r-medical-research

name	r-medical-research
description	R语言医学研究助手。自动识别研究主题并安装R包，进行数据处理、统计分析，制作发表级图表和表格，生成图注表注。适用于临床研究、流行病学、生物统计等医学数据分析场景。

R语言医学研究技能

概述

本技能提供R语言医学科学研究的完整工作流程指导，包括研究主题识别、依赖包自动安装、数据处理、统计分析、发表级图表制作及图注表注生成。

触发条件

当用户提出以下需求时，应调用此技能：

• 进行医学/临床数据分析
• 制作论文所需的统计表格（三线表）
• 进行流行病学数据分析
• 进行生存分析
• 进行meta分析
• 进行倾向性评分匹配等因果推断分析
• 询问如何用R进行医学统计
• 需要自动安装相关R包完成特定医学分析任务

---

第一部分：研究主题识别与包安装

1.1 自动识别研究类型

根据用户描述或数据特征，识别研究类型并安装相应包：

# 基础包安装（通用）
install.packages(c("tidyverse", "readxl", "openxlsx", "data.table"))

# 描述性统计与数据清洗
install.packages(c("dplyr", "tidyr", "stringr", "lubridate", "Hmisc", "mice"))

# 统计分析
install.packages(c("stats", "rstatix", "car", "lmtest", "aod"))

# 高级统计分析
install.packages(c("survival", "survminer", "glm", "lme4", "nlme", "meta", "metafor"))

# 临床研究专用
install.packages(c("tableone", "autoReg", "moonBook", "rrtable", "MatchIt", "WeightIt", "cobalt"))

# 机器学习与建模
install.packages(c("caret", "randomForest", "e1071", "glmnet", "pROC", "rms"))

# 表格制作与导出
install.packages(c("gt", "gtsummary", "flextable", "officer", "ftExtra"))

# 可视化与出版
install.packages(c("ggplot2", "ggpubr", "ggsci", "patchwork", "cowplot", "survminer"))

1.2 按研究类型的包推荐

研究类型	核心R包	辅助包
描述性分析	dplyr, Hmisc, tableone	skimr, DataExplorer
差异分析	rstatix, stats, car	ggpubr, rcompanion
回归分析	stats, glm, lm, survival	autoReg, gtsummary
生存分析	survival, survminer	survivalROC, riskRegression
Meta分析	meta, metafor, robumeta	metaBMA, netmeta
因果推断	MatchIt, WeightIt, cobalt	causalweight, causalinferencelab
诊断试验	pROC, ROCR, cutpointr	OptimalCutpoints, DiagTest
列线图	rms, nomogramEx	survival, ggplot2
机器学习	caret, randomForest, glmnet	recipes, tune, yardstick

---

第二部分：数据类型识别与处理

2.1 常见医学数据类型

# 查看数据结构
str(data)
glimpse(data)
skimr::skim(data)

# 识别变量类型
# 连续变量: 年龄、血压、血糖、实验室指标
# 分类变量: 性别、疾病分期、治疗组
# 有序变量: 病情严重程度、教育水平
# 时间变量: 随访时间、生存时间
# 事件变量: 死亡、复发、进展

2.2 数据读取

library(readxl)
library(openxlsx)

# 读取Excel
data <- read_excel("data.xlsx", sheet = 1)

# 读取CSV
data <- read.csv("data.csv", stringsAsFactors = FALSE)

# 读取SPSS/SAS/Stata
library(haven)
data <- read_spss("data.sav")
data <- read_stata("data.dta")

2.3 数据清洗与预处理

library(dplyr)
library(tidyr)

# 缺失值处理
data <- data %>% drop_na()  # 删除含缺失值行
data <- data %>% na_replace(0)  # 用0填充
data <- data %>% fill(group, .direction = "down")  # 用前值填充

