Using atrial myopathy metrics to devise a risk prediction tool for atrial fibrillation in endurance athletes
October 2025
Code Transparency Statement This document contains the complete analysis code for the development of a clinical risk prediction tool for atrial fibrillation (AF) in endurance athletes using multimodal atrial myopathy markers. All code is provided in the interest of transparency, reproducibility, and to facilitate independent verification of our findings. Researchers are welcome to use and adapt this code for their own studies.
Corresponding Author: Professor Andre La Gerche: andre.lagerche@svi.edu.au
Citation: Spencer LW, D’Ambrosio P, et al. Using atrial myopathy metrics to devise a risk prediction tool for atrial fibrillation in endurance athletes. Manuscript in submission for European Heart Journal.
AF risk prediction model: We have developed a AF risk prediction tool available at: https://heart-lab.org/atrial-fibrillation
1 Analysis Setup and Data Preparation
1.1 R packages
# Data manipulation
library(tidyverse)
library(dplyr)
library(readr)
# Statistical analysis
library(tableone)
library(broom)
library(pROC)
library(ResourceSelection)
# Visualisation
library(ggplot2)
library(ggpubr)
library(gridExtra)
library(scales)
library(patchwork)
library(viridis)
library(fmsb)
# Output
library(openxlsx)
library(writexl)
library(knitr)
library(kableExtra)
# Set options
options(scipen = 999) # To avoid scientific notation
set.seed(123) # Reproducibility1.2 Filter and Create Analysis Cohorts
# Remove Atrial Myopathy related pathology
NoCM <- filter(AM_paper, `AM related pathology` != 1)
# Main athlete cohort (excluding young/middle-age controls)
Athletes_NoCM <- filter(NoCM,
Cohort != "young control" &
Cohort != "middle age control")
# Age groups and AF status variables
Athletes_NoCM <- Athletes_NoCM %>%
mutate(
Age = as.numeric(Age),
AgeGroup = factor(ifelse(Age <= 45, "≤45", ">45"),
levels = c("≤45", ">45")),
AFgroup = factor(ifelse(AF == 1, "AF+", "AF-"),
levels = c("AF-", "AF+")),
AF_numeric = as.numeric(AF)
)
# Control and athlete groups for supplementary analysis
controls <- filter(NoCM, EICR == "Control")
athletes <- filter(NoCM, EICR == "Athlete")
# Cohort summary table
cohort_summary <- data.frame(
Group = c("Athletes (excluding cardiomyopathy)",
" AF+",
" AF-",
" Age ≤45 years",
" Age >45 years",
"Controls"),
N = c(
nrow(Athletes_NoCM),
sum(Athletes_NoCM$AF == 1, na.rm = TRUE),
sum(Athletes_NoCM$AF == 0, na.rm = TRUE),
sum(Athletes_NoCM$AgeGroup == "≤45", na.rm = TRUE),
sum(Athletes_NoCM$AgeGroup == ">45", na.rm = TRUE),
nrow(controls)
)
)
kable(cohort_summary,
col.names = c("Cohort Description", "N"),
align = c("l", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover"),
full_width = FALSE)| Cohort Description | N |
|---|---|
| Athletes (excluding cardiomyopathy) | 353 |
| AF+ | 96 |
| AF- | 257 |
| Age ≤45 years | 201 |
| Age >45 years | 152 |
| Controls | 97 |
1.3 Variable Definitions
# Table 1 variables
table1_vars <- c(
# Demographics
"Age", "sex", "bmi", "Bsa", "averagehr1", "minhr1", "brad_burden_pct1",
"htn", "rest_SBP", "rest_DBP", "high_cholesterol", "diabetes",
"ex_smoker", "etoh sd/wk",
# Medications
"bblocker", "nondihydropyridine ccb", "flec", "sotalol", "amio",
# Exercise
"vo2_kg_peak", "vo2_absolute", "Total MET hrs/wk", "Exercise Duration (yr)",
# Sport
"Rowing", "Cycling", "Running", "Other"
)
# Table 2 variables
table2_vars <- c(
# LV
"enddiastolic_volume_ml_lv", "endsystolic_volume_ml_lv",
"lvef_bip_q", "g_peak_slavg_",
# LA
"laesv_bip_ml", "LAVi", "la/lv",
"LA Resevoir Strain", "LA Conduit Strain", "LA Contractile Strain",
# RV
"RVEDV", "RVESV", "RVEF",
# Doppler
"MV E Vel (m/s)", "MV A Vel (m/s)", "MV E/A Ratio",
"Sept E' (cm/s)", "E/E'",
# ECG
"P wave Duration", "P Wave Axis", "PTFV1 (ms*mV)", "Adv IAB",
"ECTOPIES LS",
# Biomarkers
"bnp", "troponin1",
# Genetic
"AF centile", "highest quartile", "highest decile"
)
#Sport
Athletes_NoCM <- Athletes_NoCM %>%
mutate(
Rowing = as.numeric(`Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other` == 1),
Cycling = as.numeric(`Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other` == 2),
Running = as.numeric(`Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other` == 3),
Other = as.numeric(`Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other` %in% c(4, 5, 6, 7))
)
# Supplementary variables
supp_vars <- c(
"% predicted vo2", "peak rer",
"LAVi > 34", "LAVi > 48", "LAVi > 60",
"Prevalence of P-wave duration >120 ms", "High PTFV1",
"Prevalence of >=100 PACs/24 h",
"AF PRS 6238158var PGS000016", "highest quintile", "98th percentile",
"Rare Variant"
)
all_variables <- unique(c(table1_vars, table2_vars, supp_vars))
all_variables <- all_variables[all_variables %in% names(Athletes_NoCM)]
# Non-normal distributions
nonnormal_vars <- c("Age", "bmi", "Bsa", "rest_SBP", "rest_DBP", "minhr1",
"brad_burden_pct1", "Exercise Duration (yr)", "Total MET hrs/wk",
"laesv_bip_ml", "LAVi", "la/lv", "P wave Duration", "PTFV1 (ms*mV)",
"ECTOPIES LS", "bnp", "troponin1", "AF centile",
"AF PRS 6238158var PGS000016", "etoh sd/wk")
# Categorical variables
categorical_vars <- c("sex", "htn", "high_cholesterol", "diabetes", "ex_smoker",
"bblocker", "nondihydropyridine ccb", "flec", "sotalol", "amio",
"Rowing", "Cycling", "Running", "Other",
"LAVi > 34", "LAVi > 48", "LAVi > 60",
"Prevalence of P-wave duration >120 ms", "High PTFV1",
"Prevalence of >=100 PACs/24 h",
"Adv IAB", "highest quartile", "highest quintile",
"highest decile", "98th percentile", "Rare Variant")1.4 Statistical Functions
calculate_p_value <- function(data, var, group_var, is_categorical = FALSE, is_nonnormal = NULL) {
if (!var %in% names(data) || !group_var %in% names(data)) return(NA)
clean_data <- data %>% filter(!is.na(.data[[var]]) & !is.na(.data[[group_var]]))
if (nrow(clean_data) < 3) return(NA)
group_levels <- unique(clean_data[[group_var]])
if (length(group_levels) < 2) return(NA)
tryCatch({
if (is_categorical) {
tab <- table(as.factor(clean_data[[var]]), as.factor(clean_data[[group_var]]))
if (any(dim(tab) < 2) || sum(tab) < 5) return(NA)
if (any(tab < 5)) {
test_result <- fisher.test(tab, simulate.p.value = TRUE)
} else {
test_result <- chisq.test(tab)
}
return(test_result$p.value)
} else {
var_numeric <- as.numeric(clean_data[[var]])
if (all(is.na(var_numeric))) return(NA)
if (is.null(is_nonnormal)) {
n <- length(var_numeric[!is.na(var_numeric)])
if (n >= 3 && n <= 5000) {
shapiro_result <- tryCatch(shapiro.test(var_numeric),
error = function(e) list(p.value = 0.049))
is_nonnormal <- shapiro_result$p.value < 0.05
} else {
is_nonnormal <- TRUE
}
}
if (is_nonnormal) {
test_result <- wilcox.test(var_numeric ~ clean_data[[group_var]])
} else {
test_result <- t.test(var_numeric ~ clean_data[[group_var]])
}
return(test_result$p.value)
}
}, error = function(e) NA)
}
calculate_interaction_p <- function(data, var, is_categorical = FALSE) {
required_vars <- c(var, "AF_numeric", "AgeGroup")
if (!all(required_vars %in% names(data))) return(NA)
clean_data <- data %>%
filter(!is.na(.data[[var]]) & !is.na(AF_numeric) & !is.na(AgeGroup))
if (nrow(clean_data) < 20) return(NA)
tryCatch({
if (is_categorical) {
model <- glm(as.factor(clean_data[[var]]) ~ AgeGroup * AFgroup,
data = clean_data, family = binomial())
} else {
model <- lm(as.numeric(clean_data[[var]]) ~ AgeGroup * AFgroup,
data = clean_data)
}
anova_result <- anova(model)
int_row <- grep(":", rownames(anova_result))
if (length(int_row) > 0) {
p_col <- which(names(anova_result) %in% c("Pr(>F)", "Pr(>Chi)"))
if (length(p_col) > 0) return(anova_result[int_row[1], p_col[1]])
}
return(NA)
}, error = function(e) NA)
}
format_stats <- function(data, var, is_cat, is_nonnorm) {
if (!var %in% names(data) || nrow(data) == 0) return("--")
tryCatch({
if (is_cat) {
var_data <- data[[var]]
if (is.logical(var_data)) {
count <- sum(var_data == TRUE, na.rm = TRUE)
total <- sum(!is.na(var_data))
} else if (is.numeric(var_data) && all(var_data %in% c(0, 1, NA), na.rm = TRUE)) {
count <- sum(var_data == 1, na.rm = TRUE)
total <- sum(!is.na(var_data))
} else {
tab <- table(var_data, useNA = "no")
if (length(tab) > 0) {
count <- max(tab)
total <- sum(tab)
} else return("0 (0%)")
}
if (total == 0) return("0 (0%)")
return(paste0(count, " (", round(100*count/total, 1), "%)"))
} else {
values <- as.numeric(data[[var]])
values <- values[!is.na(values)]
if (length(values) == 0) return("--")
decimals <- ifelse(var == "la/lv", 2, 1)
if (is_nonnorm) {
med <- median(values)
q1 <- quantile(values, 0.25)
q3 <- quantile(values, 0.75)
return(paste0(round(med, decimals), " [",
round(q1, decimals), "-", round(q3, decimals), "]"))
} else {
m <- mean(values)
s <- sd(values)
return(paste0(round(m, decimals), " ± ", round(s, decimals)))
}
}
}, error = function(e) "--")
}
format_p_value <- function(p) {
if (is.null(p) || length(p) == 0 || is.na(p)) return("--")
p_num <- as.numeric(p)
if (is.na(p_num)) return("--")
if (p_num < 0.001) return("<0.001")
if (p_num < 0.01) return(sprintf("%.3f", p_num))
return(sprintf("%.3f", p_num))
}
calculate_sport_p_value <- function(data, group_var) {
sport_var <- "Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other"
if (!sport_var %in% names(data) || !group_var %in% names(data)) return(NA)
clean_data <- data %>%
filter(!is.na(.data[[sport_var]]) & !is.na(.data[[group_var]]))
if (nrow(clean_data) < 3) return(NA)
tryCatch({
clean_data$sport_collapsed <- ifelse(
clean_data[[sport_var]] %in% c(4, 5, 6, 7),
4,
clean_data[[sport_var]]
)
tab <- table(clean_data$sport_collapsed, clean_data[[group_var]])
if (any(dim(tab) < 2)) return(NA)
if (any(tab < 5)) {
test_result <- fisher.test(tab, simulate.p.value = TRUE)
} else {
test_result <- chisq.test(tab)
}
return(test_result$p.value)
}, error = function(e) NA)
}1.5 Helper Functions
assign_category <- function(var) {
# Table 1 categories
if (var %in% c("Age", "sex", "bmi", "Bsa", "averagehr1", "minhr1",
"brad_burden_pct1", "htn", "rest_SBP", "rest_DBP",
"high_cholesterol", "diabetes", "ex_smoker", "etoh sd/wk")) {
return("Demographics")
} else if (var %in% c("bblocker", "nondihydropyridine ccb", "flec",
"sotalol", "amio")) {
return("Anti-arrhythmic Medication use, n (%)")
} else if (var %in% c("vo2_kg_peak", "vo2_absolute", "Total MET hrs/wk",
"Exercise Duration (yr)")) {
return("Exercise Capacity")
} else if (var == "Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other" ||
var %in% c("Rowing", "Cycling", "Running", "Other")) {
return("Primary Sport, n (%)")
} else if (var %in% c("enddiastolic_volume_ml_lv", "endsystolic_volume_ml_lv",
"lvef_bip_q", "g_peak_slavg_")) {
return("Left ventricular parameters")
} else if (var %in% c("laesv_bip_ml", "LAVi", "la/lv", "LAVi > 34",
"LAVi > 48", "LAVi > 60",
"LA Resevoir Strain", "LA Conduit Strain",
"LA Contractile Strain")) {
return("Left atrial parameters")
} else if (var %in% c("RVEDV", "RVESV", "RVEF")) {
return("Right Ventricular Parameters")
} else if (var %in% c("MV E Vel (m/s)", "MV A Vel (m/s)", "MV E/A Ratio",
"Sept E' (cm/s)", "E/E'")) {
return("Doppler Parameters")
} else if (var %in% c("P wave Duration", "P Wave Axis", "PTFV1 (ms*mV)",
"Prevalence of P-wave duration >120 ms", "High PTFV1",
"Adv IAB", "ECTOPIES LS")) {
return("Electrocardiographic")
} else if (var %in% c("bnp", "troponin1")) {
return("Biochemical")
} else if (var %in% c("AF centile", "AF PRS 6238158var PGS000016",
"highest quartile", "highest quintile", "highest decile",
"98th percentile", "Rare Variant")) {
return("Genetic")
} else {
return("Other")
}
}
clean_variable_names <- function(var) {
case_when(
var == "Age" ~ "Age, years",
var == "sex" ~ "Male sex, n (%)",
var == "bmi" ~ "BMI, kg/m²",
var == "Bsa" ~ "BSA, m²",
var == "averagehr1" ~ "Average HR, bpm",
var == "minhr1" ~ "Minimum HR, bpm",
var == "brad_burden_pct1" ~ "Bradycardia burden, % (24h)",
var == "htn" ~ "Hypertension, n (%)",
var == "rest_SBP" ~ "SBP, mmHg",
var == "rest_DBP" ~ "DBP, mmHg",
var == "high_cholesterol" ~ "Dyslipidaemia, n (%)",
