Cardiovascular risk calculator estimates vary widely in people with diabetes  — ASN Events
  1. Schedule
  2. Joint Poster Session
  3. Cardiovascular risk calculator estimates vary widely in people with diabetes 

Cardiovascular risk calculator estimates vary widely in people with diabetes  (#317)

Aditia Nayak 1 2 , Hern Kai Tan Ronald 1 2 , Anna Duke 3 4 , Jenny Gunton 1 2 5
  1. The Centre for Diabetes, Obesity and Endocrinology Research, the Westmead Institute for Medical Research, Sydney, NSW, Australia
  2. Sydney Medical School, Faculty of Medicine and Health, University of Sydney, Sydney, NSW, Australia
  3. Department of Diabetes and Endocrinology, Blacktown Hospital, Sydney, NSW, Australia
  4. School of Medicine, Conjoint Lecturer, Western Sydney University, Blacktown, Sydney, NSW, Australia
  5. Department of Diabetes and Endocrinology, WestmeadHospital. Westmead, Sydney, NSW, Australia

Cardiovascular (CV) disease is a leading cause of death and disability worldwide, and people with diabetes face a 2-4-fold greater risk. CV risk estimation is important to target interventions.
Accuracy and performance of 5 different CV risk calculators (ASCVD+, AusCVD, Mesa-nhlbi, QRisk3, and StenoT1 Risk Engine) were compared in 200 patients attending diabetes clinics in Western Sydney. We hypothesised that there would be small between-calculator differences.  


We collected data from 100 patients attending diabetes clinics at Westmead and 100 at Blacktown Hospital in 2019. Risk scores were calculated for each risk engine. Future CV events were collected over the following 5 years. Calculators were compared Bland-Altman plots, and ROC analysis.  

 

Risk scores differed markedly between calculators. Average 10-year risk ranging from 7.7% to 22.1% across calculators. The greatest range of 10-year risk estimates in one person was 5.2-80.7% (75.5% range). Despite significantSpearman’s correlations for calculator-pairs, Bland-Altman analysis revealed statistically significant differences, wide limits of agreement and poor absolute agreement. Most people (83%) had a larger range of estimates than their mean/median risk estimate. 

Actual 5-year major adverse CV events (MACE) occurred in 21people over 5 years. Receiver operator characteristic (ROC) curves showed significant prediction with area under the curves for AusCVD of 0.686, ASCVD 0.657, Qrisk3 of 0.717 and Steno of 0.739. ROC was not significantly predictive for Mesa. Eighteen people died, 10 of whom had MACE. ROC were predictive of MACE or death with AUCs of 0.719 (AusCVD), 0.699 (ASCVD), 0.685 (Mesa), 0.727 (Qrisk3) and 0.747 (StenoT1D).  

 

Calculators are not interchangeable, and discriminative ability for predicting actual 5-year CV events was modest. For this group of people with diabetes, Qrisk3 and Steno performed the best. Despite use of SGLT2 inhibitors and GLP-1 receptor agonists over this time, absolute CV risk remained high.