Anticonvulsant drugs for acute and chronic pain

Anticonvulsant drugs have been used in the management of pain since the 1960s. The clinical impression is that they are useful for chronic neuropathic pain, especially when the pain is lancinating or burning.

To evaluate the analgesic effectiveness and adverse effects of anticonvulsant drugs for pain management in clinical practice and to identify a clinical research agenda. Migraine and headache studies are excluded in this revision.

Randomised trials of anticonvulsants in acute, chronic or cancer pain were identified by Medline (1966-1999), Embase (1994-1999), SIGLE (1980-1999) and the Cochrane Controlled Trials Register (CENTRAL/CCTR) (Cochrane Library Issue 3, 1999). In addition, 40 medical journals were hand searched. Additional reports were identified from the reference list of the retrieved papers, and by contacting investigators. Date of most recent search: September 1999.

Randomised trials reporting the analgesic effects of anticonvulsant drugs in patients, with subjective pain assessment as either the primary or a secondary outcome.

Data were extracted by two independent reviewers, and trials were quality scored. Numbers-needed-to-treat (NNTs) were calculated from dichotomous data for effectiveness, adverse effects and drug-related study withdrawal, for individual studies and for pooled data.

Twenty-three trials of six anticonvulsants were considered eligible (1,074 patients). The only placebo-controlled study in acute pain found no analgesic effect of sodium valproate. Three placebo-controlled studies of carbamazepine in trigeminal neuralgia had a combined NNT (95% confidence interval (CI)) for effectiveness of 2.5 (CI 2.0-3.4). A single placebo-controlled trial of gabapentin in post-herpetic neuralgia had an NNT of 3.2 (CI 2.4-5.0). For diabetic neuropathy NNTs for effectiveness were as follows: (one RCT for each drug) carbamazepine 2.3 (CI 1.6-3.8), gabapentin 3.8 (CI 2.4-8.7) and phenytoin 2.1 (CI 1.5-3.6). Numbers-needed-to-harm (NNHs) were calculated where possible by combining studies for each drug entity irrespective of the condition treated. The results were, for minor harm, carbamazepine 3.7 (CI 2.4-7.8), gabapentin 2.5 (CI 2.0-3.2), phenytoin 3.2 (CI 2.1-6.3). NNHs for major harm were not statistically significant for any drug compared with placebo. Phenytoin had no effect in irritable bowel syndrome, and carbamazepine little effect in post-stroke pain. Clonazepam was effective in one study of temporomandibular joint dysfunction.

Although anticonvulsants are used widely in chronic pain surprisingly few trials show analgesic effectiveness. No trial compared different anticonvulsants. Only one studied considered cancer pain. There is no evidence that anticonvulsants are effective for acute pain. In chronic pain syndromes other than trigeminal neuralgia, anticonvulsants should be withheld until other interventions have been tried. While gabapentin is increasingly being used for neuropathic pain the evidence would suggest that it is not superior to carbamazepine.