Adherence was not usually defined in the published studies, but referred to generally as patients following medical recommendations. Authors generally considered adherence in behavioural terms, whereby the patient had an active and informed role to play in a therapeutic situation (13,20). In this sense, adherence to prescribed medication was seen as a health-promoting behaviour (21).

The types of nonadherence were described as follows: reduced or increased amount of single dose; decreased or increased number of daily doses; extra dosing; incorrect dosing intervals; being unaware of the need for life-long regular medication; taking duplicate medication; taking discontinued medication; discontinuing prescribed medication; regularly forgetting to take medication, and incorrect use of medication (18,20,22).

Medication use was assessed by review of medical records; patient self-report; family report; pill counts; prescription refill rates, and biological markers, including serum, urine and saliva assays to quantify medications or their metabolites (2,11,12,14,23 - 26). The best indicator of adherence is believed to be serum levels of anticonvulsant drugs (18,27). Other methods of monitoring adherence, such as electronic measures are not discussed further here because of the lack of published studies in this area. In several studies, patients whose serum levels were outside the therapeutic range were classified as nonadherent (19,23,28). However, serum levels are not a perfect measure. Although blood levels of anticonvulsant medications can be measured, it is difficult to translate them into comparable measures of adherence for patients on different medications and doses. Furthermore, sub-therapeutic levels of a drug in the serum can be due either to poor compliance or the need for a higher dosage (2). Patients with impaired absorption or rapid or ultra-rapid metabolism can have low serum levels even if their intake of AEDs is regular and according to prescription (11,26,29).

Dowse et al. and Leppik et al. reported that indirect measures such as patient interview, tablet counts and prescription refill records gave no indication of the true amount of the drug present in the body and could be inaccurate or biased (18,19). However, using the measurement of drug concentration in blood alone, except in cases of extremely low adherence and variability of drug intake, is not sufficient to detect incorrect drug intake. Therefore, the use of clinical markers and self-reported adherence should also be considered (11).