########################################### Load packages ################################ #install.packages("haven") #install.packages("ggplot2") #install.packages("dplyr") #install.packages("lavaan") ################################ Read packages ################# library(ggplot2) library(dplyr) library(haven) library(lavaan) ############## Define function to calculate metabolic syndrome using IDF Adult definition ##################### ms <- function(c_data) { # Insert required variables required_vars <- c("sex","waist circumference","Fasting_triglycerides","HDL-cholesterol","systolic blood pressure","diastolic blood pressure","Fasting glucose") missing_vars <- setdiff(required_vars, names(data)) if(length(missing_vars) > 0){ stop(paste("Missing variables:", paste(missing_vars, collapse=", "))) } # CENTRAL OBESITY data$central_obesity <- ifelse( (data$sex == 1 & c_data$waist circumference>= 94) | (data$sex == 2 & c_data$waist circumference>= 80), 1, 0 ) # RISK FACTORS data$rf_trig <- ifelse(data$Fasting_triglycerides >= 1.7, 1, 0) data$rf_hdl <- ifelse( (data$sex == 1 & data$HDL-cholesterol< 1.03) | (data$sex == 2 & data$HDL-cholesterol< 1.29), 1, 0 ) data$rf_bp <- ifelse( data$systolic blood pressure>= 130 | data$diastolic blood pressure>= 85, 1, 0 ) data$rf_glucose <- ifelse(data$Fasting glucose>= 5.6, 1, 0) # COUNT data$risk_factors <- data$rf_trig + data$rf_hdl + data$rf_bp + data$rf_glucose # METS data$ms <- ifelse( data$central_obesity == 1 & data$risk_factors >= 2, 1, 0 ) return(data) } data <- ms(data) View(data) table(data$ms) prop.table(table(data$ms)) table(data$sex , data$ms) ##############################Define metabolic syndrome at 16 years using IDF 90th percentile ########################## metabolic_syndrome <- function(data) { # ---- 1. Compute sex-specific 90th percentile for Waist Circumference ---- data <- data %>% group_by(sex) %>% mutate(wc_p90 = quantile(waist circumference, 0.9, na.rm = TRUE)) %>% ungroup() # ---- 2. Define central obesity (mandatory in IDF framework) ---- data$central_obesity <- ifelse(data$waist circumference>= data$wc_p90, 1, 0) # ---- 3. Define other risk factors ---- # Elevated glucose data$rf_glucose <- ifelse(data$Fasting glucose >= 5.6, 1, 0) # Blood pressure (either systolic OR diastolic) data$rf_bp <- ifelse(data$systolic Blood Pressure >= 130 | data$diastolic Blood Pressure >= 85, 1, 0) # Triglycerides data$rf_trig <- ifelse(data$Fasting triglycerides >= 1.7, 1, 0) # HDL (sex-specific) data$rf_hdl <- ifelse( (data$sex == 1 & data$HDL-cholesterol< 1.03) | # males (data$sex == 2 & data$FS_KOL_HDL < 1.29), # females 1, 0 ) # ---- 4. Count additional risk factors (excluding WC) ---- data$risk_factors <- data$rf_glucose + data$rf_bp + data$rf_trig + data$rf_hdl # ---- 5. Define metabolic syndrome (IDF logic adapted to p90 WC) ---- data$metabolic_syndrome <- ifelse( data$central_obesity == 1 & data$risk_factors >= 2, 1, 0 ) return(data) } # Apply function data <- metabolic_syndrome(data) View(data) table(data$metabolic_syndrome) prop.table(table(data$metabolic_syndrome)) table(data$sex,data$metabolic_syndrome) ###################################ASSOCIATION BETWEEN BMI SLOPES AND METABOLIC SYNDROME (MetS) ############################################# ##################################Logistic regression model for BMI Slopes and metabolic syndrome (MetS)############## s1 <- glm(factor(MetS) ~ slope, family = binomial(link = "logit"), data = data) summary(s1) exp(coef(s1)[2]) # Extract the coefficients and standard errors coef <- coef(s1) se <- sqrt(diag(vcov(s1))) # Calculate the 95% CI for the log-odds for slopes ci_log_odds <- coef["slope"] + c(-1.96, 1.96) * se["slope"] # Exponentiate to get the 95% CI for the OR ci_or <- exp(ci_log_odds) # Print the OR and its 95% CI ci_or #Adjusting for sex s2 <- glm(factor(MetS)~slope+factor(sex), family = binomial(link = "logit"), data = data) summary(s2) exp(coef(s2)[2]) # Extract the coefficients and standard errors coef <- coef(s2) se <- sqrt(diag(vcov(s2))) # Calculate the 95% CI for the log-odds for slopes ci_log_odds <- coef["slope"] + c(-1.96, 1.96) * se["slope"] # Exponentiate to get the 95% CI for the OR ci_or <- exp(ci_log_odds) # Print the OR and its 95% CI ci_or s3 <- glm(factor(MetS)~slope+factor(sex)+birthweight+maternal