The continued high incidence of complications in patients with NIDDM [3, 7] indicates that the current therapeutic approach is inadequate for maintaining good health. At present, patients are often treated to prevent marked hyperglycemia when it induces symptoms such as thirst, whereas moderately raised blood glucose levels are often accepted. Mechanisms by which hyperglycemia might induce tissue damage have now been described, including glycosylation of tissues with formation of advanced glycation end products [8] and increased flux through the sorbitol pathway [9]. Reducing glucose concentrations in patients with NIDDM might be sensible [10]. However, it is uncertain whether hyperglycemia should be treated more intensively—with diet, oral agents, or insulin—to maintain near-normal glucose levels and prevent the onset of complications or whether these therapies have a neutral or even harmful effect [7].

Association of Blood Glucose Control with Diabetic Complications

The epidemiologic association of glycemic levels with the risk for microvascular and macrovascular disease suggests that these two types of complications are affected by differing degrees of hyperglycemia. The development of diabetic retinopathy [11, 12] and nephropathy [12] mainly occurs when the fasting plasma glucose level is 7.8 mmol/L or higher. The increased risk for these specific diabetic complications when the glucose concentration rises above 7.8 mmol/L has led to it becoming the accepted defining criterion for diabetes [13]. On the other hand, a higher incidence of cardiovascular disease [14, 15] is associated with blood glucose levels at the upper end of the normal range (approximating a fasting plasma glucose level more than 6 mmol/L). The relevance of these data, however, remains controversial because of discordant results [16] and because other associated risk factors may be involved. For instance, patients with impaired glucose tolerance may have been at risk for heart disease because of associated dyslipidemia, hypertension, and raised insulin levels [17-19]. The prospective data in patients with NIDDM suggest that hyperglycemia is a risk factor for cardiovascular disease [4, 20], but insufficient data exist to determine whether there is a glycemic threshold for macrovascular disease. These data suggest that it may be necessary to obtain stricter control of diabetes to prevent cardiovascular disease than is necessary to prevent microvascular disease, but further studies are required.

Previous Prospective Clinical Trials

The Diabetes Control and Complications Trial (DCCT) studied 1441 patients with insulin-dependent diabetes mellitus (IDDM) (mean age, 27 years) and showed that intensive therapy, which resulted in a mean hemoglobin A_1c level of 7.1%, compared with conventional therapy, which resulted in a level of 9.0%, retards the progress of diabetic microvascular disease [21]. Whether such improved glucose control will be similarly beneficial in patients with NIDDM is unknown. A similar reduction in microvascular disease might be expected, but most of the complications in patients with NIDDM are from premature macrovascular, not microvascular, disease.

The University Group Diabetes Program (UGDP) studied 1027 patients with NIDDM (mean age, 53 years) and showed no benefit from improved glucose control induced by insulin, biguanide, or sulfonylurea therapy in preventing diabetic complications over 11 years of follow-up [7].

Potential "Sting in the Tail" of Pharmacologic Therapies

The UGDP reported that a sulfonylurea (tolbutamide) [22] and a biguanide (phenformin) [23] appeared to increase cardiovascular mortality as compared with placebo (see "Exogenous Insulin Administration and Cardiovascular Risk in NIDDM and IDDM"). The results were statistically significant, but because this finding was unexpected and without a previous hypothesis, the study in effect introduced a new hypothesis rather than giving a confirmatory result. A mechanism by which sulfonylureas might increase the risk for cardiovascular disease is the presence in the myocardium of ATP (adenosine triphosphate)-sensitive potassium channels [24]. In theory, closure of these channels by sulfonylurea might prevent ischemia-induced vasodilatation and thus exacerbate ischemic episodes. In addition, ischemic pain might be lessened so that patients do not rest when ischemia becomes critical [25]. In a study done in rabbits, an interaction of sulfonylureas with ouabain induced cardiac arrhythmia with a first- but not second-generation sulfonylurea [26]. The UGDP results can possibly be explained by the use of the first-generation tolbutamide, and the results might not be applicable to second-generation sulfonylurea.

Insulin therapy has been suggested to be potentially harmful, because when given subcutaneously it induces higher peripheral insulin levels than the usual endogenous secretion into the portal vein (see "Do NIDDM and Cardiovascular Disease Share Common Antecendents?"). This event occurs because the therapy needs to provide sufficiently high insulin levels in the portal vein to reduce excess hepatic glucose output. The high peripheral levels may lead to atherogenesis, as suggested by in vitro studies of the effect of raised insulin levels on the morphologic and biochemical responses of the arterial wall [27]. The epidemiologic associations in the general population of high fasting insulin levels with an increased risk for myocardial infarction [28, 29] are in accord with this theory, although the dyslipidemia and hypertension associated with high insulin levels, rather than high insulin levels per se, may account for the association [30].