# 多重插补
library(mice)
imputed <- mice(data, m = 5, method = "pmm")
data <- complete(imputed, 1)

# 变量转换
data <- data %>%
  mutate(
    age_group = cut(age, breaks = c(0, 40, 60, 100), 
                    labels = c("青年", "中年", "老年")),
    bmi_cat = ifelse(bmi < 18.5, "偏瘦", 
              ifelse(bmi < 24, "正常", 
              ifelse(bmi < 28, "偏胖", "肥胖"))),
    log_alt = log(ALT + 1)
  )

# 数据标准化
data <- data %>%
  mutate(across(where(is.numeric), scale))

2.4 倾向性评分匹配

library(MatchIt)
library(cobalt)

# 1:1匹配
psm_result <- matchit(treatment ~ age + bmi + hypertension + diabetes,
                      data = data,
                      method = "nearest",
                      ratio = 1)

matched_data <- match.data(psm_result)

# 检查匹配平衡性
love.plot(psm_result, binary = "std")

---

第三部分：统计分析方法

3.1 描述性统计

library(tableone)
library(gtsummary)

# 创建Table 1 (基线特征表)
vars <- c("age", "sex", "bmi", "sbp", "dbp", "glucose", "treatment")
factorVars <- c("sex", "treatment", "hypertension", "diabetes")

table1 <- CreateTableOne(vars = vars, 
                         data = data,
                         factorVars = factorVars,
                         strata = "group",
                         test = TRUE)
print(table1, smd = TRUE)

# 使用gtsummary创建更美观的表格
data %>% 
  tbl_summary(by = "group", 
              statistic = list(all_continuous() ~ "{mean} ± {sd}",
                              all_categorical() ~ "{n} ({p})")) %>%
  add_p() %>%
  add_difference() %>%
  bold_labels()

3.2 差异分析

library(rstatix)

# 连续变量：t检验
data %>% t_test(value ~ group, data = .)

# 连续变量：Wilcoxon秩和检验
data %>% wilcox_test(value ~ group, data = .)

# 连续变量：方差分析
data %>% anova_test(value ~ group)

# 分类变量：卡方检验
chisq.test(table(data$group, data$outcome))

# 配对样本t检验
data %>% t_test(value ~ time, paired = TRUE, data = .)

# 多重比较校正
p_values <- c(0.001, 0.01, 0.03, 0.04)
p.adjust(p_values, method = "bonferroni")
p.adjust(p_values, method = "fdr")

3.3 回归分析

# 线性回归
model_lm <- lm(outcome ~ age + bmi + treatment, data = data)
summary(model_lm)
gtsummary::tbl_regression(model_lm)

# Logistic回归
model_logit <- glm(outcome ~ age + bmi + treatment, 
                   data = data, 
                   family = binomial())
summary(model_logit)
exp(confint(model_logit))  # OR值及置信区间

# 使用autoReg自动生成回归表格
library(autoReg)
autoReg(outcome ~ age + bmi + treatment, data = data) %>%
  myft()

# Cox比例风险模型
library(survival)
model_cox <- coxph(Surv(time, event) ~ age + bmi + treatment, data = data)
summary(model_cox)

3.4 生存分析

library(survival)
library(survminer)

# Kaplan-Meier生存曲线
fit <- survfit(Surv(time, event) ~ group, data = data)
ggsurvplot(fit, data = data,
           pval = TRUE,
           conf.int = TRUE,
           risk.table = TRUE,
           ncensor.plot = TRUE,
           palette = c("#E64B35", "#4DBBD5"))

# Log-rank检验
survdiff(Surv(time, event) ~ group, data = data)

# Cox回归
cox_model <- coxph(Surv(time, event) ~ group + age + bmi, data = data)
ggforest(cox_model, data = data)

# 竞争风险模型
library(cmprsk)
crr_model <- crr(ftime, fstatus, cov1, failcode = 1)

---

第四部分：发表级图表制作

4.1 基础图表模板

library(ggplot2)
library(ggpubr)
library(ggsci)

