrimp did not show statistically significantly associations with cardiovascular mortality in unstratified models, all relationships between food sensitizations and CV mortality except for egg were strengthened when considering only consumers. Shrimp and peanut sensitization emerged as a statistically significant risk factors for CV mortality (HR 3.7, [95% CI: 1.3–10.1], p=0.016 and HR 2.5, [1.1–5.6], p=0.029, respectively) among consumers. In contrast, shrimp sensitization was not a risk factor among non-consumers (HR 1.0, [95% CI: 0.3–3.3], p=0.96). The number of participants in the strata were too small for reliable estimates of non-consumers of milk, peanut, or egg. The sample size in MESA was too small to stratify by consumption status.

In sensitivity analyses considering higher-level IgE sensitization stratified by consumption status, all hazard ratios were larger than in analyses defining sensitization by a lower threshold among those who consume the allergen, with peanut sensitization and egg sensitization emerging as statistically significant risk factors for cardiovascular death (HR 6.6 [95% CI: 2.3–18.9], p=0.002 and HR 6.4 [95% CI: 1.3–32.1], p=0.026, respectively, Supplemental Table 3).

Discussion:

Among 4414 adults in NHANES and 960 adults in MESA followed for more than a decade, the presence of milk-specific IgE was found to be associated with higher risk of cardiovascular mortality. Aside from two recent reports linking IgE to the unusual carbohydrate allergen alpha-gal to coronary artery disease, cardiovascular disease has not previously been identified as a long-term complication of food sensitization.

Although the observation linking milk sensitization with cardiovascular mortality is novel, other allergic conditions have been previously associated with cardiovascular disease. In population based studies, atopic dermatitis, asthma and total IgE have been identified as risk factors for cardiovascular disease(16–22). Here, total IgE was associated with cardiovascular mortality in both NHANES and MESA, as has been described for a shorter follow-up time in this NHANES cohort(23). However, in both NHANES and MESA, sensitization to aeroallergens was not associated with cardiovascular mortality, and the association between milk sensitization and cardiovascular mortality was robust to inclusion of atopic dermatitis and asthma, suggesting that food sensitization may convey a risk for cardiovascular disease apart from other allergic diseases. Of note, although small amounts of alpha-gal are present in milk and dairy products, IgE to alpha-gal does not explain the current findings as the prevalence of alpha-gal IgE was very low in MESA and would also be expected to be uncommon in NHANES.

Although sensitization to common foods and cardiovascular disease have not been previously linked, there is substantial evidence for the importance of allergic-type immune pathways in normal cardiac physiology and heart disease. Mast cells are present in all layers of the heart, where they are integrally involved in cardiac development, regulation of the cardiac renin-angiotensin system, and remodeling after myocardial infarction(24–28). Like all mast cells, they express the high-affinity IgE receptor (FcεR1) and allergen-dependent cross-linkage of IgE bound to FcεR1 leads to release of inflammatory mediators. Mast cells are also found in cardiac arteries, especially in and around atherosclerotic plaques. They are activated as atherosclerosis progresses, and this may augment plaque progression and destabilization. These mast cells carry IgE, although the specificity of the IgE has not been determined(29). In addition, several murine models implicate IgE and FcεR1 in cardiac fibrosis and heart failure; these models include allergic asthma-induced cardiac fibrosis and heart failure caused by non-allergic factors such as transverse aortic constriction or angiotensin II(30–32).

It is notable that in both cohorts aeroallergen sensitization was not associated with cardiovascular mortality and that sensitization to shrimp and peanut emerged as risk factors for cardiovascular death only among those who regularly consume the food. Together, these two findings suggest that the link between food IgE and cardiovascular disease depends on chronic exposure via the gastrointestinal tract. We speculate that oral exposure to food allergens leads to proteins in the circulation(33–35) interacting with IgE on cardiac mast cells, resulting in local inflammation, progression and/or destabilization of atherosclerotic plaques, and fibrosis. A similar working model has been proposed for IgE-armed mast cells in coronary arteries to explain the association observed between alpha-gal sensitization and coronary artery disease(36). In this model, sensitization to aeroallergens would not be expected to cause cardiovascular disease because the magnitude of exposure from respiratory allergens is log orders lower than for food allergens (nanograms(37) versus grams), making circulating allergen less likely. Moreover, stronger associations would be expected for foods that are eaten in larger quantities, such as cow’s milk, which accounts for an estimated 13% of the average American adult’s daily protein intake, compared to 4% for eggs, 4% for seafood as a whole, including fish, and 2% for nuts(38).

