To determine whether red blood cell glucose-6-phosphate dehydrogenase (G6PD) activity is associated with retinopathy of prematurity (ROP). This case-control study was conducted in a Level-3 neonatal unit. Subjects were inborn boys with birth weight <2000g. "Cases" were consecutive subjects with ROP of any severity. "Controls" were consecutive unrelated subjects without ROP. Recipients of blood or exchange transfusions were excluded. Sixty cases (out of 98 screened) and 60 controls (out of 93 screened) were enrolled. G6PD activity (quantitative assay) as the candidate risk factor was evaluated. Sixty cases with 60 controls [mean (SD) gestation 28.80 (2.2) and 30.60 (2.2) wk respectively] were compared. "Cases" had a higher median (1st, 3rd quartile) G6PD activity compared to "controls" [7.39 (4.7, 11.5) vs. 6.28 (4.2,8.8) U/g Hb, p = 0.084]. G6PD activity was highest among ROP requiring treatment [8.68 (4.7, 12.3)] followed by ROP not requiring treatment [6.91 (4.4, 11.0)], followed by controls (plinear trend = 0.06). Gestation, birth weight, duration of oxygen, breastmilk feeding, and clinical sepsis were other variables associated with ROP on univariable analysis. On multivariable logistic regression, G6PD activity [Adjusted OR 1.14 (1.03, 1.25), p = 0.01] and gestation [Adjusted OR 0.74 (0.56, 0.97), p = 0.03] independently predicted ROP. C-statistic of the model was 0.76 (95% CI 0.67, 0.85). Higher G6PD activity was independently associated with ROP after adjusting for confounders. Each 1 U/g Hb increase in G6PD increased the odds of ROP by 14%. Severer forms of ROP were associated with higher levels of G6PD activity.