The Uncertainty about Therapies

The previous studies lend no support for the hypothesis that improved glucose control with currently available pharmaceutical agents helps to prevent the complications of NIDDM. The effect of available therapies on the development of cardiovascular disease is unknown, and the side effects of therapies may outweigh any potential gain from their blood glucose-lowering effect.

United Kingdom Prospective Diabetes Study

The United Kingdom Prospective Diabetes Study (UKPDS) recruited 5102 patients with newly diagnosed NIDDM in 23 centers between 1977 and 1991 [31]. Of these patients, 4209 have been included in randomized, controlled trials of different therapies to determine whether patients with NIDDM can obtain clinical benefit from intensive glycemic control and whether sulfonylurea, metformin, and insulin therapies have specific advantages or disadvantages. The study will end in 1997 when the median duration from randomization will be 11 years (range, 6 to 20 y), and the results will be published in 1998. In 1991, the study was described [31]; the glucose control achieved by the allocated therapies over 3 and 6 years has also recently been described [32, 33]. This article outlines the study's progress, the efficacy of the different therapies over 9 years in improving glucose control, the side effects of the therapies, the natural history of the disease in terms of increasing glycemia, and the overall development of complications. It also compares the UKPDS with two previously reported studies—the DCCT [21] and UGDP [7]—both of which reported data for 9 years of follow-up.

Methods

Top Methods Results Discussion Author & Article Info References

Subjects

The UKPDS recruited 5102 subjects who were 25 to 65 years old and who had newly diagnosed diabetes. The median age was 53 years, and the median fasting plasma glucose level was 11.9 mmol/L. Fifty percent of the patients were obese (> 120% of ideal body weight [34]), and 39% were hypertensive [31]. Ten percent were Asian-Indian, and 8% were Afro-Caribbean [35]. Approximately 50% of the patients already had an indication of diabetes-related tissue damage: 8% had an indication of cardiovascular disease [31], 37% had microaneurysm or more severe retinopathy in one eye, 18% had retinopathy in both eyes, and 18% had microalbuminuria (> 50 mg/L) [35]. The patients from different ethnic groups had a similar prevalence of preexisting heart disease, retinopathy, and microalbuminuria [35].

Initial Diet Therapy and Randomization

During a 3-month run-in period, patients were given intensive dietary advice at monthly clinic visits and lost an average of 5 kg of body weight. Eighty-six percent became symptom free, but only 17% achieved a fasting plasma glucose level of less than 6.0 mmol/L [36]. In two thirds of these patients, the fasting plasma glucose level increased to more than 6.0 mmol/L within 3 years. Between 1978 and 1994, 4209 asymptomatic patients whose fasting plasma glucose level remained at or rose to between 6 and 15 mmol/L were randomly allocated to different therapies. Of the 4209 patients, 3513 were at the end of the initial 3 months of diet therapy; 696 had a fasting plasma glucose level of less than 6 mmol/L after the initial diet therapy but subsequently became hyperglycemic with a fasting plasma glucose level of more than 6 mmol/L. Once patients were entered in the formal therapeutic study, nonobese patients were randomly allocated to four groups—primary therapy with diet, chlorpropamide, glibenclamide, or insulin therapy. Obese patients were randomly allocated to five groups—the same four groups as nonobese patients and primary therapy with metformin.

Four questions about prevention of complications are being asked:

1. Will improved blood glucose control with either sulfonylurea or insulin therapy, both of which increase the insulin supply, be beneficial or harmful?

This question embraces increasing endogenous insulin secretion by sulfonylurea therapy and exogenous therapy with insulin injections. Random allocation of 3867 nonobese and obese patients Figure 1 whose fasting plasma glucose level remained at 6 to 15 mmol/L after diet therapy Table 1 occurred as follows:

View larger version (33K): [in this window] [in a new window]   Figure 1. A flow diagram of randomization. The main analysis compares patients assigned to conventional therapy with diet with patients assigned to intensive therapy with sulfonylurea or insulin (dotted box). Obese patients assigned to conventional therapy are compared with those assigned to metformin therapy (marked by an asterisk). Additionally, patients assigned to sulfonylurea are compared with those assigned to insulin and patients assigned to the first-generation sulfonylurea chlorpropamide are compared with those assigned to second-generation glyburide or glipizide. fpg equals fasting plasma glucose level.

