forest_subgroups

group	level	n	n_events	HR	CI_lower	CI_upper	p_value	p_interaction	HR_label
Overall	Overall	440,045	40,199	1.42	1.35	1.49	2.40e-48	NA	1.42 (1.35, 1.49)
Age	<60 years	182,316	11,254	1.31	1.21	1.42	4.52e-11	0.038	1.31 (1.21, 1.42)
Age	>=60 years	257,729	28,945	1.5	1.41	1.59	4.31e-39	0.038	1.50 (1.41, 1.59)
Sex	Female	238,119	18,772	1.37	1.28	1.46	2.06e-20	0.214	1.37 (1.28, 1.46)
Sex	Male	201,926	21,427	1.47	1.38	1.57	1.72e-31	0.214	1.47 (1.38, 1.57)
BMI	Underweight	18,421	2,184	1.16	0.994	1.35	0.0603	0.006	1.16 (0.99, 1.35)

group

level

n_events

CI_lower

CI_upper

p_value

p_interaction

HR_label

Overall

440,045

40,199

1.42

1.35

1.49

2.40e-48

1.42 (1.35, 1.49)

Age

<60 years

182,316

11,254

1.31

1.21

1.42

4.52e-11

0.038

1.31 (1.21, 1.42)

Age

>=60 years

257,729

28,945

1.5

1.41

1.59

4.31e-39

0.038

1.50 (1.41, 1.59)

Sex

Female

238,119

18,772

1.37

1.28

1.46

2.06e-20

0.214

1.37 (1.28, 1.46)

Sex

Male

201,926

21,427

1.47

1.38

1.57

1.72e-31

0.214

1.47 (1.38, 1.57)

BMI

Underweight

18,421

2,184

1.16

0.994

1.35

0.0603

0.006

1.16 (0.99, 1.35)

#	变量	类型	缺失	统计
1	groupSubgroup variable; Overall identifies the estimate for the full cohort.	字符型	—	不同值 4BMI · Age · Sex · Overall
2	levelPopulation level represented by the estimate.	字符型	—	不同值 9Overall · <60 years · >=60 years · Female · Male · Underweight · Normal weight · Overweight
3	nParticipants in the analytic sample.	整数	—	min 18,421q1 —中位 182,316均值 195,576q3 —max 440,045
4	n_eventsIncident type 2 diabetes cases observed during follow-up.	整数	—	min 2,184q1 —中位 15,086均值 17,866q3 —max 40,199
5	HRkeySimulated hazard ratio for incident type 2 diabetes.	双精度	—	min 1q1 —中位 1.39均值 1.38q3 —max 1.82
6	CI_lowerLower bound of the 95% confidence interval.	双精度	—	min 0.903q1 —中位 1.31均值 1.28q3 —max 1.7
7	CI_upperUpper bound of the 95% confidence interval.	双精度	—	min 1.11q1 —中位 1.48均值 1.49q3 —max 1.95
8	p_valueTwo-sided Wald p value for the effect estimate.	双精度	—	min 1.3e-65q1 —中位 7.8e-25均值 0.118q3 —max 1
9	p_interactionInteraction p value shared by levels of the same subgroup variable.	双精度	1	min 0.006q1 —中位 0.022均值 0.066q3 —max 0.214
10	HR_labelPreformatted hazard ratio and 95% confidence interval.	字符型	—	不同值 91.42 (1.35, 1.49) · 1.31 (1.21, 1.42) · 1.50 (1.41, 1.59) · 1.37 (1.28, 1.46) · 1.47 (1.38, 1.57) · 1.16 (0.99, 1.35) · 1.00 (0.90, 1.11) · 1.39 (1.31, 1.48)

变量

类型

缺失

统计

groupSubgroup variable; Overall identifies the estimate for the full cohort.

字符型

—

不同值 4BMI · Age · Sex · Overall

levelPopulation level represented by the estimate.

字符型

—

不同值 9Overall · <60 years · >=60 years · Female · Male · Underweight · Normal weight · Overweight

nParticipants in the analytic sample.

整数

—

min 18,421q1 —中位 182,316均值 195,576q3 —max 440,045

n_eventsIncident type 2 diabetes cases observed during follow-up.

整数

—

min 2,184q1 —中位 15,086均值 17,866q3 —max 40,199

HRkeySimulated hazard ratio for incident type 2 diabetes.