# 论文主题设置
theme_pub <- function() {
  theme_classic() +
    theme(
      axis.text = element_text(size = 10, color = "black"),
      axis.title = element_text(size = 12, face = "bold"),
      legend.text = element_text(size = 10),
      legend.title = element_text(size = 11, face = "bold"),
      plot.title = element_text(size = 14, face = "bold", hjust = 0.5),
      panel.grid.major = element_blank(),
      panel.grid.minor = element_blank()
    )
}

4.2 常用图表类型

散点图与回归图

ggplot(data, aes(x = age, y = bmi, color = group)) +
  geom_point(alpha = 0.6, size = 2) +
  geom_smooth(method = "lm", se = TRUE) +
  stat_cor(method = "pearson", aes(color = group)) +
  scale_color_npg() +
  theme_pub() +
  labs(x = "年龄 (岁)", y = "BMI (kg/m²)", 
       title = "年龄与BMI的相关性")

箱线图/小提琴图

ggplot(data, aes(x = group, y = value, fill = group)) +
  geom_violin(alpha = 0.6, trim = FALSE) +
  geom_boxplot(width = 0.15, fill = "white") +
  stat_compare_means(method = "t.test", label = "p.format") +
  scale_fill_npg() +
  theme_pub() +
  labs(x = "分组", y = "指标值")

生存曲线

ggsurvplot(survfit(Surv(time, event) ~ group, data = data),
           data = data,
           pval = TRUE,
           conf.int = TRUE,
           risk.table = TRUE,
           palette = c("#E64B35", "#4DBBD5"),
           legend.title = "治疗组",
           xlab = "随访时间 (月)",
           ylab = "生存概率")

森林图

library(forestplot)
forestplot::forestplot(
  labeltext = as.matrix(table_data[, 1:3]),
  mean = table_data$OR,
  lower = table_data$lower,
  upper = table_data$upper,
  hrzl_lines = TRUE,
  col = fpColors(box = "#4DBBD5", line = "black"),
  xlab = "Odds Ratio (95% CI)"
)

热图

library(pheatmap)
pheatmap(mat = expression_matrix,
         scale = "row",
         cluster_rows = TRUE,
         cluster_cols = TRUE,
         show_rownames = TRUE,
         show_colnames = TRUE,
         color = colorRampPalette(c("#2166AC", "#F7F7F7", "#B2182B"))(50),
         border_color = NA)

4.3 多图组合

library(patchwork)

(p1 + p2 + p3) / (p4 + p5) +
  plot_annotation(tag_levels = "A") &
  theme(plot.tag = element_text(face = "bold"))

4.4 高质量导出

# 导出为TIFF (期刊要求)
ggsave("Figure1.tiff", 
       plot = last_plot(),
       width = 8, height = 6, units = "cm",
       dpi = 300, compression = "lzw")

# 导出为PDF (矢量图)
ggsave("Figure1.pdf",
       plot = last_plot(),
       width = 8, height = 6, units = "cm")

# 批量导出
for (i in 1:3) {
  ggsave(paste0("Figure", i, ".tiff"),
         plot = plots[[i]],
         width = 8, height = 6, units = "cm", dpi = 300)
}

---

第五部分：发表级表格制作

5.1 三线表 (Table 1 - 基线特征表)

library(tableone)
library(officer)
library(flextable)

# 创建基线特征表
vars <- c("age", "sex", "bmi", "sbp", "glucose", "outcome")
factorVars <- c("sex", "outcome")

table1 <- CreateTableOne(vars = vars, 
                         data = data,
                         strata = "group",
                         factorVars = factorVars,
                         test = TRUE)

# 导出为Word
print(table1, smd = TRUE, quote = FALSE, noSpaces = TRUE)

# 使用gtsummary创建更美观的表格
baseline_table <- data %>%
  tbl_summary(
    by = "group",
    statistic = list(
      all_continuous() ~ "{mean} ± {sd}",
      all_categorical() ~ "{n} ({p}%)"
    ),
    label = list(
      age ~ "年龄 (岁)",
      sex ~ "性别 (男)",
      bmi ~ "BMI (kg/m²)"
    )
  ) %>%
  add_p() %>%
  add_n() %>%
  bold_labels() %>%
  italicize_levels()

