Sixteen studies (1,856 patients) were included in the review. There were 905 RC patients and 951 non-RC patients.
The average quality rating score was 42.2% plus or minus 21.4%. There were no significant differences among specific drugs.
Nine studies included both RC and non-RC patients. For treatment effects, recurrence (Q=90.2, d.f.=8) and non-improvement (Q=111.5, d.f.=10) were heterogeneous across studies (both P<0.0001).
At least one recurrence was experienced in 384 out of 505 (76%) RC patients receiving any active treatment, compared with 70 out of 93 (75.3%) RC patients randomised to placebo (9 studies with 15 treatment arms). At least one recurrence was experienced in 287 out of 452 (63.5%) non-RC patients receiving any active treatment, compared with 9 out of 10 (90%) non-RC patients randomised to placebo. The crude rate-estimate of recurrence was 2.31%/month for RC patients and 1.25%/month for non-RC patients.
Risk estimates were also calculated without adjusting for exposure times. This resulted in substantial changes in the magnitude of study-specific risk estimates, with small changes in the relative sizes of risk estimates. Both methods of estimating the risk found a greater risk of treatment failure with RC patients.
Non-improvement was experienced in 264 out of 524 (50.4%) RC patients receiving any active treatment, compared with 309 out of 878 (35.2%) RC patients (11 studies). The crude rate-estimate of recurrence was 1.57%/month for RC patients and 0.48%/month for non-RC patients.
Risk estimates were also calculated without adjusting for exposure times. This resulted in minor differences between treatments in RC cases. Both methods of estimating the risk found a greater risk of treatment failure with RC patients.
The summary statistics indicated a significantly higher rate of treatment failure in RC patients compared with non-RC patients.
The pooled recurrence rate for RC and non-RC patients in studies providing a direct comparison was 1.40 (95% CI: 1.26, 1.56, P<0.0001). The pooled non-improvement rate for the same comparison was 1.45 (95% CI: 1.21, 1.74, P<0.0001).
The one direct treatment comparison possible found no statistically significant difference. In studies that compared carbamazepine with lithium, the pooled recurrence rate for RC patients was 0.93 (95% CI: 0.74, 1.18, P=0.54) and the pooled non-improvement rate for RC patients was 0.94 (95% CI: 0.81, 1.08, P=0.37).