Parenteral Nutrition and Electrolyte Balance

Evidence

This single-center retrospective observational cohort study examined preterm neonates admitted to the neonatal intensive care unit at Policlinico Umberto I, Sapienza University of Rome. Eligibility included gestational age at or below 34 weeks and/or birth weight at or below 1,500 g, with parenteral nutrition received for at least three days between January 2015 and December 2024. The comparison involved sequential practice cohorts: individualized parenteral nutrition, compounded by the hospital pharmacy, and commercially available standardized parenteral nutrition, Numeta G13%E. The unit changed from individualized to standardized parenteral nutrition in November 2022 for institutional logistic and organizational reasons, without other stated modifications to clinical protocols. The individualized cohort included 362 neonates and the standardized cohort included 33 neonates. pmid:42898319.

The investigators collected prenatal, perinatal, and postnatal information prospectively with a standardized data-collection form and extracted the data from a prospectively maintained neonatal intensive care unit database. Nutrition information was collected daily and included parenteral and total fluid volume; sodium, potassium, calcium, and phosphorus delivery; calcium-to-phosphorus ratio; cumulative parenteral protein intake; and the proportion of enteral intake from own mother’s milk or preterm formula. During the first week of life, laboratory collection included serum sodium, potassium, ionized calcium, phosphate, and acid–base status. Ionized calcium came from routine blood-gas samples; total calcium was not assessed to avoid additional blood sampling. pmid:42898319.

The study compared electrolyte disturbances, nutritional intakes, and morbidities between the two cohorts. It reported no significant differences in baseline clinical characteristics between groups. Standardized parenteral nutrition was associated with significantly lower sodium and calcium intakes during the first week of life than individualized parenteral nutrition. This intake pattern is important to the reported comparison because the study evaluated both the nutritional exposures delivered during early neonatal care and the subsequently observed laboratory-defined disturbances, rather than treating formulation type as an isolated exposure without nutrition data. The reported measures nevertheless reflect care in this particular neonatal intensive care unit and its sequential transition in practice. pmid:42898319.

For adjusted analyses, the authors built four separate binary logistic-regression models to identify independent risk factors for hyponatremia, hyperkalemia, hypocalcemia, and hyperphosphatemia. Each outcome model used a specific set of clinically relevant covariates, and categorical variables were compared with chi-square testing. Because the hyponatremia model had quasi-complete separation, it was also re-estimated with Firth-penalized logistic regression as a sensitivity analysis. In multivariable analysis, standardized parenteral nutrition was independently associated with hyponatremia, with an odds ratio of 38.9 and p < 0.001. Standardized parenteral nutrition was also associated with hyperkalemia, hypocalcemia, and hyperphosphatemia in the respective multivariable models. pmid:42898319.

The outcome pattern reported here is an association between the standardized formulation and several electrolyte or mineral disturbances in neonates receiving parenteral nutrition. It does not establish that the formulation caused those disturbances. The article also describes the broader comparative literature on individualized and standardized parenteral nutrition as inconclusive regarding safety and superiority. Thus, the study supplies a detailed single-unit observational comparison of early nutritional intake and laboratory outcomes, not a randomized comparative assessment of the two formulations. pmid:42898319.

Analysis — Observational association and interpretation

The sequential design makes the comparison clinically recognizable but analytically constrained. The cohorts were formed by a unit-wide change in parenteral nutrition practice rather than random allocation, so formulation type is inseparable from calendar period and from unmeasured differences that may have accompanied care over time. Prospectively recorded clinical and nutrition data, daily first-week laboratory measures, and outcome-specific multivariable models strengthen the description of the observed pattern. However, regression adjustment addresses only the covariates entered into each model; it cannot remove selection processes or confounding factors that were not measured or modeled. pmid:42898319.

The reported lower sodium and calcium intakes in the standardized group provide context for the electrolyte and mineral findings, but they do not convert the association into a causal mechanism. Likewise, the very large adjusted association reported for hyponatremia should be read alongside the model’s quasi-complete separation and its additional Firth-penalized sensitivity analysis. The findings are consistent with the study’s retrospective, non-randomized selection process: in this neonatal intensive care unit, disturbances were observed more often in the later standardized-parenteral-nutrition cohort after adjustment under the specified models. Statistical significance does not by itself establish clinical significance. pmid:42898319.

Limitations

The retrospective design cannot establish causation. This was a single-center, non-randomized study of sequential cohorts, which limits generalizability beyond the neonatal intensive care unit studied and leaves the comparison vulnerable to selection bias and temporal confounding. The substantially unequal cohort sizes may affect statistical power. Exclusion of neonates who lost central venous access for more than 72 consecutive hours when that prevented continuation of parenteral nutrition also defines the analyzed population and may limit applicability to neonates with interrupted parenteral access. pmid:42898319.

Uncertainty differs across outcomes. The hyperkalemia comparison had post-hoc power of 55.7% and was characterized by the authors as hypothesis-generating. The laboratory assessment measured ionized calcium from routine blood-gas samples, whereas total calcium was not assessed; interpretation should therefore remain specific to the measured calcium parameter. Finally, the study measured nutritional intake, electrolyte disturbances, and morbidities during early life, but its observational structure does not resolve whether the associations reflect the parenteral nutrition formulation, associated intake differences, the sequential care context, or residual confounding. pmid:42898319.