The isoxazolyl penicillins, including flucloxacillin, have the highest levels of plasma protein binding among the semisynthetic penicillins. Because only the free fraction of the penicillin is pharmacologically active, it would be useful to measure both protein-bound and free flucloxacillin to determine its protein binding. Until now, flucloxacillin protein binding in newborn infants has been investigated in only two studies with relatively small populations. In the present study, flucloxacillin protein binding was investigated in 56 (preterm) infants aged 3 to 87 days (gestational age, 25-41 weeks). Surplus plasma samples from routine gentamicin assays of each infant were collected and combined to obtain a sufficiently large sample for analysis. Free flucloxacillin was separated from protein-bound flucloxacillin using ultrafiltration. Reversed-phase high-performance liquid chromatography with ultraviolet detection was used to measure free flucloxacillin concentrations in ultrafiltrate and total flucloxacillin concentrations in pooled plasma. Flucloxacillin protein binding was 74.5% +/- 13.1% (mean +/- standard deviation) with a high variability among the infants (34.3% to 89.7%). High Pearson correlations were found between protein binding and the covariates-plasma albumin concentration (r = 0.804, P < 0.001, n = 18) and plasma creatinine concentration (r = -0.601, P < 0.001, n = 45). Statistically significant but less striking correlations were found between protein binding and gestational age, postconceptional age, body weight, and triglyceride concentration. Because of the high variability of protein binding among infants, it is difficult to devise a flucloxacillin dosage regimen effective for all infants. Individualized dosing, based on free flucloxacillin concentrations, might help to optimize treatment of late-onset neonatal sepsis, but practical obstacles will probably prevent analysis of free flucloxacillin concentrations in newborn infants on a routine basis.