Eighteen studies were included: three head-to-head trials (n=137 participants), six placebo controlled trials of gabapentin (n=894 participants) and nine placebo controlled trials of tricyclic antidepressants (number of participants unclear). Trials of gabapentin tended to be more recent (published from 1998 onwards) than those of tricyclic antidepressants (published before 1992). None of the trials met all quality criteria.
Direct meta-analysis showed no difference between gabapentin and tricyclic antidepressants in terms of pain relief (relative risk 0.99, 95% CI: 0.76, 1.29, I2=0%) or rates of withdrawal due to adverse events (relative risk 0.27; 95% CI: 0.03, 2.34). Pain relief was greater than placebo for both gabapentin (relative risk 2.18, 95% CI: 1.78, 2.67, I2=0% based on six trials, 894 patients) and tricyclic antidepressants (relative risk 5.27, 95% CI: 3.05, 9.11, I2=10% based on six trials, 289 patients).
Based on indirect meta-analysis, pain relief was worse with gabapentin than with tricyclic antidepressants (relative risk 0.41, 95% CI: 0.23, 0.74). This difference remained when the analysis was restricted to trials in patients with diabetic neuropathy (relative risk 0.28, 95% CI: 0.11, 0.73). A similar trend was found for studies conducted in patients with post-herpetic neuralgia but the difference was not statistically significant (relative risk 0.60, 95% CI: 0.24, 1.46). There was no difference in risk of withdrawals due to adverse events. However, there was an increased risk of somnolence/sedation (relative risk 2.61, 95% CI: 1.48, 4.62) and dizziness (relative risk 3.16, 95% CI: 1.43, 6.99) with gabapentin compared to tricyclic antidepressants.
None of the sensitivity analyses showed any significant differences between groups.
There was some evidence of publication bias for the tricyclic trials (p=0.01) but not the gabapentin trials.