The research aimed to derive and externally validate a clinical algorithm to predict suboptimal low density lipoprotein cholesterol response LDLC in patients initiated on statin therapy Research Design ID: 1046099
Download Presentation The PPT/PDF document "The Research Question One in two individ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. The Research QuestionOne in two individuals prescribed statins for primary prevention of cardiovascular disease are observed to not achieve targeted cholesterol reductionsThe research aimed to derive and externally validate a clinical algorithm to predict sub-optimal low density lipoprotein cholesterol response (LDL-C) in patients initiated on statin therapy
2. Research Design and MethodTwo population cohorts were harmonized and used to develop and validate the clinical algorithm in different primary care settingsUnited Kingdom (Clinical Practice Research Datalink) – 183,213 patients initiated on statins Hong Kong (Clinical Data Analysis Reporting Systems) – 170,904 patients initiated on statinsWe used both standard statistical approaches (multivariable logistic regression) and machine-learning algorithms (deep neural network)We assessed performance of the algorithm by area under the curve (AUC) and calibration slope (predicted vs observed risk)We then assessed 10-year risk of any major adverse CVD event, stratifying for predicted statin response and baseline CVD risk.
3. What the Research FoundThe clinical algorithm included 11 variables: age, gender, baseline LDL-C, atrial fibrillation, diabetes, dyslipidemia, statin dosage, treated hypertension, oral corticosteroids, non-statin drugs, number of concurrent medications prescribedPerformance of the logistic regression model resulted in an AUC of 0.703 in the UK cohort and an AUC of 0.683 in the HK cohort. The more complex machine-learning algorithm resulted almost identical AUC compared to the simpler logistic model (0.702 in UK cohort; 0.682 in HK cohort)Calibration was nearly perfect in the UK population (slope = 1.0), but the algorithm under-predicted risk in the HK population (slope = 0.87)Patients with predicted sub-optimal LDL-C reduction (compared to optimal LDL-C reduction) also experienced significantly increased risk of CVD in both high and low risk CVD groups over 10 years follow-up
4. What this means for Clinical PracticeDetermining an individual’s likelihood for sub-optimal LDL-C response to statins should be considered along baseline cardiovascular risk assessment to enhance cholesterol management, which could include more routine monitoring, higher dosage, combination therapies, or alternative medication if contra-indicated