United Kingdom Prospective Diabetes Study 24: A 6-Year, Randomized, Controlled Trial Comparing Sulfonylurea, Insulin, and Metformin Therapy in Patients with Newly Diagnosed Type 2 Diabetes That Could Not Be Controlled with Diet Therapy

Methods

Patients

In the parent study, we enrolled 4075 patients aged 25 to 65 years with newly diagnosed type 2 diabetes (fasting plasma glucose level > 6 mmol/L on two occasions) from 15 centers between 1977 and 1991. Patients were excluded if they had ketonuria (ketone concentration > 3 mmol/L) (2% of referrals), a history of myocardial infarction in the previous year, or current angina or heart failure; had had more than one major vascular episode; had a serum creatinine level greater than 175 µmol/L, severe retinopathy requiring photocoagulation, malignant hypertension, an uncorrected endocrine abnormality, or severe intercurrent illness likely to limit life expectancy; or required extensive systemic therapy [3]. The parent study was approved by the institutional review board of each hospital and was conducted in accord with the Declaration of Helsinki as amended in Tokyo (1975), Venice (1983), and Hong Kong (1989).

Initial Diet Therapy and Randomization

Initially, patients were seen monthly and prescribed a diet consisting of about 50% carbohydrates, low saturated fat, and moderately high fiber, with reduced total calorie content if patients were obese (>120% of ideal body weight). During this period, we identified 560 patients (13.7%) who had fasting plasma glucose levels greater than 15 mmol/L or persistent symptoms of hyperglycemia. These 560 patients (primary diet failure group) were randomly allocated to intensive treatment after we stratified for obesity [4]. Of the nonobese patients, 56% received sulfonylurea (chlorpropamide or glyburide) and 44% received insulin. Of the obese patients, 42% received sulfonylurea, 32% received insulin, and 26% received metformin (Figure 1).

At 3 months, 2769 of 4075 patients continued to receive diet therapy alone and were asymptomatic with fasting plasma glucose levels of 6 to 15 mmol/L. These 2769 patients formed the main randomization group of the parent study and were allocated to the same intensive therapies as the primary diet failure group and to a group that was allocated diet therapy alone. Seven hundred forty-six patients had fasting plasma glucose levels less than 6 mmol/L after 3 months of diet therapy (diet satisfactory) and were followed up. We compared patients in the primary diet failure group with those in the main randomization group who received the same intensive therapy allocation.

Measurements and Follow-up

Sulfonylurea doses were adjusted with the aim of obtaining a fasting plasma glucose level less than 6.0 mmol/L without triggering episodes of hypoglycemia. If fasting plasma glucose levels were 15.0 mmol/L or more at two consecutive clinic visits or if symptoms of hyperglycemia developed while patients received the maximum sulfonylurea dosage (chlorpropamide, 500 mg/d; glyburide, 20 mg/d), metformin was added to the regimen. Glyburide was added if patients met these criteria while receiving the maximum metformin dosage (2550 mg/d). Patients who met these criteria while receiving both sulfonylurea and metformin were switched to insulin. Patients who were randomized or switched to insulin received ultralente insulin (initially, beef Ultratard MC; subsequently, human Ultratard HM, Novo Nordisk, Pease Pottage, United Kingdom) [5]. These patients monitored their blood glucose levels at home if they required more than 14 U of ultralente insulin daily. If blood glucose levels before meals or at bedtime were greater than 7.0 mmol/L, additional soluble insulin, two or three times daily, or twice-daily mixtures of medium-acting and short-acting insulins were advised. We defined monotherapy as allocated therapy with sulfonylurea, metformin, or ultralente insulin alone.

At each 3-month visit, all patients were asked about hypoglycemic symptoms. The most severe hypoglycemic episode for each period was recorded; the clinician graded the episode as minor if the patient could treat the symptoms unaided or major if third-party help or medical intervention was necessary. We calculated the mean annual rates for major and all hypoglycemic episodes.

