Cardiovascular disease (CVD) risk assessment is the first and critical step for guiding preventive strategies, including for those with diabetes mellitus (DM). While DM patients may have heterogeneous risk factors, currently there is no widely used CVD risk score for the US DM population. In the current AHA/ACC and ADA guidelines, the Pooled Cohort Equation (PCE) for atherosclerotic CVD (ASCVD) derived from four prospective cohort studies is recommended in diabetic population for risk assessment.¹ The PCE has been criticized for overestimation of CVD risk compared to the actual observed absolute risk, which may lead to poor calibration.² In addition, the PCE includes diabetes as a binary factor, ignoring its heterogeneity in risks and interaction with other risk factors as well as many diabetes-specific risk factors such as HbA1c and DM duration and may contribute to the poorer discrimination ability of PCE in the diabetic population than in non-diabetic population.

The most well-known CVD risk score for type 2 diabetes mellitus (T2DM), the United Kingdom Prospective Diabetes Study (UKPDS), has limitations too. It was developed from the UK T2DM population where the predominant participants were Caucasians, making it less appropriate for use in the US population with great ethnic diversity. Further, it was derived from early cohorts in which the exposure to risk factors, disease incidence, screening tests and preventive management of CVD were notably different from more contemporary population and therefore outdated without a recalibration.³ UKPDS risk score has been updated with 30 years of follow-up and included other microvascular endpoints.⁴ Unfortunately, few studies except for UKPDS have focused on microvascular complications, mainly due to a lack of accurate data on these endpoints.

Emerging clinical trials involving diabetes patients provide us with large sample size, detailed surveillance of cardiometabolic risk factors, complete follow-up data and strictly adjudicated events, making them wonderful resources to develop new risk scores for various types of T2DM complications. In a recently published article, Basu et al.⁵ utilized data from the ACCORD trial and created a comprehensive risk scoring system called RECODe (Risk Equations for Complications Of type 2 Diabetes) for both microvascular and macrovascular complications. The study followed standard protocols of developing and validating risk scores while also imposed some leading-edge methodology including machine learning in risk factor selection steps. Multiple risk scores were developed to predict 10-year risk of each microvascular complications including nephropathy, retinopathy and neuropathy, and macrovascular endpoints of myocardial infarction, stroke, congestive heart failure and cardiovascular mortality as well as composite endpoints of ASCVD and total mortality. For the ASCVD risk score, key risk factors include age, gender, race, smoking status, CVD history, systolic blood pressure, hypertension medication, statin user, anticoagulants, HbA1c, total cholesterol, HDL-C, serum creatinine and urine albumin creatinine ratio. Internal and external validation achieved moderate discrimination and good calibration. When compared with UKPDS and PCE, statistically significant improvement in net reclassification index was observed.

The RECODe is the first risk scoring system developed for the US population to predict risk for the whole spectrum of diabetes macrovascular and microvascular complications, far extending the current risk scores for ASCVD events. The multiple risk scores may better help pick out the high-risk patients for a specific type of complications in addition to ASCVD and assist both physician and patients better target the corresponding risk reduction. Although predictors were selected from a complete list of collected risk factors, the final included ones are commonly seen in clinic and therefore may have wide and easy use without requesting additional expensive or complicated tests. Meanwhile since the RECODe ASCVD risk score has CVD history as one significant predictor, the RECODe may have the potential to be applied in those with past history of CVD, which is not the case for the ACC/AHA Pooled Cohort or UKPDS risk scores. Although risk assessment is not as commonly used in prevalent CVD population as in CVD-free population, knowing the residual risk of subsequent CVD and other diabetes complications may alert health care providers intensify treatment and dynamically evaluate treatment effect from the change of repeated risk calculation.

