We read with interest the manuscript by Gotru et al., entitled: ‘The absolute nucleated red blood cell (aNRBC) count at birth is not an indicator for retinopathy of prematurity (ROP)’.1 In their article, Gotru et al. aimed to ‘establish the reproducibility’ of our published observation ‘that infants affected by retinopathy of prematurity (ROP) had higher absolute nucleated red blood cell (aNRBC) counts than those unaffected’.2 Contrary to our findings, Gotru et al. concluded that they were ‘unable to support the previously reported relationship of aNRBCs with ROP’. Moreover, they suggested that their findings were somewhat more credible than ours, because their ‘population was three times larger’.
From our reading of Gotru et al.’s paper, we are less sure that our findings can be convincingly discarded. On the contrary, several obvious flaws in the design of Gotru et al.’s paper make their findings less convincing. To point out some of the most important ones:
-
1)
Known confounding variables that affect the number of aNRBC at birth include gestational or insulin-dependent diabetes; pregnancy-induced hypertension; intrauterine growth restriction; placenta abruption or placenta previa; maternal heart, kidney or lung chronic conditions; drug, tobacco or alcohol abuse; perinatal infection including chorioamnionitis; fetal distress; perinatal blood loss; and hemolysis and chromosomal anomalies (all referenced in Lubetzky et al.2). In our study, all the above-mentioned conditions were separate criteria of exclusion. In contrast, in Gotru et al.’s study, some of these conditions were only recorded (‘extracted’), but not excluded. The only infants that were excluded, among those who may have significantly biased the results were those who died, those who were small for gestational age (GA) and those with congenital anomalies. Moreover, the authors did not even attempt to study the effect of these potential confounders upon the aNRBC count in a multivariate type of analysis.
-
2)
In our study, matching was carefully effected in order to reach two groups, one with ROP, and the other one without that would be as similar as possible, in order to insure comparability. Each ROP child was matched systematically with the infant admitted immediately after him/her, who did not develop ROP and had the same GA (±1 week) and who had the same 1 and 5 min Apgar scores (±1). In their study, Gotru et al. matched by birth weight and GA. Apgar scores, a fundamental clinical estimate of intrauterine well being, were totally ignored in the matching system, and not even mentioned in the analyses of the group’s comparability, or in a subsequent multivariate analysis, that was not conducted.
-
3)
In their article, Gotru et al. mentioned that during the study period, a strict protocol of ROP screening consistent with the 2001 American Academy of Pediatrics guidelines was followed. We wonder how strict was the protocol because out of 7564 patients in the data warehouse, 419 (5.5%) had a missing ROP diagnosis.
-
4)
In our study, all aNRBC counts were obtained within 1 h of birth in all infants. aNRBC counts are known to drop over the first few hours and days of life, and are in most cases undetectable by day 2–5 of life. Gotru et al. did not mention when the samples were drawn, and it is likely that the retrospective approach from a data warehouse will not allow to easily determine the age of the samples.
This is a preview of subscription content, access via your institution