Secondary investigation using adjusted OR (aOR) and 95% CI for visual outcomes based on type of cerebral injury, relative to no cerebral injury
| Possible risk factor | Blindness n/N (%) | aOR* (95% CI) | P value |
| Reference group: no cerebral injury | 80/15 586 (0.5) | 1.0 | – |
| a. Ventricular size enlarged (with or without concurrent/prior blood) | 33/2141 (1.5) | 2.23 (1.49 to 3.33) | <0.0001 |
| b. Blood/echodensity in parenchyma with or without midline shift† | 0/200 (0) | N/A | |
| c. PVL | 15/294 (5.1) | 8.94 (4.60 to 17.4) | <0.0001 |
| d. Combination of ≥2 above cerebral injury criteria | 45/1149 (3.9) | 5.58 (3.33 to 9.36) | <0.0001 |
| e. Any of criteria A–D with shunt for hydrocephalus | 40/476 (8.4) | 12.7 (8.86 to 18.2) | <0.0001 |
*The adjusted ORs and p values are from a logistic regression model using GEE; the model adjusted for sex, birth weight, multiple birth, maternal race, type of cerebral injury (none; ventricular size enlarged only; PVL only; two or more of the following: ventricular size enlarged, PVL, blood/echodensity in the parenchyma; one of the above with shunt for hydrocephalus), severe ROP and centre as a cluster effect. The model does not include an interaction term between type of cerebral injury and severe ROP because we also assessed this interaction in a separate logistic regression model and found no significant interaction (p=0.15).
†Infants with blood/echodensity in parenchyma (only) were excluded from the regression since none of these infants had the outcome of blindness.
GEE, generalised estimating equation; PVL, periventricular leukomalacia; ROP, retinopathy of prematurity.