D3.157 - A study of the detailed patterns of wheeze presence and absence from infancy through to adolescence

Clinical experience and analyses in birth cohort studies indicate that children who have wheezed in the past are more likely to do so again in the future. We aimed to provide a detailed descriptive and exploratory analysis of how past and current wheeze influences future wheeze.

We used complete longitudinal information about wheeze presence at five time points (TPs) in 7719 individuals across five UK birth cohorts in the STELAR consortium: infancy (0.5-1 year); early childhood (2-3 years); pre-school to early school (4-5 years); middle childhood (8-10 years); and adolescence (14-18 years). We described the patterns of wheeze presence and absence across time using a graphical "bifurcating tree". We focus especially on positive (or negative) “forward probabilities” (PFPs)—that is, the probability of (not) having wheeze at the next TP, from the current state.

We used Correspondence Analysis (CA) to explore the associations of wheezing status at TPs 1 through 4 with wheezing at TP5. A 16×2 contingency table was constructed, with rows representing the different wheezing patterns across TPs 1-4 and columns indicating wheeze presence and absence at TP5. CA decomposes the structure of associations into a single dimension, allowing us to assess which time points contributed most to the differentiation between individuals who developed wheeze at T5 and those who did not.

The detailed descriptions of all 32 possible patterns of wheeze presence/absence across the five TPs are shown in the bifurcating tree in Figure 1. The forward probabilities show clear patterns whereby PFPs that in principle could be independent of one another exhibit marked correlations; these clearly relate to similar patterns of recent wheeze (Table 1).

CA identified a significant association between wheezing status over time and outcomes at TP5. The primary axis of differentiation (12.2% variance explained) shows that wheezing at TPs 3 and 4 was most strongly associated with wheeze at TP5, as indicated by higher category contributions. In contrast, wheezing at TPs 1 and 2 showed weaker associations with TP5 outcomes.

At every age, the probability of future wheeze depends heavily on an individual’s complete history of wheeze at specific time points. The bifurcating tree representation allows for answers to relevant clinical questions relating to wheeze history to be more easily framed and visualised. Importantly, the description which naturally emerges from this approach shows that later phase wheezing (school age) is more closely related to wheezing in adolescence than early-life wheezing, suggesting that the importance of early life wheeze as a risk factor for wheeze/asthma in adolescence and adulthood should not be overemphasised. Future work should focus on identification of factors which differentiate branches at each node and their incorporation into work about predicting wheeze development.