Reference percentiles for paired arterial and venous umbilical cord blood gases: An indirect nonparametric approach

Background: Reference intervals for arterial and venous umbilical cord blood gas (UCBG) parameters are scarce, are mainly focused on pH, pO2, pCO2 and base deficit, and are usually assessed using parametric tests, despite a generally skewed data distribution. Here, the purpose is to determine reference percentiles for nine parameters of concomitant arterial and venous UCBG (CAV-UCBG) from neonates at birth, using nonparametric tests. Methods: Results of CAV-UCBG, assayed over a 4.5-year period, were extracted from a hospital laboratory database for pH, pCO2, pO2, oxygen saturation, concentration of total oxygen, total carbon dioxide, hydrogen carbonate, total haemoglobin, and acid-base excess. Exclusion criteria were: a venous-arterial pH difference < 0.02, an arterial-venous pCO2 < 0.7 kPa, and a venous pCO2 < 2.9 kPa. Nonparametric bivariate kernel density estimations were used for the selection of plots within the 95% percentile surface of the pCO2-to-pH relationship (NBKDE-95P). Outliers from skewed data were removed using an adjusted-Tukey method, and percentiles were calculated according to the CLSI EP28-A3 nonparametric method. Results: Overall, 31% (5033/16164) of CAV-UCBG were discarded using the three exclusion criteria. Then, 6% (670/11131) of CAV-UCBG were excluded from the NBKDE-95P, and 0.1 to 3.5% outliers were subsequently removed. Depending on the parameter, the 2.5th and 97.5th percentiles from the whole group were similar or slightly narrower compared to reference intervals from other studies, while those from female and male neonates did not differ substantially. Conclusions: Using an indirect nonparametric approach, this study proposes new percentiles for parameters from concomitant arterial and venous umbilical cord blood gases.