To the Editor: Effective treatment of bipolar disorder with valproic acid is associated with serum levels of 45–125 μg/ml. Lower levels are associated with reduced clinical response, and higher levels with adverse events (1). Toxicity above 100 μg/ml is related to saturation of protein binding sites and nonlinear increases in free valproate (2). Although increased unbound valproic acid has been associated with toxicity in hypoalbuminemic patients (3), there are no reports, to our knowledge, of toxicity associated with free valproate in psychiatric patients. We report the case of a patient who experienced toxicity with increased unbound valproic acid and normal serum valproate and albumin.
Mr. A was a 49-year-old single Caucasian man who was admitted to a state hospital with a diagnosis of schizoaffective disorder, manic. His other disorders included type II diabetes mellitus and hyperlipidemia that were treated with metformin, 850 mg b.i.d., atorvastatin, 20 mg at bedtime, and regular insulin coverage. Vivid delusions and hallucinations were refractory to treatment with multiple antipsychotic agents. Agitation, irritability, and mood swings led to the addition of divalproex sodium. A regimen of 56 mg/day of perphenazine, 2000 mg/day of extended-release divalproex sodium, 6 mg/day of lorazepam, 50 mg/day of sertraline, and 2 mg/day of benztropine did not lead to symptom abatement or, initially, altered cognition or sensorium. After 2 months, the hospital staff noted an ataxic gait, confusion, and slurred speech. Reducing Mr. A’s perphenazine dose yielded no benefit. Tapering his lorazepam and benztropine led to modest sensorium clearing; however, after 5–6 days, the staff noted increasing confusion, disorientation, and ataxia. His serum valproic acid level was 86.1 μg/ml (normal range=50–125), his albumin level was 3.8 g/dl (normal range=3.2–5.2), and his liver function tests, ammonia levels, and platelet count were within normal limits. His free valproic acid was found to be 15 μg/ml (normal range=5–10), so his dose of extended-release divalproex sodium was reduced to 1000 mg/day, and his sensorium returned to normal in approximately 3 days.
"Normal" serum valproic acid ranges anticipate typical protein binding; however, atypical saturation kinetics could result in increased free valproic acid disparate from that suggested by serum total concentrations (3, 4). No specific factors appeared to account for the patient’s unexpectedly high free valproic acid level. His medical conditions and medications are not known to influence valproate binding. Our tapering of lorazepam (the levels of which may increase when valproate is added) did not resolve his cognitive impairments, which did respond to reduced valproate. Measurement of free valproate should be considered in patients with unexplained altered cognition, even when protein levels are apparently normal.