Background
Implementation
Domain | Function | Brief description | Checklist Item No. | ||
---|---|---|---|---|---|
PRISMA [11] | ISPOR-AMPC-NCA [10] | NICE-DSU [12] | |||
Data preparation | data.prep() | Prepares data for further processing | N/A | N/A | N/A |
Description of network | net.tab() | Descriptive statistics of evidence network | S4, 20 | 2, 14 | A9.1, C3.1 |
net.plot() | Plot of evidence network | S3 | 2, 14 | A9.1, C3.1, C4.1 | |
Homogeneity assessment | data.plot() | Graph of characteristics by study or treatment | 18 | 5, 6 | A7.2, B2.5, C4.2 |
pma() | Heterogeneity statistics for all direct comparisons | - | 5, 6 | B2.1, B4.1, C4.2 | |
Network meta-analysis | nma.model() | Specifies the NMA model (including meta-regression) | 23 | 7, 9 10, 12, 13, 19 | A6.3, A7.2, B2.3, B3.1, C2.1 |
nma.run() | Runs the NMA | N/A | N/A | N/A | |
Model assessment | nma.diag() | Trace plots and other convergence diagnostics | – | – | B1.1 |
nma.fit() | Leverage plots and deviance information criterion (DIC) values | 23, S5 | 8, 11, 12 | B2.2 | |
Output results | nma.forest() | A forest plot of the NMA results | 21 | 17 | B4.1 |
nma.league() | League table of the NMA results | 21 | 17 | B4.1 | |
nma.rank() | Tabular and graphical results for treatment ranking including Surface under the cumulative ranking curve (SUCRA) plot | 21 | 18 | B4.1 | |
nma.regplot() | Meta-regression only. Plot of estimated relative treatment effects (on the linear scale) as a function of covariate values | 23 | 10, 13, 19 | A6.3, A7.2, B2.3 | |
pma() | Results from pairwise comparisons | 20 | 16 | B4.1 | |
Consistency assessment | nma.compare() | Comparison of consistency and inconsistency models [17] | S5 | 8 | C4.3 |
Data preparation
Description of network
Homogeneity
Network meta-analysis
Bayesian inference
NMA models
Type of Outcome | Arm-Level Data Required | Distribution Family | Link function | Measure of effect | Assumption on follow-up time |
---|---|---|---|---|---|
Continuous | Mean & Standard Error | Normal | Identity | Mean Difference | Outcome is unrelated to follow-up time |
Dichotomous | Events & Sample Size | Binomial | Logit | Odds Ratio | |
Log | Risk Ratio | ||||
Events & Sample Size & Median Follow-Up Time | Complementary log-log | Hazard Ratio | Event rates are constant over the duration of follow-up | ||
Count | Events & Person-Time at Risk | Poisson | Log | Rate Ratio |
Meta-regression
Choice of priors
Consistency Model | Inconsistency Model | |||
---|---|---|---|---|
Parameters | Random effect | Fixed effect | Random effect | Fixed effect |
μ1, …, μM | iid N(0,(15u)2) Except when a log link is used with a binomial family, in which case | |||
d1, 2, …. ,d1, T | iid N(0,(15u)2) | NA | ||
d1, 2, …. ,d1, T, …, dT − 2, T − 1, dT − 2, T, dT − 1, T | NA | iid N(0,(15u)2) | ||
σ | U(0,u) | NA | U(0,u) | NA |
β(1, 2), …, β(1, K) (meta-regression only) | Unrelated: iid t(0, u2, df = 1) Exchangeable: iid N(b, γ2), b~ t(0, u2, df = 1), γ~U(0, u) Equal: β2 = … = βT = B, B~ t(0, u2, df = 1) |
Model assessment
Consistency
Output
Results
Study ID | Treatment | Number of Participants | Number of Events | Age (Mean) | Age (SD) |
---|---|---|---|---|---|
MRC-E | Diuretic | 1081 | 43 | 60.7 | 14.3 |
MRC-E | Placebo | 2213 | 34 | 59.2 | 13.1 |
MRC-E | blocker | 1102 | 37 | 60.2 | 14.0 |
EWPH | Diuretic | 416 | 29 | 59.0 | 15.2 |
EWPH | Placebo | 424 | 20 | 57.0 | 14.8 |
. | |||||
. | |||||
. | |||||
VALUE | CCB | 5074 | 845 | 56.4 | 13.1 |
VALUE | ARB | 5087 | 690 | 57.8 | 13.0 |
Data preparation, description of network and homogeneity
Characteristic | Value |
---|---|
Number of Interventions | 6 |
Number of Studies | 22 |
Total Number of Patients in Network | 154,176 |
Total Possible Pairwise Comparisons | 15 |
Total Number of Pairwise Comparisons With Direct Data | 14 |
Is the network connected? | TRUE |
Number of Two-arm Studies | 18 |
Number of Multi-Arms Studies | 4 |
Total Number of Events in Network | 10,962 |
Number of Studies With No Zero Events | 22 |
Number of Studies With At Least One Zero Event | 0 |
Number of Studies with All Zero Events | 0 |
Mean person follow up time | 4.06 |