var == "diabetes" ~ "Diabetes, n (%)",
var == "ex_smoker" ~ "History of smoking, n (%)",
var == "etoh sd/wk" ~ "Alcohol, drinks/week",
var == "bblocker" ~ "Beta-blockers",
var == "nondihydropyridine ccb" ~ "Non-dihydropyridine CCB's",
var == "flec" ~ "Flecainide",
var == "sotalol" ~ "Sotalol",
var == "amio" ~ "Amiodarone",
var == "vo2_kg_peak" ~ "VO₂peak, mL/kg/min",
var == "vo2_absolute" ~ "VO₂peak, L/min",
var == "% predicted vo2" ~ "% of predicted exercise capacity",
var == "peak rer" ~ "Respiratory exchange ratio",
var == "Total MET hrs/wk" ~ "Total weekly MET-hours",
var == "Exercise Duration (yr)" ~ "Yrs of endurance training",
var == "Rowing" ~ "Rowing",
var == "Cycling" ~ "Cycling",
var == "Running" ~ "Running",
var == "Other" ~ "Other",
var == "Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other" ~ "Primary Sport, n (%)",
var == "enddiastolic_volume_ml_lv" ~ "LVEDV, mL",
var == "endsystolic_volume_ml_lv" ~ "LVESV, mL",
var == "lvef_bip_q" ~ "LVEF, %",
var == "g_peak_slavg_" ~ "LV strain, %",
var == "laesv_bip_ml" ~ "LAESV, mL",
var == "LAVi" ~ "LAVi, mL/m²",
var == "LAVi > 34" ~ "Mildly abnormal >34 mL/m², n (%)",
var == "LAVi > 48" ~ "Severely abnormal >48 mL/m², n (%)",
var == "LAVi > 60" ~ "Extremely abnormal >60 mL/m², n (%)",
var == "la/lv" ~ "LA:LV",
var == "LA Resevoir Strain" ~ "Reservoir strain, %",
var == "LA Conduit Strain" ~ "Conduit strain, %",
var == "LA Contractile Strain" ~ "Contractile strain, %",
var == "RVEDV" ~ "RVEDV, mL",
var == "RVESV" ~ "RVESV, mL",
var == "RVEF" ~ "RVEF, %",
var == "MV E Vel (m/s)" ~ "E velocity, m/s",
var == "MV A Vel (m/s)" ~ "A velocity, m/s",
var == "MV E/A Ratio" ~ "E/A ratio",
var == "Sept E' (cm/s)" ~ "E′ velocity, cm/s",
var == "E/E'" ~ "E/E' ratio",
var == "P wave Duration" ~ "P wave duration, ms",
var == "Prevalence of P-wave duration >120 ms" ~ "P-wave duration >120 ms, n (%)",
var == "P Wave Axis" ~ "P wave axis, °",
var == "PTFV1 (ms*mV)" ~ "PTFV₁, mV·ms",
var == "High PTFV1" ~ "PTFV₁ ≥4 mV·ms, n (%)",
var == "Adv IAB" ~ "Advanced IAB, n (%)",
var == "ECTOPIES LS" ~ "Atrial ectopy, per 24 h",
var == "Prevalence of >=100 PACs/24 h" ~ "≥100 PACs per 24 h, n (%)",
var == "bnp" ~ "BNP, pg/mL",
var == "troponin1" ~ "cTnI, ng/L",
var == "AF centile" ~ "AF-PRS percentile",
var == "AF PRS 6238158var PGS000016" ~ "AF-PRS",
var == "highest quartile" ~ "AF-PRS highest quartile",
var == "highest quintile" ~ "AF-PRS highest quintile",
var == "highest decile" ~ "AF-PRS highest decile",
var == "98th percentile" ~ "AF-PRS ≥98th percentile",
var == "Rare Variant" ~ "Rare Variant",
TRUE ~ var
)
}1.6 Table Setup
results_table <- data.frame(Variable = character(), stringsAsFactors = FALSE)
for (var in all_variables) {
if (!var %in% names(Athletes_NoCM)) next
is_cat <- var %in% categorical_vars
is_nonnorm <- var %in% nonnormal_vars
af_overall <- Athletes_NoCM %>% filter(AF == 1)
noaf_overall <- Athletes_NoCM %>% filter(AF == 0)
young_data <- Athletes_NoCM %>% filter(AgeGroup == "≤45")
af_young <- young_data %>% filter(AF == 1)
noaf_young <- young_data %>% filter(AF == 0)
old_data <- Athletes_NoCM %>% filter(AgeGroup == ">45")
af_old <- old_data %>% filter(AF == 1)
noaf_old <- old_data %>% filter(AF == 0)
# Calculate p-values - Sport
if (var %in% c("Rowing", "Cycling", "Running", "Other")) {
if (var == "Rowing") {
p_overall <- format_p_value(calculate_sport_p_value(Athletes_NoCM, "AFgroup"))
p_young <- format_p_value(calculate_sport_p_value(young_data, "AFgroup"))
p_old <- format_p_value(calculate_sport_p_value(old_data, "AFgroup"))
p_interaction <- format_p_value(calculate_interaction_p(Athletes_NoCM,
"Sport: 1=row, 2=cyc, 3=run, 4=CCS, 5=swim, 6=tri, 7=other", is_cat = TRUE))
} else {
p_overall <- ""
p_young <- ""
p_old <- ""
p_interaction <- ""
}
} else {
p_overall <- format_p_value(calculate_p_value(Athletes_NoCM, var, "AFgroup", is_cat, is_nonnorm))
p_young <- format_p_value(calculate_p_value(young_data, var, "AFgroup", is_cat, is_nonnorm))
p_old <- format_p_value(calculate_p_value(old_data, var, "AFgroup", is_cat, is_nonnorm))
p_interaction <- format_p_value(calculate_interaction_p(Athletes_NoCM, var, is_cat))
}
results_table <- rbind(results_table, data.frame(
Variable = var,
AF_Overall = format_stats(af_overall, var, is_cat, is_nonnorm),
NoAF_Overall = format_stats(noaf_overall, var, is_cat, is_nonnorm),
P_Overall = p_overall,
AF_Young = format_stats(af_young, var, is_cat, is_nonnorm),
NoAF_Young = format_stats(noaf_young, var, is_cat, is_nonnorm),
P_Young = p_young,
AF_Old = format_stats(af_old, var, is_cat, is_nonnorm),
NoAF_Old = format_stats(noaf_old, var, is_cat, is_nonnorm),
P_Old = p_old,
P_Interaction = p_interaction,
stringsAsFactors = FALSE
))
}
results_table$Category <- sapply(results_table$Variable, assign_category)
results_table$Variable <- sapply(results_table$Variable, clean_variable_names)
n_af_overall <- sum(Athletes_NoCM$AF == 1, na.rm = TRUE)
n_noaf_overall <- sum(Athletes_NoCM$AF == 0, na.rm = TRUE)
n_af_young <- sum(Athletes_NoCM$AF == 1 & Athletes_NoCM$AgeGroup == "≤45", na.rm = TRUE)
n_noaf_young <- sum(Athletes_NoCM$AF == 0 & Athletes_NoCM$AgeGroup == "≤45", na.rm = TRUE)
n_af_old <- sum(Athletes_NoCM$AF == 1 & Athletes_NoCM$AgeGroup == ">45", na.rm = TRUE)
n_noaf_old <- sum(Athletes_NoCM$AF == 0 & Athletes_NoCM$AgeGroup == ">45", na.rm = TRUE)
# Define category orders
category_order_table1 <- c("Demographics", "Anti-arrhythmic Medication use, n (%)",
"Exercise Capacity", "Primary Sport, n (%)")
category_order_table2 <- c("Left ventricular parameters", "Left atrial parameters",
"Right Ventricular Parameters", "Doppler Parameters",
"Electrocardiographic", "Biochemical", "Genetic")
# Ordered category tables
table1_with_cat <- results_table %>%
filter(Category %in% category_order_table1) %>%
mutate(Category = factor(Category, levels = category_order_table1, ordered = TRUE)) %>%
arrange(Category)
table2_with_cat <- results_table %>%
filter(Category %in% category_order_table2) %>%
mutate(Category = factor(Category, levels = category_order_table2, ordered = TRUE)) %>%
arrange(Category)
# Create display tables
table_1 <- table1_with_cat %>% select(-Category)
table_2 <- table2_with_cat %>% select(-Category)
# Rename columns
colnames(table_1) <- c(
"Characteristic",
paste0("AF (n=", n_af_overall, ")"),
paste0("No-AF (n=", n_noaf_overall, ")"),
"p value",
paste0("AF (n=", n_af_young, ")"),
paste0("No-AF (n=", n_noaf_young, ")"),
"p value ",
paste0("AF (n=", n_af_old, ")"),
paste0("No-AF (n=", n_noaf_old, ")"),
"p value ",
"Interaction p value"
)
colnames(table_2) <- colnames(table_1)
cat("\n# Table 1\n\n")2 Table 1
kable(table_1,
caption = "Table 1. Baseline Characteristics of Endurance Athletes Stratified by Age and Atrial Fibrillation Status",
align = c("l", rep("c", ncol(table_1)-1))) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = TRUE, font_size = 10) %>%
add_header_above(c(" " = 1,
"Overall" = 3,
"Younger Athletes (≤45 years)" = 3,
"Older Athletes (>45 years)" = 3,
" " = 1)) %>%
pack_rows(index = table(table1_with_cat$Category)) %>%
add_footnote(c("Data are presented as mean ± SD, median (IQR), or n (%)",
"P-values from Wilcoxon rank-sum test (continuous) or chi-square/Fisher exact test (categorical)",
"Interaction p-values derived from two-way ANOVA testing for age×AF interaction effects"),
notation = "none")| Characteristic | AF (n=96) | No-AF (n=257) | p value | AF (n=25) | No-AF (n=176) | p value | AF (n=71) | No-AF (n=81) | p value | Interaction p value |
|---|---|---|---|---|---|---|---|---|---|---|
| Demographics | ||||||||||
| Age, years | 60 [45-68] | 31 [20-51] | <0.001 | 35 [30-43] | 22 [18.8-32] | <0.001 | 64 [57-68.5] | 60 [52-66] | 0.009 | 0.026 |
| Male sex, n (%) | 89 (92.7%) | 173 (67.3%) | <0.001 | 21 (84%) | 120 (68.2%) | 0.159 | 68 (95.8%) | 53 (65.4%) | <0.001 | 0.063 |
| BMI, kg/m² | 24.6 [23.5-26.7] | 23 [21.2-24.9] | <0.001 | 24.5 [23.8-25.6] | 22.1 [20.5-23.6] | <0.001 | 24.8 [23.4-27.4] | 25 [23.3-27.4] | 0.762 | <0.001 |
| BSA, m² | 2 [1.9-2.1] | 1.9 [1.8-2] | <0.001 | 2 [1.9-2.1] | 1.9 [1.7-2] | <0.001 | 2 [1.9-2.1] | 2 [1.9-2.2] | 0.638 | 0.004 |
| Average HR, bpm | 64.8 ± 9 | 66.9 ± 9.1 | 0.084 | 64.5 ± 8.3 | 66.6 ± 9.6 | 0.282 | 64.9 ± 9.4 | 67.5 ± 7.6 | 0.104 | 0.852 |
| Minimum HR, bpm | 45 [40-49] | 43 [39-48] | 0.332 | 44.5 [40.2-48.8] | 42 [38-47] | 0.263 | 45 [40-50] | 46 [43-50] | 0.140 | 0.091 |
| Bradycardia burden, % (24h) | 4.5 [0.2-23.4] | 7.4 [0.4-28.3] | 0.167 | 12.5 [0.4-26.7] | 16.9 [0.8-33] | 0.304 | 3.5 [0.2-20.2] | 1.6 [0.2-8.6] | 0.270 | 0.121 |
| Hypertension, n (%) | 14 (14.9%) | 17 (6.7%) | 0.029 | 2 (8.3%) | 3 (1.7%) | 0.111 | 12 (17.1%) | 14 (17.5%) | 1.000 | 0.127 |
| SBP, mmHg | 132 [124-143.2] | 123 [115-134] | <0.001 | 126 [122-135] | 122 [113-131] | 0.027 | 133 [126.5-145.5] | 130 [119-138] | 0.043 | 0.884 |
| DBP, mmHg | 76 [69.8-82.2] | 66 [60.8-75] | <0.001 | 70 [64-76] | 64 [59-70] | 0.007 | 78 [72-83] | 75 [68-80] | 0.068 | 0.305 |
| Dyslipidaemia, n (%) | 23 (24.5%) | 22 (8.6%) | <0.001 | 3 (12.5%) | 1 (0.6%) | 0.006 | 20 (28.6%) | 21 (26.2%) | 0.893 | 0.006 |
| Diabetes, n (%) | 2 (2.1%) | 0 (0%) | 0.072 | 0 (0%) | 0 (0%) | – | 2 (2.9%) | 0 (0%) | 0.216 | 1.000 |
| History of smoking, n (%) | 22 (23.4%) | 25 (9.8%) | 0.002 | 2 (8.3%) | 4 (2.3%) | 0.155 | 20 (28.6%) | 21 (26.2%) | 0.893 | 0.226 |
| Alcohol, drinks/week | 3 [0-7] | 0 [0-5] | <0.001 | 0.5 [0-3.5] | 0 [0-0] | 0.007 | 4.5 [1.2-8.8] | 7 [4-12] | 0.061 | 0.047 |
| Anti-arrhythmic Medication use, n (%) | ||||||||||
| Beta-blockers | 11 (11.7%) | 1 (0.4%) | <0.001 | 2 (8.3%) | 0 (0%) | 0.014 | 9 (12.9%) | 1 (1.2%) | 0.006 | 0.218 |
| Non-dihydropyridine CCB’s | 2 (50%) | 0 (0%) | 1.000 | 0 (0%) | 0 (0%) | – | 2 (50%) | 0 (0%) | 1.000 | – |
| Flecainide | 12 (12.5%) | 1 (0.4%) | <0.001 | 3 (12%) | 0 (0%) | 0.002 | 9 (12.7%) | 1 (1.2%) | 0.006 | 0.154 |
| Sotalol | 2 (2.1%) | 0 (0%) | 0.074 | 0 (0%) | 0 (0%) | – | 2 (2.8%) | 0 (0%) | 0.219 | 1.000 |
| Amiodarone | 1 (1%) | 0 (0%) | 0.274 | 0 (0%) | 0 (0%) | – | 1 (1.4%) | 0 (0%) | 0.470 | 1.000 |
| Exercise Capacity | ||||||||||
| VO₂peak, mL/kg/min | 41.3 ± 11.5 | 51.7 ± 14.2 | <0.001 | 51.6 ± 8.9 | 58.7 ± 10.1 | <0.001 | 37.6 ± 10 | 36.5 ± 8.9 | 0.469 | 0.002 |
| VO₂peak, L/min | 3378.2 ± 947.3 | 3778.3 ± 995.5 | <0.001 | 4209.6 ± 809 | 4148.5 ± 876.5 | 0.729 | 3085.5 ± 812.3 | 2974 ± 731 | 0.378 | 0.821 |
| Total weekly MET-hours | 74 [45-91] | 87 [55-109.5] | 0.007 | 81 [57-106] | 98 [74-122] | 0.075 | 74 [44-86.5] | 64 [41-87] | 0.535 | 0.056 |
| Yrs of endurance training | 26 [17-36.5] | 17 [8.5-26] | <0.001 | 15 [10-18] | 11 [4-18] | 0.066 | 30 [20-40] | 29 [20-39] | 0.395 | 0.567 |
| Primary Sport, n (%) | ||||||||||
| Rowing | 43 (44.8%) | 125 (48.6%) | 0.127 | 5 (20%) | 47 (26.7%) | 0.342 | 38 (53.5%) | 78 (96.3%) | <0.001 | <0.001 |
| Cycling | 23 (24%) | 77 (30%) | 9 (36%) | 75 (42.6%) | 14 (19.7%) | 2 (2.5%) | ||||
| Running | 10 (10.4%) | 26 (10.1%) | 3 (12%) | 25 (14.2%) | 7 (9.9%) | 1 (1.2%) | ||||
| Other | 20 (20.8%) | 29 (11.3%) | 8 (32%) | 29 (16.5%) | 12 (16.9%) | 0 (0%) | ||||
| Data are presented as mean ± SD, median (IQR), or n (%) | ||||||||||
| P-values from Wilcoxon rank-sum test (continuous) or chi-square/Fisher exact test (categorical) | ||||||||||
| Interaction p-values derived from two-way ANOVA testing for age×AF interaction effects | ||||||||||
cat("\n# Table 2\n\n")3 Table 2
kable(table_2,
caption = "Table 2. Cardiac Structure, Function, and Electrophysiological Parameters in Endurance Athletes by Age and Atrial Fibrillation Status",
align = c("l", rep("c", ncol(table_2)-1))) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = TRUE, font_size = 10) %>%
add_header_above(c(" " = 1,
"Overall" = 3,
"Younger Athletes (≤45 years)" = 3,
"Older Athletes (>45 years)" = 3,
" " = 1)) %>%
pack_rows(index = table(table2_with_cat$Category)) %>%
add_footnote(c("Data are presented as mean ± SD, median (IQR), or n (%)",
"Bolded values indicate statistically significant pairwise comparisons (p<0.05)",
