R使用gemtc程序包实现连续型变量的频率学网状meta分析

今天的教程主要是使用gemtc程序包实现连续型变量的贝叶斯网状meta分析，数据来源于《A systematic review and mixed treatment comparison of pharmacological interventions for the treatment of obesity》。

公众号后台回复nmacon获取训练数据。

1.程序包加载及数据导入

#install.packages("gemtc")
#install.packages("rjags")
#install.packages("coda")
library(gemtc)
data <- read.csv("diabetes.csv", sep=",", header=T)

2. 网络证据图绘制

network <-mtc.network(data)
summary(network)
plot(network,edge.color="black",vertex.color="steelblue",vertex.label.dist=2.5)

3. 模型构建

model<-mtc.model(network, 
                 type="consistency",
                 n.chain=3,
                 likelihood="normal",
                 link="identity",
                 linearModel="random",
                 dic = TRUE)
results<- mtc.run(model,
                  n.adapt = 2000,
                  n.iter = 10000, 
                  thin = 3)

4. 模型收敛度诊断

gelman.diag(results)#诊断结果

Potential scale reduction factors:

                        Point est. Upper C.I.
d.Metformin.Orlistat          1.00       1.01
d.Metformin.Placebo           1.00       1.01
d.Metformin.Rimonbant         1.00       1.00
d.Metformin.Sibutramine       1.01       1.01
d.Orlistat.OrliSibut          1.00       1.00
sd.d                          1.00       1.00

Multivariate psrf

1

gelman.plot(results)## 诊断图
plot(results)

5. 森林图绘制

forest(relative.effect(results, "Placebo"))

6 . 异质性检验

result.anohe<-mtc.anohe(network,n.adapt = 2000, n.iter = 10000)
summary.anohe=summary(result.anohe)
summary.anohe

Analysis of heterogeneity
=========================

Per-comparison I-squared:
-------------------------

          t1          t2  i2.pair  i2.cons   incons.p
1  Metformin    Orlistat       NA 75.95795 0.07919085
2  Metformin     Placebo       NA  0.00000 0.37703011
3  Metformin   Rimonbant       NA       NA         NA
4  Metformin Sibutramine       NA 34.94588 0.24997324
5  OrliSibut    Orlistat       NA 54.95407 0.18717601
6  OrliSibut     Placebo 81.94884 83.58646         NA
7  OrliSibut Sibutramine       NA  0.00000 0.62915475
8   Orlistat     Placebo 47.79664 44.49184 0.50275971
9   Orlistat Sibutramine  0.00000  0.00000 0.61098135
10   Placebo Sibutramine 41.77298 43.30047         NA

Global I-squared:
-------------------------

   i2.pair  i2.cons
1 43.66205 41.52727

7. 节点分割法

nodesplit <- mtc.nodesplit(network,
                           linearModel = "random",
                           n.adapt = 2000,
                           n.iter = 10000,
                           thin = 3)
summary(nodesplit)
plot(summary(nodesplit))

Node-splitting analysis of inconsistency
========================================

   comparison              p.value    CrI                 
1  d.Metformin.Orlistat    0.12091209                     
2  -> direct                          1.4 (0.29, 2.4)     
3  -> indirect                        -0.15 (-1.9, 1.6)   
4  -> network                         0.95 (-0.055, 1.8)  
5  d.Metformin.Placebo     0.15781578                     
6  -> direct                          1.1 (-0.58, 2.8)    
7  -> indirect                        2.6 (1.4, 3.7)      
8  -> network                         2.0 (1.0, 2.9)      
9  d.Metformin.Sibutramine 0.11956196                     
10 -> direct                          -0.67 (-2.3, 1.0)   
11 -> indirect                        0.93 (-0.26, 2.1)   
12 -> network                         0.40 (-0.62, 1.3)   
13 d.OrliSibut.Orlistat    0.02265227                     
14 -> direct                          1.5 (0.38, 2.6)     
15 -> indirect                        -1.3 (-3.3, 0.80)   
16 -> network                         0.95 (0.0028, 1.8)  
17 d.OrliSibut.Sibutramine 0.03930393                     
18 -> direct                          0.72 (-0.40, 1.8)   
19 -> indirect                        -1.8 (-3.7, 0.28)   
20 -> network                         0.40 (-0.54, 1.3)   
21 d.Orlistat.Placebo      0.87428743                     
22 -> direct                          1.2 (0.63, 1.7)     
23 -> indirect                        1.1 (-0.070, 2.2)   
24 -> network                         1.1 (0.67, 1.5)     
25 d.Orlistat.Sibutramine  0.55820582                     
26 -> direct                          -0.44 (-1.1, 0.25)  
27 -> indirect                        -0.73 (-1.5, 0.024) 
28 -> network                         -0.55 (-0.98, -0.12)
29 d.Placebo.Sibutramine   0.22832283                     
30 -> direct                          -1.7 (-2.0, -1.3)   
31 -> indirect                        -0.79 (-2.2, 0.67)  
32 -> network                         -1.6 (-1.9, -1.3)

8. 两两比较赛联表

league<-relative.effect.table(results)
league1<-round(exp(league),2)
league1

9. 排序图及累计排序图

rank <- rank.probability(results, preferredDirection = -1)
print(rank)
plot(rank, beside=TRUE, cex.names=0.5)

print(sucra(rank))
plot(rank)

- END -