forplo | 冲冲冲！这个画森林图的包好flexible哦~

news2025/7/19 6:21:40

1写在前面

我想大家肯定都用过森林图，应用比较多的场景可能是展示meta分析，回归分析结果的时候。🥳
画森林图的包还是挺多的，今天介绍一下forplo包的用法。😘

2用到的包

rm(list = ls())
library(tidyverse)
library(forplo)
library(meta)
library(autoReg)
library(survival)

3meta分析结果可视化

3.1 示例数据

这里我准备了一个随机数据，假设是我们收集的各篇文章中的某个指标。

dat <- read.csv("./forplo_meta.csv",row.names = 1)
DT::datatable(dat)

3.2 初步绘图

forplo(dat[,1:3],
       groups=dat$groups,
       font = "Arial",
       grouplabs=c('Low risk of bias',
                   'Some concerns',
                   'High risk of bias'))

3.3 分组计算meta效应值

logORs <- round(log(dat$OR),2)

SE <- round((log(dat$UCI)-logORs)/1.96,2)

meta1 <- meta::metagen(logORs[1:3],SE[1:3])

meta2 <- meta::metagen(logORs[4:9],SE[4:9])

meta3 <- meta::metagen(logORs[10:12],SE[10:12])

metatot <- meta::metagen(logORs,SE)

3.4 创建可视化需要的data

dat2 <- dat
dat2$logORs <- logORs
dat2$SE <- SE
dat2$weights <- round(metatot$w.random/sum(metatot$w.random)*100,2)
dat2 <- rbind(subset(dat2,groups==1),
      c(round(exp(meta1$TE.random),2),
        round(exp(meta1$lower.random),2),
        round(exp(meta1$upper.random),2),
        1,round(meta1$TE.random,2),round(meta1$seTE.random,2),sum(dat2$weights[1:3])),
      subset(dat2,groups==2),
      c(round(exp(meta2$TE.random),2),
        round(exp(meta2$lower.random),2),
        round(exp(meta2$upper.random),2),
        2,round(meta2$TE.random,2),round(meta2$seTE.random,2),sum(dat2$weights[4:9])),
      subset(dat2,groups==3),
      c(round(exp(meta3$TE.random),2),
        round(exp(meta3$lower.random),2),
        round(exp(meta3$upper.random),2),
        3,round(meta3$TE.random,2),round(meta3$seTE.random,2),sum(dat2$weights[10:12])),
      c(round(exp(metatot$TE.random),2),
        round(exp(metatot$lower.random),2),
        round(exp(metatot$upper.random),2),
        4,round(metatot$TE.random,2),round(metatot$seTE.random,2),100))
dat2 <- data.frame(dat2)

rownames(dat2)[c(4,11,15,16)] <- c('Effect 1','Effect 2','Effect 3','Effect 4')


DT::datatable(dat2)

3.5 森林图可视化meta结果

forplo(dat2[,1:3],
       font = "Arial",
       groups=dat2$groups,
       grouplabs=c('Low risk of bias',
                   'Some concerns',
                   'High risk of bias',
                   'Overall'),
       left.align=TRUE,
       add.columns=dat2[,5:7],
       add.colnames=c('log(OR)','SE','Weights'),
       scaledot.by = dat2$weights,
       col= '#FEBE8C',
       ci.edge= T,
       char = 15,  # 10为空心圆, 15为实心方格, 20实心圆
       diamond=c(4,11,15,16),
       diamond.col= '#EB6440',
       favorlabs=c('Favours other models','Favours midwife-led'), # x.lab
       ## 阴影参数
       shade.every=1, # 每行加背景底色
       )

4Cox回归结果可视化

本期使用survival里的cancer数据，类型为lung。😘

data("cancer")
DT::datatable(lung)

4.1 因子转换

这里我们先做个factor转换，方便后续操作。🧐

lung$sex <- factor(lung$sex, levels = unique(lung$sex))
lung$ph.ecog <- factor(lung$ph.ecog, levels = c(0,1,2,3))

4.2 Cox回归建模

把所有变量都纳入进去进行回归建模吧。👇

fit1 <- coxph(Surv(time, status) ~ sex + age + ph.ecog + ph.karno + pat.karno + meal.cal + wt.loss,
           data = lung)
summary(fit1)

4.3 整理成table

这里我们用一下之前介绍的autoreg, 看一下回归结果。🤩

Note! 由于这个dataset存在缺失值，所以使用了mice包进行填补，想要了解mice具体教程的小伙伴可以在公众号内搜索ggmice或者mice。 😏

autoReg(fit1, uni = T, threshold = 0.1, 
        final = T,imputed = T) %>%
  myft()

4.4 森林图可视化

不知道大家有没有发现, 这个结果和autoreg出的结果会有一点细微差异, 仁者见仁,智者见智吧。🤒

forplo(fit1,
       font='Arial',
       sort = T,
       #flipbelow1=T,
       left.align=T,
       ci.edge= T,
       scaledot.by=abs(coef(fit1)),
       col = '#BAD1C2',
       char = 20,
       shade.every = 1,
       shade.col = '#FDF0E0',
       shade.alpha = 0.8
       )

5logistic回归结果可视化

5.1 logistic回归建模

lung$status <- factor(lung$status, levels = rev(unique(lung$status)))
  
fit2 <- glm(status ~ sex + age + ph.karno + pat.karno + meal.cal + wt.loss,
           data = lung,
           family="binomial")
summary(fit2)

5.2 整理成table

autoReg(fit2, uni = T, threshold = 0.1, 
        final = T,imputed = T) %>% 
  myft()

5.3 森林图可视化

forplo(fit2,
       font='Arial',
       sort = T,
       #flipbelow1=T,
       left.align=T,
       ci.edge= T,
       scaledot.by=abs(coef(fit2)),
       col = '#DBA39A',
       char = 15,
       shade.every = 1,
       shade.col = '#FDF0E0',
       shade.alpha = 0.8,
       right.bar=T,
       rightbar.ticks=T,
       leftbar.ticks=T
       )