a. The number of patients assigned to conventional therapy was 1138; they were initially treated with diet therapy alone, aiming for less than 6 mmol/L, avoiding hyperglycemic symptoms and achieving the best possible fasting plasma glucose level. If after continued dietary advice the patients became symptomatic or the fasting plasma glucose level rose to more than 15 mmol/L, they were randomly allocated to additional sulfonylurea or insulin therapy (with the option of metformin for obese patients), now aiming to maintain a fasting plasma glucose level of less than 15 mmol/L.

b. The number of patients allocated to intensive therapy was 2729, with 1156 allocated to sulfonylurea and 1573 to insulin therapy, aiming, when feasible, to maintain a fasting plasma glucose level of less than 6 mmol/L. In those allocated to sulfonylurea or insulin therapy, the dose was increased until the fasting plasma glucose level, monitored every 3 months in the clinic, was less than 6 mmol/L, unless hypoglycemia occurred [20] or maximal doses were achieved (chlorpropamide, 500 mg; glibenclamide, 20 mg; or glipizide, 40 mg, with no limit on the amount of insulin). In those assigned to insulin therapy, insulin was primarily given as a basal supplement with a once-daily, evening injection of ultralente insulin [37]. When the insulin dose became more than 16 U per day, home blood glucose monitoring was instituted with the addition of either subcutaneous regular insulin therapy before meals or a switch to a regular and isophane insulin regimen; each regimen aimed for preprandial plasma glucose concentrations of 4 to 7 mmol/L [38]. When the fasting plasma glucose level reached 15 mmol/L or hyperglycemic symptoms occurred, metformin was added for patients assigned to sulfonylurea therapy. When the fasting plasma glucose level again rose to more than 15 mmol/L or the patients had recurrent symptoms, insulin therapy was substituted.

The protocol continued with sulfonylurea or biguanide (see below) monotherapy for as long as possible because a major study aim was to determine whether either therapy increased the incidence of cardiovascular disease. In 1987, the study group determined that hyperglycemia was increasing with all therapies, and two protocol modifications were made for those patients assigned to sulfonylurea whose fasting plasma glucose concentrations remained more than 6 mmol/L on maximal sulfonylurea doses. This situation was termed "sulfonylurea inadequacy," and in one third of patients assigned to sulfonylurea therapy, early addition of metformin or insulin therapy was instituted to try to maintain near-normal glycemia and to assess the efficacy of the additional therapies. These additions may improve the glycemic separation between the conventional and intensive therapy groups and thus help to improve the power of the study in determining whether improved diabetes control prevents diabetes complications.

2. Will improving glucose control with metformin be beneficial or harmful?

Obese patients (defined as more than 120% of ideal body weight), in addition to assignment to conventional therapy and intensive therapy with sulfonylurea or insulin, were also assigned to intensive therapy with metformin Table 1. Comparisons are being done between a group of 411 obese patients assigned to conventional therapy and a group of 342 assigned to metformin Figure 1. Metformin is thought to enhance insulin sensitivity, although a more specific action in reducing hepatic glucose output is also a possibility.

3. Will either insulin or sulfonylurea therapy be beneficial or harmful?

This question is being studied by the initial 15 centers in the patients assigned to intensive therapy, comparing 1233 of those assigned to sulfonylurea with 911 assigned to insulin Figure 1. This analysis is between groups who attained similar blood glucose control and is statistically independent from the main analysis. There were similar patients in the other eight study centers, but they were excluded from this analysis because insulin therapy was added early in those who developed sulfonylurea inadequacy, making comparison between monotherapies infeasible.

4. Will first- or second-generation sulfonylurea be particularly beneficial or harmful?

Those patients assigned to intensive therapy with sulfonylurea were evenly apportioned to either a first-generation sulfonylurea therapy—chlorpropamide—or a second-generation sulfonylurea therapy—glyburide—to determine whether there were any differences.

Results

Top Methods Results Discussion Author & Article Info References

Efficacy of Sulfonylurea and Insulin Therapies

The 6-year follow-up data in a cohort of 2287 patients have been published [33]. Assignment to intensive therapy reduced the fasting plasma glucose concentration from a median of 8.1 mmol/L at randomization to the median concentrations of 8.2 and 6.8 mmol/L at 1 year in the conventional and intensive groups, respectively [33]. Sulfonylurea and insulin therapy were equally effective. The hemoglobin A_1c values, measured with the same BioRad HPLC analyzer (BioRad Laboratories, Hemel Hempstead, United Kingdom) used in the DCCT, were 6.8% for sulfonylurea and 6.1% for insulin therapy at 1 year. The age-matched normal range is 4.5% to 6.2%. The fasting plasma glucose values remained higher than 6.0 mmol/L in most patients assigned to intensive therapy, either because maximum sulfonylurea doses were insufficiently effective or because the doses of the therapies were limited by hypoglycemic reactions.

