American Heart Association PREVENT Equations & ASCVD Risk in Arab Women & Men
Okay, here’s a breakdown of the statistical analysis methods used in the study, based on the provided text. I’ll organize it into sections for clarity:
1. Handling Missing Data:
* Multiple Imputation: Missing values were addressed using multiple imputation.
* Variables with Missing Data: The following variables had missing data:
* Total Cholesterol (TC) (<1%)
* High-Density Lipoprotein Cholesterol (HDL-C) (<1%)
* HbA1c (4.7%)
* Creatinine (<1%)
* Systolic Blood Pressure (SBP) (<1%)
* Body Mass Index (BMI) (<1%)
* Assumption: The study assumed that the missing data were “missing at random” (MAR).
* Imputation Method: Predictive mean-matching was used to generate five imputed datasets.
* Combining Results: Rubin’s rule was used to combine the results from the five imputed datasets. (Reference 21)
2. Cohort Comparison:
* Comparison Groups: The Emirati validation cohort was compared to the original PREVENT study cohort.
* Statistical Tests:
* Continuous variables: Student’s t-test
* Categorical Variables: Chi-square test (two-sided)
* Stratification: All comparisons were stratified by sex (male/female). This was done to align with the original PREVENT Equation Study, which developed separate risk equations for men and women.
3. Risk Prediction & Equations Used:
* Risk Equations: Two equations were used to calculate 10-year ASCVD risk:
* PREVENT-ASCVD Base: Includes age, sex, TC, HDL-C, SBP, diabetes, smoking status, BMI, eGFR, and medication use (blood pressure and cholesterol).
* PREVENT-ASCVD HbA1c: Adds HbA1c levels to the PREVENT-ASCVD Base equation.
* Purpose of HbA1c Equation: The HbA1c equation is intended to provide a more precise risk assessment, notably for individuals with diabetes.
4. Evaluating Predictive Accuracy:
* Measures Used: Both calibration and discrimination were assessed.
* Discrimination:
* Measure: C-index (proposed by Harrell et al. – Reference 22). This is comparable to the Area under the Receiver Operating Characteristic Curve (AUC).
* Interpretation: A C-index > 0.75 was considered indicative of good discrimination.
* Calibration: Assesses the alignment between observed and predicted risk. (The text stops mid-sentence here, so details on how calibration was assessed are missing.)
* Analysis by Sex: The performance of both equations was compared separately for men and women.
In essence, the study used robust statistical methods to validate existing risk prediction equations in a new population (Emirati cohort), carefully addressing missing data and considering sex-specific differences.