Probability‐Based Compatibility Curves for Calcium and Phosphates in Parenteral Nutrition Formulations

Abstract

The information content of the calcium phosphate compatibility curves for adult parenteral nutrition (PN) solutions may benefit from a more sophisticated statistical treatment. Binary logistic regression analyses were evaluated as part of an alternate method for generating formulation compatibility curves. A commercial PN solution was challenged with a systematic array of calcium and phosphate concentrations. These formulations were then characterized for particulates by visual inspection, light obscuration, and filtration followed by optical microscopy. Logistic regression analyses of the data were compared with traditional treatments for generating compatibility curves. Assay-dependent differences were observed in the compatibility curves and associated probability contours; the microscopic method of precipitate detection generated the most robust results. Calcium and phosphate compatibility data generated from small-volume glass containers reasonably predicted the observed compatibility of clinically relevant flexible containers. The published methods for creating calcium and phosphate compatibility curves via connecting the highest passing or lowest failing calcium concentrations should be augmented or replaced by probability contours of the entire experimental design to determine zones of formulation incompatibilities. We recommend researchers evaluate their data with logistic regression analysis to help build a more comprehensive probabilistic database of compatibility information.