After 6 years, 86% of the patients assigned to sulfonylurea therapy were still receiving it, including 19% who were also receiving metformin and 11% who were also receiving insulin. Of those patients assigned to insulin, 77% were still receiving insulin, but 24% were on a more complex insulin regimen than ultralente insulin alone. The mean dose in those allocated to insulin was 30 U per day at 6 years. Because only patients who could be treated with diet therapy were included in this part of the study, the mean dose was lower than in patients customarily treated with insulin who have often developed high blood glucose concentrations despite maximal sulfonylurea therapy. By 6 years, 45% of the patients assigned to conventional therapy were on diet therapy alone, and the remainder required therapy other than diet, with the aim of continued prevention of symptoms rather than achievement of near-normal glucose levels.

Figure 2 shows the median fasting plasma glucose (panel A) and hemoglobin A_1c [panel B] levels as cross-sectional data for all patients who have reached each year of follow-up up to 9 years. The data for the cohorts of 1108 patients on conventional therapy and 2662 patients on intensive therapy who were followed for 9 years are similar Figure 3 and Figure 4. From the point-of-view of a clinical trial, the study has achieved its primary goal of maintaining improved glucose control in patients assigned to the intensive therapy group over a prolonged period. Over 9 years, the difference in median fasting plasma glucose concentrations between conventional and intensive therapies Figure 2, panel A) was 1.7 mmol/L [P < 0.0001; median, 9.0 and 7.3 mmol/L, respectively]; the difference for hemoglobin A_1c levels Figure 2, panel B) was 0.8% (P < 0.0001; median, 7.5% and 6.7%, respectively) [33].

View larger version (18K): [in this window] [in a new window]   Figure 2. The left-hand panels show the median fasting plasma glucose (A) and hemoglobin A1c (B) in patients assigned to conventional therapy (\#9679;) and those assigned to intensive therapy with sulfonylurea or insulin (fill diamond) in all patients studied each year up to 9 years of follow-up. The histograms show the total number of patients with data at each year in this cross-sectional analysis. The dotted lines for fasting plasma glucose show 7.8 and 6.0 mmol/L, the apparent thresholds from epidemiologic studies for microvascular and macrovascular complications, and for hemoglobin A1c, the upper end of the normal range, 6.2%. The right-hand panels shows the median fasting plasma glucose level (C) and hemoglobin A1c level (D) in obese patients assigned to conventional therapy and those assigned to intensive therapy with metformin (+) and with sulfonylurea or insulin (open diamond). FPG equals fasting plasma glucose level; HbA1c equals hemoglobin A1c level.

View larger version (24K): [in this window] [in a new window]   Figure 3. A 9-year comparison of hemoglobin A1c levels in the conventional and intensive therapy groups in patients with insulin-dependent diabetes in the Diabetes Control and Complications Trial (DCCT) [21] and in the cohorts of patients with non–insulin-dependent diabetes mellitus in the United Kingdom Prospective Diabetes Study (UKPDS). Both studies used the Biorad HPLC analyzer method and were analyzed by assignment to therapy. For the first 3 years, hemoglobin A1c levels in the UKPDS patients had a lower range than in the DCCT patients, but after 7 years, UKPDS levels were in the lower DCCT range.

View larger version (25K): [in this window] [in a new window]   Figure 4. A 9-year comparison of fasting plasma glucose (FPG) levels in the patients of the University Group Diabetes Program (UGDP) [7] who were taking either placebo or variable-insulin doses and in the cohorts of the conventional and insulin therapy groups of the United Kingdom Prospective Diabetes Study (UKPDS) studied for 9 years [33] , both analyzed by assignment to therapy. The UGDP measured blood glucose levels and a correction of x 1.1 was used to give equivalent fasting plasma glucose values. The two studies have shown similar deterioration of blood glucose control over 9 years.

The proportion of patients assigned to conventional and intensive therapy who obtained a median fasting plasma glucose level of less than 7.8 mmol/L at 6 years was 24% and 50%, respectively (P < 0.001); at 9 years, the proportion was 24% and 47%, respectively (P < 0.0001).

Progressive Deterioration of Diabetes Control from Decreasing Beta-Cell Function

For patients in all therapy groups, glycemic control steadily deteriorated over 6 years of follow-up [33]. In patients assigned to conventional therapy with diet, fasting plasma glucose levels steadily increased from a median of 8.2 to 9.5 mmol/L, and hemoglobin A_1c levels increased from 6.8% to 8.0%. A similar but lower level increase occurred in patients assigned to intensive therapy; fasting plasma glucose levels rose from a median of 6.8 to 7.8 mmol/L, and hemoglobin A_1c levels rose from a median of 6.1% to 7.1% [33]. The data at 9-year follow-up show that the increase lessened in the last 3 years, in part because of additional therapies given to prevent marked hyperglycemia.

