Prevention of T2DM: selection criteria
Diabetes risk scores provide individuals with an estimate of the likelihood of getting diabetes in the course of a certain number of years. Risk scores can be helpful to select persons who should undergo further medical examination or who could benefit from lifestyle modification.
What is the idea behind risk scores?
There is strong evidence from RCTs like the DPS (Diabetes Prevention Study) and the DPP (Diabetes Prevention Program) that people who are at higher risk of type 2 diabetes (T2DM) benefit from lifestyle interventions . In both studies, the risk of developing diabetes was reduced by around 60% in participants with impaired glucose tolerance who managed to reduce their weight by 5 to 7%, engaged in regular moderate physical activity, and reduced their intake of fat. However, many people who might profit from such interventions do not know that they are at high risk of T2DM.
For example, in a Dutch study subjects were asked to express their risk of having T2DM before they underwent screening: among subjects who turned out to have unknown diabetes, more than 50% said they did not know, and about 20% considered that their personal risk was zero . In another Dutch study, it was found that many people had a basic understanding of the main risk factors of diabetes, but did not translate their own risk profile into an appropriate perceived diabetes risk . Therefore, scores providing estimates of the individual risk of diabetes might be helpful to identify high risk subjects who might benefit from interventions.
What do risk scores look like?
In principle, there are two kinds of risk scores:
First, there are “paper and pencil” scores which do not need invasive measurements and which include familiar risk factors such as age, sex, overweight and obesity, family history of diabetes, smoking, or use of antihypertensive medication. People can use such risk scores on their own, and some of these scores are available on the internet.
A well-known example is the FINDRISK score which was developed in Finland . People have to answer eight simple questions to get an estimate of their T2DM risk for the next ten years. Another score which is available online is the German Diabetes Risk Score which gives more weight to patterns of nutrition .
Second, there are clinical scores which additionally include blood parameters such as, for example, levels of glucose, triglycerides or uric acid. These give somewhat more valid risk estimates than scores without invasive parameters . Glucose levels are particularly strong predictors of T2DM, and subjects with increased HbA1c values, or higher levels of (fasting or 2-hour plasma) glucose, have a larger risk of developing T2DM than subjects with normal glucose regulation.
It is important that people are also offered an interpretation of their individual scores along the lines of “currently no increased risk”, “slightly increased risk”, or “need to take action” which could be combined with recommendations for lifestyle changes or medication.
How are scores developed?
To develop risk scores, it is necessary to have data from a so-called cohort study. This is a study in which the participants do not have diabetes at the beginning (” baseline”), and are followed up for many years. At baseline, risk factors of diabetes have to be assessed carefully. The study must be representative of the whole population and should not focus on high risk groups. Moreover, during follow-up a sufficient number of new (“incident”) cases of diabetes must occur. From these data, statistical models can be derived to estimate individual risks of developing diabetes from the individual patterns of risk factors. These models are the basis of diabetes risk scores.
What is a good risk score?
A risk score must have good prognostic properties, and there are various statistical criteria to assess the predictive ability of the scores . These statistical criteria give information how well a score discriminates subjects who will develop diabetes from those who will not develop diabetes. The most often used measure of discrimination is the so called “area under the receiver operating curve” (AROC). An AROC of 0.5 means that a score gives no predictive information at all, and, judging in a rough manner, 0.7 is moderate, 0.8 is fairly good, 0.9 is excellent, and 1.0 is the maximum value. An important caveat is that risk scores can only be used for practical purposes when they have been validated externally. This means that the prognostic ability of a score must not only be assessed in the study population which was used to develop the score but also in another (“external”) cohort study. Often scores perform considerably poorer in such external study groups, and, thus, cannot be recommended for practical use.
However, risk scores cannot be judged by statistical criteria alone. It is also important that the data which are needed to calculate the score are easily accessible and that the costs of measurements are low. Furthermore, it depends on the context which kind of score does best. To select subjects who might be invited for a more thorough biochemical examination, paper and pencil scores like FINDRISK are suitable. In a clinical context where blood parameters like glucose, HbA1c or lipid levels are available, clinical scores like KORA should be applied.