# 导出
as_flex_table(baseline_table) %>%
  save_as_docx(path = "Table1.docx")

5.2 回归分析结果表

library(gtsummary)

# Logistic回归表格
data %>%
  mutate(outcome = factor(outcome)) %>%
  tbl_uvregression(
    method = glm,
    y = outcome,
    method.args = list(family = binomial),
    exponentiate = TRUE,
    pvalue_fun = function(x) style_pvalue(x, digits = 3)
  ) %>%
  add_global_p() %>%
  bold_p() %>%
  inline_text("variable", row = 1, column = "estimate")

# 导出到Word
tbl_regression(model_logit, exponentiate = TRUE) %>%
  as_flex_table() %>%
  save_as_docx(path = "Table2.docx")

5.3 亚组分析表

library(metafor)

# 创建亚组分析表
subgroup_results <- data %>%
  group_by(subgroup) %>%
  summarise(
    n = n(),
    events = sum(outcome),
    or = NA,
    lower = NA,
    upper = NA
  )

for (i in 1:nrow(subgroup_results)) {
  subset_data <- data %>% filter(subgroup == subgroup_results$subgroup[i])
  model <- glm(outcome ~ treatment, data = subset_data, family = binomial)
  or <- exp(coef(model)["treatment"])
  ci <- exp(confint(model)["treatment", ])
  subgroup_results$or[i] <- or
  subgroup_results$lower[i] <- ci[1]
  subgroup_results$upper[i] <- ci[2]
}

---

第六部分：图注与表注生成

6.1 自动生成图注

generate_figure_caption <- function(fig_num, type, description, data_info) {
  caption <- paste0(
    "图", fig_num, " ", description, "。",
    "A: ", data_info$group_a, "; B: ", data_info$group_b, "。",
    "数据以均值±标准误表示。",
    "组间比较采用", data_info$test_method, "检验。"
  )
  return(caption)
}

# 示例
fig1_caption <- generate_figure_caption(
  fig_num = "1",
  type = "boxplot",
  description = "两组患者血糖水平比较",
  data_info = list(
    group_a = "治疗组 (n=100)",
    group_b = "对照组 (n=100)",
    test_method = "t"
  )
)
cat(fig1_caption)
# 输出: 图1 两组患者血糖水平比较。 A: 治疗组 (n=100); B: 对照组 (n=100)。 数据以均值±标准误表示。 组间比较采用t检验。

6.2 自动生成表注

generate_table_notes <- function(table_num, n, events, methods, software) {
  notes <- paste0(
    "表", table_num, " 注: ",
    "数据以均数±标准差或频数(%)表示。",
    "P值由", methods, "计算。",
    "统计分析使用R软件 (", software, ")完成。",
    "n=", n, "; 事件数=", events, "。"
  )
  return(notes)
}

# 示例
table1_notes <- generate_table_notes(
  table_num = "1",
  n = 200,
  events = 45,
  methods = "t检验或卡方检验",
  software = "version 4.3.1"
)
cat(table1_notes)
# 输出: 表1 注: 数据以均数±标准差或频数(%)表示。 P值由t检验或卡方检验计算。 统计分析使用R软件 (version 4.3.1)完成。 n=200; 事件数=45。

