Joint modelling of wheeze and lung function from childhood to early adulthood: four population-based birth cohorts

Wheeze and lung function (LF) during childhood are key indicators of respiratory health, yet their trajectories are usually examined separately. We aimed to identify joint developmental patterns of wheeze and LF.

We used data from four unselected birth cohorts established between 1989 and 1996 with repeated assessments of wheeze from infancy and spirometry from early school-age to early adulthood. We used group-based multi-trajectory modelling to derive trajectories based on joint modelling of current wheeze and forced expiratory volume in 1 s/forced vital capacity ratio (FEV₁/FVC).

In the discovery analysis (n = 4645), we identified 6 trajectories: (1) Never/infrequent wheeze with normal LF (NIFW-NLF, 2925/4645 [62.97%]); (2) Never/infrequent wheeze with reduced LF (NIFW-RLF, 475/4645 [10.22%]); (3) Early-transient wheeze with normal LF (ETW-NLF, 559/4645 [12.03%]); (4) Late-onset wheeze with NLF (LOW-NLF, 335/4645 [7.21%]); (5) Persistent wheeze with NLF (PEW-NLF, 202/4645 [4.34%]); and (6) PEW with RLF (PEW-RLF, 149/4645 [3.21%]). Risk profiles of two trajectories characterised by persistent wheeze but differentiated by normal or reduced LF differed significantly. Elevated fractional exhaled nitric oxide (FeNO) and allergic sensitisation were highly prevalent in both, but only PEW-RLF was significantly associated with perinatal and early-life factors/exposures (prematurity; lower gestational age: RRRs [95% CI] 2.21 [1.49-3.28], low birth weight: 2.60 [1.47-4.60]: and exposure to smoking during gestation: 2.00 [1.49-2.63]). Two low lung function trajectories (with and without symptoms; PEW-RLF and NIFW-RLF) had similar LF impairment, but divergent clinical and risk factor profiles. PEW-RLF was associated with high rates of asthma diagnosis, high FeNO, bronchodilator reversibility, and family history of atopy. In contrast, those in NIFW-RLF trajectory had no elevation in inflammatory biomarkers and low prevalence of airway hyperreactivity, and were characterised by much higher rates of prenatal tobacco smoke exposure, and greater active adolescent smoking (2.06 [1.41-3.01]), and higher body fat mass in adolescence (1.02 [1.01-1.03], p = 0.01) with no difference in birth weight or preterm birth. Replication analyses in independent cohorts (n = 3388) were consistent with the discovery.

The disconnect between symptoms and lung function, along with the differences in risk profiles, has important implications for respiratory health intervention strategies.

UK MRC grant MR/S025340/1.