This study examined a cohort of 2,500 patients at a Swiss hospital and identified 23 patients with hypoalbuminemia (albumin ≤35 g/L). These patients were divided into two groups: mild hypoalbuminemia (albumin 25–35 g/L; 11 patients) and severe hypoalbuminemia (albumin <25 g/L; 12 patients). All patients underwent testing for total and free phenytoin sodium levels, the former being measured by the EMIT method and the latter by HPLC. Free phenytoin levels were calculated using the Sheiner–Tozer formula shown in Figure 1. These results were then compared with the actual measured free phenytoin data using statistical analyses, including Spearman's correlation, Passing–Bablok regression and Bland–Altman plot analysis.

Figure 1，The Sheiner–Tozer formula. Calculating free phenytoin requires knowledge of the total phenytoin concentration and the mean serum albumin level.

The results showed that the measured and calculated values of free phenytoin generally agreed closely. The mean measured and calculated values were 1.1 and 1.2 mg/L, respectively, and the correlation coefficient was r = 0.907 (P < 0.001), indicating a nearly perfect correlation. When patients with mild and severe hypoalbuminemia were analyzed separately, no difference was observed between the calculated and measured free phenytoin results, regardless of hypoalbuminemia severity. The difference was 0.10 mg/L in the mild group and 0.13 mg/L in the severe group, with a P-value of 0.78 from the t-test, indicating no statistically significant difference. Regression and consistency analysis revealed that the calculated values were equivalent to the measured values (Passing-Bablok regression, Figure 2), with a slope close to 1 and an intercept close to 0. The Bland–Altman plot (Figure 3) revealed a mean bias of only −0.11 mg/L with a narrow 95% confidence interval, indicating no systematic or proportional bias. Thus, in patients with low albumin levels, free phenytoin sodium levels calculated using the Sheiner–Tozer formula showed high agreement with gold-standard measured results.

Figure2，Passing–Bablok regression analysis shows that the calculated and measured free concentrations are comparable.

Figure 3，The Bland–Altman plot showed that the mean difference between the calculated and measured free concentrations was −0.11 (SD = 0.28), with no statistically significant difference from zero (t(22) = 0.07).

The study demonstrated that the Sheiner–Tozer algorithm can be safely and accurately used to estimate free phenytoin sodium concentrations in patients with hypoalbuminemia in the absence of direct measurement, thereby supporting precise clinical dose adjustment.

For critically ill or hypoalbuminemia patients, this implies that it is possible to rapidly assess true drug efficacy using total concentration and albumin levels without relying on expensive free phenytoin sodium testing. This is particularly beneficial for patients receiving rapid loading doses in the emergency department or during surgery, as it reduces waiting times and the risks of overdose or underdosing.