"Interaction p-values test whether AF associations differ between age groups"),
notation = "none")| Characteristic | AF (n=96) | No-AF (n=257) | p value | AF (n=25) | No-AF (n=176) | p value | AF (n=71) | No-AF (n=81) | p value | Interaction p value |
|---|---|---|---|---|---|---|---|---|---|---|
| Left ventricular parameters | ||||||||||
| LVEDV, mL | 157.4 ± 40.5 | 170.4 ± 46.5 | 0.011 | 173.7 ± 45.3 | 183.3 ± 44.2 | 0.328 | 151.6 ± 37.2 | 142.3 ± 38.5 | 0.136 | 0.094 |
| LVESV, mL | 64 ± 17.8 | 69.9 ± 24.1 | 0.013 | 68.7 ± 21.1 | 75.5 ± 24.5 | 0.147 | 62.4 ± 16.4 | 57.9 ± 18.2 | 0.113 | 0.051 |
| LVEF, % | 58.1 ± 5.4 | 58.1 ± 5.6 | 0.986 | 60.2 ± 4.8 | 57.9 ± 5.7 | 0.040 | 57.4 ± 5.4 | 58.6 ± 5.5 | 0.182 | 0.020 |
| LV strain, % | -18.2 ± 2.8 | -19.2 ± 2.1 | 0.001 | -18.7 ± 2 | -19.1 ± 2.1 | 0.291 | -18 ± 3 | -19.3 ± 2.1 | 0.003 | 0.173 |
| Left atrial parameters | ||||||||||
| LAESV, mL | 90 [77.8-118.5] | 79 [66-95] | <0.001 | 84 [73-100] | 79 [66-93] | 0.116 | 97 [78.5-122.5] | 81 [66-98] | <0.001 | 0.058 |
| LAVi, mL/m² | 46.2 [38.7-55.9] | 41.9 [34.5-48.4] | <0.001 | 41.8 [37.5-49.6] | 41.9 [34.8-48.4] | 0.694 | 47.7 [39.9-58.6] | 41.9 [32.7-47.8] | <0.001 | 0.004 |
| LA:LV | 0.59 [0.49-0.75] | 0.46 [0.39-0.56] | <0.001 | 0.5 [0.45-0.6] | 0.44 [0.38-0.49] | 0.001 | 0.64 [0.52-0.84] | 0.59 [0.5-0.69] | 0.034 | 0.391 |
| Reservoir strain, % | 25.8 ± 8.9 | 32.2 ± 5.9 | <0.001 | 33.2 ± 6.3 | 33.6 ± 5.5 | 0.770 | 23.4 ± 8.4 | 29.2 ± 5.4 | <0.001 | 0.003 |
| Conduit strain, % | 15.2 ± 6.3 | 20.9 ± 6.5 | <0.001 | 21.6 ± 5.5 | 23.6 ± 5.2 | 0.127 | 13.1 ± 5.1 | 15.2 ± 5.2 | 0.015 | 0.908 |
| Contractile strain, % | 10.7 ± 5.6 | 11.4 ± 4.3 | 0.311 | 11.5 ± 3.3 | 10 ± 3.4 | 0.064 | 10.4 ± 6.2 | 14.1 ± 4.9 | <0.001 | <0.001 |
| Mildly abnormal >34 mL/m², n (%) | 87 (90.6%) | 196 (76.3%) | 0.004 | 20 (80%) | 137 (77.8%) | 1.000 | 67 (94.4%) | 59 (72.8%) | <0.001 | 0.030 |
| Severely abnormal >48 mL/m², n (%) | 41 (42.7%) | 66 (25.7%) | 0.003 | 7 (28%) | 46 (26.1%) | 1.000 | 34 (47.9%) | 20 (24.7%) | 0.005 | 0.107 |
| Extremely abnormal >60 mL/m², n (%) | 18 (18.8%) | 8 (3.1%) | <0.001 | 2 (8%) | 7 (4%) | 0.310 | 16 (22.5%) | 1 (1.2%) | <0.001 | 0.039 |
| Right Ventricular Parameters | ||||||||||
| RVEDV, mL | 144.1 ± 42.5 | 152.3 ± 48.7 | 0.132 | 163.5 ± 47.7 | 165.1 ± 47.4 | 0.877 | 137 ± 38.5 | 124.8 ± 39.3 | 0.059 | 0.250 |
| RVESV, mL | 78.7 ± 26.6 | 76 ± 25.5 | 0.417 | 86 ± 28.1 | 79.7 ± 24.3 | 0.300 | 76 ± 25.7 | 68.2 ± 26.4 | 0.075 | 0.844 |
| RVEF, % | 45.4 ± 8.3 | 49.5 ± 8.9 | <0.001 | 47 ± 8.4 | 51.2 ± 8.9 | 0.031 | 44.8 ± 8.3 | 45.8 ± 7.9 | 0.470 | 0.173 |
| Doppler Parameters | ||||||||||
| E velocity, m/s | 0.6 ± 0.2 | 0.7 ± 0.2 | <0.001 | 0.7 ± 0.2 | 0.8 ± 0.2 | 0.003 | 0.6 ± 0.2 | 0.6 ± 0.1 | 0.226 | 0.001 |
| A velocity, m/s | 0.5 ± 0.1 | 0.4 ± 0.1 | 0.019 | 0.4 ± 0.1 | 0.4 ± 0.1 | 0.184 | 0.5 ± 0.2 | 0.5 ± 0.1 | 0.029 | 0.009 |
| E/A ratio | 1.6 ± 1.3 | 1.9 ± 0.9 | 0.012 | 1.8 ± 0.5 | 2.3 ± 0.8 | <0.001 | 1.5 ± 1.6 | 1.2 ± 0.4 | 0.122 | <0.001 |
| E′ velocity, cm/s | 9.3 ± 2.9 | 11.4 ± 3.4 | <0.001 | 11.5 ± 2.7 | 13 ± 2.5 | 0.015 | 8.6 ± 2.5 | 8 ± 2.2 | 0.104 | 0.001 |
| E/E’ ratio | 6.2 ± 2.4 | 5.7 ± 1.4 | 0.055 | 5 ± 1.1 | 5.2 ± 1 | 0.362 | 6.6 ± 2.6 | 6.6 ± 1.7 | 0.913 | 0.692 |
| Electrocardiographic | ||||||||||
| P wave duration, ms | 124 [114-131] | 111 [101-121] | <0.001 | 117 [98-129] | 106 [96.8-115.2] | 0.017 | 126 [117-132] | 120 [112-129] | 0.053 | 0.220 |
| P wave axis, ° | 51.3 ± 32.4 | 49.6 ± 23.9 | 0.667 | 50.1 ± 35.1 | 48.4 ± 24.8 | 0.822 | 51.8 ± 31.6 | 52 ± 21.5 | 0.956 | 0.790 |
| PTFV₁, mV·ms | 2.3 [1-4] | 1.9 [1-3] | 0.049 | 2 [1.2-3.2] | 2 [1.3-3.1] | 0.775 | 2.4 [0-4.3] | 1.6 [0-2.4] | <0.001 | 0.016 |
| Advanced IAB, n (%) | 12 (15%) | 16 (6.3%) | 0.026 | 0 (0%) | 2 (1.1%) | 1.000 | 12 (21.8%) | 14 (17.5%) | 0.687 | 0.412 |
| Atrial ectopy, per 24 h | 118 [20-521] | 15 [6-47] | <0.001 | 15.5 [7-35.2] | 10 [5-34] | 0.299 | 239 [69-919] | 28 [12-98] | <0.001 | 0.001 |
| P-wave duration >120 ms, n (%) | 51 (62.2%) | 74 (28.8%) | <0.001 | 12 (48%) | 32 (18.2%) | 0.002 | 39 (68.4%) | 42 (51.9%) | 0.077 | 0.208 |
| PTFV₁ ≥4 mV·ms, n (%) | 21 (25.6%) | 23 (8.9%) | <0.001 | 3 (12%) | 19 (10.8%) | 0.742 | 18 (31.6%) | 4 (4.9%) | <0.001 | 0.011 |
| Biochemical | ||||||||||
| BNP, pg/mL | 24.8 [12.7-55.8] | 13.9 [8.5-24] | <0.001 | 11 [7.9-21.3] | 10 [7-17.5] | 0.540 | 38.4 [18.1-63.5] | 22.4 [14-37.8] | 0.002 | 0.004 |
| cTnI, ng/L | 4 [2-7] | 3 [2-8.2] | 0.025 | 3 [2-4] | 3 [2-12] | 0.285 | 6 [3-8] | 2 [2-4] | <0.001 | 0.504 |
| Genetic | ||||||||||
| AF-PRS percentile | 64.5 [33-87.2] | 51 [29-73] | 0.016 | 73 [53-78] | 49.5 [29-76.2] | 0.065 | 62 [30.5-88.5] | 52 [29-71] | 0.092 | 0.698 |
| AF-PRS highest quartile | 34 (35.4%) | 60 (23.3%) | 0.032 | 7 (28%) | 45 (25.6%) | 0.987 | 27 (38%) | 15 (18.5%) | 0.012 | 0.145 |
| AF-PRS highest decile | 22 (22.9%) | 26 (10.1%) | 0.003 | 5 (20%) | 19 (10.8%) | 0.318 | 17 (23.9%) | 7 (8.6%) | 0.018 | 0.512 |
| AF-PRS | 30.2 [30.1-30.3] | 30.2 [30.1-30.2] | 0.015 | 30.2 [30.2-30.3] | 30.2 [30.1-30.3] | 0.066 | 30.2 [30.1-30.3] | 30.2 [30.1-30.2] | 0.091 | 0.989 |
| AF-PRS highest quintile | 29 (30.2%) | 53 (20.6%) | 0.079 | 6 (24%) | 40 (22.7%) | 1.000 | 23 (32.4%) | 13 (16%) | 0.030 | 0.175 |
| AF-PRS ≥98th percentile | 8 (8.5%) | 6 (2.4%) | 0.022 | 2 (8.3%) | 4 (2.3%) | 0.155 | 6 (8.6%) | 2 (2.5%) | 0.147 | 0.960 |
| Rare Variant | 17 (18.1%) | 32 (12.5%) | 0.251 | 4 (16.7%) | 16 (9.1%) | 0.273 | 13 (18.6%) | 16 (20%) | 0.989 | 0.303 |
| Data are presented as mean ± SD, median (IQR), or n (%) | ||||||||||
| Bolded values indicate statistically significant pairwise comparisons (p<0.05) | ||||||||||
| Interaction p-values test whether AF associations differ between age groups | ||||||||||
4 Clinical Risk Prediction Models
# Binary variables for clinical model
clinical_mapping <- list(
"LAVi_34" = "LAVi > 34",
"LAVi_60" = "LAVi > 60",
"P_wave_120" = "Prevalence of P-wave duration >120 ms",
"PTFV1_high" = "High PTFV1",
"PAC_100" = "Prevalence of >=100 PACs/24 h",
"highest_decile" = "highest decile"
)
Athletes_NoCM$Age_decades <- Athletes_NoCM$Age / 10
Athletes_NoCM$High_normal_BP <- as.integer(Athletes_NoCM$rest_SBP >= 130)
for (new_name in names(clinical_mapping)) {
old_name <- clinical_mapping[[new_name]]
if (old_name %in% names(Athletes_NoCM)) {
Athletes_NoCM[[new_name]] <- as.integer(Athletes_NoCM[[old_name]] == 1)
} else {
Athletes_NoCM[[new_name]] <- NA_integer_
}
}
# Exclude athletes with AF during monitoring
Athletes_modelling <- Athletes_NoCM %>%
filter(is.na(`AF during ECG`) | `AF during ECG` != 1) %>%
filter(is.na(`AF Throughout`) | `AF Throughout` != 1)
# Filter complete cases for modelling
Athletes_analysis <- Athletes_modelling %>%
filter(!is.na(AF) & !is.na(Age) & !is.na(`AF centile`)) %>%
filter(complete.cases(select(., Age_decades, High_normal_BP, LAVi_34,
LAVi_60, P_wave_120, PTFV1_high, PAC_100)))#Exclusions
# Calculate exclusions for Model B (Clinical-Only)
n_excluded_monitoring <- nrow(Athletes_NoCM) - nrow(Athletes_modelling)
n_excluded_monitoring_af <- sum((Athletes_NoCM$`AF during ECG` == 1 |
Athletes_NoCM$`AF Throughout` == 1) &
Athletes_NoCM$AF == 1, na.rm = TRUE)
n_excluded_monitoring_noaf <- n_excluded_monitoring - n_excluded_monitoring_af
n_excluded_missing_clinical <- nrow(Athletes_modelling) - nrow(Athletes_analysis)
n_excluded_missing_clinical_af <- sum(Athletes_modelling$AF == 1, na.rm = TRUE) -
sum(Athletes_analysis$AF == 1)
n_excluded_missing_clinical_noaf <- n_excluded_missing_clinical - n_excluded_missing_clinical_af
# Calculate Model A (Genetic-Only) - needs PRS data
Athletes_model_A <- Athletes_modelling %>%
filter(!is.na(AF) & !is.na(`highest decile`))
n_model_a <- nrow(Athletes_model_A)
n_model_a_af <- sum(Athletes_model_A$AF == 1)
n_model_a_noaf <- sum(Athletes_model_A$AF == 0)
# Calculate Model C (Clinical + Genetic) - needs both clinical and PRS
Athletes_model_C <- Athletes_modelling %>%
filter(!is.na(AF) & !is.na(Age) & !is.na(`highest decile`)) %>%
filter(complete.cases(select(., Age_decades, High_normal_BP, LAVi_34,
LAVi_60, P_wave_120, PTFV1_high, PAC_100, highest_decile)))
n_model_c <- nrow(Athletes_model_C)
n_model_c_af <- sum(Athletes_model_C$AF == 1)
n_model_c_noaf <- sum(Athletes_model_C$AF == 0)
# Calculate how many excluded due to missing PRS (difference between B and C)
n_excluded_prs <- nrow(Athletes_analysis) - n_model_c
n_excluded_prs_af <- sum(Athletes_analysis$AF == 1) - n_model_c_af
n_excluded_prs_noaf <- n_excluded_prs - n_excluded_prs_af
# Exclusion Table
exclusion_flow <- data.frame(
Step = c("Starting cohort",
"Excluded: AF during ECG/Holter",
"Excluded: Missing clinical data",
"Model A: Genetic-Only",
"Model B: Clinical-Only",
"Excluded: Missing PRS data",
"Model C: Clinical + Genetic"),
N = c(
nrow(Athletes_NoCM),
n_excluded_monitoring,
n_excluded_missing_clinical,
n_model_a,
nrow(Athletes_analysis),
n_excluded_prs,
n_model_c
),
AF_cases = c(
sum(Athletes_NoCM$AF == 1, na.rm = TRUE),
n_excluded_monitoring_af,
n_excluded_missing_clinical_af,
n_model_a_af,
sum(Athletes_analysis$AF == 1),
n_excluded_prs_af,
n_model_c_af
),
Non_AF = c(
sum(Athletes_NoCM$AF == 0, na.rm = TRUE),
n_excluded_monitoring_noaf,
n_excluded_missing_clinical_noaf,
n_model_a_noaf,
sum(Athletes_analysis$AF == 0),
n_excluded_prs_noaf,
n_model_c_noaf
)
)
kable(exclusion_flow,
col.names = c("Cohort Stage", "Total (n)", "AF cases (n)", "Non-AF (n)"),
align = c("l", "c", "c", "c"),
caption = "Study Population Flow for Prediction Models") %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE) %>%
row_spec(c(4, 5, 7), bold = TRUE, background = "#f0f0f0") # Highlight the three models| Cohort Stage | Total (n) | AF cases (n) | Non-AF (n) |
|---|---|---|---|
| Starting cohort | 353 | 96 | 257 |
| Excluded: AF during ECG/Holter | 15 | 15 | 0 |
| Excluded: Missing clinical data | 10 | 4 | 6 |
| Model A: Genetic-Only | 338 | 81 | 257 |
| Model B: Clinical-Only | 328 | 77 | 251 |
| Excluded: Missing PRS data | 0 | 0 | 0 |
| Model C: Clinical + Genetic | 328 | 77 | 251 |
# What data was missing
incomplete_only <- Athletes_modelling %>%
filter(!complete.cases(select(., AF, Age, `AF centile`, Age_decades, High_normal_BP,
LAVi_34, LAVi_60, P_wave_120, PTFV1_high, PAC_100)))
missing_summary <- data.frame(
Variable = c("Age", "AF-PRS", "SBP ≥130 mmHg", "LAVi >34 mL/m²",
"LAVi >60 mL/m²", "P-wave >120 ms", "PTFV₁ ≥4 mV·ms", "PACs ≥100/24h"),
Missing_n = c(
sum(is.na(incomplete_only$Age)),
sum(is.na(incomplete_only$`AF centile`)),
sum(is.na(incomplete_only$High_normal_BP)),
sum(is.na(incomplete_only$LAVi_34)),
sum(is.na(incomplete_only$LAVi_60)),
sum(is.na(incomplete_only$P_wave_120)),
sum(is.na(incomplete_only$PTFV1_high)),
sum(is.na(incomplete_only$PAC_100))
)
)
missing_summary$Pct_Excluded <- round((missing_summary$Missing_n / nrow(incomplete_only)) * 100, 1)
kable(missing_summary,
col.names = c("Variable", "Missing (n)", "% of Excluded Cases"),
align = c("l", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)| Variable | Missing (n) | % of Excluded Cases |
|---|---|---|
| Age | 0 | 0 |
| AF-PRS | 0 | 0 |
| SBP ≥130 mmHg | 1 | 10 |
| LAVi >34 mL/m² | 0 | 0 |
| LAVi >60 mL/m² | 0 | 0 |
| P-wave >120 ms | 0 | 0 |
| PTFV₁ ≥4 mV·ms | 0 | 0 |
| PACs ≥100/24h | 9 | 90 |
4.1 Model A: Genetic-Only
model_A <- glm(AF ~ highest_decile, data = Athletes_analysis, family = binomial())
Athletes_analysis$predicted_prob_A <- predict(model_A, type = "response")
roc_A <- pROC::roc(Athletes_analysis$AF, Athletes_analysis$predicted_prob_A, quiet = TRUE)
auc_A <- as.numeric(pROC::auc(roc_A))
ci_A <- pROC::ci.auc(roc_A)
model_A_summary <- broom::tidy(model_A, conf.int = TRUE, exponentiate = TRUE) %>%
filter(term != "(Intercept)")
cat("Model A:Genetic Risk (Highest Decile)\n\n")## Model A:Genetic Risk (Highest Decile)cat("AUC:", round(auc_A, 3), "(95% CI:", round(ci_A[1], 3), "-", round(ci_A[3], 3), ")\n")## AUC: 0.561 (95% CI: 0.51 - 0.611 )cat("OR:", round(model_A_summary$estimate, 2),
