A new model for 5-year risk of cardiovascular disease in type 2 diabetes, from the Swedish National Diabetes Register (NDR)

https://doi.org/10.1016/j.diabres.2011.05.037Get rights and content

Abstract

Aim

We assessed the association between risk factors and cardiovascular disease (CVD) in an observational study of type 2 diabetes patients from the Swedish National Diabetes Register.

Methods

A derivation sample of 24,288 patients, aged 30–74 years, 15.3% with previous CVD, baseline 2002, 2488 CVD events when followed for 5 years until 2007. A separate validation data set of 4906 patients, baseline 2003, 522 CVD events when followed for 4 years.

Results

Adjusted hazard ratios at Cox regression for fatal/nonfatal CVD were: onset-age 1.59, diabetes duration 1.55, total-cholesterol-to-HDL–cholesterol ratio 1.20, HbA1c 1.12, systolic BP 1.09, BMI 1.07 (1 SD increase in natural log continuous variables); males 1.41, smoker 1.35, microalbuminuria 1.27, macroalbuminuria 1.53, atrial fibrillation 1.50, previous CVD 1.98 (all p < 0.001 except BMI p = 0.0018). All 12 variables were used to elaborate an equation for 5-year CVD risk in the derivation dataset: mean 5-year risk 11.9 ± 8.4%.

Calibration in the validation dataset was adequate: ratio predicted 4-year risk/observed rate 0.97. Discrimination was sufficient: C statistic 0.72, sensitivity 51% and specificity 78% for top quartile.

Conclusion

This CVD risk model from a large observational study of patients in routine care showed adequate calibration and discrimination, and can be useful for clinical practice.

Introduction

Cardiovascular disease (CVD) risk estimates can be used as prognostic information and support for the choice of therapeutic strategies. Physicians engaged in diabetes care should have an interest in the assessment of risk of developing any major CVD event using a global CVD risk assessment tool. Estimates of 5-year risk are considered more accurate than 10-year risk estimates [1].

Several CVD risk models have been developed for the general population using diabetes as a dichotomous variable [2], [3], [4], but with poor calibration of absolute risk in type 2 diabetes [5], [6]. A few scores have been specifically designed for type 2 diabetes, estimating coronary heart disease (CHD) risk [7], [8] or CVD risk [9], [10], [11]. The UKPDS risk model [8], [9] is most widely used and has been recommended for practice guidelines [12], [13], although nowadays considered to have poor calibration in more recent samples of patients with type 2 diabetes [5], [6], [14], [15], [16], [17], [18]. To enhance prediction of CVD in type 2 diabetes patients, there is a need to update or construct a new and improved diabetes-specific model with better performance and, more importantly, better validity [17], [18].

Against this background, we assessed the association between risk factors and the risk of CVD in a large sample of patients with type 2 diabetes, aiming to develop a risk equation for the absolute and modifiable 5-year risk of CVD based on these risk predictors.

Section snippets

The Swedish National Diabetes Register (NDR)

The Swedish NDR was initiated as a tool for quality improvement in diabetes care with local feedback. Annual reporting to the NDR is carried out by trained physicians and nurses via the Internet or via clinical records databases, with information collected during patient visits at hospital outpatient clinics and primary health care clinics nationwide. All included patients have agreed by informed consent to register before inclusion. The Regional Ethics Review Board at the University of

Results

Baseline characteristics in the derivation dataset of 24,288 patients with type 2 diabetes registered in 2002 and followed for 5 years, and in the validation dataset of 4906 patients registered in 2003 and followed for 4 years, are given as mean values (SD) or frequencies (%) in Table 1.

Main findings, calibration and discrimination

In this report we present a risk model with twelve predictors for estimation of the absolute 5-year risk of fatal/nonfatal CVD and modifiable risk percent in type 2 diabetes patients. It was developed from a large sample of patients in routine diabetes care nationwide covered by the Swedish NDR, and validated in a separate sample of patients with baseline one year after the derivation data set. Calibration assessed as the P/O ratio (predicted risk/observed rate) was 1.0 in the derivation

Conflict of interest

The authors declare that they have no conflict of interest.

Acknowledgements

We thank all regional NDR coordinators, contributing nurses, physicians, and patients. The patient organization Swedish Diabetes Association, and the Swedish Society of Diabetology support the NDR. The Swedish Association of Local Authorities and Regions funds the NDR. Results and views of the present study represent the authors and not necessarily any official views of the Medical Products Agency where one author is employed (BZ).

References (38)

  • R. Coleman et al.

    The Oxford risk engine: a cardiovascular risk calculator for individuals with or without type 2 diabetes

    Diabetes

    (2007)
  • J. Cederholm et al.

