forest_models · Folio

exposure	term	model	n	n_events	person_years	HR	CI_lower	CI_upper	p_value	HR_label
bmi_category	Underweight	Unadjusted	440,045	40,199	5,752,833	1.22	1.08	1.39	0.00222	1.22 (1.08, 1.39)
bmi_category	Underweight	Age and sex adjusted	440,045	40,199	5,752,833	1.2	1.04	1.38	0.0105	1.20 (1.04, 1.38)
bmi_category	Underweight	Fully adjusted	440,045	40,199	5,752,833	1.18	1.02	1.37	0.0273	1.18 (1.02, 1.37)
bmi_category	Overweight	Unadjusted	440,045	40,199	5,752,833	1.53	1.47	1.6	5.18e-78	1.53 (1.47, 1.60)
bmi_category	Overweight	Age and sex adjusted	440,045	40,199	5,752,833	1.47	1.4	1.54	3.95e-55	1.47 (1.40, 1.54)
bmi_category	Overweight	Fully adjusted	440,045	40,199	5,752,833	1.42	1.35	1.49	1.87e-41	1.42 (1.35, 1.49)

exposure

term

model

n_events

person_years

CI_lower

CI_upper

p_value

HR_label

bmi_category

Underweight

Unadjusted

440,045

40,199

5,752,833

1.22

1.08

1.39

0.00222

1.22 (1.08, 1.39)

bmi_category

Underweight

Age and sex adjusted

440,045

40,199

5,752,833

1.2

1.04

1.38

0.0105

1.20 (1.04, 1.38)

bmi_category

Underweight

Fully adjusted

440,045

40,199

5,752,833

1.18

1.02

1.37

0.0273

1.18 (1.02, 1.37)

bmi_category

Overweight

Unadjusted

440,045

40,199

5,752,833

1.53

1.47

1.6

5.18e-78

1.53 (1.47, 1.60)

bmi_category

Overweight

Age and sex adjusted

440,045

40,199

5,752,833

1.47

1.4

1.54

3.95e-55

1.47 (1.40, 1.54)

bmi_category

Overweight

Fully adjusted

440,045

40,199

5,752,833

1.42

1.35

1.49

1.87e-41

1.42 (1.35, 1.49)

#	变量	类型	缺失	统计
1	exposureExposure represented by all rows in the table.	字符型	—	不同值 1bmi_category
2	termBMI category compared with the normal-weight reference group.	字符型	—	不同值 3Underweight · Overweight · Obese
3	modelAdjustment level used by the Cox model.	字符型	—	不同值 3Unadjusted · Age and sex adjusted · Fully adjusted
4	nParticipants in the analytic sample.	整数	—	min 440,045q1 —中位 440,045均值 440,045q3 —max 440,045
5	n_eventsIncident type 2 diabetes cases observed during follow-up.	整数	—	min 40,199q1 —中位 40,199均值 40,199q3 —max 40,199
6	person_yearsTotal follow-up time, in person-years.	整数	—	min 5.8e+6q1 —中位 5.8e+6均值 5.8e+6q3 —max 5.8e+6
7	HRkeySimulated hazard ratio for incident type 2 diabetes.	双精度	—	min 1.18q1 —中位 1.47均值 1.57q3 —max 2.2
8	CI_lowerLower bound of the 95% confidence interval.	双精度	—	min 1.02q1 —中位 1.4均值 1.46q3 —max 2.09
9	CI_upperUpper bound of the 95% confidence interval.	双精度	—	min 1.37q1 —中位 1.54均值 1.69q3 —max 2.32
10	p_valueTwo-sided Wald p value derived from the estimate and standard error.	双精度	—	min 4.0e-210q1 —中位 4.0e-55均值 0.00445q3 —max 0.0273
11	HR_labelPreformatted hazard ratio and 95% confidence interval.	字符型	—	不同值 91.22 (1.08, 1.39) · 1.20 (1.04, 1.38) · 1.18 (1.02, 1.37) · 1.53 (1.47, 1.60) · 1.47 (1.40, 1.54) · 1.42 (1.35, 1.49) · 2.20 (2.09, 2.32) · 2.02 (1.92, 2.14)

变量

类型

缺失

统计

exposureExposure represented by all rows in the table.

字符型

—

不同值 1bmi_category

termBMI category compared with the normal-weight reference group.