The deterioration of glucose control over 6 years was shown to be associated with diminishing ß-cell function assessed from fasting glucose and insulin values by homeostasis model assessment [39]. In patients treated with diet therapy alone for 6 years, ß-cell function decreased from 51% at randomization to 28% at 6 years [33]. In patients treated with sulfonylurea alone for 6 years, improved glycemia was associated with an increase in ß-cell function (46% at randomization to 78% at 1 year), but then ß-cell function deteriorated at a rate similar to that in the group treated with diet therapy alone (52% at 6 years). Thus, the failure of sulfonylurea therapy appeared not to be caused by a diminishing effect of sulfonylurea but by an underlying progressive decrease in ß-cell function.

Efficacy of Therapy with Metformin

The 6-year follow-up of obese patients treated with metformin showed that it reduced fasting plasma glucose concentrations over the first year; levels for the conventional and intensive therapy groups were 8.1 and 7.2 mmol/L, respectively [33], and hemoglobin A_1c levels were 7.0% and 6.4%, respectively. Figure 2 shows the fasting plasma glucose (panel C) and hemoglobin A_1c (panel D) levels for 9 years in obese patients assigned to conventional and metformin therapies. Over 9 years, the difference between the median fasting plasma glucose concentrations in the group receiving conventional therapy and the group receiving intensive therapy with metformin was 1.3 mmol/L (P < 0.0001; median, 9.3 and 8.0 mmol/L, respectively); the difference between hemoglobin A_1c levels for these groups was 0.7% (P < 0.0001; median, 7.7% and 7.0%, respectively) [33].

After 6 years of follow-up, 87% and 84% of patients assigned to sulfonylurea and metformin therapies, respectively, were still taking their assigned therapeutic agent; 27% and 14% were taking both sulfonylurea and metformin; and 10% on sulfonylurea and 6% on metformin transferred to insulin therapy. By 6-year follow-up, 60% of patients assigned to conventional therapy required therapy other than diet.

The proportion of patients who obtained a median fasting plasma glucose level of less than 7.8 mmol/L at 6 years in the conventional as compared with intensive therapy with metformin groups was 20% and 40%, respectively (P < 0.0001); at 9 years, the proportion was 24% and 26%, respectively.

Body Weight

Over 9 years, body weight increased in patients assigned to sulfonylurea or insulin therapy (mean increase, 5 kg and 7 kg, respectively, compared with 3 kg in patients assigned to conventional therapy). On the other hand, obese patients assigned to metformin and to conventional therapy only had an increase of 1 kg.

Fasting Plasma Insulin Levels

Fasting plasma insulin levels at 9 years were higher in patients assigned to sulfonylurea and insulin therapy (geometric mean, 12.1 mU/L and 15.6 mU/L, respectively) than in those assigned to conventional therapy (mean, 11.4 mU/L). In obese subjects assigned to metformin, fasting plasma insulin levels were lower (12.5 mU/L) than in those assigned to conventional therapy (14.3 mU/L).

Hypoglycemic Reactions

At each 3-month visit, patients were asked if they had had any hypoglycemic episodes or "funny turns," and the physician, on the basis of the patient's description of events, decided whether he or she thought the patient had had any hypoglycemic events. Over 6 years, the percentage of patients reporting one or more hypoglycemic events per year was 17% for those assigned to and receiving sulfonylurea and 37% for those assigned to and receiving insulin compared with 0.9% for those assigned to and continuing on diet therapy [33]. Patients treated with diet alone had predominantly reactive hypoglycemic events. The cumulative percentages over 6 years were 45%, 76%, and 3%, respectively, for patients receiving sulfonylurea, insulin, and diet therapies. Major hypoglycemic events requiring third-party assistance or hospitalization occurred per year in 0.7% of patients assigned to sulfonylurea, 2.3% of those assigned to insulin, and 0.03% of those assigned to and remaining on diet therapy.

The percentage of obese patients assigned to and receiving metformin and assigned to and remaining on diet therapy who reported one or more hypoglycemic events per year were 5% and 1%, respectively. Major hypoglycemic events occurred in 0.3% and 0.1%, respectively [33]. The cumulative percentage of any episode over 6 years was 17.6% and 2.8% on metformin and diet therapy, respectively, and, for major episodes, 2.4% and 0.4%, respectively.