"(95% CI:", round(model_A_summary$conf.low, 2), "-", round(model_A_summary$conf.high, 2), ")\n")## OR: 2.56 (95% CI: 1.28 - 5.03 )cat("P-value:", format_p_value(model_A_summary$p.value), "\n")## P-value: 0.0074.2 Model B: Clinical-Only
# Fit model
model_B <- glm(AF ~ Age_decades + High_normal_BP + LAVi_34 + LAVi_60 +
P_wave_120 + PTFV1_high + PAC_100,
data = Athletes_analysis, family = binomial())
Athletes_analysis$predicted_prob_B <- predict(model_B, type = "response")
# ROC analysis
roc_B <- pROC::roc(Athletes_analysis$AF, Athletes_analysis$predicted_prob_B, quiet = TRUE)
auc_B <- as.numeric(pROC::auc(roc_B))
ci_B <- pROC::ci.auc(roc_B)
cat("Model B Performance\n\n")## Model B Performancecat("- AUC:", round(auc_B, 3), "(95% CI:", round(ci_B[1], 3), "-", round(ci_B[3], 3), ")\n\n")## - AUC: 0.844 (95% CI: 0.791 - 0.897 )# Derive point system
beta_values <- coef(model_B)[-1]
min_beta <- min(abs(beta_values))
derived_points <- round(abs(beta_values) / min_beta)
# Create coefficient table
model_B_coefs <- broom::tidy(model_B, conf.int = TRUE, exponentiate = FALSE)
model_B_or <- broom::tidy(model_B, conf.int = TRUE, exponentiate = TRUE)
model_B_summary <- model_B_coefs %>%
filter(term != "(Intercept)") %>%
mutate(
Variable = case_when(
term == "Age_decades" ~ "Age (per decade)",
term == "High_normal_BP" ~ "SBP ≥130 mmHg",
term == "LAVi_34" ~ "LAVi 35-60 mL/m²",
term == "LAVi_60" ~ "LAVi >60 mL/m²",
term == "P_wave_120" ~ "P-wave >120 ms",
term == "PTFV1_high" ~ "PTFV₁ ≥4 mV·ms",
term == "PAC_100" ~ "PACs ≥100/24h"
),
OR = model_B_or$estimate[-1],
OR_lower = model_B_or$conf.low[-1],
OR_upper = model_B_or$conf.high[-1],
Points = derived_points,
Points_Display = if_else(term == "LAVi_60",
paste0("+", derived_points[4], " (", derived_points[3] + derived_points[4], " total)"),
as.character(derived_points))
) %>%
select(Variable, `β` = estimate, SE = std.error, OR,
`OR CI lower` = OR_lower, `OR CI upper` = OR_upper,
`P value` = p.value, `Points Assigned` = Points_Display)
# Calculate risk scores
Athletes_analysis$clinical_score_B <- with(Athletes_analysis,
Age_decades * derived_points[1] +
High_normal_BP * derived_points[2] +
if_else(LAVi_60 == 1, derived_points[3] + derived_points[4],
if_else(LAVi_34 == 1, derived_points[3], 0)) +
P_wave_120 * derived_points[5] +
PTFV1_high * derived_points[6] +
PAC_100 * derived_points[7]
)
# Model calibration
hl_test <- hoslem.test(Athletes_analysis$AF, Athletes_analysis$predicted_prob_B, g = 10)
cat("Calibration: Hosmer-Lemeshow χ² =", round(hl_test$statistic, 2),
", p =", round(hl_test$p.value, 3),
if_else(hl_test$p.value > 0.05, "(good fit)\n\n", "(poor fit)\n\n"))## Calibration: Hosmer-Lemeshow χ² = 6.56 , p = 0.584 (good fit)# Optimal cutoff
coords_result <- coords(roc_B, "best", ret = c("threshold", "sensitivity", "specificity", "ppv", "npv"))
cat("Optimal Cutoff (Youden): Threshold =", round(coords_result$threshold, 3), "\n")## Optimal Cutoff (Youden): Threshold = 0.22cat("Sensitivity:", round(coords_result$sensitivity * 100, 1),
"% | Specificity:", round(coords_result$specificity * 100, 1), "%\n")## Sensitivity: 80.5 % | Specificity: 77.3 %cat("PPV:", round(coords_result$ppv * 100, 1),
"% | NPV:", round(coords_result$npv * 100, 1), "%\n\n")## PPV: 52.1 % | NPV: 92.8 %# Performance at clinical thresholds
thresholds <- c(0.10, 0.40, 0.70)
threshold_names <- c("Low-Moderate (10%)", "Moderate-High (40%)", "High-Very High (70%)")
perf_table <- map_dfr(1:length(thresholds), function(i) {
predicted_class <- if_else(Athletes_analysis$predicted_prob_B >= thresholds[i], 1, 0)
tp <- sum(predicted_class == 1 & Athletes_analysis$AF == 1)
tn <- sum(predicted_class == 0 & Athletes_analysis$AF == 0)
fp <- sum(predicted_class == 1 & Athletes_analysis$AF == 0)
fn <- sum(predicted_class == 0 & Athletes_analysis$AF == 1)
tibble(
Threshold = threshold_names[i],
Sensitivity = paste0(round(100 * tp/(tp+fn), 1), "%"),
Specificity = paste0(round(100 * tn/(tn+fp), 1), "%"),
PPV = paste0(round(100 * tp/(tp+fp), 1), "%"),
NPV = paste0(round(100 * tn/(tn+fn), 1), "%")
)
})
kable(perf_table, caption = "Model Performance at Clinical Thresholds") %>%
kable_styling(bootstrap_options = c("striped", "hover"), full_width = FALSE)| Threshold | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|
| Low-Moderate (10%) | 92.2% | 55.4% | 38.8% | 95.9% |
| Moderate-High (40%) | 58.4% | 90.4% | 65.2% | 87.6% |
| High-Very High (70%) | 29.9% | 98.4% | 85.2% | 82.1% |
# Risk stratification
prob_thresholds <- c(0.10, 0.40, 0.70)
cutoffs <- round(map_dbl(prob_thresholds, ~min(Athletes_analysis$clinical_score_B[Athletes_analysis$predicted_prob_B >= .x], na.rm = TRUE)))
score_breaks <- c(-Inf, cutoffs, Inf)
Athletes_analysis <- Athletes_analysis %>%
mutate(
clinical_score_B_rounded = round(clinical_score_B),
risk_category = cut(clinical_score_B_rounded, breaks = score_breaks,
labels = c("Low", "Moderate", "High", "Very High"),
include.lowest = TRUE)
)
risk_summary <- Athletes_analysis %>%
group_by(risk_category) %>%
summarise(
`Score Range` = paste0(min(clinical_score_B_rounded), "–", max(clinical_score_B_rounded)),
N = n(),
`AF Cases` = sum(AF),
`AF Prevalence (%)` = round(mean(AF) * 100, 1),
`Mean Age (years)` = round(mean(Age), 1),
`High BP (%)` = round(mean(High_normal_BP) * 100, 1),
`LAVi >34 (%)` = round(mean(LAVi_34 | LAVi_60) * 100, 1),
`P-wave >120ms (%)` = round(mean(P_wave_120) * 100, 1),
`High PTFV1 (%)` = round(mean(PTFV1_high) * 100, 1),
`PACs ≥100/24h (%)` = round(mean(PAC_100) * 100, 1),
.groups = "drop"
) %>%
rename(`Risk Category` = risk_category)4.3 Model C: Clinical-Genetic
model_C <- glm(AF ~ Age_decades + High_normal_BP + LAVi_34 + LAVi_60 +
P_wave_120 + PTFV1_high + PAC_100 + highest_decile,
data = Athletes_analysis, family = binomial())
Athletes_analysis$predicted_prob_C <- predict(model_C, type = "response")
roc_C <- pROC::roc(Athletes_analysis$AF, Athletes_analysis$predicted_prob_C, quiet = TRUE)
auc_C <- as.numeric(pROC::auc(roc_C))
ci_C <- pROC::ci.auc(roc_C)
delong_BC <- pROC::roc.test(roc_B, roc_C, method = "delong")
cat("Model C: Clinical + Genetic\n\n")## Model C: Clinical + Geneticcat("- AUC:", round(auc_C, 3), "(95% CI:", round(ci_C[1], 3), "-", round(ci_C[3], 3), ")\n")## - AUC: 0.85 (95% CI: 0.798 - 0.902 )cat("- Comparison to Model B: p =", format(delong_BC$p.value, digits = 3), "\n")## - Comparison to Model B: p = 0.3245 Table 3
kable(model_B_summary, digits = 3,
caption = paste0("Table 3. Multivariable Logistic Regression Model Predicting Atrial Fibrillation (n=", nrow(Athletes_analysis), ")")) %>%
kable_styling(bootstrap_options = c("striped", "hover"))| Variable | β | SE | OR | OR CI lower | OR CI upper | P value | Points Assigned |
|---|---|---|---|---|---|---|---|
| Age (per decade) | 0.243 | 0.104 | 1.276 | 1.044 | 1.570 | 0.019 | 1 |
| SBP ≥130 mmHg | 0.971 | 0.338 | 2.640 | 1.372 | 5.190 | 0.004 | 4 |
| LAVi 35-60 mL/m² | 0.858 | 0.468 | 2.359 | 0.989 | 6.290 | 0.066 | 4 |
| LAVi >60 mL/m² | 0.878 | 0.642 | 2.406 | 0.673 | 8.473 | 0.171 | +4 (8 total) |
| P-wave >120 ms | 0.993 | 0.348 | 2.699 | 1.371 | 5.392 | 0.004 | 4 |
| PTFV₁ ≥4 mV·ms | 0.861 | 0.446 | 2.367 | 0.978 | 5.666 | 0.053 | 4 |
| PACs ≥100/24h | 1.535 | 0.358 | 4.644 | 2.311 | 9.444 | 0.000 | 6 |
6 Table 4
kable(risk_summary,
caption = "Table 4. Clinical Risk Stratification Categories") %>%
kable_styling(bootstrap_options = c("striped", "hover"),
full_width = FALSE, font_size = 10)| Risk Category | Score Range | N | AF Cases | AF Prevalence (%) | Mean Age (years) | High BP (%) | LAVi >34 (%) | P-wave >120ms (%) | High PTFV1 (%) | PACs ≥100/24h (%) |
|---|---|---|---|---|---|---|---|---|---|---|
| Low | 2–10 | 150 | 6 | 4.0 | 27.3 | 20.0 | 65.3 | 10.0 | 4.7 | 0.7 |
| Moderate | 11–17 | 110 | 26 | 23.6 | 44.8 | 58.2 | 87.3 | 47.3 | 11.8 | 21.8 |
| High | 18–24 | 46 | 26 | 56.5 | 59.9 | 60.9 | 95.7 | 76.1 | 23.9 | 63.0 |
| Very High | 25–32 | 22 | 19 | 86.4 | 67.7 | 86.4 | 95.5 | 90.9 | 40.9 | 100.0 |
7 Central Figure:
AF Polygenic Risk Score Distribution
if ("AF PRS 6238158var PGS000016" %in% names(Athletes_NoCM)) {
ggplot(Athletes_NoCM, aes(x = factor(AF), y = `AF PRS 6238158var PGS000016`,
fill = factor(AF))) +
geom_violin(trim = FALSE, alpha = 0.8, color = NA) +
geom_boxplot(width = 0.1, color = "black", fill = "white") +
stat_compare_means(method = "t.test", label = "p.format",
comparisons = list(c("0", "1")),
bracket.size = 0.7, size = 5, tip.length = 0.02) +
scale_fill_manual(values = c("0" = "#0072B2", "1" = "#990000")) +
scale_x_discrete(labels = c("No AF", "AF")) +
labs(x = "", y = "AF Polygenic Risk Score") +
theme_classic(base_size = 20) +
theme(legend.position = "none")
} else {
cat("AF PRS variable not found in dataset\n")
}
Legend = AF Polygenic Risk Score Distribution by AF Status
8 Figure 1
Figure 1. Bar Plot showing the prevalence of abnormal atrial myopathy markers in athletes with (n=96) and without (n=257) Atrial Fibrillation
# Ensure AF_status exists
Athletes_NoCM$AF_status <- factor(Athletes_NoCM$AF, levels = c(0, 1), labels = c("No AF", "AF"))
# Panel A: Individual criterion prevalence
criteria_prevalence <- Athletes_NoCM %>%
group_by(AF_status) %>%
summarise(
`SBP ≥130 mmHg` = mean(High_normal_BP, na.rm = TRUE) * 100,
`LAVi 35-60 mL/m²` = mean(LAVi_34 & !LAVi_60, na.rm = TRUE) * 100,
`LAVi >60 mL/m²` = mean(LAVi_60, na.rm = TRUE) * 100,
`P-wave >120 ms` = mean(P_wave_120, na.rm = TRUE) * 100,
`PTFV₁ ≥4 mV·ms` = mean(PTFV1_high, na.rm = TRUE) * 100,
`PACs ≥100/24h` = mean(PAC_100, na.rm = TRUE) * 100,
`AF-PRS (top decile)` = mean(highest_decile, na.rm = TRUE) * 100,
.groups = "drop"
) %>%
pivot_longer(cols = -AF_status, names_to = "Criterion", values_to = "Percentage")
fig1a <- ggplot(criteria_prevalence, aes(x = reorder(Criterion, -Percentage),
y = Percentage, fill = AF_status)) +
geom_bar(stat = "identity", position = position_dodge(width = 0.9),
color = "black", alpha = 0.8, width = 0.8) +
scale_x_discrete(labels = function(x) {
x <- gsub("PTFV₁", "PTFV1", x) # Replace subscript with regular 1
return(x)
}) +
scale_fill_manual(values = c("No AF" = "#0072B2", "AF" = "#990000")) +
labs(x = "Atrial Myopathy Criterion", y = "Prevalence (%)", fill = "") +
theme_minimal(base_size = 12) +
theme(axis.text.x = element_text(angle = 35, hjust = 1),
legend.position = "bottom")
# Panel B: Number of criteria met (overlap)
Athletes_NoCM$criteria_count <- rowSums(
select(Athletes_NoCM, High_normal_BP, LAVi_34, LAVi_60,
P_wave_120, PTFV1_high, PAC_100, highest_decile),
na.rm = TRUE
)
Athletes_NoCM$criteria_group <- factor(
case_when(
Athletes_NoCM$criteria_count == 0 ~ "0",
Athletes_NoCM$criteria_count == 1 ~ "1",
Athletes_NoCM$criteria_count == 2 ~ "2",
Athletes_NoCM$criteria_count == 3 ~ "3",
Athletes_NoCM$criteria_count == 4 ~ "4",
Athletes_NoCM$criteria_count >= 5 ~ "≥5"
),
levels = c("0", "1", "2", "3", "4", "≥5")
)
criteria_overlap <- Athletes_NoCM %>%
count(AF_status, criteria_group) %>%
group_by(AF_status) %>%
mutate(percentage = n / sum(n) * 100) %>%
ungroup() %>%
complete(AF_status, criteria_group, fill = list(n = 0, percentage = 0))
fig1b <- ggplot(criteria_overlap, aes(x = criteria_group, y = percentage,
fill = AF_status)) +
geom_bar(stat = "identity", position = position_dodge(width = 0.9),
color = "black", alpha = 0.8, width = 0.8) +
geom_text(aes(label = paste0(round(percentage, 1), "%")),
position = position_dodge(width = 0.9), vjust = -0.5, size = 3) +
scale_fill_manual(values = c("No AF" = "#0072B2", "AF" = "#990000")) +
scale_y_continuous(expand = expansion(mult = c(0, 0.15))) +
labs(x = "Number of Criteria Met", y = "Percentage (%)", fill = "") +
theme_minimal(base_size = 12) +
theme(legend.position = "bottom")
# Combine panels
ggarrange(fig1a, fig1b, ncol = 2, labels = c("A", "B"),
common.legend = TRUE, legend = "bottom")
Legend = Abnormal Atrial Myopathy Markers – AF-PRS = atrial fibrillation polygenic risk score; LAVi = left atrial volume index; PAC = premature atrial complex; PTFV1 = P wave terminal force velocity in lead V1; P-wave = P wave duration; SBP = systolic blood pressure.