    Risk prediction of cardiovascular disease in type 2 diabetes: a risk equation from the Swedish National Diabetes Register

    Diabetes Care

    (2008)
  • C.R. Elley et al.

    Derivation and validation of a new cardiovascular risk score for people with type 2 diabetes. The New Zealand diabetes cohort study

    Diabetes Care

    (2010)
  • National Collaborating Centre for Chronic Conditions.Type 2 diabetes: national clinical guideline for management in...
  • Guidelines & Protocols Advisory Committee. Cardiovascular disease—primary prevention; 2010 [accessed 7 June 2010],...
  • J.W. Stephens et al.

    Cardiovascular risk and diabetes. Are the methods of risk prediction satisfactory?

    Eur J Cardiovasc Prev Rehabil

    (2004)
  • A.A. van der Heijden et al.

    Prediction of coronary heart disease risk in a general, pre-diabetic, and diabetic population during 10 years of follow-up: accuracy of the Framingham, SCORE, and UKPDS risk functions: the Hoorn study

    Diabetes Care

    (2009)
  • A.P. Kengne 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)
  • P. Chamnan et al.

    Cardiovascular risk assessment scores for people with diabetes: a systematic review

    Diabetologia.

    (2009)
  • Cited by (59)

    • Population median imputation was noninferior to complex approaches for imputing missing values in cardiovascular prediction models in clinical practice

      2022, Journal of Clinical Epidemiology
      Citation Excerpt :

      Examples are the reduced model methods [14], hybrid model method [14], conditional single imputation, mean/median imputation [15], and the naïve approach [16]. The objective of the present study is to compare the validity and robustness of individual CVD risk prediction using the before mentioned five methods for handling missing patient characteristics in a real-world clinical setting, using the development of the Swedish National Diabetes Registry (NDR) risk prediction model and its validation in Scottish data for patients with T2DM as an example [17]. Using data from nationwide health registers from the Swedish NDR and the Scottish Care Information-Diabetes (SCI-Diabetes) database linked to hospital admission and death records.

    • Quantifying the Risk Continuum for Cardiovascular Death in Adults with Type 2 Diabetes

      2021, Canadian Journal of Diabetes
      Citation Excerpt :

      Risk-prediction models may be valuable in this context by identifying the highest-risk patients most likely to have events averted with new therapies while minimizing the number needed to treat for benefit which helps control cost (23). Multiple risk-prediction models for CV endpoints applicable to T2D have been developed and validated over the last few decades (24–37). Application of these models to contemporary T2D populations is challenged by a predominance of models developed strictly for primary prevention, development of models in highly selected clinical trial populations and the inclusion of specialized predictors seldom measured in clinical practice.

    • 2021 Consensus Pathway of the Taiwan Society of Cardiology on Novel Therapy for Type 2 Diabetes

      2021, JACC: Asia
      Citation Excerpt :

      We considered several factors in our selection of risk equations: 1) the risk equations should be based on the same derivative cohorts that can predict 10-year risk of both ASCVD and HF; 2) the total risk of ASCVD, including coronary heart disease and stroke, should be included in the calculation; 3) the risk equations have been widely adopted in major treatment guidelines; and 4) the risk equations have web-based tools. Several risk equations have been developed specifically for diabetes population, including the UK Prospective Diabetes Study risk engine (80), the model from the Swedish National Diabetes Register (81), and an Australian cardiovascular risk equation (82). However, these prediction equations can predict the risk of coronary heart disease, but not the risks of stroke and HF.

    • Paleolithic diet fraction in post hoc data analysis of a randomized cross-over study comparing Paleolithic diet with diabetes diet

      2021, Clinical Nutrition Open Science
      Citation Excerpt :

      The consistent associations between higher PDF for energy and weight and beneficial changes in glucometabolic and cardiovascular outcome measures in this study and our previous study comparing a Paleolithic diet with a Mediterranean-like diet indicate that a diet with relatively higher PDF for energy and weight could confer a lowered risk of associated diseases. To exemplify, using data from this study only and entering mean participant baseline values for age, diabetes duration and height together with mean participant outcome values after the diabetes diet for an otherwise assumed healthy non-smoking male without macro- or microalbuminuria into the 5-year cardiovascular disease risk model from the Swedish National Diabetes Registry (NDR), result in a total absolute 5-year risk of cardiovascular disease of 9.4% [6,14,15]. Entering again all values unchanged except for beneficial changes in HbA1c, total cholesterol and weight projected from associations with a 40% increase in PDF for energy, result in a total absolute 5-year risk of cardiovascular disease of 8.5%.

    View all citing articles on Scopus
    View full text