Answers are grounded in this paper and cite their sources.
Low gestational age (GA) and birthweight (BW), the strongest risk factors for retinopathy of prematurity (ROP), are routinely adjusted for without consideration of variable structure. We explored how to model them, using sepsis, oxygen exposure, and small for gestational age (SGA) as illustrative exposures. We analyzed data from the prospective ELGAN cohort (23–27 weeks’ gestation; 14 US hospitals, 2002–2004), excluding infants without eye examination or missing data, leaving 1,259 of 1,506 infants. Any ROP was modeled by logistic regression, with GA and BW as continuous, categorical, or indicator variables. Parameterizations were compared by likelihood ratio test and Akaike information criterion; effect modification was assessed by interaction terms. ROP prevalence was 67%, declining from 91% at 23 to 47% at 27 weeks. A linear GA term fitted less well than week-specific indicators (p = 0.03). The crude oxygen OR, 4.3 (3.1–6.0), decreased to 2.2 (1.4–3.5) after GA adjustment, but exposure was confined to the lower GA group (331/638 vs 10/621). Sepsis estimates were stable (1.6 to 1.5; interaction p = 0.2). SGA was associated with ROP (1.8, 1.3–2.7), but only in the more mature half (3.0, 1.9–4.7 vs 1.1, 0.6–2.0; interaction p = 0.01). GA and BW are not interchangeable proxies for immaturity; uncritical adjustment can attenuate, mask or reverse associations. As screening and prevention rely on such estimates, studies should state the assumed role of GA and BW, justify their functional form and report GA-specific estimates.
No takes yet. Share an insight, caveat, or question.
Dieken et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: