Screening for type 2 diabetes may be cost effective if it is initiated between the ages of 30 and  45 and repeated every 3-5 years, according to Archimedes, a sophisticated computer modelling program (and not the ancient Greek scientist).

Richard Kahn and colleagues (including David Eddy, the subject of a recent article in Wired magazine) used Archimedes to evaluate the value of 8 screening strategies for type 2 diabetes. The authors reported that screening strategies would result in reductions in MI  and microvascular events (for each, 3-9 events would be prevented per 1000 people screened) and would increase the number of quality-adjusted life-years over 50 years. There was little or no effect on stroke incidence.

The authors acknowledge that “there is no way to validate our results until a real trial is done on all points of interest” but point out:

Our analysis suggests that a randomized clinical trial to compare different screening strategies for type 2 diabetes would not be feasible. Even a trial with 325 000 people in each group followed for 50 years with perfect follow-up and ideal performance and compliance to screening would probably not show significant differences between strategies. Thus, the only formal way to analyse screening for type 2 diabetes is through a mathematical model.

Richard Kahn answered several questions sent to him by CardioBrief:

Can you elaborate a bit on the underlying assumptions of your model?

The core of the model doesn’t really work on “assumptions”. It takes data from the literature on physiology, pathology, procedures, tests, events, costs–and puts everything into equations. These equations (scores of them) are tested to be sure they accurately reflect the science from which they are derived. Then the entire model is tested (ie. validated) to be certain it captures the real world as it is, ie, by faithfully replicating real clinical trials.

Specifically, how do you calculate the actual benefits of screening?

We created virtual people who get screened at different ages/intervals, and as in a real clinical trial, we follow them in the real world. Some die soon from “other causes”, some don’t die until age 90, some are found to get diabetes and are treated just as they would be in the real world. We count up all the “events” (MI, stokes, death, etc) and everyone’s individual costs and then see which screening strategies works best. It’s just like the game Sim City. Archimedes is a very accurate virtual world of health care.

Is there any evidence that early screening leads to better treatment that then leads to better outcomes?

We only focused on various screening strategies, keeping health care for each of them constant. In other studies we looked at what would happen if we improved health care ie, prevention , over a lot of different interventions. That study was published in Circulation.

As we have just seen demonstrated recently, more treatment does not necessarily lead to better outcomes, and sometimes it may even lead to worse outcomes.

That can happen, but I do think there is great evidence that better treatment for those with diabetes who are not meeting guidelines will in fact result in much better outcomes. For example, see the paper in Circulation. I’m not advocating taking everyone with diabetes much lower than guidelines suggest. Rather, simply taking those with very poor control of lipids, glucose and/or BP to values that are “average” would do quite a lot.

Here is the Lancet press release:

TYPE 2 DIABETES SCREENING IN US POPULATION IS COST EFFECTIVE WHEN STARTED BETWEEN AGE 30—45 YEARS, WITH SCREENING REPEATED EVERY 3—5 YEARS

New research published in Article Online First (www.thelancet.com) and in an upcoming edition of The Lancet shows that, in the US population, screening for type 2 diabetes is cost effective when started between the ages of 30 years and 45 years, with screening repeated every 3–5 years. The study is written by Dr Richard Kahn, American Diabetes Association, Alexandria, VA, USA, and colleagues.

To date, no clinical trials have assessed the effects or cost-effectiveness of sequential screening strategies to detect new cases of type 2 diabetes. In this study, the authors used a representative sample of the US population to create a simulated population of 325 000 people aged 30 years without diabetes. They then used computer modelling (the Archimedes Model*) to compare eight simulated screening strategies* for type 2 diabetes with a no-screening control strategy. Strategies differed in terms of age at initiation and frequency of screening, and also as to whether patients were visiting their doctor specifically for diabetes screening or as part of a visit to monitor high blood pressure. The study is the first to address the effects of sequential, rather than one-off, screening. The effects of each strategy on incidence of type 2 diabetes, myocardial infarction, stroke, and microvascular complications were calculated, in addition to quality of life, costs, and cost per quality-adjusted life-year (QALY).

Compared with no screening, all simulated screening strategies reduced the incidence of heart attacks (3–9 events prevented per 1000 people screened) and diabetes-related microvascular complications (3–9 events prevented per 1000 people), and increased the number of QALYs (93–194 undiscounted QALYs) added over 50 years. Most strategies prevented a significant number of simulated deaths (2–5 events per 1000 people) over 50 years. There was little or no effect of screening on incidence of stroke (0–1 event prevented per 1000 people). Five of the screening strategies had costs per QALY of about US$10 500 or less, whereas costs were much higher for screening started at 45 years of age and repeated every year ($15 509), screening started at 60 years of age and repeated every 3 years ($25 738), or a maximum screening strategy (screening started at 30 years of age and repeated every 6 months; $40 778). Several strategies differed substantially in the number of QALYs gained.

Each screening strategy also resulted in an earlier diagnosis. The mean lead time in diagnosis, compared with control, varied from 1.8 years (starting at age 60, every 3 years) to 7.8 years (maximum screening). Despite the ability to detect diabetes sooner when screening is conducted more frequently, the authors found that the greater lead time for the most feasible screening strategies did not significantly improve long-term outcomes.

The five most cost-effective simulated screening strategies varied in their expected degree of benefit. Initiation of screening at 30 years or 45 years of age provided the most benefit. The appropriate choice of strategy would deliver the greatest benefit, while having a low cost per QALY.

The authors conclude: “We therefore recommend starting screening between the ages of 30 years and 45 years, with screening repeated every 3–5 years. The cost per QALY would be further improved if screening were combined with screening events for other disorders, such as screening for hypertension. Similarly, initiation of screening at 45 years with follow-up every 5 years would have the best cost per QALY if screening were done at the time of a visit for lipid testing.”

In an accompanying Comment, Dr Guy Rutten University Medical Center Utrecht, Julius Center for Health Sciences and Primary Care, Utrecht, Netherlands, points out that the population used in Kahn and colleagues’ study was representative of the US population, and differences in race or ethnic origin or differences in behaviours between northern American and European or Asian people might make the results less generalisable to countries other than the USA—as might the variation in health-care costs between the USA’s insurance based model compared to a system where the one system is responsible for almost all costs, such as the UK’s National Health Service.

He concludes: “Today’s paper provides further evidence that screening for diabetes should be combined with screening for hypertension and lipid tests. This recommendation is also in line with the current guideline for screening from the American Diabetes Association. Further input into the model of information on screen-detected people with type 2 diabetes, and separate analyses of different populations or health-care systems, might strengthen the role of the Archimedes model to provide further useful information for future guidelines about screening for diabetes.”