9 Figure 2
Figure 2. Receiver-operating characteristic (ROC) curves for Model A (Genetic-Only) Model B (Clinical-Only), and Model C (Clinical-Genetic).
# Model A ROC
p2a <- ggplot(data.frame(fpr = 1 - roc_A$specificities,
tpr = roc_A$sensitivities),
aes(x = fpr, y = tpr)) +
geom_line(color = "#e9c46a", size = 1.2) +
geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
annotate("text", x = 0.6, y = 0.2,
label = paste0("AUC = ", round(auc_A, 3),
"\n95% CI: ", round(ci_A[1], 3), "–",
round(ci_A[3], 3)),
size = 4) +
labs(title = "Model A: Genetic-Only",
x = "1 - Specificity", y = "Sensitivity") +
coord_equal() +
theme_minimal(base_size = 14)
# Model B ROC
p2b <- ggplot(data.frame(fpr = 1 - roc_B$specificities,
tpr = roc_B$sensitivities),
aes(x = fpr, y = tpr)) +
geom_line(color = "#a7c957", size = 1.2) +
geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
annotate("text", x = 0.6, y = 0.2,
label = paste0("AUC = ", round(auc_B, 3),
"\n95% CI: ", round(ci_B[1], 3), "–",
round(ci_B[3], 3)),
size = 4) +
labs(title = "Model B: Clinical-Only",
x = "1 - Specificity", y = "Sensitivity") +
coord_equal() +
theme_minimal(base_size = 14)
# Model C ROC
p2c <- ggplot(data.frame(fpr = 1 - roc_C$specificities,
tpr = roc_C$sensitivities),
aes(x = fpr, y = tpr)) +
geom_line(color = "#6a994e", size = 1.2) +
geom_abline(intercept = 0, slope = 1, linetype = "dashed") +
annotate("text", x = 0.6, y = 0.2,
label = paste0("AUC = ", round(auc_C, 3),
"\n95% CI: ", round(ci_C[1], 3), "–",
round(ci_C[3], 3)),
size = 4) +
labs(title = "Model C: Clinical-Genetic",
x = "1 - Specificity", y = "Sensitivity") +
coord_equal() +
theme_minimal(base_size = 14)
# Combine plots
gridExtra::grid.arrange(p2a, p2b, p2c, ncol = 3)
Legend = AUC values (with 95% CIs) are annotated for each model. “Genetic Only” refers to AF centile as a continuous percentile predictor. “Clinical-Only” utilises clinical markers (systolic blood pressure ≥ 130 mmHg; P wave duration > 120 ms; P terminal force in V1 ≥ 4.0 mV·ms; left atrial volume index > 34 mL/m² and > 60 mL/m²; and PACs ≥ 100/24 h. “Clinical Genetic” combines Clinical markers) with the AF-PRS dichotomised at the top decile (≥ 90th percentile).
10 Supplementary Tables
controls <- filter(NoCM, EICR == "Control")
athletes <- filter(NoCM, EICR == "Athlete")
# Helper function for age-adjusted p-values
calculate_age_adjusted_p <- function(data, var, group_var, is_categorical = FALSE) {
if (!all(c(var, group_var, "Age") %in% names(data))) return(NA)
clean_data <- data %>% filter(!is.na(.data[[var]]) & !is.na(.data[[group_var]]) & !is.na(Age))
if (nrow(clean_data) < 10 || var == "Age") return(NA)
tryCatch({
if (is_categorical) {
model <- glm(as.factor(clean_data[[var]]) ~ clean_data[[group_var]] + Age,
data = clean_data, family = binomial())
} else {
model <- lm(as.numeric(clean_data[[var]]) ~ clean_data[[group_var]] + Age,
data = clean_data)
}
coef_summary <- summary(model)$coefficients
group_row <- grep(group_var, rownames(coef_summary))[1]
if (is.na(group_row)) group_row <- 2
return(coef_summary[group_row, 4])
}, error = function(e) NA)
}
# Create comparison table
supp_table_athletes_controls <- data.frame(Variable = character(), stringsAsFactors = FALSE)
combined_data <- rbind(athletes %>% mutate(group = "Athletes"), controls %>% mutate(group = "Controls"))
for (var in all_variables) {
if (!var %in% names(athletes) || !var %in% names(controls)) next
is_cat <- var %in% categorical_vars
is_nonnorm <- var %in% nonnormal_vars
supp_table_athletes_controls <- rbind(supp_table_athletes_controls, data.frame(
Variable = var,
Athletes = format_stats(athletes, var, is_cat, is_nonnorm),
Controls = format_stats(controls, var, is_cat, is_nonnorm),
P_value = format_p_value(calculate_p_value(combined_data, var, "group", is_cat, is_nonnorm)),
P_age_adj = format_p_value(calculate_age_adjusted_p(combined_data, var, "group", is_cat)),
stringsAsFactors = FALSE
))
}
supp_table_athletes_controls$Category <- sapply(supp_table_athletes_controls$Variable, assign_category)
supp_table_athletes_controls$Variable <- sapply(supp_table_athletes_controls$Variable, clean_variable_names)
# Define correct order within each category
variable_order <- c(
# Demographics
"Age, years", "Male sex, n (%)", "BMI, kg/m²", "BSA, m²", "Average HR, bpm", "Minimum HR, bpm",
"Bradycardia burden, % (24h)", "Hypertension, n (%)", "SBP, mmHg", "DBP, mmHg",
"Dyslipidaemia, n (%)", "Diabetes, n (%)", "History of smoking, n (%)", "Alcohol, drinks/week",
# Exercise Capacity
"VO₂peak, mL/kg/min", "VO₂peak, L/min", "% of predicted exercise capacity", "Respiratory exchange ratio",
"Total weekly MET-hours", "Years of endurance training",
# Left ventricular parameters
"LVEDV, mL", "LVESV, mL", "LVEF, %", "LV strain, %",
# Left atrial parameters
"LAESV, mL", "LAVi, mL/m²", "Mildly abnormal >34 mL/m², n (%)", "Severely abnormal >48 mL/m², n (%)",
"Extremely abnormal >60 mL/m², n (%)", "LA:LV", "Reservoir strain, %", "Conduit strain, %", "Contractile strain, %",
# Right Ventricular Parameters
"RVEDV, mL", "RVESV, mL", "RVEF, %",
# Doppler Parameters
"E velocity, m/s", "A velocity, m/s", "E/A ratio", "E′ velocity, cm/s", "E/E' ratio",
# Electrocardiographic
"P wave duration, ms", "P-wave duration >120 ms, n (%)", "P wave axis, °", "PTFV₁, mV·ms",
"PTFV₁ ≥4 mV·ms, n (%)", "Advanced IAB, n (%)", "PACs per 24 h", "≥100 PACs per 24 h, n (%)",
# Biochemical
"BNP, pg/mL", "cTnI, ng/L"
)
# Order the table by the defined sequence
supp_table_athletes_controls$Variable <- factor(supp_table_athletes_controls$Variable,
levels = variable_order)
supp_table_athletes_controls <- supp_table_athletes_controls %>% arrange(Variable)
supp_table_athletes_controls$Variable <- as.character(supp_table_athletes_controls$Variable)
# Split and display with correct ordering
table1_cats <- c("Demographics", "Exercise Capacity")
# For table 2, create nested structure for echocardiographic parameters
supp_1_cat <- supp_table_athletes_controls %>%
filter(Category %in% table1_cats) %>%
mutate(Category = factor(Category, levels = table1_cats))
# Create hierarchical grouping for Table 2
supp_2_cat <- supp_table_athletes_controls %>%
filter(Category %in% c("Left ventricular parameters", "Left atrial parameters",
"Right Ventricular Parameters", "Doppler Parameters",
"Electrocardiographic", "Biochemical"))
# Add parent category for nested structure
supp_2_cat <- supp_2_cat %>%
mutate(ParentCategory = case_when(
Category %in% c("Left ventricular parameters", "Left atrial parameters") ~ "Echocardiographic",
TRUE ~ Category
))
supp_1 <- supp_1_cat %>% select(-Category)
supp_2 <- supp_2_cat %>% select(-Category, -ParentCategory)
colnames(supp_1) <- c("Characteristic", paste0("Athletes (n=", nrow(athletes), ")"),
paste0("Controls (n=", nrow(controls), ")"), "p-value", "Age-adjusted p value")
colnames(supp_2) <- colnames(supp_1)
cat("\n## Supplementary Table 1\n\n")10.1 Supplementary Table 1
print(kable(supp_1, caption = "Supplementary Table 1. Baseline Demographics: Endurance Athletes versus Non-athletic Controls (n=450)",
align = c("l", "c", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = TRUE, font_size = 10) %>%
pack_rows(index = table(supp_1_cat$Category)[table1_cats]) %>%
add_footnote(c("Data are presented as mean ± standard deviation, median (interquartile range), or n (%). P values were calculated using unpaired t-tests for normally distributed continuous variables, Mann-Whitney U tests for non-normally distributed continuous variables, and chi-square tests for categorical variables. Age-adjusted p values were calculated using multivariable regression models with age as a covariate to account for age differences between groups.\nAbbreviations: BMI, body mass index; BSA, body surface area; DBP, diastolic blood pressure; HR, heart rate; SBP, systolic blood pressure; VO₂peak, peak oxygen uptake."),
notation = "none"))| Characteristic | Athletes (n=353) | Controls (n=97) | p-value | Age-adjusted p value |
|---|---|---|---|---|
| Demographics | ||||
| Age, years | 41 [22-60] | 34 [22-54] | 0.315 | – |
| Male sex, n (%) | 262 (74.2%) | 62 (63.9%) | 0.061 | 0.071 |
| BMI, kg/m² | 23.5 [21.9-25.5] | 24.5 [22.3-27] | 0.023 | <0.001 |
| BSA, m² | 1.9 [1.8-2.1] | 1.9 [1.7-2] | 0.123 | 0.122 |
| Average HR, bpm | 66.4 ± 9.1 | 73.6 ± 8 | <0.001 | <0.001 |
| Minimum HR, bpm | 44 [39-49] | 51 [45-55] | <0.001 | <0.001 |
| Bradycardia burden, % (24h) | 6.4 [0.4-26.9] | 0 [0-3] | <0.001 | <0.001 |
| Hypertension, n (%) | 31 (8.9%) | 1 (1.6%) | 0.067 | 0.847 |
| SBP, mmHg | 126 [117-135] | 121 [115-129] | 0.006 | 0.106 |
| DBP, mmHg | 69 [62-77.2] | 69 [65-76] | 0.656 | 0.045 |
| Dyslipidaemia, n (%) | 45 (12.9%) | 0 (0%) | <0.001 | 0.985 |
| Diabetes, n (%) | 2 (0.6%) | 0 (0%) | 1.000 | 0.998 |
| History of smoking, n (%) | 47 (13.5%) | 10 (14.5%) | 0.972 | 0.021 |
| Alcohol, drinks/week | 1 [0-6] | 3 [3-3] | 0.531 | 0.467 |
| Exercise Capacity | ||||
| VO₂peak, mL/kg/min | 48.9 ± 14.3 | 35.8 ± 9.2 | <0.001 | <0.001 |
| VO₂peak, L/min | 3669.5 ± 997.4 | 2758.8 ± 865.9 | <0.001 | <0.001 |
| % of predicted exercise capacity | 126.4 ± 20.8 | 97.1 ± 22 | <0.001 | <0.001 |
| Respiratory exchange ratio | 1.3 ± 0.1 | 1.3 ± 0.1 | <0.001 | <0.001 |
| Total weekly MET-hours | 81 [49-104] | – | – | – |
| Years of endurance training | 20 [10-30] | – | – | – |
|
Data are presented as mean ± standard deviation, median
(interquartile range), or n (%). P values were calculated using unpaired
t-tests for normally distributed continuous variables, Mann-Whitney U
tests for non-normally distributed continuous variables, and chi-square
tests for categorical variables. Age-adjusted p values were calculated
using multivariable regression models with age as a covariate to account
for age differences between groups. Abbreviations: BMI, body mass index; BSA, body surface area; DBP, diastolic blood pressure; HR, heart rate; SBP, systolic blood pressure; VO₂peak, peak oxygen uptake. |
||||
cat("\n## Supplementary Table 2\n\n")10.2 Supplementary Table 2
# Create pack_rows structure for Table 2 - all categories at same level
# Use the same approach as Table 1 for consistency
# Define the order of categories for Table 2
table2_cat_order <- c("Left ventricular parameters", "Left atrial parameters",
"Right Ventricular Parameters", "Doppler Parameters",
"Electrocardiographic", "Biochemical")
# Ensure categories are in the right order
supp_2_cat <- supp_2_cat %>%
mutate(Category = factor(Category, levels = table2_cat_order))
table2_output <- kable(supp_2, caption = "Supplementary Table 2. Cardiac Parameters: Athletes versus Controls (n=450)",
align = c("l", "c", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = TRUE, font_size = 10) %>%
pack_rows(index = table(supp_2_cat$Category)[table2_cat_order])
print(table2_output %>%
add_footnote(c("Data are presented as mean ± standard deviation, median (interquartile range), or n (%). Statistical comparisons as per Supplementary Table 1.\nAbbreviations: BNP, brain natriuretic peptide; cTnI, cardiac troponin I; E/A, mitral inflow early/late diastolic velocity ratio; E′, mitral annular early diastolic velocity; EDV, end-diastolic volume; ESV, end-systolic volume; IAB, interatrial block; LA, left atrium; LAVi, left atrial volume index; LA:LV, left atrial to left ventricular end-diastolic volume ratio; LV, left ventricle; LVEF, left ventricular ejection fraction; PAC, premature atrial contraction; PTFV₁, P-wave terminal force in lead V1; RV, right ventricle; RVEF, right ventricular ejection fraction."),
