Three-Year Efficacy of Complex Insulin Regimens in Type 2 Diabetes

Patients

The study design and 1-year results have been reported previously.⁴ Briefly, men and women 18 years of age or older who had at least a 12-month history of type 2 diabetes mellitus and who had not been treated with insulin were recruited in 58 clinical centers in the United Kingdom and Ireland. All patients had glycated hemoglobin levels of 7.0 to 10.0% while receiving maximally tolerated doses of metformin and sulfonylurea for at least 4 months; 5% of the patients were taking only one of these drugs, since the other was not tolerated. All patients had a body-mass index (the weight in kilograms divided by the square of the height in meters) of 40 or less. Exclusion criteria included a history of thiazolidinedione therapy or triple oral antidiabetic therapy.

All patients provided written informed consent and confirmed their willingness to inject insulin and perform glucose self-monitoring. The protocol was approved by local and national ethics and regulatory agencies and was implemented in accordance with the provisions of the Declaration of Helsinki⁷ and Good Clinical Practice guidelines.⁸

Study Design

Patients were randomly assigned to receive twice-daily biphasic insulin aspart (NovoMix 30), three-times-daily prandial insulin aspart (NovoRapid), or once-daily (twice if required) basal insulin detemir (Levemir). Patients injected doses of biphasic and prandial insulin immediately before meals and basal insulin at bedtime. All three preparations were supplied by Novo Nordisk in 3-ml disposable-pen devices (FlexPen).

During the first year of the study, sulfonylurea therapy was replaced by a second type of insulin if hyperglycemia became unacceptable (a glycated hemoglobin level of >10.0% or two consecutive values of ≥8.0% at or after 24 weeks of therapy) or subsequently if glycated hemoglobin levels were more than 6.5%.⁴ For the biphasic-based regimen, midday prandial insulin was added, starting with 10% of the current total daily biphasic insulin dose and limited to a minimum of 4 units and a maximum of 6 units. For the prandial-based regimen, basal insulin (10 units) was added at bedtime. For the basal-based regimen, prandial insulin was added at breakfast, lunch, and dinner, starting with 10% of the current total daily dose of basal insulin at each time point and limited to a minimum of 4 units and a maximum of 6 units.

First-year visits with patients were scheduled at 2, 6, 12, 24, 38, and 52 weeks, with interim telephone contact. After the first year, visits were scheduled every 3 months, with patients asked in advance to perform three daily capillary glucose profiles (Medisense Optium, Abbott).⁴ Using these profiles and data regarding self-reported hypoglycemia, the trial-management system⁴ suggested changes in the insulin dose, aiming for glucose values before meals of 72 to 99 mg per deciliter (4.0 to 5.5 mmol per liter) and values 2 hours after meals of 90 to 126 mg per deciliter (5.0 to 7.0 mmol per liter). Investigators and patients were encouraged to vary suggested insulin doses, as clinically appropriate, and to amend the doses between visits. Hypoglycemia was categorized as grade 1 (symptoms only) if a patient had symptoms with a self-measured capillary glucose level of 56 mg per deciliter (3.1 mmol per liter) or more, grade 2 (minor) if the patient had symptoms with a self-measured capillary glucose level of less than 56 mg per deciliter, or grade 3 (major) if third-party assistance was required.⁴

The trial steering committee consisted of five academic members, one lay member, and three representatives of Novo Nordisk, the sponsor. Only academic members had access to the nonsafety data. All authors vouch for the accuracy, integrity, and completeness of the reported data, which were collected and analyzed by the Diabetes Trials Unit.

