Performance of cardiovascular risk scores in South Asian populations- a systematic review of the literature
The problem
People of South Asian descent have double the prevalence of known cardiovascular risk factors, such as hypertension and diabetes mellitus, and higher incidence of ischaemic heart disease compared to Caucasians. In addition, South Asian people affected by type-2 diabetes mellitus experience ischaemic cardiovascular events up to 7.4 years before Europeans. Including the South Asian diaspora and the native population, 1 in 5 people in 2050 will be of South Asian descent. Over recent years, cardiovascular risk scores have been incorporated into clinical guidelines and risk management programmes worldwide. However, there is no consensus on the most accurate model for estimating cardiovascular risk in South Asian populations. In this study we aimed to review existing cardiovascular risk models that have been applied to South Asian populations to guide use of risk models in this population and direct further research.
The approach
We performed a systematic review of the literature specifically targeting 4 search streams for terms affiliated to and including South Asian, cardiovascular, risk/score and specific risk scores for inclusion. We defined South Asian as those who reside or have ancestry belonging to the Indian subcontinent.
Findings
The search of 5 literature databases (MEDLINE, EMBASE, HMIC, AMED, PsychINFO) gathered 7560 records. Screening narrowed down 76 eligible papers for full-text assessment with 6 suitable for inclusion within the study. These 6 papers included data on 11 cardiovascular risk models (UKPDS 10-year, ETHRISK 10-year, QRISK2, Framingham 10-year coronary heart disease, Framingham 10-year cardiovascular disease and 6 modified versions of the Framingham 10-year coronary heart disease model). Objective measures of performance were only provided for modified versions of the Framingham score and QRISK2. Both had reasonable discrimination and accuracy with the area under the ROC curve ranging from 0.73 - 0.98 (n=7), sensitivity from 0.77 - 0.89 (n=5), and specificity from 0.96 to 0.98 (n=5). Calibration was similar across both models and better in men than women (ratio of observed to expected events 0.71-0.93 and 0.43-0.51 respectively).
Consequences
Considering the number of South Asians and prevalence of cardiovascular disease, we identified very few studies reporting performance of risk scores in South Asian populations and most were modified versions of the Framingham risk score. Furthermore, it was difficult to make comparisons as many did not provide measures of discrimination, accuracy and calibration. This work identifies a need for further research to evaluate risk scores in South Asians, and ideally to derive and validate a cardiovascular risk score within a South Asian population.
Credits
- Dipesh Gopal
- Juliet Usher-Smith