Therapeutic monitoring of the pharmacologically active (free drug) fraction of protein-bound medications (e.g., phenytoin) represents a major diagnostic challenge in clinical and laboratory medicine. While free drug levels may be beneficial in many clinical situations, current methods for predicting free phenytoin concentrations are unreliable and not recommended for general use. The authors have demonstrated a linear relationship (r2 = 0.98) between serum levels of total and bound phenytoin in 56 patients with seizure disorders. No significant correlations were observed when total phenytoin and albumin levels were compared independently to measured concentrations of free phenytoin or percent free phenytoin. A good correlation (r2 = 0.89) existed between free phenytoin levels and [total phenytoin]/[albumin] ratios in patients with normal or elevated albumin levels, but significantly weaker correlations were found in patients with hypoalbuminemia. Thus, [total phenytoin]/[albumin] ratios may have clinical value in predicting free phenytoin levels in uncomplicated patients without hypoalbuminemia.