Biochemical and Clinical Measurements

Glycated hemoglobin levels were measured at baseline; at 12, 24, 38, and 52 weeks; and then every 12 weeks. Plasma creatinine was measured at baseline; at 2, 6, and 12 weeks; and then every 12 weeks. Blood pressure was measured and the ratio of urinary albumin to creatinine was determined at baseline and then every 26 weeks. Plasma lipid and alanine aminotransferase levels were measured and a health-status questionnaire (EuroQol Group 5-Dimension Self-Report Questionnaire) was administered at baseline, at 12 and 52 weeks, and then annually.⁹

Primary and Secondary Outcomes

The primary 3-year outcome was the glycated hemoglobin level. Secondary outcomes were the proportion of patients with a glycated hemoglobin level of 6.5% or less, the proportion of patients with a glycated hemoglobin level of 6.5% or less but without hypoglycemia of grade 2 or more, weight gain, self-measured capillary glucose profiles, the proportion of patients requiring a second type of insulin, the ratio of albumin to creatinine, and quality of life.

Statistical Analysis

Five imputations for missing data were performed with the use of the Bayesian Markov chain Monte Carlo multiple-imputation technique.¹⁰ To account for center-level clustering, the study center was included as a random effect in all regression models. For normal continuous variables, mixed linear regression models¹¹ were used, with respective baseline values, type of baseline oral antidiabetic therapy, study group, and baseline glycated hemoglobin level as covariates. Mixed-effect logistic models were used for patients with glycated hemoglobin levels of 6.5% or less or 7.0% or less. Calculations were repeated for patients with baseline glycated hemoglobin levels of 8.5% or less, with the type of oral antidiabetic therapy and glycated hemoglobin level at baseline as potential covariates. The proportion of patients with hypoglycemia was analyzed with the use of a generalized binomial model without adjustment for covariates. For hypoglycemia rates, generalized mixed models with negative binomial distributions were used. Repeatedly observed, self-measured capillary glucose profiles were analyzed with the use of an unstructured covariance matrix and random study-center effects, with the usual covariates.

The ratio of urinary albumin to creatinine and insulin doses were analyzed with the use of generalized mixed-effect models with gamma distribution, adjusted for baseline values, including glycated hemoglobin level, and type of oral antidiabetic therapy. Quality-of-life data are presented as Winsorized means with 95% confidence intervals, with treatment comparisons at median levels based on quantile regression. For skewed data, the median with 95% confidence intervals is presented.¹²

A prespecified closed-test procedure allowed for a pairwise comparison between groups. A two-sided P value of less than 0.05 was considered to indicate statistical significance; all P values are based on adjusted analyses but have not been adjusted for multiple testing.

Patients

Figure 1. Figure 1. Enrollment and Outcomes.

From November 1, 2004, to July 31, 2006, we recruited 708 patients and randomly assigned 235 to the biphasic group, 239 to the prandial group, and 234 to the basal group (Figure 1). The patients' mean (±SD) age was 61.7±9.8 years, and the median duration of disease was 9 years. Most of the patients were white and overweight, without significant differences in baseline variables among the groups.⁴ Overall 130 patients (18.4%) did not complete the 3-year evaluation, with no significant between-group differences in the biphasic group (14.5%), the prandial group (21.3%), and the basal group (19.2%) (P=0.15 for the overall comparison). However, the proportions of patients who withdrew from the study differed significantly among the groups (5.1%, 11.7%, and 8.5% respectively; P=0.04). There were no significant differences in baseline variables between patients who withdrew from the study and those who completed the study.

Primary Outcome

Table 1. Table 1. Outcomes and Changes from Baseline at 3 Years.

Figure 2. Figure 2. Primary and Secondary Outcomes at 3 Years.

Panel A shows median levels of glycated hemoglobin in the three study groups, with a kernel-density plot of the distribution of values for patients in each group at 3 years, as compared with the distribution of values for all patients at baseline, shown in Panel B. Panel C shows mean body weight, with a kernel-density plot of the distribution of values for patients in each group at 3 years, as compared with the distribution of values for all patients at baseline, shown in Panel D. Panel E shows median insulin doses. Panel F shows the proportions of patients in the three study groups reporting grade 2 or grade 3 hypoglycemic events over time. The I bars indicate 95% confidence intervals.

