Fourteen trials (n=3,197, range 41 to 381) were included in the analyses. All studies were prospective, randomised, double-blinded and placebo controlled. Jadad scores ranged from 3 to 5 out of five.
Pooled data from 13 trials favoured NEDAs over placebo using the CGI-I scale (RR 1.36, 95% CI 1.24 to 1.49, RD 0.18, 95% CI 0.13 to 0.23, NNT=6, 95% CI 5 to 8). Similarly pooled data from 10 trials favoured NEDAs over placebo using the IRLS score (WMD -4.93, 95% CI -6.42 to -3.43; score out of 40)
Withdrawal rates in NEDA arms were higher than in placebo arms (RR 1.35, 95% CI 1.00 to 1.81); this increase was statistically significantly greater in patients who took ropinirole but not other NEDAs. Overall NEDAs were associated with statistically significant increases in the adverse events of nausea, dizziness, somnolence and fatigue. Pramipexole was associated with a greater nausea risk. Ropinirole was associated with nausea, dizziness, somnolence and fatigue risk. Sumanirole was associated with headache risk.
The review reported that visual inspection of funnel plots, the trim-and-fill method, and Egger's weighted regression all indicated low risk of publication bias.
Pooling results from each type of NEDA separately indicated that sumanirole was not statistically significantly different from placebo (RR 0.96, 95% CI 0.71 to 1.30). Sensitivity analyses for IRLS and CGI-I scores did not significantly alter the results. IRLS estimates based only on studies with shorter duration (less than 12 weeks) indicated a statistically substantially increased effects size; this relationship with duration was also seen in a random-effects meta-regression.
Heterogeneity was assessed to be statistically significant for the IRLS scores (Q test p<0.001, I2=69%) but not for CGI-I scores (Q test p<0.12, I2=33%).