So far, more than hundred diabetes risk scores have been published, and many of them are tailored to specific populations . However, many scores have not yet been validated externally, and often the authors have not specified in which context the score should be used. An up-to-date topic in epidemiological research is to add novel biomarkers (like metabolites, proteins and peptides, RNA transcripts etc.) as well as diabetes genes to conventional risk scores, but so far, addition of these more advanced risk factors has hardly led to any improvement of the predictive power of conventional scores.
There are several open questions beyond optimization of statistical properties of risk scores. Risk scores developed to predict cardiovascular diseases (CVD) have a much longer tradition than diabetes risk scores, and surveys have shown that most doctors do not use them in clinical practice. So far, there has been little research in barriers for the use of diabetes risk scores by doctors, but there are some findings on CVD risk scores which might be transferred to diabetes risk scores. In a survey among Swiss general practitioners, 58% distrusted the validity of CVD scores stating that the complex situation of a patient cannot be reflected adequately by a single risk value . One in three doctors stated that he had no overview of existing scores, and 54% were afraid that CVD risk scores might lead to overtreatment. Thus, there is a strong need for research on how to make risk scores more appealing to doctors. This might include the implementation of computerized support systems, visual presentation of the results, or linking lifestyle recommendations to the risk estimates.
So far, there is a lack of studies on two important questions. First, it has not yet been assessed whether diabetes risk scores actually predict diabetes significantly better than subjective risk assessments by doctors. This would be a strong argument to convince doctors of the advantage of risk scores. Second, it has not yet been investigated whether use of diabetes risk scores in clinical practice actually improves health outcomes. However, such research requires large cohort studies with a specific design and a long follow-up.
^ Knowler WC, Barrett-Connor E, Fowler SE, et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002; 346: 393-403
^ Tuomilehto J, Lindström J, Eriksson JG, et al. Finnish Diabetes Prevention Study Group. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med. 2001; 344: 1343-1350
^ Adriaanse MC, Snoek FJ, Dekker JM, et al. Perceived risk for type 2 diabetes in participants in a stepwise population-screening programme. Diabet Med 2003; 20: 210-215
^ Claassen L, Henneman L, Nijpels G, Dekker J, Marteau T, Timmermans D. Causal beliefs and perceptions of risk for diabetes and cardiovascular disease, The Netherlands, 2007. Prev Chronic Dis 2011; 8: A130
^ Schwarz PE, Li J, Lindstrom J, Tuomilehto J. Tools for predicting the risk of type 2 diabetes in daily practice. Horm Metab Res. 2009; 41: 86-97
^ Schulze MB, Hoffmann K, Boeing H, et al. An accurate risk score based on anthropometric, dietary, and lifestyle factors to predict the development of type 2 diabetes. Diabetes Care 2007; 30: 510–515
^ Rathmann W, Kowall B, Heier M, et al. Prediction models for incident type 2 diabetes mellitus in the older population: KORA S4/F4 cohort study. Diabet Med 2010; 27: 1116-1123
^ Wilson PWF, Meigs JB, Sullivan L, Fox CS, Nathan DM, D´Agostino RB Sr. Prediction of incident diabetes mellitus in middle-aged adults. The Framingham Offspring Study. Arch Intern Med 2007; 167: 1068 – 1074
^ Kahn HS, Cheng YJ, Thompson TJ, Imperatore G, Gregg EW. Two risk-scoring systems for predicting incident diabetes mellitus in U.S. adults age 45 to 64 years. Ann Intern Med 2009; 150: 741 – 751
^ Moons KGM, Kengne AP, Woodward M, et al. Risk prediction models I: Development, internal validation, and assessing the incremental value of a new (bio)marker. Heart 2012a; 98: 683-690.
^ Moons KGM, Kengne AP, Grobbee DE, et al. Risk prediction models: II. External validation, model updating, and impact assessment. Heart 2012b; 98: 691-698
^ Noble D, Mathur R, Dent T, Meads C, Greenhalgh T. Risk models and scores for type 2 diabetes: systematic review. BMJ 2011; 343:d7163. doi:10.1136/bmj.d7163
^ Eichler K, Zoller M, Tschudi P, Steurer J. Barriers to apply cardiovascular prediction rules in primary care: a postal survey. BMC Fam Pract 2007; 8:1