The median glycated hemoglobin levels converged after 1 year and remained stable in all groups, with an overall value at 3 years of 6.9% (95% confidence interval [CI], 6.8 to 7.1); these values did not differ significantly in the three groups (P=0.28 for the overall comparison) (Table 1 and Figure 2A). At 3 years, the mean reduction from baseline was 1.3% in the biphasic group, 1.4% in the prandial group, and 1.2% in the basal group (Figure 2A and 2B).

Secondary Outcomes

Fewer patients in the biphasic group (31.9%) achieved glycated hemoglobin levels of 6.5% or less than in either the prandial group (44.7%, P=0.006) or the basal group (43.2%, P=0.03) (Table 1). The corresponding proportions of patients with a glycated hemoglobin level of 7.0% or less also differed significantly between the biphasic group (49.4%) and each of the two other groups, with 67.4% in the prandial group (P<0.001) and 63.2% in the basal group (P=0.02).

Among patients with a baseline glycated hemoglobin level of 8.5% or less, those in the biphasic group were less likely to achieve values of 6.5% or less, as compared with either the prandial group (odds ratio, 0.48; 95% CI, 0.28 to 0.82; P=0.007) or with the basal group (odds ratio, 0.46; 95% CI, 0.27 to 0.78; P=0.004).

The proportions of patients who replaced sulfonylurea with a second type of insulin differed significantly among the three groups, with 67.7% in the biphasic group, 73.6% in the prandial group, and 81.6% in the basal group (P=0.002 for the overall comparison).

Figure 3. Figure 3. Changes from Baseline to 3 Years in Glycated Hemoglobin, Fasting Plasma Glucose, Postprandial Glucose, and Body Weight and the Rate of Hypoglycemia.

Panel A shows the mean (±SE) percentage changes in key outcome measures, with P values adjusted for baseline values (except hypoglycemia), study center, baseline glycated hemoglobin level, and type of oral antidiabetic therapy, where appropriate. Missing data were imputed with the use of a multiple-imputation technique.¹⁰ To convert values for glucose to millimoles per liter, multiply by 0.05551. Panel B shows the median number of hypoglycemic events per patient per year in the three groups. The I bars indicate 95% confidence intervals.

Self-measured capillary glucose levels at all time points except 3 a.m. were significantly lower in the prandial group than in the biphasic group (P=0.001) but were not significantly lower than in the basal group (P=0.06). No significant differences were seen in fasting glucose values in the three groups (Figure 3A). However, a greater mean reduction in postprandial glucose values was seen in the prandial group than in either the biphasic group (P<0.001) or the basal group (P=0.007), with a greater reduction in the basal group than in the biphasic group (P=0.04). The reduction in 3 a.m. glucose values was significantly greater in the basal group than in the prandial group (P=0.02)

Patients gained weight in all three groups; increases in the biphasic group and the prandial group were similar and were more than those in the basal group (Figure 2C and 2D and 3A). Waist circumference increased less in the basal group than in either the biphasic group or the prandial group.

The median daily insulin dose per kilogram of body weight increased steadily during the second and third years of the study (Figure 2E). The dose was similar at 3 years in the prandial group and the basal group but lower in the biphasic group (P=0.02 for the overall comparison). Patients who required a second type of insulin had higher median daily insulin doses, with a similar pattern but substantially different ratios of prandial insulin to total insulin.

Rates of hypoglycemia of grade 2 or more converged among the groups during the second and third years of the study and did not differ significantly in the third year (P=0.44) (Figure 2F). However, the overall hypoglycemia rates remained highest in the prandial group and lowest in the basal group (Figure 3B). The median number of hypoglycemic events per patient per year during the trial was 3.0 in the biphasic group, 5.5 in the prandial group, and 1.7 in the basal group; among patients with a glycated hemoglobin level of 6.5% or less, the corresponding numbers were similar, with 3.0, 5.5, and 2.0 events, respectively (P<0.001 for the overall comparison).