6.3 综合图注表注模板

# 为不同类型图表生成对应的标准图注

# 生存曲线图注
survival_caption <- "图X 不同治疗组的Kaplan-Meier生存曲线比较。采用log-rank检验比较两组生存差异。"

# 森林图图注
forest_caption <- "图X 亚组分析的森林图。横轴表示风险比(HR)，垂直虚线代表HR=1。"

# 热图图注
heatmap_caption <- "图X 基因表达热图。行表示基因，列表示样本。颜色表示标准化后的表达量(按行Z-score)。"

# 相关性热图图注
corr_caption <- "图X 变量相关性热图。颜色表示Pearson相关系数，蓝色为正相关，红色为负相关。"

---

第七部分：自动化工作流

7.1 一键分析模板

# 完整的临床研究分析流程
clinical_analysis <- function(data, outcome_var, group_var, covariables) {
  
  # 1. 描述性统计
  cat("=== 描述性统计 ===\n")
  baseline <- CreateTableOne(
    vars = c(covariables, outcome_var),
    data = data,
    strata = group_var,
    factorVars = outcome_var
  )
  print(baseline, smd = TRUE)
  
  # 2. 单因素分析
  cat("\n=== 单因素Logistic回归 ===\n")
  univ <- autoReg(outcome_var ~ ., data = data, uni = TRUE) %>%
    myft()
  print(univ)
  
  # 3. 多因素分析
  cat("\n=== 多因素Logistic回归 ===\n")
  multiv <- autoReg(outcome_var ~ ., data = data, multi = TRUE) %>%
    myft()
  print(multiv)
  
  # 4. 亚组分析
  cat("\n=== 亚组分析 ===\n")
  # (根据需要添加亚组代码)
  
  # 5. 图表输出
  cat("\n=== 生成图表 ===\n")
  # (添加图表生成代码)
  
  return(list(baseline = baseline, univ = univ, multiv = multiv))
}

# 使用示例
# results <- clinical_analysis(
#   data = mydata,
#   outcome_var = "outcome",
#   group_var = "treatment",
#   covariables = c("age", "sex", "bmi")
# )

7.2 批量导出报告

library(rmarkdown)
library(officer)

generate_report <- function(data, output_file = "analysis_report.docx") {
  
  # 创建Word文档
  doc <- read_docx()
  
  # 添加标题
  doc <- doc %>% 
    body_add_par("临床研究统计分析报告", style = "heading 1")
  
  # 添加表格
  doc <- doc %>% 
    body_add_par("表1 基线特征", style = "heading 2") %>%
    body_add_flextable(ft_baseline)
  
  # 添加图表
  doc <- doc %>%
    body_add_par("图1 生存曲线", style = "heading 2") %>%
    body_add_gg(p_survival, width = 5, height = 4)
  
  # 保存
  print(doc, target = output_file)
}

---

第八部分：常见期刊要求

8.1 图表规范

期刊类型	宽度要求	高度比例	分辨率
单栏图	8-9 cm	自由	300-600 dpi
双栏图	12-17 cm	自由	300-600 dpi
半版图	≤8 cm	自由	300-600 dpi
PDF矢量图	建议8-17 cm	自由	无需dpi

8.2 表格规范

# 标准三线表格式
# - 顶线、底线1.5pt
# - 栏目线0.5pt  
# - 无竖线
# - 字体Times New Roman 8-10pt
# - 数字右对齐
# - 中文用宋体

---

参考资源

1. R for Data Science: https://r4ds.had.co.nz/
2. gtsummary: https://www.danieldsjoberg.com/gtsummary/
3. survminer: http://www.sthda.com/english/rsurvival-survminer
4. tableone: https://cran.r-project.org/web/packages/tableone/
5. autoReg: https://cran.r-project.org/web/packages/autoReg/
6. ** ggplot2**: https://ggplot2.tidyverse.org/
7. STHDA: http://www.sthda.com/english/

---

注意事项

1. 数据安全: 医学数据需注意隐私保护，脱敏处理后再进行分析
2. 统计方法选择: 根据数据分布和研究设计选择合适的统计方法
3. 多重比较: 进行多次检验时需校正P值
4. 效应量报告: 除P值外，应同时报告效应量及其95%置信区间
5. 结果可重复: 记录数据处理和分析的所有参数和步骤