The central laboratory monitored fasting plasma glucose levels at individual clinics by using a monthly quality assurance scheme in which a coefficient of variation of less than 4% was maintained. In the central laboratory, glycosylated hemoglobin (hemoglobin A_1c) was assayed by using high-pressure liquid chromatography (normal range, 0.045 to 0.062), and plasma insulin was assayed by using radioimmunoassay with 100% cross-reactivity to proinsulin [6]. Levels of high-density lipoprotein cholesterol and total cholesterol were measured directly by precipitation methods [7].

The Homeostasis Model Assessment [8-10], a structural model of glucose-insulin interactions, was used to estimate the percentages of β-cell function and insulin sensitivity that would induce the measured fasting levels of insulin and glucose of each participant. Percentages of β-cell function and insulin sensitivity were expressed in relation to a reference group of 40 normoglycemic persons (age, 18 to 25 years; body mass index, 18 to 25 kg/m²) and were assessed in all patients at randomization. They were also assessed in patients who continued to receive sulfonylurea or metformin therapy at 1 year. Diabetic retinopathy was assessed by a modified Wisconsin grading of color retinal photography of four horizontal, 30-degree fields per eye [3, 11]. The presence of retinopathy was defined as grade 20-20 or worse (that is, microaneurysm in both eyes or more severe retinopathy in one eye).

Statistical Analysis

Statistical analyses were done by using SAS software [12] according to allocated therapy on an intention-to-treat basis, except where stated. For description of the groups, the mean ±SD, the geometric mean (1-SD interval), or the median (interquartile range) is given. For differences across time, data are expressed as the mean with a 95% CI. Percentages are quoted with 95% CIs except where the actual numbers of patients are given. We used the Mann-Whitney U test, t-test, or analysis of variance with paired comparisons using a Bonferroni correction to compare groups, and we used chi-square tests to compare proportions. Repeated-measures analysis of variance was used to compare data between groups across time. Adjustment for baseline values used analysis of covariance. We calculated estimates of survival functions for time to requirement of additional therapy by using the product-limit (Kaplan-Meier) method. For this analysis, patients were placed in three groups according to tertiles of the distributions of age or body mass index at randomization. For all analyses, a P value less than 0.05 was considered significant.

Annual data for body weight and fasting plasma glucose level were calculated as the median of three consecutive visits for each patient (the annual visit and the preceding and subsequent 3-month visit). Hemoglobin A_1c concentrations and plasma levels of insulin and lipids were assessed annually.

Role of Industry Sponsors

Our study was supported in part by pharmaceutical companies, including Novo Nordisk; Bayer; Bristol-Myers Squibb; Hoechst; Lilly; Lipha; and Farmitalia Carlo Erba. Additional assistance was granted by Boehringer Mannheim, Becton Dickinson, Owen Mumford, Securicor, Kodak, and Cortecs Diagnostics. None of the funding sources had a role in the collection, analysis, or interpretation of the data or in the decision to publish this report.

Results

We present 6-year follow-up data for 458 of 560 patients in the primary diet failure group and 1620 of 2769 patients in the main randomization group. The remaining patients were lost to follow-up (main randomization group, 96 [3%]; primary diet failure group, 10 [5%]), had died (main randomization group, 144 [5%]; primary diet failure group, 11 [5%]), or had not reached the 6-year time point (main randomization group, 909 [33%]; primary diet failure group, 81 [36%]). At baseline, patients who were followed did not differ from those who were not. The median time spent receiving diet therapy alone from recruitment to randomization was 42 days (95% CI, 22 to 63 days) in the primary diet failure group and 103 days (CI, 91 to 119 days) in the main randomization group (P < 0.001).

Clinical and Biochemical Characteristics at Baseline

Patients in the primary diet failure group were significantly younger and less obese than patients in the main randomization group (Table 1). Eighty-five percent of patients in the primary diet failure group and 81% of patients in the main randomization group were white. Retinopathy was more common at diagnosis in the primary diet failure group than in the main randomization group (28% and 19%; P < 0.001) and men in the primary diet failure group reported impotence more frequently (25% and 18%; P = 0.014). The proportions of patients who had hypertension (37% in the primary diet failure group and 40% in the main randomization group), macrovascular disease, or peripheral neuropathy did not differ significantly. As expected, hemoglobin A_1c concentrations were significantly higher in the primary diet failure group. In this group, percentages of β-cell function and insulin sensitivity were lower and plasma levels of cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and triglycerides were significantly elevated (Table 1).

