Back grey_arrow_rt.gif
Diabetes experts debate CV risk prediction in type 2 diabetes September 14, 2011 Shelley Wood
Lisbon, Portugal - Researchers meeting here at the European Association for the Study of Diabetes (EASD) 2011 Meeting are debating the best way to predict risk of major complications-particularly cardiovascular complications-of diabetes.
As a number of presentations Tuesday morning pointed out, the gold standard for risk prediction is what session comoderator Dr Michael Stumvoll (University Hospital Leipzig, Germany) called "the mother of all studies" in this field: the United Kingdom Prospective Diabetes Study (UKPDS). Conducted out of Oxford during the mid- to late 1970s, the UKPDS was, at the time, the largest prospective diabetes study ever conducted, and the outcomes simulation model derived from these data have been widely used in research, guideline development, and clinical practice.
But they are also out of date, noted Dr Alison Hayes (Sydney School of Public Health, Australia), who presented an "improved" risk-prediction model using additional patient-level data from the 10-year posttrial monitoring phase of UKPDS to reestimate risk equations for diabetes complications and deaths. More patient-years of data, she said, strengthen the model by adding more events and therefore more power, take into account the impact of longer diabetes duration, and add important new information on risks in older patients.
The new model also adds new end points, most notably second MI, second stroke, second amputation, and ulcer. It also adds new risk factors that have emerged since UKPDS was launched: estimated glomerular filtration rate (eGFR), microalbuminuria, white-blood-cell count, heart rate, and hemoglobin.
Using the new, expanded model, said Hayes, allows users to simulate a much wider range of long-term outcomes and also more reliably estimate risk, particularly in older subjects. For example, risk prediction in men using both the old and new model produces similar results, but in women, the newer model predicted much higher numbers of future events.
"Improved prediction of likely clinical outcomes will permit more accurate future economic evaluation of interventions in type 2 diabetes," Hayes predicted.
What's old is still old
Whether what's old can be new again, however, is up for debate. In a separate presentation, Dr Nanne Kleefstra (Diabetes Centre, Zwolle, the Netherlands) argued that a more complex risk prediction tool based on the UKPDS may not be the way forward, simply because the patients on whom the UKPDS prediction tool was derived are very different from diabetic patients today.
Dr Nanne Kleefstra
Kleefstra et al's analysis, also presented Tuesday morning, described a new risk-prediction tool drawn from a group of 1353 diabetes patients, dubbed the ZODIAC cohort, followed for almost 10 years in the Netherlands.
In this group, said Kleefstra, application of the UKPDS risk engine at baseline predicted a 29% mortality rate; at follow-up, however, only 18% of patients had died (half from cardiovascular disease). A key reason for the overestimate, according to Kleefstra, was the fact that blood pressure and lipids in more modern-day diabetic patients are much better controlled then they were at the time of the UKPDS.
Using their Dutch cohort, Kleefstra and colleagues have created the ZODIAC risk performance tool using only age, HbA1c, smoking, serum creatinine, urinary albumin/creatinine ratio, and macrovascular complications. Applying this tool within the group from which it was derived was-not surprisingly-much more accurate in predicting future events.
Simplicity and complexity
In an interview with heartwire, Kleefstra stressed that the ZODIAC tool needs to be validated externally, in different diabetes populations, and that it would be far too soon to apply their risk algorithm in clinical practice. But he stressed the need for a simpler tool and one that takes into account the strides made in controlling key cardiovascular risk factors found in diabetes patients, as well as their earlier age of diagnosis.
Also speaking with heartwire, Stumvoll pointed out that the ZODIAC tool, along with the updated UKPDS effort, are just two of many ongoing attempts to hone risk prediction, and the duplication of efforts is just splintering the field. "Ideally what you'd use is a 'metamodel,' looking at similar-type studies, which is in itself very difficult, and different tools and really derive a general model that could be used anywhere in the world," he said. "We are not quite there yet, but with more risk engines being developed by different groups, maybe they can get their act together and create this kind of meta-risk engine." Comoderator Dr Helen Colhoun (University of Dundee, Scotland), stressed the need for generalizability to other populations, calling the ZODIAC tool "day one of model development" with "a long way to go before they can say that their model is or isn't superior to UKPDS."
She also pointed out that a simpler tool is an attractive idea, particularly in situations where computer-based risk engines aren't available.
"A lot of the guidelines now do reflect a simpler approach, not just risk counting, but looking at a few risk factors that physicians can parse easily in their minds, and I think that that's sensible."
But in places where computerized risk engines can be embedded, for example, in primary-care software, the complexity of the model is less important, since the computer does the work, and "the utilization of it can be very simple and quite useful. To a certain extent, it's not the complexity of the model that matters, it's the complexity of the implementation of model."
  icon paper stack View Older Articles   Back to Top