Excessive Unbalanced Meat Consumption in the First Year of Life Increases Asthma Risk in the PASTURE and LUKAS2 Birth Cohorts

Abstract

A higher diversity of food items introduced in the first year of life has been inversely related to subsequent development of asthma. In the current analysis, we applied latent class analysis (LCA) to systematically assess feeding patterns and to relate them to asthma risk at school age. PASTURE (N=1133) and LUKAS2 (N=228) are prospective birth cohort studies designed to evaluate protective and risk factors for atopic diseases, including dietary patterns. Feeding practices were reported by parents in monthly diaries between the 4^th and 12^th month of life. For 17 common food items parents indicated frequency of feeding during the last 4 weeks in 4 categories. The resulting 153 ordinal variables were entered in a LCA. The intestinal microbiome was assessed at the age of 12 months by 16S rRNA sequencing. Data on feeding practice with at least one reported time point was available in 1042 of the 1133 recruited children. Best LCA model fit was achieved by the 4-class solution. One class showed an elevated risk of asthma at age 6 as compared to the other classes (adjusted odds ratio (aOR): 8.47, 95% CI 2.52-28.56, p = 0.001) and was characterized by daily meat consumption and rare consumption of milk and yoghurt. A refined LCA restricted to meat, milk, and yoghurt confirmed the asthma risk effect of a particular class in PASTURE and independently in LUKAS2, which we thus termed unbalanced meat consumption (UMC). The effect of UMC was particularly strong for non-atopic asthma and asthma irrespectively of early bronchitis (aOR: 17.0, 95% CI 5.2-56.1, p < 0.001). UMC fostered growth of iron scavenging bacteria such as Acinetobacter (aOR: 1.28, 95% CI 1.00-1.63, p = 0.048), which was also related to asthma (aOR: 1.55, 95% CI 1.18-2.03, p = 0.001). When reconstructing bacterial metabolic pathways from 16S rRNA sequencing data, biosynthesis of siderophore group nonribosomal peptides emerged as top hit (aOR: 1.58, 95% CI 1.13-2.19, p = 0.007). By a data-driven approach we found a pattern of overly meat consumption at the expense of other protein sources to confer risk of asthma. Microbiome analysis of fecal samples pointed towards overgrowth of iron-dependent bacteria and bacterial iron metabolism as a potential explanation.

Keywords: Infancy; asthma; cow's milk; gut microbiome; introduction of solid foods; latent class analysis; meat; nutritional immunity.

Figure 1

Selection of study populations.

Figure 2

Food introduction styles in PASTURE as defined by a latent class analysis of consumption of 17 food items over 9 months.

Figure 3

Food introduction styles stratified by macronutrients and fruits/vegetables. The asthma risk class LC2 is contrasted with a pool of the all other class (LC1, LC3, and LC4).

Figure 4

Prediction of asthma by protein and carbohydrate sources. The 10 most important prediction variables representing contrasts between food items are shown.

Figure 5

Food introduction styles in PASTURE as defined by a latent class analysis of consumption of meat, milk, and yoghurt over 9 months from month 4 to 12.

Figure 6

Distribution of latent classes across study centers in PASTURE.

Figure 7

Food introduction styles in LUKAS2 as defined by a latent class analysis of consumption of meat, milk, and yoghurt over 9 months.

Figure 8

Effects of UMC on asthma (panel A) and wheeze (panel B) phenotypes in PASTURE.

Figure 9

Associations of asthma with breastfeeding and formula feeding stratified by UMC.

Figure 10

Vulnerable window for the effect of UMC on asthma at age 6. UMC is coded as daily meat consumption without milk or yoghurt consumption at least on a weekly level.

Figure 11

Sensitivity analyses on intensity of meat and milk or yoghurt consumption and the role of industrial processing. Shown is the prevalence of asthma in relation to the intensity of meat and milk/yoghurt consumption during month 10 and 11 (A, B, respectively). (C) presents prevalence of asthma in relation to industrial food processing; for meat, excessive consumption (daily over both months 10 and 11) and for milk, moderate consumption (weekly either during month 10 or 11) is displayed. For bivariate comparisons, we used the Chi-square or Fisher’s exact test, for trends we used the Cochran-Armitage trend test.

Figure 12

Microbial genera and microbial metabolic pathways associated with UMC. (A) Odds ratios (OR) with 95%-confidence intervals for the associations of UMC with relative abundance of bacterial genera in fecal samples. (B) Odds ratios (OR) with 95%-confidence intervals for the associations of UMC with relative abundance of metabolic pathways in fecal samples.