Response to Treatment

Glycemic Control in the Primary Diet Failure Group

In nonobese and obese patients in the primary diet failure group, the mean decrease in fasting plasma glucose levels over 6 years was greater in patients allocated to insulin than in patients allocated to sulfonylurea (change in the nonobese and obese patients combined, −7.9 mmol/L [CI, −7.3 to −8.5 mmol/L] and −6.1 mmol/L [CI, −5.5 to −6.7 mmol/L], respectively [P < 0.001]). Absolute values are presented in Table 2, Table 4, and Figure 2. In the primary diet failure group, 47% of patients allocated to insulin and 32% of patients allocated to sulfonylurea had fasting plasma glucose levels less than 7.8 mmol/L at 6 years (P < 0.001). The mean decrease in hemoglobin A_1c concentration over 6 years in the primary diet failure group was similar for patients who were allocated to insulin (change, −2.5%[CI, −2.0% to −2.9%])and patients who were allocated to sulfonylurea (change, −2.6%[CI, −2.3% to −3.0%])(Figure 2).

In the primary diet failure group, the mean decrease in fasting plasma glucose levels over 6 years did not significantly differ between obese patients allocated to metformin and obese patients allocated to sulfonylurea (change, −5.2 mmol/L [CI, −4.1 to −6.3 mmol/L] and −5.4 mmol/L [CI, −4.3 to −6.5 mmol/L], respectively). The median fasting plasma glucose levels in patients who were allocated to metformin and patients who were allocated to sulfonylurea were 8.9 mmol/L (CI, 6.9 to 10.9 mmol/L) and 9.0 mmol/L (CI, 7.1 to 11.9 mmol/L), respectively, at 1 year and 9.8 mmol/L (CI, 8.2 to 12.5 mmol/L) and 10.0 mmol/L (CI, 7.6 to 11.9 mmol/L), respectively, at 6 years. The mean change in hemoglobin A_1c concentration in patients allocated to metformin was −2.0%(CI, −1.4% to −2.7%)and −1.9%(CI, −1.2% to −2.6%)in patients allocated to sulfonylurea.

Comparison with Main Randomization Group

At 1 year, the median fasting plasma glucose levels in patients allocated to insulin were similar in the primary diet failure group and the main randomization group, whereas the fasting plasma glucose level in patients allocated to sulfonylurea was higher in patients in the primary diet failure group (Table 2 and Table 4). The proportions of patients in the primary diet failure and main randomization groups who achieved fasting plasma glucose levels less than 7.8 mmol/L at 6 years were 48% (CI, 37% to 58%) and 52% (CI, 47% to 58%), respectively, in those allocated to insulin and 32% (CI, 21% to 42%) and 45% (CI, 40% to 50%), respectively, in those allocated to sulfonylurea (P < 0.001).

The decrease in hemoglobin A_1c concentrations was greater in the primary diet failure group than in the main randomization group, but initial concentrations in the primary diet failure group were considerably higher (Table 1). Among patients who were allocated to insulin, absolute hemoglobin A_1c concentrations were higher in the primary diet failure group than in the main randomization group at 1 year; similar results were observed for patients who were allocated to sulfonylurea (Table 2 and Table 4). At 6 years, the median hemoglobin A_1c concentration for patients allocated to insulin or sulfonylurea in the primary diet failure group was higher than that in the main randomization group (Table 2 and Table 4). The proportions of patients in whom hemoglobin A_1c concentrations were less than 0.08 at 6 years in the primary diet failure and main randomization groups were 49% (CI, 38% to 59%) and 66% (CI, 61% to 70%), respectively (P < 0.001), for patients allocated to insulin and 55% (CI, 46% to 64%) and 67% (CI, 63% to 71%), respectively (P < 0.001), for patients allocated to sulfonylurea.