Clinical End Points

Twenty specific macrovascular and microvascular clinical end points are being assessed prospectively Table 2 [31] and are aggregated into three categories: 1) diabetes-related mortality—death from heart attacks, sudden death, stroke, complications from peripheral vascular disease or amputations, renal failure, hyperglycemic or hypoglycemic coma; 2) total mortality; and 3) diabetes-related mortality and major clinical end points. The Kaplan-Meier analysis shows the high proportion of patients who develop clinical end points Figure 5 and provides estimates of the proportion of the 5102 patients recruited who will have a clinical end point by 9-year follow-up (Table 2). By 9 years, 8% of patients had died from a diabetes-related end point, and 32% had had a diabetes-related fatal or nonfatal event. Macrovascular end points occurred in 20%, and microvascular end points, in 9%, with a fatal outcome 70 times less frequent than for macrovascular disease because of fewer deaths from renal failure.

View larger version (30K): [in this window] [in a new window]   Figure 5. Proportion of patients developing diabetes-related clinical end points over 9 years, as shown by a Kaplan-Meier plot.*.

Discussion

Top Methods Results Discussion Author & Article Info References

Redefining the Questions Posed in Practical Terms

From the perspective of a clinical trial, the UKPDS has been successful: Over 9 years, it has maintained a sustained difference of 1.7 mmol/L in fasting plasma glucose levels in intensive and conventional therapies (median, 7.3 and 9.0 mmol/L, respectively), and a difference of 0.8% in hemoglobin A_1c levels (median, 6.7% and 7.5%, respectively). The study will thus be able to determine whether this degree of improved glycemia by sulfonylurea or insulin therapy, which can be obtained in routine clinical practice, will help to maintain patient health. Whereas the DCCT studied patients with IDDM who had mean hemoglobin A_1c levels of 9.0% and 7.1%, respectively, in patients assigned to conventional and intensive insulin therapies [21], the UKPDS, over the initial 3 years of therapy, studied patients with less severe diabetes control assessed by hemoglobin A_1c levels (Figure 3). Thereafter, with the increasing glycemia that occurred in the UKPDS patients, hemoglobin A_1c levels in the intensive therapy group became similar to those in the DCCT's intensive therapy group, and the UKPDS conventional group developed levels midway between the intensive and conventional therapy groups of the DCCT. On the other hand, the UKPDS has shown, similar to the DCCT experience with insulin therapy, that intensive therapy with insulin or sulfonylurea increased the risk for hypoglycemic episodes and increased body weight [32, 33]. Although, in theory, insulin therapy should make it possible to maintain near-normal glycemia, in practice, this was often not feasible when the severity of diabetes required insulin before meals as well as a basal insulin supplement. The study included a representative selection of middle-aged patients, some of whom were unable or unwilling to undertake and monitor closely a multiple insulin injection regimen. The risk for hypoglycemia became a limiting factor, with 70% of those assigned to insulin therapy having had one or more hypoglycemic events in 6 years and 11% having had a major hypoglycemic event requiring medical assistance.

The study is also comparing whether therapy with sulfonylurea or insulin is specifically advantageous and whether, with sulfonylurea therapy, a first- (chlorpropamide) or second-generation (glyburide) agent may be advisable. Both sulfonylurea and insulin treatments increased fasting insulin levels, which might possibly contribute to atherogenesis in the long term [27]. The different therapies have each similarly improved the glycemia in UKPDS patients, so it will be possible to determine whether any specific therapy is advantageous or disadvantageous in patients who were otherwise similar for both baseline data and glucose control.

A major advantage of the UKPDS over the UGDP is that the UKPDS has randomly assigned 4209 patients (as compared with 1027), and the follow-up will be longer, with a median of 11 years (range, 6 to 20 years) when the study ends in 1997. Clinical trial methods have become more rigorous, with stricter monitoring of studies with stringent validation of data, including clinical end points and clear definition of principal outcome measurements and of the required levels of significance. In addition to the primary clinical end points, many risk factors for complications are being assessed, including smoking, exercise, obesity and central obesity, plasma triglyceride and cholesterol subfractions, plasma insulin concentrations, and presence of microalbuminuria. It will therefore be possible to determine the epidemiologic associations of these risk factors with different diabetic complications.

..Tx.-

Deterioration of Glycemic Control on All Therapies

The deterioration of diabetes control over 9 years of the study is a marked feature, although close attention has been paid to obtaining near-normal glucose levels by increasing doses of sulfonylurea, insulin, or metformin and by combination therapy [33]. The increase in fasting glucose levels is similar to that found in the UGDP groups assigned to placebo or variable insulin therapy (7; (Figure 4). Clearly, the pharmacologic therapies improved glucose control but had no effect on the underlying factors contributing to deterioration of control. Assessment of ß-cell function has indicated that deterioration in glucose control was probably caused by progressive deterioration of ß-cell function [33].

