Fig. 3.
Post hoc analysis of serum electrolyte concentrations during the initial 72 postoperative hours: (A) sodium, (B) chloride, (C) blood urea nitrogen, (D) potassium, (E) bicarbonate, (F) creatinine. Lines and bands represent means and 95% confidence intervals, respectively. Plots were generated with the use of locally weighted scatterplot smoothing. The P values in the figure represent the overall difference between the two study groups, calculated with the use of linear mixed models (for modeling purposes, we assumed linearity despite of curvilinear smoothing in the plots), which allowed adjusting for the correlation present among repeated measurements obtained from a given patient (using unstructured correlation). Over time, the separation between groups changed for sodium (P < 0.001 for interaction) and chloride (P < 0.001 for interaction); interaction terms for the other variables were not significant (significance criterion for interaction was P > 0.10). All serum electrolyte concentration measurements before surgery were similar for the two study groups (table 1). LR, lactated Ringer’s.