Frequency of Acute Kidney Injury and Association With Mortality Among Extremely Preterm Infants

Abstract

ImportanceNeonatal acute kidney injury (AKI) is common and associated with morbidity and mortality. The temporal relationship between AKI and critical illness, as well as the frequency of AKI definition components (urine output and serum creatinine [sCr] concentration change), are unknown in extremely low-birth-weight (ELBW) (<1000 g), extremely preterm (<29 weeks’ completed gestational age [GA]) infants.ObjectiveTo measure the frequency of AKI from birth to death or discharge with attention to the definition components as well as the temporal relationship of AKI to critical illness and death.Design, Setting, and ParticipantsA single-center, multiyear, retrospective cohort study was conducted at an academic level IV neonatal intensive care unit between January 1, 2012, and January 1, 2020. Participants included inborn ELBW and infants at 22 to 28 weeks’ completed GA with confirmed congenital anomalies who survived 12 hours or more.ExposuresExtremely preterm birth and ELBW.Main Outcomes and MeasuresThe primary outcome was AKI frequency. The timing, severity, and criteria for AKI were measured. The temporal relationship between AKI, organ dysfunction, and outcomes were quantified using odds ratios (ORs), logistic regression, and Shapley Additive Explanations. Acute kidney injury recognition, imaging, pediatric nephrology consultation, and follow-up were determined.ResultsA total of 436 infants (52% male; 44% Black) met the inclusion criteria (median BW, 725 g; median GA, 25.7 wk). Acute kidney injury was common in the first week of life (44%), primarily based on the change in the sCr concentration criterion (88%), and negatively associated with GA (OR, 0.69; 95% CI, 0.60-0.78), but positively associated with antecedent critical illness (OR, 1.17; 95% CI, 1.12-1.23), severe intraventricular hemorrhage (OR, 1.86; 95% CI, 1.12-3.08), late-onset sepsis (OR, 1.03; 95% CI, 1.02-1.03), and mortality (OR, 2.77; 95% CI, 1.63-4.72). Acute kidney injury had negligible clinical contribution to death within the model (Shapley Additive Explanation, <0.5% change to outcome) relative to antecedent patient-concentration organ dysfunction (6%-15% change). Among infants with severe AKI, recognition (32%), nephrology inpatient consultation (16%), and outpatient follow-up (9%) were not common.Conclusions and RelevanceIn this cohort study of ELBW infants, AKI was common in the first week of life, inversely associated with GA, and followed organ (primarily cardiovascular) dysfunction. Acute kidney injury considered as the primary pathway to mortality was rare, and amelioration of AKI to modify death was not well supported.

Neonatal Acute Kidney Injury Association With Mortality—Culprit, Innocent Bystander, or Canary in the Coal Mine?

“…8 Further work to understand how different neonatal AKI endotypes (hypoperfusion, nephrotoxicity, sepsis, and ischemia) impact outcomes is greatly needed as outlined in recent consensus guidelines on pediatric AKI in JAMA Network Open in October 2022. 8 A third possible reason for the findings of Aziz et al 1 suggesting that AKI has a negligible influence on death is the possibility of biases and collinearity in the statistical approach. Machine learning processes are not a panacea and are still subject to bias through incorrectly classified variables and models or incorrectly collected data points.…”

“…In their study on acute kidney injury (AKI) in extremely preterm infants, Aziz et al 1 provide supportive information about the prevalence of AKI and its association with mortality in extremely low-birthweight (ELBW) neonates. Using a contemporary neonatal AKI definition, this retrospective, observational, single-center cohort study of 436 ELBW infants shows a prevalence of early AKI of 44%, and those with any AKI had a 2.77 (95% CI, 1.63-4.72) higher odds of death compared with those without AKI (odds ratio for severe AKI, 3.52; 95% CI, 2.15-5.76).…”

“…This study then addresses another critical question. 1 Does AKI directly cause death or is it an innocent bystander (ie, not contributory)? Using the Shapley Additive Explanations (SHAP) analysis, a machine learning approach that shows the relative association of each variable being measured with a given outcome, they found that, after controlling for vasoactive inotrope score (VIS) and the neonatal Sequential Organ Failure Assessment (nSOFA) measure, AKI was not associated with mortality.…”

“…This study then addresses another critical question . Does AKI directly cause death or is it an innocent bystander (ie, not contributory)?…”

“…Finally, recent studies require a minimum sCr level of 0.4 mg/dL to eliminate a diagnosis of AKI based on small changes that could be due to variability in laboratory error (ie, sCr increase from 0.2 to 0.3 mg/dL is a 50% increase) . The AKI definition used by Aziz et al comprised any prior sCr value (including day 0 and 1) to define the baseline, did not use minimum sCr values, and in most cases, met the sCr percent change criterion, resulting in very low peak sCr values overall in infants with mild AKI (mean sCr, 0.68 mg/dL; IQR, 0.32-1.08 mg/dL). For these reasons there is a possibility for misclassification bias.…”