At 6 years, the median hemoglobin A_1c concentrations in obese patients in the primary diet failure and main randomization groups were 0.082 (CI, 0.071 to 0.094) and 0.074 (CI, 0.066 to 0.089), respectively, in patients allocated to metformin (P = 0.0013) and 0.081 (CI, 0.068 to 0.097) and 0.073 (CI, 0.064 to 0.089), respectively, in patients allocated to sulfonylurea (P = 0.041). At 6 years, the proportion of patients in the primary diet failure and main randomization groups whose hemoglobin A_1c concentrations were less than 0.08 was 53% (CI, 34% to 72%) and 61% (CI, 53% to 68%), respectively, for patients allocated to insulin; 49% (CI, 33% to 66%) and 64% (CI, 58% to 70%), respectively (P = 0.028), for patients allocated to sulfonylurea; and 43% (CI, 22% to 64%) and 61% (CI, 53% to 69%), respectively (P = 0.027), for patients allocated to metformin.

Requirement for Additional Therapy

Figure 3 shows the therapies administered to nonobese and obese patients in the primary diet failure group at 6 years. The proportion of patients who continued to receive sulfonylurea or metformin at 6 years was smaller than the proportion of patients in the main randomization group. At 6 years, 38% (CI, 23% to 53%) and 72% (CI, 66% to 77%) (P < 0.001) of patients allocated to chlorpropamide in the primary diet failure group and main randomization group, respectively, were receiving chlorpropamide only and 31% (CI, 16% to 45%) and 60% (CI, 54% to 66%) (P < 0.001), respectively, of patients allocated to glyburide were receiving glyburide only. Thirty-five percent (CI, 12% to 57%) of patients in the primary diet failure group and 69% (CI, 63% to 77%) of patients in the main randomization group (P < 0.001) allocated to metformin were receiving metformin alone. Of patients allocated to insulin, 49% (CI, 38% to 60%) of patients in the primary diet failure group and 75% (CI, 71% to 80%) of patients in the main randomization group (P < 0.001) received ultralente insulin alone.

A Kaplan-Meier survival analysis of time to requirement of additional therapy showed that more patients in the primary diet failure group required additional therapy earlier than did patients in the main randomization group (log-rank test, P < 0.001) (Figure 4). Over 6 years, the proportion of patients in the primary diet failure group who continued to receive monotherapy was lower among patients younger than 46 years of age at diagnosis than among patients in older age groups (log-rank test, P < 0.001) (Figure 4, top). After 1 year, significantly more patients in the primary diet failure group in the lower third of the distribution for body mass index (<23 kg/m²) required additional therapy (P < 0.001). This trend was observed through 4 years, although the proportion of patients who required additional therapy in all three groups was similar after 4 years (Figure 4, bottom). Women did not differ significantly from men overall during the first 3 years, although more women required additional therapy during 3 to 6 years of follow-up (data not shown).

Dosages

By 6 years, most patients in the primary diet failure group who continued to receive sulfonylurea required the maximum dosage (67% [CI, 48% to 86%] for chlorpropamide, 500 mg/d, and 73% [CI, 57% to 88%] for glyburide, 20 mg/d). In the main randomization group, a smaller but not significantly different proportion of patients who continued to receive sulfonylurea required the maximum dose (58% [CI, 50% to 66%] for chlorpropamide, 500 mg/d, and 58% [CI, 50% to 65%] for glyburide, 20 mg/d).

At 6 years, the median total daily dose of insulin given to patients in the primary diet failure group who continued to receive insulin was significantly greater than that given to patients in the main randomization group who continued to receive insulin (nonobese patients, 36 U/d and 26 U/d, respectively [P < 0.001]; obese patients, 53 U/d and 40 U/d, respectively [P = 0.0016]).

Of obese patients in the primary diet failure group who were allocated to and continued to receive metformin therapy, 59% (CI, 28% to 89%) were receiving maximum dosages (2550 mg/d) at 6 years compared with 61% (CI, 52% to 71%) in the main randomization group.