The increasing glycemia of patients on all assigned therapies implied that no single therapy was able to maintain near-normal glycemia as intended. Whereas sulfonylurea, metformin, or insulin therapies improved fasting plasma glucose levels, by 4 to 5 years after initiation, hemoglobin A_1c levels have generally returned to the higher values that existed before adding therapy [33]. Patients continued with each assigned therapy until fasting plasma glucose levels increased to more than 15 mmol/L or symptoms ensued. However, in 1987, when the increasing hyperglycemia in patients receiving sulfonylurea became apparent, two additional therapies were included in a subset of one third of the patients: Metformin or insulin therapy was added early when fasting plasma glucose levels became more than 6 mmol/L in patients taking maximal sulfonylurea doses. Thus, the study will assess the degree to which early institution of sulfonylurea-metformin and sulfonylurea-insulin combinations in patients with "sulfonylurea inadequacy" will improve glucose control and help to prevent progressive hyperglycemia.

Increasing postprandial glucose excursions contribute to the increase in hemoglobin A_1c levels. To evaluate whether they can be reduced by an -glucosidase inhibitor, acarbose therapy [40] was added in 1994 in a double-blind, placebo-controlled study in 1946 eligible patients to determine how much hemoglobin A_1c levels are improved by this therapy.

The natural history of deteriorating glycemic control limits patients' and physicians' ability to maintain near-normal glucose levels. Therefore, it is uncertain whether sufficiently improved control can be obtained long enough to prevent complications. The UKPDS will determine whether the long-term benefit in preventing diabetic complications that can be obtained from improved glycemic control is sufficiently great to counteract the inconvenience of hypoglycemic attacks and their associated risk for injury, the inconvenience of the close monitoring required to maintain improved blood glucose control, and any potential harmful effects from the therapies. The quality of life of a patient subgroup is being studied; questionnaires are given to each patient and a close relative or friend to assess the effect of different therapies.

Therapy with Metformin

The study is also assessing whether intensive therapy with metformin, which enhances insulin sensitivity, will be advantageous or disadvantageous compared with conventional therapy with diet alone. Metformin induced improved glucose control similar to that seen with sulfonylurea and insulin therapy [33] but did not increase body weight or the incidence of hypoglycemic episodes and decreased rather than increased fasting plasma insulin levels [32]. These aspects may all be advantageous. On the other hand, the UGDP found an increased incidence of myocardial infarctions in those assigned to phenformin [23]. Prejudgment of the UPKDS outcome is not possible while the study continues. Increasing glycemia, assessed by fasting plasma glucose or hemoglobin A_1c levels, was similar in patients treated with metformin and those treated with sulfonylurea or insulin [33].

Clinical End Points in the Study

The study objective is to determine whether the therapy aiming for near-normal glycemia will prevent the onset of clinical complications. The study has shown that by 9 years from diagnosis of NIDDM, 29% of patients have had a diabetes-related end point, and 9% have had a fatal end point. These proportions are high when one considers that the median age of diagnosis at study entry was 53 years and that at that stage most patients were thought to be clinically fit. Macrovascular end points were more common than microvascular end points, occurring by 9 years in 20% and 9% of the patients, respectively. A fatal outcome occurred 70 times more frequently with macrovascular than with microvascular disease. These data underline that the major study outcome will depend on whether the assigned therapies affect macrovascular disease.

At diagnosis, 37% of the patients had microaneurysms or more severe retinopathy in one eye; 18% had retinopathy in both eyes. Thus, although the patients were newly diagnosed, they must have had undiagnosed diabetes for a considerable period. Although the primary analyses to be done will include data on all patients assigned to each therapy, an additional, secondary analysis will be done in the subgroups that did or did not have evidence of diabetic tissue damage at study entry. Thus, as for the DCCT, the study is both a primary and secondary prevention study. The DCCT was designed to detect differences in the progress of microvascular disease using subclinical measurements, whereas clinical end points constitute the UKPDS outcome. Nevertheless, the UKPDS is also assessing retinopathy by triennial color retinal photographs with grading of retinopathy and by visual acuity; nephropathy by annual assessment of urine albumin excretion and plasma creatinine values; and neuropathy by triennial assessment of ankle and knee tendon reflexes. The progress of macrovascular disease is also being assessed with triennial 12-lead electrocardiograms with Minnesota coding and by a clinical assessment of whether both dorsalis pedis or both posterior tibial pulses were impalpable [31, 33]. All these indices showed progression over 6 years [33].