The RECODe risk scores should be considered as one of the most updated, well-developed and comprehensive risk prediction tools using contemporary cohort data. The next questions is: should we consider including it in the next guidelines of diabetes management? It may be still early to apply it to clinic given the following consideration. First of all, the RECODe risk scores have been externally validated in two clinical trial cohorts but not in real-world observational cohort yet. The generalizability of risk score derived from clinical trial samples to the general diabetic population depends on the similarity between the two.⁶ Given that increasing number of diabetes patients are now receiving multiple CVD prevention therapies, clinical trial samples could actually be good representations of more contemporary cohorts. However, whether or not this assumption is true needs to be examined in real-world cohorts. Secondly, most risk scores in the RECODe showed moderate discrimination (as measured by c-statistics), indicating less than perfect discrimination ability. Deep machine learning technology might be useful to improve internal validity however it can also bring problem of overfitting and limit its generalizability.

How to improve the performance of current RECODe scores requires further exploration. Another potential problem is that significantly positive NRI is driven by non-event NRI, which means the RECODe is superior to UKPDS and PCE mainly because it can correctly reclassify those without event into low categories. The researchers used the original UKPDS and PCE without recalibration so the overly high risk from old scores main explain why only non-event NRI is significant. NRI after recalibration of old scores and weighted NRI should be considered as alternative measures to evaluate the relative discrimination ability of RECODe.

Nevertheless, this work a big step in diabetes complication risk assessment and has the potential to be a useful risk prediction tools in the future. New research on risk scores for complications needs to focus on using advanced methodology and incorporated diverse data sources. Pooling of multiple diabetes cohorts can be considered in risk score development stage. Novel biomarkers and subclinical atherosclerosis measures is the next to be considered including in the risk model.⁷ And validation of risk score with necessary recalibration using real-world data, including medical records, should be done more than once to dynamically evaluate the performance of existing risk score and alarm the need of new risk scoring system.

References

1. Goff DC, Lloyd-Jones DM, Bennett G, et al. 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2014;63:2935-59.
2. Rana JS, Tabada GH, Solomon MD, et al. Accuracy of the atherosclerotic cardiovascular risk equation in a large contemporary, multiethnic population. J Am Coll Cardiol 2016;67:2118-30.
3. Kegne AP, Patel A, Colagiuri S, et al. The Framingham and UK Prospective Diabetes Study (UKPDS) risk equations do not reliably estimate the probability of cardiovascular events in a large ethnically diverse sample of patients with diabetes: the Action in Diabetes and Vascular Disease: Preterax and Diamicron-MR Controlled Evaluation (ADVANCE) Study. Diabetologia 2010;53:821-31.
4. Hayes AJ, Leal J, Gray AM, Holman RR, Clarke PM. UKPDS outcomes model 2: a new version of a model to simulate lifetime health outcomes of patients with type 2 diabetes mellitus using data from the 30 year United Kingdom Prospective Diabetes Study: UKPDS 82. Diabetologia 2013;56:1925-33.
5. Basu S, Sussman JB, Berkowitz SA, Hayward RA, Yudkin JS. Development and valudation of Risk Equations for Complications Of type 2 Diabetes (RECODe) using individual participant data from randomised trials. Lancet Diabetes Endocrinol 2017;5:788-98.
6. Blaha MJ. The critical importance of risk score calibration: time for transformative approach to risk score validation? J Am Coll Cardiol 2016;67:2131-4.
7. Zhao Y. Cardiovascular risk assessment and screening in diabetes. Cardiovasc Endocrinol 2017;6:17-22.

Keywords: Risk Factors, Hydroxymethylglutaryl-CoA Reductase Inhibitors, Diabetes Mellitus, Type 2, Prospective Studies, Glycated Hemoglobin A, Blood Pressure, Follow-Up Studies, Risk Assessment, Diabetes Complications, Prevalence, Atherosclerosis, Stroke, Myocardial Infarction, Anticoagulants, Risk Reduction Behavior, Medical Records, Heart Failure, Health Personnel, Biomarkers, Cholesterol, Hypertension, Metabolic Syndrome

< Back to Listings