Changes in Body Weight

In the primary diet failure group, the mean increase in body weight was greater in patients allocated to insulin than in patients allocated to sulfonylurea (9.9 kg [CI, 8.6 to 11.1 kg] and 5.3 kg [CI, 4.0 to 6.6 kg], respectively [P < 0.001]). Among obese patients in the primary diet failure group, those allocated to insulin had a greater mean increase in body weight (10.4 kg [CI, 7.3 to 13.5 kg]) than those allocated to sulfonylurea (3.7 kg [CI, −0.5 to 7.9 kg]; P = 0.013). Obese patients in the primary diet failure group who were allocated to metformin had a nonsignificant decrease in body weight (change, −1.3 kg [CI, −5.8 to 3.2 kg]; P < 0.001).

The increase in body weight was greater in patients in the primary diet failure group who were allocated to insulin than in patients in the main randomization group who were allocated to insulin (7.8 kg [CI, 7.2 to 8.4 kg]; P = 0.0024). This increase remained significant when adjusted for initial weight (10 kg [CI, 8.7 to 11.0 kg] in the primary diet failure group and 7.8 kg [CI, 7.3 to 8.4 kg] in the main randomization group; P < 0.001). Among patients allocated to sulfonylurea, the mean increase in body weight in the primary diet failure group did not significantly differ from that in the main randomization group (5.3 kg [CI, 3.6 to 7.0 kg] and 4.9 kg [CI, 4.4 to 5.3 kg]). Most of the weight gain in all groups occurred during the first year, with a slower gain during subsequent years (Figure 5).

Plasma Insulin Levels, β-Cell Function, and Insulin Sensitivity

The geometric mean of fasting plasma levels of insulin increased over 6 years in patients in the primary diet failure and main randomization groups who received insulin. The increase was less marked in patients who received sulfonylurea. Among patients in the primary diet failure group who received metformin, plasma insulin levels decreased initially, then returned to the starting level (Figure 5); this occurred because 65% of these patients (CI, 49% to 82%) were receiving sulfonylurea in addition to metformin or had switched to insulin by the 6-year follow-up point. Fasting plasma insulin levels over 6 years differed significantly among the allocated therapies, as assessed by repeated-measures analysis of variance (P < 0.001).

Among the 158 patients in the primary diet failure group who were allocated to and continued to receive sulfonylurea therapy at 1 year of follow-up, β-cell function increased from a geometric mean of 13% (CI, 6% to 26%) at randomization to 50% (CI, 26% to 99%) (P < 0.001). Among the 38 obese patients in the primary diet failure group who were allocated to and continued to receive metformin, β-cell function increased over 1 year from 18% (CI, 10% to 31%) to 48% (CI, 27% to 89%) (P < 0.001). Among 739 patients in the main randomization group who were allocated to and continued to receive sulfonylurea at 1 year of follow-up, β-cell function increased from 44% (CI, 25% to 77%) to 73% (CI, 42% to 127%) (P < 0.001). In 204 obese patients in the main randomization group who were allocated to and continued to receive metformin, β-cell function increased from 50% (CI, 29% to 85%) to 60% (CI, 36% to 100%) (P < 0.001). Similar calculations were not done at 6 years in either group because of the numbers of patients receiving combination therapy or exogenous insulin.

Hypoglycemia

Table 3 shows the average annual proportions of patients who reported hypoglycemic episodes and continued to receive monotherapy or were allocated to ultralente insulin but received a combination of short-acting and longer-acting insulins during the 6 years. Overall, patients in the primary diet failure group who received ultralente insulin (52%) reported significantly more episodes of hypoglycemia (52%) than did patients who received other therapies (chlorpropamide, 16%; glyburide, 21%) (P < 0.001), but the oral therapies did not differ significantly. Reported rates of any hypoglycemic episode among patients in the primary diet failure group who received intensive therapy did not differ significantly from those for patients in the main randomization group for any therapy. Hypoglycemic episodes were more common in patients in the primary diet failure group who were allocated to ultralente insulin and received combination insulin than in patients in the main randomization group who received the same therapies (70% and 54%; P < 0.001). In addition, more major hypoglycemic episodes occurred in the primary diet failure group than in the main randomization group (chlorpropamide, 1.5% and 0.4% [P = 0.016]; glyburide, 2.5% and 0.9% [P < 0.001]; ultralente insulin, 4.3% and 2.2% [P = 0.008]; combination insulin, 11.7% and 2.1% [P < 0.001]). However, this did not occur with metformin (0.0% and 0.4%).