notation = "none"))| Characteristic | Athletes (n=353) | Controls (n=97) | p-value | Age-adjusted p value |
|---|---|---|---|---|
| Left ventricular parameters | ||||
| LVEDV, mL | 166.9 ± 45.3 | 118.2 ± 36 | <0.001 | <0.001 |
| LVESV, mL | 68.3 ± 22.7 | 50.1 ± 17 | <0.001 | <0.001 |
| LVEF, % | 58.1 ± 5.5 | 57.4 ± 4.6 | 0.639 | 0.714 |
| LV strain, % | -18.9 ± 2.3 | -19 ± 2 | 0.662 | 0.796 |
| Left atrial parameters | ||||
| LAESV, mL | 83 [69-98.2] | 55.4 [47.2-68.5] | <0.001 | <0.001 |
| LAVi, mL/m² | 42.9 [35.7-49.8] | 29.7 [25.4-35.7] | <0.001 | <0.001 |
| Mildly abnormal >34 mL/m², n (%) | 283 (80.2%) | 27 (28.7%) | <0.001 | <0.001 |
| Severely abnormal >48 mL/m², n (%) | 107 (30.3%) | 4 (4.3%) | <0.001 | <0.001 |
| Extremely abnormal >60 mL/m², n (%) | 26 (7.4%) | 1 (1.1%) | 0.025 | 0.071 |
| LA:LV | 0.49 [0.42-0.62] | 0.48 [0.41-0.58] | 0.508 | 0.857 |
| Reservoir strain, % | 30.4 ± 7.4 | 34 ± 6.3 | <0.001 | <0.001 |
| Conduit strain, % | 19.3 ± 6.9 | 21 ± 5.7 | 0.021 | 0.385 |
| Contractile strain, % | 11.2 ± 4.7 | 13 ± 3.8 | <0.001 | <0.001 |
| Right Ventricular Parameters | ||||
| RVEDV, mL | 150.2 ± 47.2 | 90.9 ± 35.7 | <0.001 | <0.001 |
| RVESV, mL | 76.7 ± 25.8 | 51.7 ± 20.9 | <0.001 | <0.001 |
| RVEF, % | 48.4 ± 8.9 | 42.6 ± 8.4 | <0.001 | <0.001 |
| Doppler Parameters | ||||
| E velocity, m/s | 0.7 ± 0.2 | 0.7 ± 0.2 | 0.452 | 0.824 |
| A velocity, m/s | 0.4 ± 0.1 | 0.5 ± 0.1 | 0.032 | <0.001 |
| E/A ratio | 1.8 ± 1 | 1.7 ± 0.7 | 0.089 | 0.010 |
| E′ velocity, cm/s | 10.9 ± 3.4 | 10.8 ± 3.4 | 0.962 | 0.021 |
| E/E’ ratio | 5.8 ± 1.7 | 6.3 ± 2.1 | 0.032 | <0.001 |
| Electrocardiographic | ||||
| P wave duration, ms | 114 [102.5-125] | 107.5 [102-112.2] | <0.001 | <0.001 |
| P-wave duration >120 ms, n (%) | 125 (36.9%) | 12 (12.6%) | <0.001 | <0.001 |
| P wave axis, ° | 50 ± 26.1 | 54.3 ± 31 | 0.237 | 0.178 |
| PTFV₁, mV·ms | 2 [1-3.2] | 1.8 [0.8-3] | 0.987 | 0.944 |
| PTFV₁ ≥4 mV·ms, n (%) | 44 (13%) | 7 (8.6%) | 0.376 | 0.344 |
| Advanced IAB, n (%) | 28 (8.4%) | 0 (0%) | 0.022 | 0.989 |
| PACs per 24 h | 20 [7-91.5] | 6 [2-24.5] | <0.001 | 0.345 |
| ≥100 PACs per 24 h, n (%) | 76 (23%) | 7 (8.3%) | 0.004 | 0.047 |
| Biochemical | ||||
| BNP, pg/mL | 15.7 [9.6-29.2] | 10 [5.5-13.4] | <0.001 | 0.465 |
| cTnI, ng/L | 3 [2-8] | 2 [1-3] | <0.001 | 0.266 |
|
Data are presented as mean ± standard deviation, median
(interquartile range), or n (%). Statistical comparisons as per
Supplementary Table 1. Abbreviations: BNP, brain natriuretic peptide; cTnI, cardiac troponin I; E/A, mitral inflow early/late diastolic velocity ratio; E′, mitral annular early diastolic velocity; EDV, end-diastolic volume; ESV, end-systolic volume; IAB, interatrial block; LA, left atrium; LAVi, left atrial volume index; LA:LV, left atrial to left ventricular end-diastolic volume ratio; LV, left ventricle; LVEF, left ventricular ejection fraction; PAC, premature atrial contraction; PTFV₁, P-wave terminal force in lead V1; RV, right ventricle; RVEF, right ventricular ejection fraction. |
||||
# Define risk factors and their thresholds based on your document
risk_factors <- list(
"Age" = list(var = "Age", threshold = 45, type = "continuous", cutoff_type = ">"),
"SBP" = list(var = "rest_SBP", threshold = 130, type = "continuous", cutoff_type = "≥"),
"LAVi_normal" = list(var = "LAVi", threshold = c(16, 34), type = "range", label = "Normal Range (16-34)"),
"LAVi_severe" = list(var = "LAVi", threshold = 48, type = "continuous", cutoff_type = "≥", label = "Severely dilated (≥48)"),
"LAVi_extreme" = list(var = "LAVi", threshold = 60, type = "continuous", cutoff_type = "≥", label = "Extremely dilated (≥60)"),
"P_wave" = list(var = "P wave Duration", threshold = 120, type = "continuous", cutoff_type = ">"),
"PTFV1" = list(var = "PTFV1 (ms*mV)", threshold = 4, type = "continuous", cutoff_type = "≥"),
"PACs" = list(var = "ECTOPIES LS", threshold = 100, type = "continuous", cutoff_type = "≥"),
"PRS_Q4" = list(var = "highest quartile", threshold = 1, type = "binary"),
"PRS_D10" = list(var = "highest decile", threshold = 1, type = "binary"),
"PRS_P98" = list(var = "98th percentile", threshold = 1, type = "binary")
)
# Function to calculate OR and CI
calculate_or_ci <- function(data, var_info) {
clean_data <- data %>% filter(!is.na(AF) & !is.na(.data[[var_info$var]]))
if (nrow(clean_data) < 10) return(NULL)
tryCatch({
# Create binary exposure variable based on threshold
if (var_info$type == "range") {
clean_data$exposure <- as.numeric(clean_data[[var_info$var]] >= var_info$threshold[1] &
clean_data[[var_info$var]] <= var_info$threshold[2])
} else if (var_info$type == "binary") {
clean_data$exposure <- as.numeric(clean_data[[var_info$var]] == var_info$threshold)
} else {
if (var_info$cutoff_type == "≥" | var_info$cutoff_type == ">") {
clean_data$exposure <- as.numeric(clean_data[[var_info$var]] >= var_info$threshold)
} else {
clean_data$exposure <- as.numeric(clean_data[[var_info$var]] > var_info$threshold)
}
}
# Fit logistic regression
model <- glm(AF ~ exposure, data = clean_data, family = binomial())
# Get results
results <- broom::tidy(model, conf.int = TRUE, exponentiate = TRUE) %>%
filter(term == "exposure")
# Count cases
af_exposed <- sum(clean_data$AF == 1 & clean_data$exposure == 1, na.rm = TRUE)
noaf_exposed <- sum(clean_data$AF == 0 & clean_data$exposure == 1, na.rm = TRUE)
return(list(
n_af = af_exposed,
pct_af = round(100 * af_exposed / sum(clean_data$AF == 1, na.rm = TRUE), 0),
n_noaf = noaf_exposed,
pct_noaf = round(100 * noaf_exposed / sum(clean_data$AF == 0, na.rm = TRUE), 0),
or = results$estimate,
ci_lower = results$conf.low,
ci_upper = results$conf.high,
p_value = results$p.value
))
}, error = function(e) NULL)
}
# Build table
supp_table_3 <- data.frame(
Criterion = character(),
Threshold = character(),
AF = character(),
NoAF = character(),
OR_CI = character(),
P_value = character(),
stringsAsFactors = FALSE
)
# Process each risk factor
for (rf_name in names(risk_factors)) {
rf <- risk_factors[[rf_name]]
if (!rf$var %in% names(Athletes_NoCM)) next
result <- calculate_or_ci(Athletes_NoCM, rf)
if (!is.null(result)) {
# Format criterion name
criterion <- case_when(
rf_name == "Age" ~ "Age",
rf_name == "SBP" ~ "Systolic Blood Pressure",
rf_name == "LAVi_normal" ~ "Left Atrial Volume Index",
rf_name == "LAVi_severe" ~ "Left Atrial Volume Index",
rf_name == "LAVi_extreme" ~ "Left Atrial Volume Index",
rf_name == "P_wave" ~ "P-wave duration",
rf_name == "PTFV1" ~ "PTFV₁",
rf_name == "PACs" ~ "Premature Atrial Contractions",
rf_name == "PRS_Q4" ~ "AF-PRS",
rf_name == "PRS_D10" ~ "AF-PRS",
rf_name == "PRS_P98" ~ "AF-PRS",
TRUE ~ rf_name
)
# Format threshold
threshold <- if (!is.null(rf$label)) {
rf$label
} else if (rf$type == "range") {
paste0(rf$threshold[1], "-", rf$threshold[2], " mL/m²")
} else if (rf_name == "Age") {
">45 years"
} else if (rf_name == "SBP") {
"≥130 mmHg"
} else if (rf_name == "P_wave") {
">120 ms"
} else if (rf_name == "PTFV1") {
"≥4 mV·ms"
} else if (rf_name == "PACs") {
"≥100/24h"
} else if (rf_name == "PRS_Q4") {
"Highest Quartile (Q4)"
} else if (rf_name == "PRS_D10") {
"Highest Decile (D10)"
} else if (rf_name == "PRS_P98") {
"98th Percentile"
} else {
paste0(rf$cutoff_type, rf$threshold)
}
supp_table_3 <- rbind(supp_table_3, data.frame(
Criterion = criterion,
Threshold = threshold,
AF = paste0(result$n_af, " (", result$pct_af, "%)"),
NoAF = paste0(result$n_noaf, " (", result$pct_noaf, "%)"),
OR_CI = paste0(sprintf("%.1f", result$or), " (",
sprintf("%.1f", result$ci_lower), "–",
sprintf("%.1f", result$ci_upper), ")"),
P_value = format_p_value(result$p_value),
stringsAsFactors = FALSE
))
}
}
# Rename columns
colnames(supp_table_3) <- c(
"Criterion",
"Threshold",
paste0("AF (n=", sum(Athletes_NoCM$AF == 1, na.rm = TRUE), ")"),
paste0("No AF (n=", sum(Athletes_NoCM$AF == 0, na.rm = TRUE), ")"),
"Odds Ratio (95% CI)",
"p value"
)
cat("\n## Supplementary Table 3\n\n")10.3 Supplementary Table 3
print(kable(supp_table_3,
caption = "Supplementary Table 3. Univariable Risk Factors for Atrial Fibrillation Development",
align = c("l", "l", "c", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = TRUE, font_size = 10) %>%
add_footnote(c("Legend = Data are presented as n (%). Odds ratios with 95% confidence intervals represent the increased odds of atrial fibrillation for athletes meeting each threshold criterion compared to those below threshold. P values were calculated using logistic regression. Abbreviations: AF, atrial fibrillation; CI, confidence interval; LAVi, left atrial volume index; OR, odds ratio; PAC, premature atrial contraction; PRS, polygenic risk score; PTFV₁, P-wave terminal force in lead V1; Q4, fourth quartile; D10, tenth decile."), notation = "none"))| Criterion | Threshold | AF (n=96) | No AF (n=257) | Odds Ratio (95% CI) | p value |
|---|---|---|---|---|---|
| Age | >45 years | 74 (77%) | 85 (33%) | 6.8 (4.0–11.9) | <0.001 |
| Systolic Blood Pressure | ≥130 mmHg | 58 (60%) | 95 (37%) | 2.6 (1.6–4.2) | <0.001 |
| Left Atrial Volume Index | Normal Range (16-34) | 9 (9%) | 62 (24%) | 0.3 (0.1–0.7) | 0.003 |
| Left Atrial Volume Index | Severely dilated (≥48) | 41 (43%) | 66 (26%) | 2.2 (1.3–3.5) | 0.002 |
| Left Atrial Volume Index | Extremely dilated (≥60) | 18 (19%) | 8 (3%) | 7.2 (3.1–18.1) | <0.001 |
| P-wave duration | >120 ms | 51 (62%) | 74 (29%) | 4.1 (2.4–6.9) | <0.001 |
| PTFV₁ | ≥4 mV·ms | 21 (26%) | 23 (9%) | 3.5 (1.8–6.7) | <0.001 |
| Premature Atrial Contractions | ≥100/24h | 45 (56%) | 33 (13%) | 8.3 (4.7–14.9) | <0.001 |
| AF-PRS | Highest Quartile (Q4) | 34 (35%) | 60 (23%) | 1.8 (1.1–3.0) | 0.023 |
| AF-PRS | Highest Decile (D10) | 22 (23%) | 26 (10%) | 2.6 (1.4–4.9) | 0.002 |
| AF-PRS | 98th Percentile | 8 (9%) | 6 (2%) | 3.9 (1.3–12.0) | 0.015 |
| Legend = Data are presented as n (%). Odds ratios with 95% confidence intervals represent the increased odds of atrial fibrillation for athletes meeting each threshold criterion compared to those below threshold. P values were calculated using logistic regression. Abbreviations: AF, atrial fibrillation; CI, confidence interval; LAVi, left atrial volume index; OR, odds ratio; PAC, premature atrial contraction; PRS, polygenic risk score; PTFV₁, P-wave terminal force in lead V1; Q4, fourth quartile; D10, tenth decile. |
# Use the already-calculated Model C dataset from the exclusions section
Athletes_complete <- Athletes_model_C
cat("Complete cases for multivariable model:", nrow(Athletes_complete), "\n")Complete cases for multivariable model: 328
cat("AF cases:", sum(Athletes_complete$AF), "\n\n")AF cases: 77
if (nrow(Athletes_complete) >= 50) {
# Fit multivariable model
mv_model <- glm(AF ~ Age_decades + High_normal_BP + LAVi_34 + LAVi_60 +
P_wave_120 + PTFV1_high + PAC_100 + highest_decile,
data = Athletes_complete,
family = binomial())
# Get coefficients
mv_coefs <- broom::tidy(mv_model, conf.int = TRUE, exponentiate = FALSE) %>%
filter(term != "(Intercept)")
mv_or <- broom::tidy(mv_model, conf.int = TRUE, exponentiate = TRUE) %>%
filter(term != "(Intercept)")
# Calculate points (from beta coefficients)
beta_values <- mv_coefs$estimate
min_beta <- min(abs(beta_values))
points <- round(abs(beta_values) / min_beta)
# Create summary table
supp_table_4 <- data.frame(
Variable = c("Age (per 10 years)", "SBP ≥130 mmHg", "LAVi >34 mL/m²",
"LAVi >60 mL/m²", "P-wave >120 ms", "PTFV₁ ≥4 mV·ms",
"PACs ≥100/24h", "AF-PRS top decile"),
Beta = sprintf("%.3f", mv_coefs$estimate),
SE = sprintf("%.3f", mv_coefs$std.error),
OR = sprintf("%.2f", mv_or$estimate),
