The concept that tight control might prevent the microangiopathic, neuropathic, and macrovascular complications in both insulin-dependent and non-insulin-dependent patients certainly isn't new; Elliott Joslin, among others, preached this message more than half a century ago. Our greater focus on present rather than future outcomes therefore can't be attributed to a lack of ideas about what to do. What we've lacked is the confidence to implement a tight control strategy in the face of all odds. Although we've been willing to give tight control the benefit of the doubt, the odds against achieving the last measure of tight control were, and still are, heavy. In patients with NIDDM, the seeming intractability of obesity; the frequent failure of sulfonylureas, if not primary then secondary; the serious toxicity, and ultimate removal from the market, of phenformin; and, ultimately, the stresses of insulin therapy, including the indignities and dangers of insulin reactions, are among the snags and obstacles that have blocked our path at every turn.

And confidence in tight control, in turn, was lacking mainly because, until recently, the evidence that tight control prevented chronic complications simply wasn't convincing. The results of the Diabetes Control and Complications Trial have, fortunately, now given us an important part of that missing evidence (and have helped raise the standards for scientific evidence generally [1] into the bargain). For patients with NIDDM, however, the bad news is that only patients with insulin-dependent diabetes mellitus (IDDM) were included in that trial, and evidence of benefit was established only for microvascular and neuropathic complications; the data on macrovascular complications were, understandably, inconclusive.

Non-insulin-dependent diabetes mellitus, by far the more prevalent disease, differs from IDDM in many important ways—for example, genetically, etiologically, pathophysiologically. Therefore, the challenge now is to decide whether we can safely and appropriately extrapolate the conclusions of the Diabetes Control and Complications Trial to the much more common problems of NIDDM patients. Extrapolation itself is only part of the problem: As Clement McDonald points out in this issue [2], we extrapolate the results of limited research findings all the time in medicine, almost without thinking about it. In his view, the difficulties are, first, that we don't extrapolate consistently, sometimes generalizing freely (for example, in treating hypertension), sometimes holding back rigidly (for example, in using ß-blockers after myocardial infarction); and, second, that we haven't developed a way of studying and improving the process of extrapolation itself.

Perhaps more to the point, we need to recognize that a decision to push for tighter control is like most decisions we make about changing our clinical practices—namely, it depends more on extension or refinement of existing knowledge than on true "paradigm shifts" in our mental models. Thus, by the time we encounter a clinical problem, we've most often already put together, from a variety of sources, a reasonable mental model of its causes and management and have developed a degree of belief in that model [3]. New information therefore only rarely changes the model itself; what it changes, rather, is our sense of the likelihood that the model or an element of the model, such as the effectiveness of tight metabolic control, is true. In a word, what changes is our confidence, and the new information is helpful not because it is qualitatively new, but because it quantitatively raises or lowers our confidence in existing knowledge—that is, it helps confirm or deny it. The Reverend Thomas Bayes, an obscure 19th century English cleric, formalized this way of thinking about the impact of information, and this approach still bears his name.

Thus, if the results from the definitive trial of tight control in NIDDM were to appear tomorrow, the Bayesians among us would argue that the true impact of this information would be an abrupt, major increase in our confidence in the value of tight metabolic control, presumably to the point where we would change our clinical practices and the systems for delivering that care [4]. Lacking such "hard" evidence, the organizers of the Fifth Regenstrief Conference set about more than a year ago to pull together the best existing evidence on diabetic complications and their prevention, however fragmentary and indirect. The proceedings of the conference appear in the Supplement to this issue. The skeptics among us will deny that we can draw meaningful conclusions from such bits and pieces of evidence; they will hold out for more definitive trials on the question. (Fortunately for them, the results of one such trial—the Kumamoto study [5], a small, 6-year randomized controlled trial from Japan—were reported after the Regenstrief Conference, and at least one other is under way in the United Kingdom.) But the organizers of the conference argue that the sum of the indirect evidence existing at the time of the meeting may already be sufficient, in the best Bayesian tradition, to increase significantly our confidence that tight metabolic control is effective in NIDDM, at least in preventing the microvascular and neuropathic complications of diabetes. The risk of hypoglycemia, a significant concern in the Diabetes Control and Complications Trial, appears to be acceptably low in those patients with NIDDM who require insulin to achieve tight control—probably less than 1%.

The conference organizers also agree that, unfortunately, the jury is still out on macrovascular disease. But, as readers of the Supplement will discover, the focus of the conference fortunately extended well beyond the narrow question of tight metabolic control and into "comprehensive management" of risk factors, with a strong preventive thrust. Importantly, much of this broader management is directed at macrovascular disease, including aggressive control of associated hypertension, hyperlipidemia, and clotting abnormalities, which demonstrably reduces cardiovascular mortality in patients with NIDDM. And, fortunately, conference participants were acutely conscious that diabetes is not treated in a vacuum. Achieving tight control is a type of care that involves not only the patients themselves but also their families and an impressive array of providers, in active ways that require uncompromising support from the health care system and that differ significantly from the care of patients with most other diseases. The conference therefore explored the implications of achieving tight control in a complex, rapidly changing health care system under many fiscal and administrative constraints.

We still have a great deal to learn about the chronic complications of IDDM and NIDDM. Hyperglycemia itself probably plays a role in their pathogenesis but may not be the whole story, and future discoveries in the biology of diabetes may well change dramatically the entire preventive and therapeutic picture. For now, however, the most pressing question remains a pragmatic one: the value of tight metabolic control in NIDDM. The Regenstrief conference has done its best to present the state of the art. Readers will have to decide for themselves whether the results increase their confidence in the effectiveness of such control, and, if so, by how much.