The Power of the Study

Long-term studies of chronic disease—for example, studies on the effect of improving blood pressure control or of using ß-blockers after heart attacks—have often shown a 15% to 20% protection. In NIDDM, it has been generally accepted that a 15% protection would be clinically important because the UKPDS data indicate that for every 1000 patients treated for 12 years, 60 would be well who would otherwise have had a major event and 20 would be alive who would otherwise be dead. It may be overly optimistic to expect a 15% benefit within a few years of improving blood glucose control because by the time diabetes is diagnosed, it may have been present for up to 10 years and 50% of patients already have evidence of diabetic tissue damage [31]. Complications can arise 20 to 25 years after the onset of diabetes, and it would not be surprising if it takes time in such a long-term disease process to show benefit from improved control. Long-term follow-up may be needed to overcome the effects of preceding tissue damage as well as to prevent new pathologic occurrences. On the other hand, longer exposure to therapies, including induced raised fasting plasma insulin levels, may be found to be deleterious with long-term follow-up.

Both microvascular and macrovascular end points are aggregated together for the main analyses because the combination of these represents the clinical burden of the disease to patients. In subsidiary analyses, the question of the different effects of the therapies on the progress of microvascular and macrovascular disease will be studied. By 1997, 1750 of 3867 patients studied are expected to have had a fatal or nonfatal diabetes-related macrovascular or microvascular end point. By the study's end, the power for detecting a 15% advantage or disadvantage at equals 0.01 in the primary conventional therapy as compared with intensive therapy with sulfonylurea or insulin analysis is 81%. The patients will have had a median of 11 years since randomization (range, 6 to 20 years). If intensive therapy for 11 years were found to be either neutral or disadvantageous, some might conclude that the potential degree of benefit would not be worth the effort of intensive therapy, including the increased risk for hypoglycemia. Quality of life has been assessed by a questionnaire given to each patient and a relative, partner, or friend. Health economics data about the costs of both therapy and treatment of complications are being assessed; both the cost–benefit and cost–utility aspects of the various therapeutic options will be available.

Hypertension in Diabetes Study

Hypertension was present in 39% of the patients at diagnosis of diabetes [35], with an associated increased age-adjusted risk of 1.8 for developing a diabetes-related death and 1.6 for a major clinical complication [41]. Eligible UKPDS patients with borderline to mild hypertension were therefore randomly assigned, in a factorial design, to "tight blood pressure control" (aiming for blood pressure less than 150 mm Hg systolic/less than 85 mm Hg diastolic, using an angiotensin-converting enzyme inhibitor [captopril] or a ß-blocker [atenolol]) or to "less tight control" (aiming for less than 180 mm Hg systolic/less than 105 mm Hg diastolic) [42]. This Hypertension in Diabetes Study (HDS) began in 1987 and aims to determine whether improved blood pressure control will prevent diabetic complications and to assess whether either angiotensin-converting enzyme inhibitor or ß-blocker antihypertensive therapies have specific advantages or disadvantages [42].

Clinical Relevance of the Study

The UKPDS is the only long-term study of the clinical effectiveness of different therapies for NIDDM that is large enough to determine whether improved control will prevent major complications. If the UKPDS shows that any diabetes therapy is advantageous, this finding will suggest that diabetes should be diagnosed at a preclinical stage, when patients have only modest hyperglycemia and have not yet developed diabetic complications. At presentation of diabetes at a median age of 53 years, only 17% of the patients could attain near-normal fasting plasma glucose levels on diet therapy alone, and in two thirds of these, the glucose levels subsequently increased. This finding indicates that if clinical benefit exists from aiming for near-normal glycemia, many patients will need pharmacologic therapy in addition to reinforced dietary and exercise advice. On the other hand, these agents may be found to be contraindicated if side effects outweigh any clinical advantage. Until firm data are available, some care that is currently given in diabetes clinics and general practice may be misdirected. The UKPDS has potential clinical relevance in three areas:

1. Any demonstrably effective therapy would be worth introducing routinely for both health and economic purposes because many of the complications of NIDDM (for example, amputations, strokes, coronary artery disease requiring coronary artery bypass grafting or angioplasty, blindness, and renal failure) seriously affect quality of life and consume huge health care resources.

2. If therapy to improve glucose control is shown to be ineffective in maintaining health, then the huge current costs of trying to improve control, including regular monitoring and multiple therapies, could be used for other effective therapies (for example, screening for incipient eye or foot disease, for retinal photocoagulation and foot care, respectively) or for more attention to other risk factors such as hypertension, smoking, and dyslipidemia.

3. Individual therapies might be shown to be more hazardous than beneficial.