LAUSR

The relationship between childhood wheeze and increased risk of asthma/lung disease is well established, and latent class analysis (LCA) of wheezing patterns has done an excellent job of categorizing childhood wheeze into different subcategories with differing risks of asthma development and associations with higher or lower levels of pulmonary function (1–3). However, the description of these categories often varies between studies because of sample size, timing, frequency of data collection, and potentially imprecise classification of an individual into a category. In this issue of the Journal, Haider and colleagues (pp. 883–893) use a data-driven, novel, dynamic “spellbased” approach and apply partition around medoids (PAM) clustering on these spell indicators to derive wheezing phenotypes, as opposed to applying LCA to the binary question of wheeze versus no wheeze (4). This spell-based approach incorporated six multidimensional longitudinal indicators, including the duration of wheezing episodes. It was applied to data from 7,719 participants from the STELAR (Study Team for Early Life Asthma Research) consortium, an established consortium of five birth cohorts with harmonized data (3). Using this spell-based approach, the authors categorized wheeze into five classes: never (NWZ; 54.1%); early-transient (ETW; 23.7%); late onset (LOW; 6.9%); persistent (PEW; 8.3%); and a novel phenotype, intermittent wheeze (INT; 6.9%). This differed from previous analyses that used LCA, which most often identified four consistent wheeze phenotypes: NWZ, ETW, LOW, and PEW (1–3). The wheezing phenotypes derived from the dynamic spell-based approach on their surface appear similar to those derived from LCA, but addition of the INT group created more homogeneous phenotypes. For instance, the spell-based NWZ group had no wheezing, whereas in the LCANWZ group, 10% had occasional wheezing. The spell-based ETW group reported no wheezing after 10 years of age, whereas in the LCA ETW group, some children reported wheeze to age 18 years. Similarly, in the spell-based LOW group, there was no wheezing before the age of 8, whereas in the LCA LOW group, there was wheezing at earlier ages. The PAM spell-based grouping also appeared more stable in terms of children’s assignments and predicting asthma and pulmonary function. The spell-based approach demonstrated that all four wheeze phenotypes compared with the NWZphenotype were associated with a higher risk of asthma diagnosis andmedication use in adolescence, with 3.4% of the ETW class having an asthma diagnosis in the PAM spell-basedmodel versus 8.4% in the LCA binarymodel, illustrating increased homogeneity in the spell-based approach. Importantly, the majority of asthma risk fell in the LOW, PEW, and INT spell-based classes and relatively little in the NWZ and ETWgroups. All four spell-based classes of wheeze were significantly associated with decreases in FEV1/FVC in adolescence, with the PEW and INT classes being consistently lower than the ETWand LOWgroups. Interestingly, the LOWgroup still showed increased risk of asthma, despite decreases in FEV1/FVC less than those observed in the INT and PEWgroups. The FEV1/FVCwas significantly lower in all wheeze phenotypes compared with the NWZ group, supporting the premise that the children with wheezing phenotypes likely did not reach normalmaximum lung function in early adulthood. The FEV1/FVC was consistently lower in the PEW and INT groups than in the ETW and LOWphenotypes and thus could have the greatest predisposition to chronic obstructive pulmonary disease over time. Of all wheeze phenotypes, the novel INT groupwas the only phenotype to have a decline in their z-score for FEV1/FVC from ages 8 to 24 in both cohorts, suggesting an early decline in lung function for this group. Genetic analysis showed that 17q12–21 and CDHR3 polymorphisms were associated with higher odds of PEW and INT but not with LOW, thus suggesting that LOW is quite different from PEW and INT (4). This is also in contrast to findings from other investigators that demonstrated similar associations between multiple SNPs in the 17q12–21 locus and all LCA wheezing phenotypes (5). This suggests that a more thorough slicing and dicing of wheeze phenotypes will facilitate better genetic linkage and explanations of cause. Thus, a key finding of this paper is that better wheeze classification will potentially allow better correlation with genetic and epigenetic mechanisms as well as prenatal and postnatal exposures. This can lead to better prevention strategies by prioritizing and modifying the many potential risk factors for wheeze. Although complex, some of these potentially modifiable prenatal factors include preterm deliveries, intrauterine growth restriction, smoking in pregnancy, toxin exposure, placental abnormalities, maternal nutrition, inflammation, stress, obesity, and delivery mode (1). Potential postnatal factors include breastfeeding duration, viral infections, allergen exposures, and the lung–gut microbiome, among others. Greater understanding of wheeze phenotypes will facilitate development of targeted therapies, potentially in utero or early in postnatal life, such as vitamin C supplementation during pregnancy, antigenic exposures, and vitamin D supplementation (6–9). Understanding the interaction of specific wheeze phenotypes with specific genotypes has particular promise (10). This study raises a number of unanswered questions. Will different approaches be needed to prevent the different types of wheeze? Howmuch heterogeneity still resides in these five classes? What are the relative prenatal versus postnatal causes of these classes and the role of environmental pollution?Which of these phenotypes may predispose individuals to chronic obstructive pulmonary disease? How will the novel INT phenotype evolve beyond early adulthood? This article is open access and distributed under the terms of the Creative Commons Attribution Non-Commercial No Derivatives License 4.0. For commercial usage and reprints, please e-mail Diane Gern ([email protected]).