Discussion

Five hundred sixty of 4075 patients (14%) with newly diagnosed type 2 diabetes in the United Kingdom Prospective Diabetes Study had primary diet failure (that is, they had persistent symptoms of hyperglycemia or fasting plasma glucose levels >15.0 mmol/L). This rate of primary diet failure is similar to that reported for patients 40 to 69 years of age with newly diagnosed diabetes in the United States [13] and underlines the fact that the general population contains persons with undiagnosed, untreated diabetes who have marked hyperglycemia. Our study patients had a greater prevalence of retinopathy and impotence than did patients with less hyperglycemia. There has been uncertainty about whether oral hypoglycemic agents are suitable therapies in patients with primary diet failure or whether immediate insulin therapy would be more appropriate. Overall, we found similar decreases in hemoglobin A_1c concentrations in patients who received sulfonylurea, insulin, or metformin. Although insulin induced lower fasting plasma glucose levels than did other therapies, the responses of the hemoglobin A_1c concentrations were similar. This suggests that the greater reduction in fasting plasma glucose levels with insulin therapy was not accompanied by similar improvement in postprandial glucose levels so that with each therapy, the overall control achieved was similar. Although increasing the dosage of shorter-acting insulin to cover meals may further reduce hemoglobin A_1c concentrations, in practice a greater risk for hypoglycemia limits the degree of control that can be achieved.

As may have been expected, fewer patients in the primary diet failure group than in the main randomization group who received oral hypoglycemic agents had hemoglobin A_1c concentrations less than 0.08 after 1 year. The protocol specified that maximum dosages of each oral hypoglycemic agent should be given to try to attain levels less than 6.0 mmol/L. If the fasting plasma glucose level was 15.0 mmol/L or more or if the patient became symptomatic while receiving the maximum dosage, a change in therapy was indicated (additional oral hypoglycemic agents or switch to insulin). In the course of 6 years, we observed a failure rate of 60% for sulfonylurea defined in relation to these criteria for hyperglycemia; this rate is similar to that reported in unselected patients with type 2 diabetes with a failure rate of 13% per year during the first 6 months and a failure rate of 5% per year thereafter [14]. The higher incidence of failure of sulfonylurea therapy in nonobese patients compared with obese patients has been noted by others [15]. Metformin therapy in obese patients was as effective as other agents in reducing glycemia in the short term, and patients who were allocated to metformin had a requirement for additional sulfonylurea that was similar to the requirement that patients who were allocated to sulfonylurea had for additional metformin. Initially, metformin reduced fasting plasma insulin levels, but in some patients these levels increased from 4 years when additional sulfonylurea or insulin therapies were begun.

One advantage of our study is that we included newly diagnosed patients who were referred directly from general practitioners in 15 centers; the patients are therefore likely to be representative of the general population. On the other hand, most analyses were done on an intention-to-treat basis and thus included patients who did not take their allocated therapies. Because our study examined in part the glycemic and clinical end point responses to individual therapies, the addition of other therapies when the fasting plasma glucose reached 15 mmol/L was conservative. In practice, it may be possible to obtain better glucose control than the control that we described by greater use of polypharmacy at an earlier stage of the progressive hyperglycemia associated with type 2 diabetes [16].

The absolute decrease in fasting plasma glucose levels or hemoglobin A_1c concentrations with insulin therapy in the primary diet failure group was greater than that in the main randomization group, but the hemoglobin A_1c concentration reached was higher than that in the main randomization group. More hypoglycemic reactions occurred with insulin than with other therapies, and insulin therapy induced more weight gain. This suggests that even in patients with primary diet failure, it may not be advantageous to proceed directly to insulin therapy. If a goal-oriented approach is used (such as the reduction of hemoglobin A_1c concentrations to below a certain level), it is reasonable to initiate therapy with oral agents and proceed to insulin if the goal is not achieved. In obese patients, initial therapy with metformin may be a reasonable option: It is as effective as other agents in the short term and induces less weight gain and fewer hypoglycemic reactions than other therapies.