The associations we have reported relate to food sensitization rather than clinical food allergy. Sensitization to common food allergens is very frequent in the U.S. population, peaking in prevalence at more than 25% in early childhood and then declining throughout adulthood(39). Although sensitization is definitionally required for IgE-mediated food allergy, it is a poor predictor of clinical allergy, as the majority of individuals with food IgE detected in blood testing do not have clinical allergy. Allergy is fundamentally a clinical diagnosis, based on reproducible acute reactions and the absence of tolerating the food(1–3). Although we did not have access to information about clinical food allergy in either cohort, we expect that individuals who report regularly consuming a food allergen on food frequency questionnaires are not clinically allergic. Thus, our findings that associations were strengthened when excluding those who avoid the food suggest that these findings are most relevant to those who have not been diagnosed with food allergy. Although still preliminary, these results raise questions about whether these apparently non-allergic individuals may have long-term consequences from oral exposure to foods to which they are sensitized. Unfortunately, the MESA sub-cohort was too small to validate the consumption findings observed in NHANES, and more data is needed on how and whether consumption affects the risk of food sensitization. At this point, the idea that chronic oral exposure to food allergens induces chronic inflammation remains hypothetical, but if true, may also be relevant for treatments for food allergy such as oral immunotherapy, which has the goal of suppressing acute reactivity to food allergens through ongoing oral exposure in the context of a persistent food-specific IgE response(40); long-term follow-up of non-allergic consequences of oral immunotherapy may be necessary.

The magnitude of the identified associations are similar to or exceed those previously associated with smoking, diabetes, and rheumatoid arthritis(41–43). Investigating MESA, McEvoy et al. reported that compared to non-smokers, current smokers had an adjusted HR of 1.7 (95%CI: 1.3–2.2) for all-cause cardiovascular disease(44). Yeboah et al. found that diabetes was associated with CVD incidence with a HR of 1.9 (95%CI: 1.5–2.4)(45).While it may be unexpected to find a novel risk factor of this magnitude, the relationship between food sensitization and long-term outcomes, to our knowledge, has not previously been explored.

These analyses do not establish causality, and it is possible that another environmental or genetic factor could jointly influence both food sensitization and cardiovascular disease. For example, we note that milk sensitization was associated with diabetes, however, our results were robust to adjustment for diabetes, hypertension, and obesity. Reverse causation is also possible; cardiovascular disease could increase the risk for food sensitization through a mechanism such as increased gastrointestinal permeability or as an inflammatory response to myocardial injury(46). Arguing against this possibility is the temporal relationship between food sensitization and the cardiovascular outcomes, which in many cases occurred more than 10 years after sensitization was measured. Further, there is no evidence that food sensitization increases with aging, as would be expected if subclinical cardiovascular disease was a major risk factor for food sensitization(39).

Limitations of this analysis include that food consumption was by self-report from instruments that were not designed specifically for allergen assessment and only offer a snapshot in time. For milk and egg, we did not include other types of foods that likely or definitely have the allergen in them, such as cheese, yogurt, etc. for milk, and baked goods for egg. For peanut, the question included nuts and seeds, so there is likely misclassification. We expect that the result of any of these dietary misclassifications would be to bias our estimates in analyses stratified by consumption to the null. Small numbers in many of the stratified groups limited our ability to generate estimates and the MESA-Wake Forest cohort was too small to analyze consumption. NHANES cause of death was subject to perturbations to protect privacy, but we expect the effect of this intentional random misclassification to be small and bias to the null. A number of statistical comparisons were made in this paper, which could lead to false positive statistical significance. However, these tests were not all independent, and the pattern of findings is reinforcing. Specifically, the association between cardiovascular death and cow’s milk sensitization was replicated in both studies, and associations were stronger among those who consumed the allergens and at higher levels of IgE. All of these aspects of our multiple tests tend to reinforce our interpretation, though additional research is needed. Finally, as discussed above, these associations do not establish causality.

In summary, IgE sensitization to cow’s milk was identified as a risk factor for cardiovascular mortality across two large cohorts. In NHANES, which was large enough to stratify some analyses by consumption of the allergen, it appeared that consumption was an important modifier of cardiovascular risk and sensitization to other foods emerged as a cardiovascular risk in these analyses. Although these findings challenge the current paradigm that food-specific IgE in the absence of overt allergic symptoms is benign, there is ample evidence that chronic activation of allergic immune pathways can contribute to inflammation, tissue remodeling, and disease. These data raise intriguing questions about the relevance of food sensitization and diet in cardiovascular disease development, but further studies - including replication of the finding of modification by consumption, investigation of the cardiovascular subtypes implicated, and identification of biological mechanisms - will be important before any changes to medical practice can be considered.

Clinical Implications:

IgE to cow’s milk was associated with higher cardiovascular mortality in both NHANES and MESA cohorts. Further study is needed to understand whether sensitization to cow’s milk and other food allergens are causal in the pathogenesis of cardiovascular disease.

Abbreviations.

CV

Cardiovascular

CM

Cow’s Milk

NHANES

National Health and Nutrition Examination Survey

MESA

Multi-ethnic Study of Atherosclerosis

HR

Hazard Ratio

CI

Confidence Interval

CAD

Coronary artery disease

CDC

Centers for Disease Control and Prevention

FFQ

Food frequency questionnaire

ADA

American Diabetes Association

PSU

Population Sampling Unit