双精度

—

min 1q1 —中位 1.39均值 1.38q3 —max 1.82

CI_lowerLower bound of the 95% confidence interval.

双精度

—

min 0.903q1 —中位 1.31均值 1.28q3 —max 1.7

CI_upperUpper bound of the 95% confidence interval.

双精度

—

min 1.11q1 —中位 1.48均值 1.49q3 —max 1.95

p_valueTwo-sided Wald p value for the effect estimate.

双精度

—

min 1.3e-65q1 —中位 7.8e-25均值 0.118q3 —max 1

p_interactionInteraction p value shared by levels of the same subgroup variable.

双精度

min 0.006q1 —中位 0.022均值 0.066q3 —max 0.214

HR_labelPreformatted hazard ratio and 95% confidence interval.

字符型

—

不同值 91.42 (1.35, 1.49) · 1.31 (1.21, 1.42) · 1.50 (1.41, 1.59) · 1.37 (1.28, 1.46) · 1.47 (1.38, 1.57) · 1.16 (0.99, 1.35) · 1.00 (0.90, 1.11) · 1.39 (1.31, 1.48)

library(readr) forest_subgroups <- read_csv( "https://assets.evanzhou.org/tessera/csv/forest_subgroups.csv", col_types = cols( group = col_character(), level = col_character(), n = col_integer(), n_events = col_integer(), HR = col_double(), CI_lower = col_double(), CI_upper = col_double(), p_value = col_double(), p_interaction = col_double(), HR_label = col_character() ) )

Source

为亚组森林图模拟的 9 个效应估计：全队列 1 行，年龄和性别各 2 行，BMI 4 行。CSV 只保存分析结果；分组标题与缩进由绘图代码根据 group 生成。

Use cases

绘制带 Overall、Age、Sex 和 BMI 层级的亚组森林图
演示组间交互作用 p 值的展示
在绘图阶段自动插入组标题并生成缩进

# Generate the subgroup toy data used by the forest figure recipes. script_dir <- local({ arg <- grep("^--file=", commandArgs(trailingOnly = FALSE), value = TRUE) path <- if (length(arg)) sub("^--file=", "", arg[[1L]]) else sys.frame(1)$ofile dirname(normalizePath(path, mustWork = TRUE)) }) out_csv <- file.path(script_dir, "..", "csv", "forest_subgroups.csv") dir.create(dirname(out_csv), recursive = TRUE, showWarnings = FALSE) forest_subgroups <- data.frame( group = c("Overall", "Age", "Age", "Sex", "Sex", "BMI", "BMI", "BMI", "BMI"), level = c( "Overall", "<60 years", ">=60 years", "Female", "Male", "Underweight", "Normal weight", "Overweight", "Obese" ), n = c(440045L, 182316L, 257729L, 238119L, 201926L, 18421L, 156804L, 168735L, 96085L), n_events = c(40199L, 11254L, 28945L, 18772L, 21427L, 2184L, 10591L, 15086L, 12338L), HR = c(1.42, 1.31, 1.50, 1.37, 1.47, 1.16, 1.00, 1.39, 1.82), se = c(0.024, 0.041, 0.031, 0.034, 0.033, 0.079, 0.052, 0.032, 0.035), p_interaction = c(NA, 0.038, 0.038, 0.214, 0.214, 0.006, 0.006, 0.006, 0.006) ) log_hr <- log(forest_subgroups$HR) forest_subgroups$CI_lower <- exp(log_hr - 1.96 * forest_subgroups$se) forest_subgroups$CI_upper <- exp(log_hr + 1.96 * forest_subgroups$se) forest_subgroups$p_value <- 2 * pnorm(-abs(log_hr / forest_subgroups$se)) forest_subgroups$se <- NULL forest_subgroups$HR <- round(forest_subgroups$HR, 4) forest_subgroups$CI_lower <- round(forest_subgroups$CI_lower, 4) forest_subgroups$CI_upper <- round(forest_subgroups$CI_upper, 4) forest_subgroups$HR_label <- sprintf( "%.2f (%.2f, %.2f)", forest_subgroups$HR, forest_subgroups$CI_lower, forest_subgroups$CI_upper ) forest_subgroups <- forest_subgroups[, c( "group", "level", "n", "n_events", "HR", "CI_lower", "CI_upper", "p_value", "p_interaction", "HR_label" )] write.csv(forest_subgroups, out_csv, row.names = FALSE)