CI_lower = sprintf("%.2f", mv_or$conf.low),
CI_upper = sprintf("%.2f", mv_or$conf.high),
P_value = sapply(mv_coefs$p.value, format_p_value),
Points = points,
stringsAsFactors = FALSE
)
colnames(supp_table_4) <- c(
"Variable",
"β (coefficient)",
"SE",
"OR",
"CI lower",
"CI upper",
"p value",
"Points Assigned"
)
cat("\n## Supplementary Table 4\n\n")
print(kable(supp_table_4,
caption = paste0("Supplementary Table 4. Multivariable Logistic Regression Model Predicting Atrial Fibrillation in Endurance Athletes (n=", nrow(Athletes_complete), ")"),
align = c("l", "c", "c", "c", "c", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = TRUE, font_size = 10) %>%
add_footnote(c("Extended model incorporating genetic risk assessment. Statistical presentation as per Table 3. Abbreviations: AF, atrial fibrillation; CI, confidence interval; LAVi, left atrial volume index; OR, odds ratio; PAC, premature atrial contraction; PRS, polygenic risk score; PTFV₁, P-wave terminal force in lead V1; SBP, systolic blood pressure."),
notation = "none"))
# Print model performance
cat("\n**Model Performance:**\n")
cat("- AIC:", round(AIC(mv_model), 1), "\n")
cat("- Deviance:", round(deviance(mv_model), 1), "\n")
} else {
cat("Insufficient complete cases for multivariable modelling\n")
}10.4 Supplementary Table 4
| Variable | β (coefficient) | SE | OR | CI lower | CI upper | p value | Points Assigned |
|---|---|---|---|---|---|---|---|
| Age (per 10 years) | 0.247 | 0.105 | 1.28 | 1.05 | 1.58 | 0.018 | 1 |
| SBP ≥130 mmHg | 0.970 | 0.341 | 2.64 | 1.36 | 5.23 | 0.004 | 4 |
| LAVi >34 mL/m² | 0.857 | 0.465 | 2.36 | 0.99 | 6.24 | 0.065 | 3 |
| LAVi >60 mL/m² | 0.983 | 0.658 | 2.67 | 0.72 | 9.68 | 0.135 | 4 |
| P-wave >120 ms | 0.970 | 0.354 | 2.64 | 1.32 | 5.33 | 0.006 | 4 |
| PTFV₁ ≥4 mV·ms | 0.731 | 0.458 | 2.08 | 0.83 | 5.08 | 0.111 | 3 |
| PACs ≥100/24h | 1.555 | 0.365 | 4.74 | 2.33 | 9.78 | <0.001 | 6 |
| AF-PRS top decile | 0.948 | 0.459 | 2.58 | 1.04 | 6.33 | 0.039 | 4 |
| Extended model incorporating genetic risk assessment. Statistical presentation as per Table 3. Abbreviations: AF, atrial fibrillation; CI, confidence interval; LAVi, left atrial volume index; OR, odds ratio; PAC, premature atrial contraction; PRS, polygenic risk score; PTFV₁, P-wave terminal force in lead V1; SBP, systolic blood pressure. |
Model Performance: - AIC: 266.9 - Deviance: 248.9
# Find matching data
matched_cases <- matched_controls <- data.frame()
for (dataset in list(Athletes_NoCM, NoCM, AM_paper)) {
if ("Matching" %in% names(dataset)) {
temp <- dataset %>% filter(!is.na(Matching) & Matching %in% c("Case", "Control"))
if (nrow(temp) > 0) {
matched_cases <- temp %>% filter(Matching == "Case")
matched_controls <- temp %>% filter(Matching == "Control")
break
}
}
}
if (nrow(matched_cases) > 0 && nrow(matched_controls) > 0) {
supp_matched <- data.frame(Variable = character(), stringsAsFactors = FALSE)
matched_combined <- rbind(matched_cases %>% mutate(group = "AF"), matched_controls %>% mutate(group = "No AF"))
for (var in all_variables) {
if (!var %in% names(matched_cases) || !var %in% names(matched_controls)) next
is_cat <- var %in% categorical_vars
is_nonnorm <- var %in% nonnormal_vars
supp_matched <- rbind(supp_matched, data.frame(
Variable = var,
AF = format_stats(matched_cases, var, is_cat, is_nonnorm),
NoAF = format_stats(matched_controls, var, is_cat, is_nonnorm),
P = format_p_value(calculate_p_value(matched_combined, var, "group", is_cat, is_nonnorm)),
stringsAsFactors = FALSE
))
}
supp_matched$Category <- sapply(supp_matched$Variable, assign_category)
supp_matched$Variable <- sapply(supp_matched$Variable, clean_variable_names)
# Order variables the same way as Tables 1 and 2
supp_matched$Variable <- factor(supp_matched$Variable, levels = variable_order)
supp_matched <- supp_matched %>% arrange(Variable)
supp_matched$Variable <- as.character(supp_matched$Variable)
# Table 5: Demographics and Exercise Capacity
table5_cats <- c("Demographics", "Exercise Capacity")
supp_5_cat <- supp_matched %>%
filter(Category %in% table5_cats) %>%
mutate(Category = factor(Category, levels = table5_cats))
# Table 6: Cardiac Parameters with same structure as Table 2
table6_cat_order <- c("Left ventricular parameters", "Left atrial parameters",
"Right Ventricular Parameters", "Doppler Parameters",
"Electrocardiographic", "Biochemical")
supp_6_cat <- supp_matched %>%
filter(Category %in% table6_cat_order) %>%
mutate(Category = factor(Category, levels = table6_cat_order))
supp_5 <- supp_5_cat %>% select(-Category)
supp_6 <- supp_6_cat %>% select(-Category)
colnames(supp_5) <- c("Characteristic", paste0("AF (n=", nrow(matched_cases), ")"),
paste0("No AF (n=", nrow(matched_controls), ")"), "p value")
colnames(supp_6) <- colnames(supp_5)
cat("\n## Supplementary Table 5\n\n")
print(kable(supp_5, caption = "Supplementary Table 5. Age- and Sex-Matched Analysis: Baseline Characteristics (n=160)",
align = c("l", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = TRUE, font_size = 10) %>%
pack_rows(index = table(supp_5_cat$Category)[table5_cats]) %>%
add_footnote(c("Data presentation and statistical methods as per Supplementary Table 1. Cases and controls were matched 1:1 on age (±5 years) and sex.\nAbbreviations: As per Supplementary Table 1; AF, atrial fibrillation."),
notation = "none"))
cat("\n## Supplementary Table 6\n\n")
print(kable(supp_6, caption = "Supplementary Table 6. Age- and Sex-Matched Analysis: Cardiac Parameters (n=160)",
align = c("l", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = TRUE, font_size = 10) %>%
pack_rows(index = table(supp_6_cat$Category)[table6_cat_order]) %>%
add_footnote(c("Data presentation, statistical methods and abbreviations as per Supplementary Table 2."),
notation = "none"))
} else {
cat("Warning: No matched data available\n")
}10.5 Supplementary Table 5
| Characteristic | AF (n=80) | No AF (n=80) | p value |
|---|---|---|---|
| Demographics | |||
| Age, years | 56.5 [43.8-64.2] | 55 [43-65] | 0.868 |
| Male sex, n (%) | 73 (91.2%) | 73 (91.2%) | 1.000 |
| BMI, kg/m² | 24.5 [23.2-25.7] | 24.6 [22.9-26.7] | 0.929 |
| BSA, m² | 2 [1.9-2.1] | 2.1 [1.9-2.2] | 0.328 |
| Average HR, bpm | 63.5 ± 6.8 | 65.9 ± 7.1 | 0.048 |
| Minimum HR, bpm | 45 [40-48.5] | 46 [41-49] | 0.322 |
| Bradycardia burden, % (24h) | 5.6 [0.4-23.5] | 2.5 [0.3-21.8] | 0.474 |
| Hypertension, n (%) | 9 (11.5%) | 14 (17.5%) | 0.403 |
| SBP, mmHg | 131 [124-140.5] | 130 [122-137] | 0.581 |
| DBP, mmHg | 74.5 [69-81] | 74 [65-79.2] | 0.384 |
| Dyslipidaemia, n (%) | 17 (21.8%) | 20 (25%) | 0.774 |
| Diabetes, n (%) | 2 (2.6%) | 0 (0%) | 0.242 |
| History of smoking, n (%) | 16 (20.5%) | 13 (16.2%) | 0.627 |
| Alcohol, drinks/week | 3 [0-7.8] | 5.5 [1.8-12] | 0.057 |
| Exercise Capacity | |||
| VO₂peak, mL/kg/min | 43.9 ± 10.2 | 44.1 ± 14.3 | 0.921 |
| VO₂peak, L/min | 3570.1 ± 888.2 | 3590.7 ± 961.9 | 0.888 |
| % of predicted exercise capacity | 124.3 ± 17.9 | 123.8 ± 23.2 | 0.875 |
| Respiratory exchange ratio | 1.3 ± 0.1 | 1.3 ± 0.1 | 0.179 |
| Total weekly MET-hours | 78 [56-94] | 71 [44-94] | 0.391 |
| Years of endurance training | 26 [16.5-36] | 22 [15-32] | 0.421 |
|
Data presentation and statistical methods as per
Supplementary Table 1. Cases and controls were matched 1:1 on age (±5
years) and sex. Abbreviations: As per Supplementary Table 1; AF, atrial fibrillation. |
|||
10.6 Supplementary Table 6
| Characteristic | AF (n=80) | No AF (n=80) | p value |
|---|---|---|---|
| Left ventricular parameters | |||
| LVEDV, mL | 162.7 ± 39.8 | 166.4 ± 44.1 | 0.585 |
| LVESV, mL | 65.8 ± 17.3 | 69 ± 21.7 | 0.303 |
| LVEF, % | 58.4 ± 4.8 | 57.4 ± 5.3 | 0.213 |
| LV strain, % | -18.5 ± 2.2 | -18.8 ± 2.1 | 0.387 |
| Left atrial parameters | |||
| LAESV, mL | 89.5 [76.8-111.8] | 84 [68.5-97] | 0.010 |
| LAVi, mL/m² | 45.8 [38.1-54.6] | 41.9 [32.7-47.9] | 0.003 |
| Mildly abnormal >34 mL/m², n (%) | 72 (90%) | 57 (71.2%) | 0.005 |
| Severely abnormal >48 mL/m², n (%) | 32 (40%) | 20 (25%) | 0.063 |
| Extremely abnormal >60 mL/m², n (%) | 12 (15%) | 1 (1.2%) | 0.002 |
| LA:LV | 0.56 [0.48-0.71] | 0.51 [0.41-0.62] | 0.010 |
| Reservoir strain, % | 27.1 ± 8 | 29.5 ± 5.9 | 0.047 |
| Conduit strain, % | 16 ± 6.3 | 16.3 ± 6 | 0.763 |
| Contractile strain, % | 11.2 ± 4.7 | 13.4 ± 5.1 | 0.008 |
| Right Ventricular Parameters | |||
| RVEDV, mL | 147.7 ± 43.5 | 144 ± 44 | 0.595 |
| RVESV, mL | 80.3 ± 27.2 | 76.9 ± 25.9 | 0.433 |
| RVEF, % | 45.7 ± 8.5 | 46.3 ± 8.6 | 0.688 |
| Doppler Parameters | |||
| E velocity, m/s | 0.6 ± 0.2 | 0.6 ± 0.2 | 0.290 |
| A velocity, m/s | 0.4 ± 0.1 | 0.5 ± 0.1 | 0.194 |
| E/A ratio | 1.6 ± 1.4 | 1.4 ± 0.6 | 0.271 |
| E′ velocity, cm/s | 9.6 ± 2.8 | 8.7 ± 2.4 | 0.034 |
| E/E’ ratio | 5.9 ± 1.9 | 6.1 ± 1.7 | 0.680 |
| Electrocardiographic | |||
| P wave duration, ms | 124 [112-131] | 120 [109.8-128] | 0.246 |
| P-wave duration >120 ms, n (%) | 45 (60.8%) | 41 (51.2%) | 0.302 |
| P wave axis, ° | 50.3 ± 33.8 | 50 ± 22.9 | 0.953 |
| PTFV₁, mV·ms | 2.4 [1.2-4.1] | 2 [0.9-3] | 0.020 |
| PTFV₁ ≥4 mV·ms, n (%) | 21 (28.4%) | 6 (7.5%) | 0.001 |
| Advanced IAB, n (%) | 11 (15.3%) | 13 (16.2%) | 1.000 |
| PACs per 24 h | 107 [20-480] | 25.5 [9-73.5] | <0.001 |
| ≥100 PACs per 24 h, n (%) | 38 (54.3%) | 16 (20.5%) | <0.001 |
| Biochemical | |||
| BNP, pg/mL | 22.5 [11.2-52.8] | 15.8 [7.3-29.1] | 0.005 |
| cTnI, ng/L | 4 [2-7] | 3 [2-5] | 0.012 |
| Data presentation, statistical methods and abbreviations as per Supplementary Table 2. | |||
For Tables 7 & 8 (Genetic Analysis):
# Filter extreme PRS groups
high_prs <- Athletes_NoCM %>% filter(!is.na(`AF decile`) & `AF decile` == 10)
low_prs <- Athletes_NoCM %>% filter(!is.na(`AF decile`) & `AF decile` == 1)
if (nrow(high_prs) > 0 && nrow(low_prs) > 0) {
# Build comparison table
supp_prs <- data.frame(Variable = character(), stringsAsFactors = FALSE)
prs_combined <- rbind(
high_prs %>% mutate(PRS_group = "High"),
low_prs %>% mutate(PRS_group = "Low")
)
for (var in all_variables) {
if (!var %in% names(high_prs) || !var %in% names(low_prs)) next
is_cat <- var %in% categorical_vars
is_nonnorm <- var %in% nonnormal_vars
supp_prs <- rbind(supp_prs, data.frame(
Variable = var,
High = format_stats(high_prs, var, is_cat, is_nonnorm),
Low = format_stats(low_prs, var, is_cat, is_nonnorm),
P = format_p_value(calculate_p_value(prs_combined, var, "PRS_group", is_cat, is_nonnorm)),
stringsAsFactors = FALSE
))
}
# Assign categories and clean names
supp_prs$Category <- sapply(supp_prs$Variable, assign_category)
supp_prs$Variable <- sapply(supp_prs$Variable, clean_variable_names)
# Order variables the same way as other tables
supp_prs$Variable <- factor(supp_prs$Variable, levels = variable_order)
supp_prs <- supp_prs %>% arrange(Variable)
supp_prs$Variable <- as.character(supp_prs$Variable)
# Table 7: Demographics and Exercise Capacity
table7_cats <- c("Demographics", "Exercise Capacity")
supp_7_cat <- supp_prs %>%
filter(Category %in% table7_cats) %>%
mutate(Category = factor(Category, levels = table7_cats))
# Table 8: Cardiac Parameters with same structure as Tables 2 and 6
table8_cat_order <- c("Left ventricular parameters", "Left atrial parameters",
"Right Ventricular Parameters", "Doppler Parameters",
"Electrocardiographic", "Biochemical")
supp_8_cat <- supp_prs %>%
filter(Category %in% table8_cat_order) %>%
mutate(Category = factor(Category, levels = table8_cat_order))
# Remove category column for display
supp_7 <- supp_7_cat %>% select(-Category)
supp_8 <- supp_8_cat %>% select(-Category)
# Rename columns
colnames(supp_7) <- c("Characteristic",
