Abstract

Background: Probiotic supplementation is among the most widely adopted preventive interventions in neonatal intensive care, on the strength of metaanalyses reporting a roughly halved risk of necrotising enterocolitis (NEC). The certainty of that evidence has been graded low, and the largest synthesis reports funnel plot asymmetry consistent with publication bias. Methods: Systematic reviews with meta-analysis of randomised trials in very preterm or very low birth weight infants reporting a pooled risk ratio for NEC were eligible. Review-level risk ratios were pooled on the log scale in a DerSimonian- Laird random-effects model with Cochran’s Q, I², ² and 95% prediction intervals. The all-trials and low-risk-of-bias estimates reported within the largest review were compared by a ratio of risk ratios. Secondary outcomes were extracted as reported. Analyses were performed in Python 3. Results: Three reviews contributing NEC estimates and representing 81 trials in 19,221 infants were pooled. Probiotics reduced NEC: pooled risk ratio 0.51 (95% CI 0.43-0.60; z = –7.84, p < 0.001), with little heterogeneity (Q = 2.22, df = 2, p = 0.330; I² = 9.7%) and a 95% prediction interval of 0.33-0.79. Restricting the largest synthesis to trials at low risk of bias moved the risk ratio from 0.54 (0.46-0.65) to 0.70 (0.55- 0.89)—a loss of approximately 35% of the relative risk reduction—and raised the number needed to treat from 33 to 50 (ratio of risk ratios 1.30, 95% CI 0.96- 1.74, p = 0.086). The effect attenuated steadily as outcomes broadened: NEC 0.51, fungal sepsis 0.57, all-cause mortality 0.77, bacterial sepsis 0.82, any lateonset sepsis 0.89. Probiotic sepsis occurred in 8 of 20,323 exposed infants (0.04%). Conclusions: Probiotics reduce necrotising enterocolitis in very preterm infants, but the headline halving overstates what the best-conducted trials support: roughly a third of the relative effect disappears under restriction to low-riskof-
bias trials, and the mortality benefit is smaller than the NEC benefit. The case for routine use rests less on the precision of the effect estimate than on an exceptionally favourable benefit-to-harm ratio, with fewer than one probiotic sepsis case per twenty thousand infants exposed.

Keywords: Probiotics, Necrotising enterocolitis; preterm infant, Risk of bias, Publication bias, Number needed to treat, Second-order meta-analysis