Dear Editor, W e read with interest the study by Abdelghani et al.,1 reporting that high-dose erythropoietin (EPO; 1000 U/kg on days 1, 2, 3, 5, and 7) significantly reduced the composite of death or neurodevelopmental impairment (NDI) at 24 months (16.6% vs. 50%; odds ratio OR: 0.07, 95% confidence interval CI: 0.03–0.83; P = 0.038) and magnetic resonance imaging (MRI)-defined brain injury (29.1% vs. 77.2%; P = 0.039) in neonates with hypoxic–ischemic encephalopathy (HIE). While the work is regionally significant, several statistical and methodological concerns warrant clarification. Multiple Comparisons and Absence of Correction This study simultaneously tests numerous outcomes—mortality, NDI, brain injury, seizures, and multiple biomarkers—using a fixed α value of 0.05 with no adjustment for multiple comparisons described. In the absence of a Bonferroni correction or Benjamini–Hochberg false discovery rate control, the probability of at least one spurious significant finding increases substantially, which may affect the interpretation of borderline P values reported across secondary outcomes. Imprecise Effect Estimates and Absent Clinical Metrics The reported OR of 0.07 (95% CI: 0.03–0.83) spans nearly two orders of magnitude, reflecting marked statistical imprecision attributable to small group sizes. Furthermore, the authors do not report absolute risk reduction (ARR), relative risk reduction, or number needed to treat (NNT)—metrics indispensable for clinicians translating trial findings into bedside practice. Based on the published data, the approximate NNT to prevent one death or NDI event is 3 (ARR ≈ 33.4%), a clinically compelling figure that the authors have omitted. Outcome Definition Discrepancy Cognitive impairment is defined as a Bayley Scales of Infant and Toddler Development-III (Bayley-III) cognitive score of <90. However, the prevailing evidence recommends a threshold of <85 (−1 standard deviation from the normative mean) as the validated benchmark for clinically meaningful neurodevelopmental delay.2 Employing a higher cutoff may increase the proportion classified as NDI across both arms and could influence the apparent treatment effect. Missing Neuromonitoring and Unresolved Data Gaps Amplitude-integrated electroencephalography —now a standard-of-care neuromonitoring tool in term infants with moderate-to-severe HIE, with established prognostic accuracy—was not utilized.3,4 Additionally, 20% of enrolled infants (28/140) did not undergo follow-up MRI; the absence of any formal missing data framework (e.g., multiple imputation) or sensitivity analysis, which may introduce attrition bias in the neuroimaging outcomes. No inter-rater reliability statistic (Cohen’s κ or intraclass correlation coefficient) is reported for the two independent radiologists who scored the MRIs, despite this being a requisite standard for imaging outcome studies. Design and Comparator Context The nonrandomized allocation—driven by clinician discretion—produced significant baseline imbalances: the EPO group had lower 1-min Apgar scores (median: 3 vs. 5; P = 0.047) and higher pCO2 (67.7 vs. 57.6 cmH2O; P = 0.008), yet paradoxically achieved superior outcomes. Propensity score methods could have strengthened the adjustment for baseline imbalance beyond logistic regression alone. The unequal group ratio (96:44) further reduces statistical power in the control arm. These concerns are amplified by the null findings of the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) trial (n = 500), where identical EPO dosing added to therapeutic hypothermia yielded no improvement in composite 2-year outcomes (52% vs. 43%; P = 0.08).5 The divergence between these results warrants further discussion. A recent meta-analysis of 10 RCTs does support a modest EPO signal (relative risk: 0.62, 95% CI: 0.48–0.81),5 lending qualified optimism, but the evidence hierarchy demands randomized replication. We urge the authors to report NNT and ARR, apply an appropriate correction for multiple testing, justify their Bayley-III threshold relative to published norms, address the missing MRI data with formal imputation, and provide inter-rater κ statistics. These additions would substantially strengthen the interpretability and credibility of an otherwise clinically important contribution. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest. Author contributions AC performed writing the original draft. RK performed the editing of the final draft, and KG performed the supervision.
CHHABRA et al. (Fri,) studied this question.