RStudio中meta包留一法分析森林图绘制报错求助

问题场景 #

首次使用RStudio的meta包开展HR的Meta分析：

1. 主分析（metagen函数）正常完成，森林图可正常生成
2. 留一法分析（metainf函数）计算结果正常，但调用forest(l1o_hfh)绘制森林图时触发上述报错

主分析代码与输出 #

hfh.m <- metagen(TE = hr, upper = upper, lower = lower,
            n.e = n.e, n.c = n.c,
                      data=Question,
                      studlab=author,
                      method.tau="REML",
                      sm="HR",
                      transf = F)

hfh.m
# 输出结果：
Number of studies: k = 7
Number of observations: o = 26400 (o.e = 7454, o.c = 18946)

                         HR           95%-CI     z  p-value
Common effect model  0.5875 [0.4822; 0.7158] -5.28 < 0.0001
Random effects model 0.5656 [0.4471; 0.7154] -4.75 < 0.0001

Quantifying heterogeneity (with 95%-CIs):
 tau^2 = 0.0161 [0.0000; 0.2755]; tau = 0.1270 [0.0000; 0.5249]
 I^2 = 0.0% [0.0%; 70.8%]; H = 1.00 [1.00; 1.85]

Test of heterogeneity:
    Q d.f. p-value
 5.54    6  0.4769

Details of meta-analysis methods:
- Inverse variance method
- Restricted maximum-likelihood estimator for tau^2
- Q-Profile method for confidence interval of tau^2 and tau
- Calculation of I^2 based on Q

留一法分析代码与输出 #

l1o_hfh <- metainf(hfh.m,
              pooled="random")

l1o_hfh
# 输出结果：
Leave-one-out meta-analysis

                         HR           95%-CI  p-value  tau^2    tau  I^2
Omitting 1           0.5610 [0.4389; 0.7170] < 0.0001 0.0198 0.1407 9.7%
Omitting 2           0.6167 [0.4992; 0.7618] < 0.0001      0      0   0%
Omitting 3           0.5186 [0.3747; 0.7177] < 0.0001 0.0450 0.2121 6.4%
Omitting 4           0.5670 [0.4418; 0.7276] < 0.0001 0.0197 0.1405 7.3%
Omitting 5           0.5058 [0.3834; 0.6673] < 0.0001 0.0058 0.0760   0%
Omitting 6           0.5780 [0.4532; 0.7371] < 0.0001 0.0155 0.1244 0.7%
Omitting 7           0.6054 [0.4932; 0.7432] < 0.0001 0.0010 0.0310   0%
                                                                        
Random effects model 0.5656 [0.4471; 0.7154] < 0.0001 0.0161 0.1270   0%

Details of meta-analysis methods:
- Inverse variance method
- Restricted maximum-likelihood estimator for tau^2
- Calculation of I^2 based on Q

报错信息 #

forest(l1o_hfh)
# 报错：
Error in round(x, digits) : non-numerical argument to mathematical function

使用的数据集 #

解决方案 #

方案1：转换author列为字符型 #

报错核心原因之一是author列是数值型，metainf生成的对象中研究标签（studlab）为数值，导致forest函数处理时出现类型冲突。修改代码如下：

# 先将author列转为字符型
Question$author <- as.character(Question$author)

# 重新运行主分析与留一法分析
hfh.m <- metagen(TE = hr, upper = upper, lower = lower,
            n.e = n.e, n.c = n.c,
                      data=Question,
                      studlab=author,
                      method.tau="REML",
                      sm="HR",
                      transf = F)

l1o_hfh <- metainf(hfh.m, pooled="random")

# 再绘制森林图
forest(l1o_hfh)

方案2：修复metainf对象中的I^2字段 #

metainf输出的I^2是带百分号的字符型，部分版本的forest.metainf函数会尝试对其做数值运算导致报错。手动转换为数值：

# 将I^2列的百分号去除并转为数值
l1o_hfh$I2 <- as.numeric(sub("%", "", l1o_hfh$I2))

# 绘制森林图
forest(l1o_hfh)

方案3：更新meta包 #

若上述方法无效，可能是旧版本meta包的bug，更新到最新版本：

install.packages("meta")
library(meta)

内容的提问来源于stack exchange，提问作者Rafael Miranda