Dueling medical mathematical models- to choose- An act of faith?

The August 16, 2005 issue of the Annals of Internal Medicine published an article by Dr. David M. E. Eddy et al using a mathematical model called the Archimedes model. This computer based model was applied to the question of is it more cost effective to give metformin or to employ the methods of the Diabetes Prevention program (DPP) of lifestyle modification to prevent or delay the onset of diabetes in high-risk individuals.

 The DPP had shown that over the almost three years of the trial there was a relative risk reduction of 58% from life-style and 31% for metformin. An earlier article in the same journal used a different model (Markow model) to attempt to answer the same question. The authors concluded that the lifestyle modification was more cost effective. The Eddy article found the opposite to be the case.

The views of two editorialists were sought and published in the August Annals issue in an apparent attempt to place things in context.

One editorial author was Dr. Michael M. Engelgau who was one of the researchers in earlier study that found lifestyle modifications more cost effective. Not surprisingly he criticizes the Eddy model saying in part " I believe Eddy and colleagues' criticisms [of his model] are based on inaccurate descriptions of the DPPRG model". Eddy et al in their introduction state their method is " more thorough,clinically realistic and independently validated".
So whose model is better? Whose conclusion is correct? Is there a "correct"? Dueling experts with opposing conclusions.

The Archimedes model also found that it would be more cost effective to wait until a person developed diabetes before instituting either metformin or lifestyle modification. It will be interesting to see what the folks at the America Diabetes Association or Joslin Clinic say about that.
In real life, of course, a physician will attempt to convince the "pre-diabetic" to loose weight and exercise and may even in addition suggest metformin. But these models are less used to aid individuals docs and individual patients decide what to do that they are to make policy decisions for large organizations in regard to how to allocate funds.

The line between "pre diabetes" and diabetes is arbitrary and the definitions of these two conditions change with the edicts of the ADA and the physiological processes that Eddy claims to model with differential equations are poorly understood and always in flux as new biological insight are gleaned. The output of these models is quality-adjusted life-years (QALY) in which researchers claim to be able to measure and average the "quality" of a person's life. I have posted before on the topic of QALY and cost effectiveness articles.

These models have assumptions on top of assumptions, offer little transparency, are essentially mysterious black boxes to all but a few of the researchers who are expert in such matters and for physicians , or for that matter, policy makers to accept these at face value ( and here how can you when two studies are so contradictory?) requires faith.