Statistical analyses and quality of individual participant data network meta-analyses were suboptimal: a cross-sectional study

Table 3 Reporting information of statistical analyses of included IPD-NMAs

Items	Frequency	Proportion (%)
Fixed- or random-effects?
Fixed-effects	6	28.6
Random-effects	6	28.6
Fixed- and random-effects	6	28.6
Not reported	3	14.3
Bayesian or Frequentist method?
Bayesian	12	57.1
Frequentist	9	42.9
1-stage or 2-stage process?
1-stage	14	66.7
2-stage	7	33.3
How was the model fit assessed?
Deviance information criterion	8	38.1
Deviance information criterion + residual deviance	2	9.5
Not reported	2	9.5
Not applicable	9	42.9
Were the prior distributions reported?
Yes	8	38.1
Noninformative prior^a	5	23.8
Informative prior^a	1	4.8
Noninformative prior + informative prior	2	9.5
Not reported	4	19
Not applicable	9	42.9
Was the convergence assessed?
Yes	8	38.1
Gelman-Rubin statistic	4	19
Visual plot inspection	1	4.8
Gelman-Rubin statistic + visual plot inspection	3	14.3
Not reported	4	19
Not applicable	9	42.9
Statistical techniques used for missing participant data
LOCF	3	14.3
ACA	1	4.8
MCMC multiple imputations	1	4.8
Not reported	16	76.2
Was the heterogeneity assessed?
Yes	15	71.4
Not reported	6	28.6
Was the consistency assessed?
Yes	9	42.9
Loop-specific approach	2	9.5
Node-splitting	2	9.5
Lu and Ades	1	4.8
Lumley	1	4.8
Informal approaches^b	3	14.3
Not reported	12	57.1
Was the transitivity assessed? (yes)	0	0
Subgroup analysis conducted? (yes)	10	47.6
Sensitivity analysis conducted? (yes)	16	76.2
Meta-regression analysis conducted? (yes)	8	38.1
With GRADE used	2	9.5
Software used
WinBUGS/OpenBUGS	7	33.3
R	3	14.3
WinBUGS + R	3	14.3
WinBUGS + Stata	1	4.8
SAS + Review Manager^c	2	9.5
SAS + Stata	1	4.8
R + Stata	1	4.8
Not reported	3	14.3

IPD-NMAs individual participant data network meta-analyses, LOCF last observation carried forward, ACA available case analysis, MCMC Markov chain Monte Carlo, GRADE Grading of Recommendations Assessment, Development and Evaluation
^aPriors are based on the detailed methods of prior distribution reported in IPD-NMAs
^bInformal approaches are the comparison of IPD-NMA with meta-regression IPD-NMA results, comparison of IPD-NMA with aggregate data NMA results, and comparison of NMA results with pairwise meta-analyses results
^cReview Manager was used for pairwise meta-analyses

ISSN: 1741-7015