Five trials were included in the main analyses: two of pramipexole (n=689) and three of ropinirole (n=931). Two polysomnographic studies (one of each drug) were included only in the sensitivity analyses (pramipexole n=109, ropinirole n=73). Sample sizes ranged from 266 to 380. All of the pramipexole studies had multiple arms. Heterogeneity for the pramipexole trials was noted to be moderate (range I2=26.8% to 54.3%) for the outcomes IRLS score, withdrawals for any reason, withdrawals for adverse events and incidence of nausea or nasopharyngitis. All outcomes in the ropinirole trials were low heterogeneity except withdrawal due to adverse events (I2=54.1%).
Direct Comparisons:
For both drugs, there was a significantly greater change in IRLS scores for the active intervention compared with placebo (WMD for pramipexole -5.45, 95% CI -7.70 to -3.20 and WMD for ropinirole was -3.16, 95% CI -4.26 to -2.05).
Proportions of responders according to CGI rating were significantly higher for active interventions versus placebo (OR for pramipexole 2.98, 95% CI 2.08 to 4.26 and OR for ropinirole 1.99, 95% CI 1.52 to 2.60). Based on the CGI response data, for pramipexole NNT=4 (95% CI 3 to 5) and for ropinirole NNT=6 (95% CI 4 to 10).
Indirect comparisons:
Overall there was a greater change in IRLS scores for pramipexole patients versus ropinirole (mean change -2.33, CrI -4.23 to -0.41). The probability that pramipexole was non-inferior to ropinirole was 1 and that the treatment effect was clinically meaningful was 0.43. The odds ratio of being a responder according to CGI scores was 1.50 (95% CrI 0.97 to 2.32) in favour of pramipexole over ropinirole. The probabilities of non-inferiority and a clinically meaningful treatment effect were 0.99 and 0.92.
Tolerability/safety: Incidences of nausea, vomiting and dizziness were significantly increased when being treated with pramipexole compared with ropinirole (from the direct comparisons).
Sensitivity: Adding polysomnographic studies or varying prior distributions did not significantly alter the results.