paste0("AF-PRS ≥90th percentile (n=", nrow(high_prs), ")"),
paste0("AF-PRS <10th percentile (n=", nrow(low_prs), ")"),
"p value")
colnames(supp_8) <- colnames(supp_7)
# Display tables
cat("\n## Supplementary Table 7\n\n")
print(kable(supp_7,
caption = "Supplementary Table 7. Extreme Genetic Risk Analysis: Baseline Characteristics",
align = c("l", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = TRUE, font_size = 10) %>%
pack_rows(index = table(supp_7_cat$Category)[table7_cats]) %>%
add_footnote(c("Comparison of athletes in highest (≥90th percentile) versus lowest (<10th percentile) deciles of AF-PRS. Data presentation and statistical methods as per Supplementary Table 1.\nAbbreviations: As per Supplementary Table 1; AF, atrial fibrillation; PRS, polygenic risk score."),
notation = "none"))
cat("\n## Supplementary Table 8\n\n")
print(kable(supp_8,
caption = "Supplementary Table 8. Extreme Genetic Risk Analysis: Cardiac Parameters",
align = c("l", "c", "c", "c")) %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"),
full_width = TRUE, font_size = 10) %>%
pack_rows(index = table(supp_8_cat$Category)[table8_cat_order]) %>%
add_footnote(c("Data presentation and statistical methods as per Supplementary Table 2.\nAbbreviations: As per Supplementary Table 2."),
notation = "none"))
} else {
cat("Insufficient PRS data for extreme risk analysis\n")
}10.7 Supplementary Table 7
| Characteristic | AF-PRS ≥90th percentile (n=48) | AF-PRS <10th percentile (n=29) | p value |
|---|---|---|---|
| Demographics | |||
| Age, years | 45.5 [22.2-65.2] | 42 [22-59] | 0.666 |
| Male sex, n (%) | 40 (83.3%) | 24 (82.8%) | 1.000 |
| BMI, kg/m² | 24.2 [22.4-26.4] | 24 [22.3-26.3] | 0.785 |
| BSA, m² | 2 [1.9-2.1] | 2 [1.8-2.1] | 0.829 |
| Average HR, bpm | 67.9 ± 10.4 | 64.8 ± 7.4 | 0.153 |
| Minimum HR, bpm | 45 [40.2-49] | 45 [39.5-48.5] | 0.575 |
| Bradycardia burden, % (24h) | 2.1 [0.4-14] | 4.5 [0.3-26] | 0.351 |
| Hypertension, n (%) | 6 (13.3%) | 2 (7.1%) | 0.702 |
| SBP, mmHg | 130 [120.5-135] | 123 [118-131] | 0.161 |
| DBP, mmHg | 74 [62.5-79] | 71 [62-76] | 0.401 |
| Dyslipidaemia, n (%) | 7 (15.6%) | 5 (17.9%) | 1.000 |
| Diabetes, n (%) | 1 (2.2%) | 0 (0%) | 1.000 |
| History of smoking, n (%) | 6 (13.3%) | 3 (10.7%) | 1.000 |
| Alcohol, drinks/week | 0 [0-3] | 3 [0-8] | 0.132 |
| Exercise Capacity | |||
| VO₂peak, mL/kg/min | 45.5 ± 15 | 47.1 ± 13.2 | 0.624 |
| VO₂peak, L/min | 3548.8 ± 1089.1 | 3633.6 ± 988.2 | 0.727 |
| % of predicted exercise capacity | 119.6 ± 19.9 | 121.2 ± 20.7 | 0.750 |
| Respiratory exchange ratio | 1.3 ± 0.1 | 1.3 ± 0.1 | 0.291 |
| Total weekly MET-hours | 77 [46-105.8] | 84 [58-104] | 0.874 |
| Years of endurance training | 20 [9.2-30] | 19 [15-20] | 0.994 |
|
Comparison of athletes in highest (≥90th percentile) versus
lowest (<10th percentile) deciles of AF-PRS. Data presentation and
statistical methods as per Supplementary Table 1. Abbreviations: As per Supplementary Table 1; AF, atrial fibrillation; PRS, polygenic risk score. |
|||
10.8 Supplementary Table 8
| Characteristic | AF-PRS ≥90th percentile (n=48) | AF-PRS <10th percentile (n=29) | p value |
|---|---|---|---|
| Left ventricular parameters | |||
| LVEDV, mL | 160.9 ± 45.4 | 164.2 ± 32.6 | 0.714 |
| LVESV, mL | 68 ± 21.8 | 65 ± 17.5 | 0.518 |
| LVEF, % | 56.6 ± 6.1 | 58.9 ± 5.4 | 0.108 |
| LV strain, % | -18.2 ± 2.6 | -18.7 ± 2.1 | 0.412 |
| Left atrial parameters | |||
| LAESV, mL | 84.5 [71-98.2] | 81 [72-95] | 0.829 |
| LAVi, mL/m² | 43.6 [35.8-49.1] | 42.7 [35.2-48.6] | 0.946 |
| Mildly abnormal >34 mL/m², n (%) | 38 (79.2%) | 22 (75.9%) | 0.956 |
| Severely abnormal >48 mL/m², n (%) | 14 (29.2%) | 8 (27.6%) | 1.000 |
| Extremely abnormal >60 mL/m², n (%) | 4 (8.3%) | 1 (3.4%) | 0.645 |
| LA:LV | 0.53 [0.43-0.69] | 0.49 [0.43-0.64] | 0.444 |
| Reservoir strain, % | 28.8 ± 8.9 | 31 ± 7.3 | 0.262 |
| Conduit strain, % | 18.1 ± 6.7 | 19.5 ± 7.6 | 0.458 |
| Contractile strain, % | 10.7 ± 5.8 | 11.6 ± 4.8 | 0.496 |
| Right Ventricular Parameters | |||
| RVEDV, mL | 145.2 ± 40.9 | 131.6 ± 33.2 | 0.120 |
| RVESV, mL | 77.3 ± 20.9 | 68.7 ± 16.3 | 0.050 |
| RVEF, % | 45.8 ± 9.2 | 46.8 ± 9.8 | 0.666 |
| Doppler Parameters | |||
| E velocity, m/s | 0.7 ± 0.2 | 0.7 ± 0.2 | 0.321 |
| A velocity, m/s | 0.4 ± 0.1 | 0.4 ± 0.1 | 0.733 |
| E/A ratio | 1.7 ± 0.8 | 1.8 ± 1 | 0.813 |
| E′ velocity, cm/s | 10.2 ± 3.3 | 10 ± 3.2 | 0.783 |
| E/E’ ratio | 6 ± 1.6 | 5.7 ± 1.4 | 0.465 |
| Electrocardiographic | |||
| P wave duration, ms | 114 [104-129] | 118 [110-125] | 0.991 |
| P-wave duration >120 ms, n (%) | 20 (46.5%) | 12 (41.4%) | 0.851 |
| P wave axis, ° | 56.1 ± 21 | 49.7 ± 31.8 | 0.359 |
| PTFV₁, mV·ms | 2 [1.2-3.4] | 1.4 [0-2.8] | 0.107 |
| PTFV₁ ≥4 mV·ms, n (%) | 9 (20.9%) | 4 (13.8%) | 0.541 |
| Advanced IAB, n (%) | 5 (12.2%) | 0 (0%) | 0.075 |
| PACs per 24 h | 34.5 [11.8-109.8] | 16.5 [6-67.5] | 0.127 |
| ≥100 PACs per 24 h, n (%) | 12 (27.3%) | 6 (22.2%) | 0.846 |
| Biochemical | |||
| BNP, pg/mL | 12.6 [9.4-34.7] | 12.4 [5.7-30.1] | 0.438 |
| cTnI, ng/L | 3 [2-9] | 2 [2-4] | 0.489 |
|
Data presentation and statistical methods as per
Supplementary Table 2. Abbreviations: As per Supplementary Table 2. |
|||
11 Supplementary Figures
11.1 Supplementary Figure 1
Supplementary Figure 1. Distribution of Clinical Risk Scores by AF Status
if (!exists("cutoffs")) {
prob_thresholds <- c(0.10, 0.40, 0.70)
cutoffs <- round(map_dbl(prob_thresholds,
~min(Athletes_analysis$clinical_score_B[Athletes_analysis$predicted_prob_B >= .x], na.rm = TRUE)))
}
ggplot(Athletes_analysis, aes(x = clinical_score_B_rounded,
fill = factor(AF, labels = c("No AF", "AF")))) +
geom_histogram(binwidth = 1, position = "dodge",
alpha = 0.8, color = "black") +
# Use dynamic cutoffs instead of hardcoded values
geom_segment(aes(x = cutoffs[1], xend = cutoffs[1], y = 0, yend = Inf),
linetype = "dashed", color = "black", linewidth = 0.8) +
geom_segment(aes(x = cutoffs[2], xend = cutoffs[2], y = 0, yend = Inf),
linetype = "dashed", color = "black", linewidth = 0.8) +
geom_segment(aes(x = cutoffs[3], xend = cutoffs[3], y = 0, yend = Inf),
linetype = "dashed", color = "black", linewidth = 0.8) +
annotate("text",
x = c(cutoffs[1]/2,
mean(c(cutoffs[1], cutoffs[2])),
mean(c(cutoffs[2], cutoffs[3])),
cutoffs[3] + 3),
y = -2,
label = c("Low", "Moderate", "High", "Very High"),
size = 3.5, fontface = "italic") +
scale_fill_manual(values = c("No AF" = "#0072B2", "AF" = "#990000")) +
scale_y_continuous(expand = expansion(mult = c(0.08, 0.12))) +
labs(x = "Clinical Risk Score",
y = "Number of Athletes",
fill = "AF Status") +
coord_cartesian(clip = "off") +
theme_minimal(base_size = 14) +
theme(legend.position = "bottom",
plot.margin = margin(t = 10, r = 10, b = 25, l = 10))
Legend = Distribution of Clinical Risk Scores by AF Status
11.2 Supplementary Figure 2
Supplementary Figure 2. AF prevalence by risk category
risk_prev <- Athletes_analysis %>%
group_by(risk_category) %>%
summarise(
N = n(),
AF_Cases = sum(AF),
AF_Prevalence = mean(AF) * 100,
CI_lower = binom.test(AF_Cases, N)$conf.int[1] * 100,
CI_upper = binom.test(AF_Cases, N)$conf.int[2] * 100,
.groups = "drop"
)
ggplot(risk_prev, aes(x = risk_category, y = AF_Prevalence)) +
geom_col(fill = "#6a994e", alpha = 0.8) +
geom_errorbar(aes(ymin = CI_lower, ymax = CI_upper),
width = 0.2, size = 0.6) +
geom_text(aes(y = CI_upper, label = paste0(round(AF_Prevalence, 1), "%")),
vjust = -2.5, size = 4, color = "black") +
geom_text(aes(y = CI_upper, label = paste0("(", AF_Cases, "/", N, ")")),
vjust = -1.0, size = 4, color = "black") +
scale_y_continuous(expand = expansion(mult = c(0, 0.15))) +
labs(x = "Risk Category",
y = "AF Prevalence (%)") +
theme_minimal(base_size = 14) +
theme(panel.grid.major.x = element_blank())
11.3 Supplementary Figure 3
Supplementary Figure 3. Profile of Atrial Remodelling Markers by Clinical Risk Category
if (!"risk_category" %in% names(Athletes_analysis)) {
# Define the probability thresholds
prob_thresholds <- c(0.10, 0.40, 0.70)
# Calculate the corresponding clinical score cutoffs dynamically
cutoffs <- round(purrr::map_dbl(prob_thresholds,
~min(Athletes_analysis$clinical_score_B[Athletes_analysis$predicted_prob_B >= .x], na.rm = TRUE)))
# Define the breaks for the cut function
score_breaks <- c(-Inf, cutoffs, Inf)
# Create the risk categories using the rounded integer score for consistency
Athletes_analysis <- Athletes_analysis %>%
mutate(
clinical_score_B_rounded = round(clinical_score_B),
risk_category = cut(clinical_score_B_rounded, breaks = score_breaks,
labels = c("Low", "Moderate", "High", "Very High"),
include.lowest = TRUE)
)
}
# Prepare radar chart data with Age >45 as binary prevalence
radar_data_C <- Athletes_analysis %>%
group_by(risk_category) %>%
summarise(
`Age >45 years (%)` = mean(Age > 45, na.rm = TRUE) * 100,
`SBP ≥130 mmHg (%)` = mean(High_normal_BP, na.rm = TRUE) * 100,
`LAVi 35-60 mL/m² (%)` = mean(LAVi_34, na.rm = TRUE) * 100,
`LAVi >60 mL/m² (%)` = mean(LAVi_60, na.rm = TRUE) * 100,
`P wave >120ms (%)` = mean(P_wave_120, na.rm = TRUE) * 100,
`PTFV1 ≥4.0 mV·ms (%)` = mean(PTFV1_high, na.rm = TRUE) * 100,
`PACs ≥100/24h (%)` = mean(PAC_100, na.rm = TRUE) * 100,
.groups = "drop"
)
# Convert to data frame with row names
radar_df <- as.data.frame(radar_data_C[, -1])
rownames(radar_df) <- radar_data_C$risk_category
# Add max and min rows for fmsb package (all variables now 0-100%)
max_row <- rep(100, ncol(radar_df))
min_row <- rep(0, ncol(radar_df))
radar_chart_data <- rbind(max_row, min_row, radar_df)
# Define colors for each risk category
colors_border <- c("#0072B2", "#67A9CF", "#F4A582", "#C00000")
colors_fill <- adjustcolor(colors_border, alpha.f = 0.3)
# Set up 2x2 plotting layout
par(mfrow = c(2, 2), mar = c(1, 1, 3, 1))
# Create radar chart for each risk category
for (i in 1:4) {
radarchart(
radar_chart_data[c(1, 2, i + 2), ],
axistype = 1,
pcol = colors_border[i],
pfcol = colors_fill[i],
plwd = 2,
cglcol = "grey",
cglty = 1,
cglwd = 0.8,
axislabcol = "grey",
vlcex = 0.9,
title = rownames(radar_chart_data)[i + 2],
title.col = "black",
cex.main = 1.5
)
}
# Reset plotting parameters
par(mfrow = c(1, 1))11.4 Supplementary Figure 4
Supplementary Figure 4. Clinical risk score cutoffs corresponding to predicted AF probability thresholds
# Define the probability thresholds used for risk stratification in Table 4
prob_thresholds <- c(0.10, 0.40, 0.70)
# Calculate the corresponding clinical score cutoffs dynamically
# This ensures the vertical lines in the plot match the table's risk categories
cutoffs <- round(purrr::map_dbl(prob_thresholds,
~min(Athletes_analysis$clinical_score_B[Athletes_analysis$predicted_prob_B >= .x], na.rm = TRUE)))
# Generate the plot
ggplot(Athletes_analysis, aes(x = clinical_score_B,
y = predicted_prob_B * 100)) +
geom_point(aes(color = factor(AF, labels = c("No AF", "AF"))),
alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "black") +
# Use the thresholds variable for the horizontal lines
geom_hline(yintercept = prob_thresholds * 100,
linetype = "dashed", alpha = 0.5) +
# Use the calculated cutoffs variable for the vertical lines
geom_vline(xintercept = cutoffs,
linetype = "dashed", alpha = 0.5) +
scale_color_manual(values = c("No AF" = "#0072B2", "AF" = "#990000")) +
labs(x = "Clinical Risk Score",
y = "Predicted AF Probability (%)",
color = "AF Status") +
theme_minimal(base_size = 14) +
theme(legend.position = "bottom")
Legend: Predicted AF probabilities were stratified into four probability thresholds: <10%, 10-40%, 40-70%, and >70%.