Division of Chronic Disease Research Across the Lifecourse, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA, USA.
Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
Paediatr Perinat Epidemiol. 2021 Sep;35(5):569-577. doi: 10.1111/ppe.12760. Epub 2021 Mar 22.
Obesity and asthma in childhood often co-occur. Few studies have examined this relationship using repeated measures of body mass index (BMI) or asthma symptoms (such as wheeze).
We compared two analytic approaches for repeated measures data to investigate this relationship.
Our baseline sample consisted of 1277 children enrolled in a Boston-area cohort with BMI or wheeze at age 1 year and no missing covariates. We used latent class growth models (LCGM) and inverse probability weighting (IPW) of marginal structural models to examine the extent to which presence of overweight across childhood was associated with early adolescent current asthma, and conversely of repeated measures of wheeze across childhood with early adolescent obesity.
Using LCGM, a "persistent" childhood overweight class (vs "never") was associated with higher risk of asthma in early adolescence (RR 1.9; 95% CI 1.1, 3.0), while "persistent" childhood wheeze (vs "never") was associated with higher risk of obesity in early adolescence (RR 2.7; 95% CI 1.0, 6.4) after adjusting for baseline covariates. An IPW analysis treating childhood overweight and wheeze as time-varying exposures and adjusting for baseline and time-varying covariates resulted in weaker and less precise associations of "persistent" (vs "never") overweight with adolescent asthma (RR 1.3; 95% CI 0.3, 3.0), and of "persistent" (vs "never") wheeze with adolescent obesity (RR 2.3; 95% CI 0.4, 5.3).
Our point estimates from both approaches suggest an association between "persistent" childhood overweight and adolescent asthma, and between "persistent" childhood wheeze and adolescent obesity. LCGM results were stronger and more precise, whereas IPW results were less conclusive with wider 95% confidence intervals containing the null. The precision gained from LCGM may be at the expense of bias, and the use of both approaches helps to shed some light on this tradeoff.
儿童肥胖症和哮喘常同时发生。很少有研究使用身体质量指数（BMI）或哮喘症状（如喘息）的重复测量来检验这种关系。
我们比较了两种用于重复测量数据的分析方法，以调查这种关系。
我们的基线样本包括 1277 名儿童，他们参加了波士顿地区的一个队列研究，在 1 岁时测量了 BMI 或喘息，且没有缺失的协变量。我们使用潜在类别增长模型（LCGM）和边际结构模型的逆概率加权（IPW）来检验儿童期超重与青少年早期当前哮喘之间的关系，反之，儿童期喘息的重复测量与青少年早期肥胖之间的关系。
使用 LCGM，与“持续”儿童超重（与“从不”相比）相比，“持续”儿童期超重与青少年早期哮喘的风险较高（RR 1.9；95%CI 1.1，3.0），而“持续”儿童期喘息（与“从不”相比）与青少年早期肥胖的风险较高（RR 2.7；95%CI 1.0，6.4），调整了基线协变量。一项将儿童期超重和喘息视为时间变化暴露，并调整了基线和时间变化协变量的 IPW 分析，导致“持续”（与“从不”相比）超重与青少年哮喘（RR 1.3；95%CI 0.3，3.0）以及“持续”（与“从不”相比）喘息与青少年肥胖（RR 2.3；95%CI 0.4，5.3）之间的关联更弱且更不精确。
我们从两种方法得到的点估计值都表明，“持续”儿童超重与青少年哮喘之间存在关联，“持续”儿童喘息与青少年肥胖之间存在关联。LCGM 结果更有力和更精确，而 IPW 结果则不太确定，95%置信区间更宽，包含了无效结果。LCGM 获得的精度可能是以偏倚为代价的，同时使用两种方法有助于阐明这种权衡。
