Prospective Cohort Study for a
Childhood Asthma Question

RESEARCH QUESTION

What the Study Set Out to Ask

This prospective cohort study example follows a pediatric epidemiology team that has followed a birth cohort of children to age eight. The team asks whether antibiotic treatment in the first year of life is associated with a later diagnosis of asthma, a question that a randomized trial cannot realistically answer.

Antibiotics are often prescribed for early respiratory symptoms that may already be the first sign of asthma, so a simple comparison can show an association that runs backward. The question is therefore framed as an association, to be estimated with its uncertainty. The plan fixes the causal reasoning, the exposure window and the outcome definition before any asthma diagnoses are linked to exposure.

STUDY DESIGN AND DATA

A Birth Cohort of About 3,000 Children

A prospective birth cohort of about 3,000 children recruited at birth and followed to age eight, with questionnaires and clinic visits at fixed ages and linked primary care prescription records for the first year of life. About a quarter of the children are lost to follow-up before age eight, and a few hundred children have a sibling in the cohort. The dataset holds antibiotic courses with dates, age at first doctor-diagnosed asthma, parental asthma, maternal smoking, birth mode, siblings, daycare attendance and early wheeze episodes.

METHOD AND ASSUMPTIONS

Why a Causal Diagram, and What Must Hold

A causal diagram drawn before the analysis decides which variables are adjusted for. Parental asthma, maternal smoking, birth mode and siblings are confounders to adjust for, while variables that follow the exposure and lie on the path to asthma, such as later wheeze, are left out. Exposure is defined from antibiotic courses in the first year with a lag: courses given within a stated period before a child's first recorded wheeze do not count, so that treatment of early symptoms is not counted as exposure, and a longer lag is run as a sensitivity analysis to guard against reverse causation. Follow-up for asthma starts at age one, when the exposure window closes, so no child is classified by antibiotics received after follow-up has begun. Incidence is reported as rates per 1,000 person-years, a different quantity from the cumulative risk by age eight and from a hazard ratio, and each is labeled as such.

The primary model is a time-to-event model with age as the time scale, and proportional hazards are checked with scaled Schoenfeld residuals read in context. Children lost to follow-up are compared with those retained, and inverse probability weights built from baseline and early-life predictors of dropout make the analysis represent the whole cohort, under the assumption that dropout depends on those predictors and not on later asthma status once they are known. Standard errors allow for siblings through family-level clustering.

OUTPUT AND INTERPRETATION

How the Results Are Read

The output has a flow diagram from the children recruited to those analyzed, a table of baseline characteristics by exposure, incidence rates per 1,000 person-years by exposure group with 95% confidence intervals, cumulative incidence curves, and the adjusted hazard ratio with its 95% confidence interval beside the unadjusted one. The causal diagram is included so that a reader can see why each variable is, or is not, in the model.

Sensitivity analyses cover the longer exposure lag, the weighted and unweighted estimates, and a negative control outcome that the exposure should not affect, such as an injury diagnosis, to look for residual confounding by health-care use. The interpretation is worded as an association in this cohort, states the confounding and reverse causation that the design cannot rule out, and makes no recommendation about prescribing. Whether any finding should change practice is a decision for clinicians and guideline groups.

REPORTING

What the Researcher Received

A written report structured to the STROBE checklist for cohort studies, with statistical wording for the methods, matched to the output, for the team to check and incorporate. The research question, clinical interpretation and conclusions remain with the research team.