The UKPS is assessing the relative efficacy of the different therapies in terms of preventing complications of diabetes, including the possibility (raised by University Group Diabetes Program) that sulfonylurea or biguanides may increase the incidence of fatal myocardial infarction [17]. A subgroup analysis of data from the University Group Diabetes Program suggested that heart attacks in patients treated with sulfonylureas may have occurred particularly in patients whose fasting plasma glucose levels were greater than 11.2 mmol/L [18]. However, the study had too few end points to allow inferences to be made, especially because unplanned subgroup analyses without previous hypotheses have little credence. To determine whether different therapies have specific effects on the development of complications, each therapy in the UKPS was maintained as monotherapy with a moderately high threshold for the fasting plasma glucose level before different agents were added. If the UKPS shows that improved glucose control is beneficial in maintaining the health of diabetic patients, a lower threshold for adding additional therapies would apply, such as fasting plasma glucose levels less than 7.0 or 8.0 mmol/L or hemoglobin A_1c concentrations less than 0.07 or 0.08. Nevertheless, the high incidence of hypoglycemia and weight gain with sulfonylurea or insulin emphasizes that obtaining nearnormal glucose levels with available therapies is difficult. Indeed, despite visits every 3 months in which diabetes specialist nurses, dietitians, and physicians were involved, home monitoring of blood glucose levels, and combination insulin regimens, patients often did not achieve normoglycemia. This was in part because of the high incidence of insulin-induced hypoglycemia, which is a limitation in treating patients with type 2 diabetes, just as it is in treating patients with type 1 diabetes [19].

By clinical criteria, only patients with type 2 diabetes were recruited to the UKPS. Patients who had ketonuria greater than 3 mmol/L that was suggestive of type 1 diabetes were excluded. Some patients in the primary diet failure group may have had late-onset, slowly progressive type 1 diabetes [20]. However, the average age of the patients in the primary diet failure group was only 3 years less than that of the patients in the main randomization group. The greater reduction in β-cell function in patients in the primary diet failure group compared with that in patients in the main randomization group is compatible both with a more severe type 2 diabetic β-cell defect and an occult type 1 diabetic process. Our study shows that patients with primary diet failure can be managed in the same way as most patients with type 2 diabetes: with initial use of diet and oral agents, as long as the efficacy of therapy is assessed in relation to glycemic goals. However, the progressive deterioration of β-cell function and increasing hyperglycemia make it essential to review therapy regularly and add other therapies when indicated.

Appendix

This paper was prepared by Alex Wright, FRCP; Carole Cull, PhD; Rury Holman, FRCP; and Robert Turner, FRCP.

The following are the physicians and centers participating in UKPS.

Professor Robert Turner, Radcliffe Infirmary, Oxford; Dr. Lilian Murchison, Aberdeen Royal Infirmary, Aberdeen; Dr. A.D. Wright, General Hospital, Birmingham; Dr. Nigel Oakley, St. George's Hospital, London; Professor Eva Kohner, Hammersmith Hospital, London; Professor Randal Hayes, Belfast City Hospital, Belfast; Dr. John Scarpello, North Staffordshire Royal Infirmary, Stoke-on-Trent; Professor David Hadden, Royal Victoria Hospital, Belfast; Dr. A.G. Spathis, St. Helier Hospital, Carshalton; Professor John Yudkin, Whittington Hospital, London; Dr. Richard Greenwood, Norfolk & Norwich Hospital, Norwich; Dr. Les Borthwick, Lister Hospital, Stevenage; Dr. John Day, The Ipswich Hospital, Ipswich; Dr. Ray Newton, Ninewells Hospital, Dundee; Dr. Charles Fox, Northampton General Hospital, Northampton; Dr. Richard Paisey, Torbay Hospital, Torquay; Dr. Jonathan Roland, Peterborough District Hospital, Peterborough; Dr. David Humphriss, Scarborough Hospital, Scarborough; Dr. Ian Peacock, Derbyshire Royal Infirmary, Derby; Dr. Andrew Boulton, Manchester Royal Infirmary, Manchester; Dr. Tim Dorman, Hope Hospital, Salford; Dr. Felix Burden, Leicester General Hospital, Leicester; and Professor John Tooke, Royal Devon & Exeter Hospital, Exeter.