字符型

—

不同值 3Underweight · Overweight · Obese

modelAdjustment level used by the Cox model.

字符型

—

不同值 3Unadjusted · Age and sex adjusted · Fully adjusted

nParticipants in the analytic sample.

整数

—

min 440,045q1 —中位 440,045均值 440,045q3 —max 440,045

n_eventsIncident type 2 diabetes cases observed during follow-up.

整数

—

min 40,199q1 —中位 40,199均值 40,199q3 —max 40,199

person_yearsTotal follow-up time, in person-years.

整数

—

min 5.8e+6q1 —中位 5.8e+6均值 5.8e+6q3 —max 5.8e+6

HRkeySimulated hazard ratio for incident type 2 diabetes.

双精度

—

min 1.18q1 —中位 1.47均值 1.57q3 —max 2.2

CI_lowerLower bound of the 95% confidence interval.

双精度

—

min 1.02q1 —中位 1.4均值 1.46q3 —max 2.09

CI_upperUpper bound of the 95% confidence interval.

双精度

—

min 1.37q1 —中位 1.54均值 1.69q3 —max 2.32

p_valueTwo-sided Wald p value derived from the estimate and standard error.

双精度

—

min 4.0e-210q1 —中位 4.0e-55均值 0.00445q3 —max 0.0273

HR_labelPreformatted hazard ratio and 95% confidence interval.

字符型

—

不同值 91.22 (1.08, 1.39) · 1.20 (1.04, 1.38) · 1.18 (1.02, 1.37) · 1.53 (1.47, 1.60) · 1.47 (1.40, 1.54) · 1.42 (1.35, 1.49) · 2.20 (2.09, 2.32) · 2.02 (1.92, 2.14)

library(readr) forest_models <- read_csv( "https://assets.evanzhou.org/tessera/csv/forest_models.csv", col_types = cols( exposure = col_character(), term = col_character(), model = col_character(), n = col_integer(), n_events = col_integer(), person_years = col_integer(), HR = col_double(), CI_lower = col_double(), CI_upper = col_double(), p_value = col_double(), HR_label = col_character() ) )

Source

为森林图模型比较示例模拟的 Cox 结果。一个 BMI 分类暴露包含三个非参考水平，每个水平分别展示未校正、年龄与性别校正及完全校正模型，共 9 行。所有数值均为模拟值。

Use cases

比较同一暴露在三种校正程度下的效应估计
演示 assoc_coxph() 风格结果表直接传入森林图
展示分类暴露的多个非参考水平

# Generate the model-comparison 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)$ofile dirname(normalizePath(path, mustWork = TRUE)) }) out_csv <- file.path(script_dir, "..", "csv", "forest_models.csv") dir.create(dirname(out_csv), recursive = TRUE, showWarnings = FALSE) terms <- data.frame( term = c("Underweight", "Overweight", "Obese"), n = c(440045L, 440045L, 440045L), n_events = c(40199L, 40199L, 40199L), person_years = c(5752833L, 5752833L, 5752833L), hr_full = c(1.18, 1.42, 1.91), se_full = c(0.075, 0.026, 0.029) ) models <- data.frame( model = c("Unadjusted", "Age and sex adjusted", "Fully adjusted"), effect_mul = c(1.22, 1.09, 1.00), se_mul = c(0.88, 0.94, 1.00) ) forest_models <- do.call(rbind, lapply(seq_len(nrow(terms)), function(i) { do.call(rbind, lapply(seq_len(nrow(models)), function(j) { log_hr <- log(terms$hr_full[i]) * models$effect_mul[j] se <- terms$se_full[i] * models$se_mul[j] data.frame( exposure = "bmi_category", term = terms$term[i], model = models$model[j], n = terms$n[i], n_events = terms$n_events[i], person_years = terms$person_years[i], HR = exp(log_hr), CI_lower = exp(log_hr - 1.96 * se), CI_upper = exp(log_hr + 1.96 * se), p_value = 2 * pnorm(-abs(log_hr / se)) ) })) })) forest_models$HR <- round(forest_models$HR, 4) forest_models$CI_lower <- round(forest_models$CI_lower, 4) forest_models$CI_upper <- round(forest_models$CI_upper, 4) forest_models$HR_label <- sprintf( "%.2f (%.2f, %.2f)", forest_models$HR, forest_models$CI_lower, forest_models$CI_upper ) write.csv(forest_models, out_csv, row.names = FALSE)