Sir: The authors of a recent paper (Reference Taylor, Starkey and GinaryTaylor et al, 2000) have drawn much-needed attention to the high prevalence of subtherapeutic treatment of conditions such as mood disorders with carbamazepine or valproate. However, the authors have given much emphasis to the use of drug serum levels in monitoring the use of both these drugs in such conditions.

Personal experience suggests that there are potential pitfalls for the unwary in the interpretation of valproate levels. In particular, in a minority of patients it may be impossible to reach the stated trough threshold level of 50 mg/l, even at daily doses in excess of the licensed maximum and despite supervised compliance.

One possible reason for this is that valproate is highly (about 90%) bound to plasma protein at low body levels, but with increasing doses the binding sites become saturated and free levels rise. More active drug is thereby delivered to the central nervous system (CNS) (Reference DolleryDollery, 1991) without in some cases more than a marginal rise in measured (bound plus unbound) serum levels. This effect should be borne in mind if measured levels do not rise in proportion to a dose increase in a particular case.

Other factors confounding the interpretation of valproate levels include the fact that: the effects in the CNS considerably outlive the detectable presence of the drug in the body, possibly because of active metabolites; there is variable displacement from serum binding sites by free fatty acids; blood samples taken just before a dose may not consistently represent trough levels, because of possible entero-hepatic recirculation and diurnal variation in elimination (Reference Chapman, Keane and MeldrumChapman et al, 1982; Reference DolleryDollery, 1991). These factors explain, at least in part, why valproate measurements tend to be unreliable for some clinical purposes.