Skip to main content

Advertisement

Table 4 Methodological characteristics of identified empirical networks, including unpublished data provided by study authors. Figures are no. (%) of studies

From: A scoping review of indirect comparison methods and applications using individual patient data

Characteristic IPD-NMA studiesa MAIC studiesa Totala
Design of studies included in analyses
 RCTs 21 (70) 9 (30) 30 (91)
 RCTs + observational 1 (100) 0 (0) 1 (3)
 RCTs + quasi-RCTs 1 (100) 0 (0) 1 (3)
 RCTs, non-RCTs, CBA 1 (100) 0 (0) 1 (3)
 Total 24 (73) 9 (27) 33 (100)
Fixed- or random-effects model    
 Random-effects model 10 (100) 0 (0) 10 (30)
 Fixed-effect model 7 (100) 0 (0) 7 (21)
 Fixed- and random-effects models 5 (100) 0 (0) 5 (15)
 Not reported/not applicable 2 (18) 9 (82) 11 (33)
 Total 24 (73) 9 (27) 33 (100)
Between-study variance estimator/prior    
 Non-informative prior 10 (100) 0 (0) 10 (67)
 Informative prior 1 (100) 0 (0) 1 (7)
 Minimally informative prior 1 (100) 0 (0) 1 (7)
 DL [72] 1 (100) 0 (0) 1 (7)
 REML [73] 1 (100) 0 (0) 1 (7)
 Not reported 1 (100) 0 (0) 1 (7)
 Total 15 (100) 0 (0) 15 (100)
Methods used to compare different models
 DIC [74] 10 (100) 0 (0) 10 (30)
 Statistical significance of regression coefficients and between-study variance 3 (100) 0 (0) 3 (9)
 DIC and residual deviance [74] 2 (100) 0 (0) 2 (6)
 Comparison of point estimates and their CIs 0 (0) 2 (100) 2 (6)
 AIC and Hosmer–Lemeshow [75, 76] 0 (0) 1 (100) 1 (3)
 DIC and AIC [74, 75] 1 (100) 0 (0) 1 (3)
 Not applicable 8 (57) 6 (43) 14 (42)
 Total 24 (73) 9 (27) 33 (100)
Statistical techniques used for missing participant data
 LOCF 2 (67) 1 (33) 3 (9)
 MCMC multiple imputations 2 (100) 0 (0) 2 (6)
 ACA 1 (100) 0 (0) 1 (3)
 LOCF and ACA 0 (0) 1 (100) 1 (3)
 Not reported/unclear 19 (73) 7 (27) 26 (79)
 Total 24 (73) 9 (27) 33 (100)
Methods used to rank treatment effectiveness/safety
 Probability of being the best 11 (100) 0 (0) 11 (33)
 Not reported/not applicable 13 (59) 9 (41) 22 (67)
 Total 24 (73) 9 (27) 33 (100)
Assessment of consistency assumption
 Yes 13 (100) 0 (0) 13 (39)
 No/unclear 6 (75) 2 (25) 8 (24)
 Not applicable 5 (42) 7 (58) 12 (36)
 Total 24 (73) 9 (27) 33 (100)
Methods used to assess consistency assumption
 Informal approachesb 8 (100) 0 (0) 8 (62)
 Loop-specific approach [68, 77] 1 (100) 0 (0) 1 (8)
 Loop-specific approach and back-calculation [68, 77, 78] 1 (100) 0 (0) 1 (8)
 Lu and Ades [79] 1 (100) 0 (0) 1 (8)
 Lumley [70] 1 (100) 0 (0) 1 (8)
 Node-splitting [78] 1 (100) 0 (0) 1 (8)
 Total 13 (100) 0 (0) 13 (100)
Inclusion of different treatment doses
 No 18 (75) 6 (25) 24 (73)
 Yes 6 (67) 3 (33) 9 (27)
 Total 24 (73) 9 (27) 33 (100)
Approaches used to account for different treatment doses
 Lumping doses 4 (80) 1 (20) 5 (56)
 Splitting doses 2 (50) 2 (50) 4 (44)
 Total 6 (67) 3 (33) 9 (100)
Software    
 WinBUGS [55] 7 (100) 0 (0) 7 (21)
 SAS [80] 2 (33) 4 (67) 6 (18)
 WinBUGS and R [55, 81] 5 (100) 0 (0) 5 (15)
 OpenBUGS [54] 2 (100) 0 (0) 2 (6)
 WinBUGS and Stata [55, 82] 2 (100) 0 (0) 2 (6)
 JAGS and R [53, 81] 1 (100) 0 (0) 1 (3)
 Stata [82] 1 (100) 0 (0) 1 (3)
 Not reported 4 (44) 5 (56) 9 (27)
 Total 24 (73) 9 (27) 33 (100)
  1. ACA available case analysis, AIC Akaike information criterion, CBA controlled before-and-after, CI confidence interval, DIC deviance information criterion, DL DerSimonian and Laird, IPD-NMA individual patient data network meta-analysis, LOCF last observation carried forward, MAIC matching adjusted indirect comparison, MCMC Markov chain Monte Carlo, RCT randomized clinical trial, REML restricted maximum likelihood
  2. aPercentages were calculated across the row for IPD-NMA and MAIC/STC, but down the column for the “Total” column. Total number of included studies n = 37. Total number of empirical networks n = 33. Please note that the empirical networks include 8 methodological and 1 review papers
  3. bInformal approaches are comparison of NMA results with results previously published, comparison of NMA results with pairwise meta-analysis results, comparison of IPD-NMA with meta-regression IPD-NMA results, comparison of IPD-NMA with aggregated data NMA results