BMI+maternal education+maternal smoking+maternal occupation, family = binomial(link = "logit"), data = data) summary(s3) exp(coef(s3)[2]) # Extract the coefficients and standard errors coef <- coef(s3) se <- sqrt(diag(vcov(s3))) # Calculate the 95% CI for the log-odds ci_log_odds <- coef["slope"] + c(-1.96, 1.96) * se["slope"] # Exponentiate to get the 95% CI for the OR ci_or <- exp(ci_log_odds) # Print the OR and its 95% CI ci_or ###################################################PATH Analysis of BMI SLOPES & METABOLIC SYNDROME ######################################################### ##### Exposure: BMI slopes at 3 time points (slope1=0 to 5 months, slope2=1.5 to 3.5 years and slope3=6.5 to 10 years) ##### Outcome: Metabolic syndrome at 31 and 16 years #######################Date preparation############################### #Standardize variables if needed# library(lavaan) pathanalysis_data= data[, c("project_ID", "sex","slope1", "slope2","slope3","metabolic_syndrome")] colSums(is.na(pathanalysis_data)) sum(is.na(pathanalysis_data)) ################################################MULTI GROUP SEX SPECIFIC path analysis ################################################# ##################################### Path A : slope1 (0-5months),slope3 (6.5-10 years) & Metabolic Syndrome(MetS) ########################### model_multi_group <- ' ################################ # STRUCTURAL MODEL ################################ slope3 ~ c(f_b2, m_b2)*slope1 # Ordered outcome (probit under WLSMV) metabolic_syndrome ~ c(f_cp1, m_cp1)*slope1 + c(f_cp3, m_cp3)*slope3 ################################ # INDIRECT EFFECTS ################################ # slope1 -> slope3 -> MetS ind_s3_f := f_b2 * f_cp3 ind_s3_m := m_b2 * m_cp3 # Total effects slope 3 on MetS total_s3_f := f_cp3 total_s3_m := m_cp3 #Total effect of slope1 on MetS # direct + indirect total_s1_f := f_cp1 + (f_b2 * f_cp3) total_s1_m := m_cp1 + (m_b2 * m_cp3) ################################ # GENDER DIFFERENCE TESTS ################################ # Wald tests for group differences diff_direct := f_cp1 - m_cp1 diff_ind_s3 := ind_s3_f - ind_s3_m ' # MODEL FITTING fit_multi_group <- sem( model = model_multi_group, data = pathanalysis_data, group = "gender_scaled", ordered = "metabolic_syndrome_scaled", estimator = "WLSMV", parameterization = "theta" ) # MODEL SUMMARY summary( fit_multi_group, standardized = TRUE, rsquare = TRUE, ci = TRUE ) # FIT INDICES fitMeasures( fit_multi_group, c( "chisq", "df", "pvalue", "cfi", "tli", "rmsea", "rmsea.ci.lower", "rmsea.ci.upper", "srmr" ) ) ####################################################### Path B: slope2 (1.5 to 3.5 years), slope3 (6.5-10 years) & MetS ################### model_m<- ' ################################ # STRUCTURAL MODEL ################################ slope3 ~ c(f_b1, m_b1)*slope2 # Ordered outcome (probit under WLSMV) metabolic_syndrome ~ c(f_cp2, m_cp2)*slope2 + c(f_cp3, m_cp3)*slope3 ################################ # INDIRECT EFFECTS: SLOPE 2 ################################ # slope2 -> slope3 -> MetS ind_s3_f := f_b1 * f_cp3 ind_s3_m := m_b1 * m_cp3 # Total effect of slope3 total_s3_f := f_cp3 total_s3_m := m_cp3 # Total effect of slope1 on MetS total_s2_f := f_cp2 + (f_b1 * f_cp3) total_s2_m := m_cp2 + (m_b1 * m_cp3) ################################ # GENDER DIFFERENCE TESTS ################################ # Wald tests for group differences diff_direct := f_cp2 - m_cp2 diff_ind_s3 := ind_s3_f - ind_s3_m ' # MODEL FITTING fit_m <- sem( model = model_m, data = pathanalysis_data, group = "gender_scaled", ordered = "metabolic_syndrome_scaled", estimator = "WLSMV", parameterization = "theta" ) # MODEL SUMMARY summary( fit_m, standardized = TRUE, rsquare = TRUE, ci = TRUE ) # FIT INDICES fitMeasures( fit_m, c( "chisq", "df", "pvalue", "cfi", "tli", "rmsea", "rmsea.ci.lower", "rmsea.ci.upper", "srmr" ) ) ############################################################################################################ #Correlation between slopes and metabolic syndrome cor(pathanalysis_data[, c("slope1", "slope2", "slope3","metabolic_syndrome")], use = "complete.obs", method = "spearman") pathanalysis_data$metabolic_syndrome <- as.numeric(as.character(pathanalysis_data$metabolic_syndrome))