At 3 years, no clinically relevant between-group differences were seen in changes from baseline in either systolic or diastolic blood pressure, high-density lipoprotein or low-density lipoprotein cholesterol, triglycerides, or the ratio of urinary albumin to creatinine, although the differences in high-density lipoprotein cholesterol were significant (P=0.03). In addition, no significant differences were seen in changes from baseline with respect to Winsorized mean scores on the EuroQol Group 5-Dimension Self-Report Questionnaire.

Adverse Events

Table 2. Table 2. Adverse Events.

During the study period, 19 patients died (7 in the biphasic group, 9 in the prandial group, and 3 in the basal group; P=0.23); of these patients, 14 died from cardiovascular disease (4 in the biphasic group, 9 in the prandial group, and 1 in the basal group; P=0.002). The proportion of patients with any type of serious adverse event differed among the groups, with the highest proportion in the biphasic group (P=0.01). No significant between-group differences were seen in the proportion of patients with individual serious adverse events or in the numbers of nonserious adverse events (Table 2).

No clinically relevant changes occurred in levels of plasma creatinine or alanine aminotransferase in any group. According to the protocol, 14 patients discontinued metformin therapy (4 in the biphasic group, 6 in the prandial group, and 4 in the basal group) after two successive plasma creatinine measurements of more than 1.7 mg per deciliter (150 μmol per liter).¹³

In our 3-year evaluation of three different analogue insulin regimens, the median achieved glycated hemoglobin was similar in all three groups, but the distributions differed, with fewer patients achieving glycemic targets in the biphasic group than in either the prandial group or the basal group. The substantially improved glycated hemoglobin levels that were achieved at the beginning of the trial were generally maintained, although a majority of patients required intensification to a more complex insulin regimen.

There were important clinical differences among the three strategies. There was less weight gain and a smaller increase in waist circumference in the basal group than in either the biphasic group or the prandial group. Rates of hypoglycemia also differed and were lowest in the basal group and highest in the prandial group.

We used a clinically relevant, pragmatic protocol with clinic visits every 3 months, a schedule that was commensurate with routine management in primary care; the 3-year retention rate was 82%. Insulin titration to glycemic targets, which was guided by a computerized algorithm on the basis of self-monitored glucose profiles, was consistent among the three groups. We believe that our findings may be generalizable, since short- and long-acting analogues have efficacies similar to those of human insulin,^14-17 and meta-analysis has demonstrated only a minor benefit for short-acting analogues.¹⁸

A strength of our trial was its long duration, with overall maintenance of glycemic control with low rates of hypoglycemia. In routine clinical practice, even moderate glycemic control remains an elusive goal, characterized by delays in intensifying oral therapies and in the initiation of insulin, as evidenced by a retrospective study between 1995 and 2005.¹⁹

The results of our trial support current guidelines, which suggest that basal and prandial insulin regimens should be considered if adequate glycemic control is not achieved with initial regimens.^20,21 Although there is evidence for the advantages of the present approach in patients with type 1 diabetes, data supporting such a strategy in those with type 2 diabetes have been sparse, apart from a recent nonrandomized subgroup observational analysis of the effects of transferring from a biphasic regimen to various basal–prandial regimens.²²

The achieved glycated hemoglobin level⁵ and the proportions of patients who had glycated hemoglobin levels of 7.0% or less and those who had levels of 6.5% or less²³ are consistent with previous trials of complex insulin regimens. Approximately two thirds of patients in the two groups in which intensification led to a basal–prandial regimen reached the 7.0% glycated hemoglobin target, which showed that the tight glycemic control that was achieved in short-term studies of treat-to-target insulin initiation^15,24 can be maintained. The lower success rate for the biphasic-based regimen and the lower median insulin dose achieved may reflect the decreased flexibility of a fixed-ratio insulin formulation, as compared with a basal–prandial regimen.

During a 3-year period, the median daily insulin doses rose progressively to become higher than those reported in short-term insulin trials^24,25 and greater in the basal and prandial regimens than in the biphasic regimen. Although daily insulin doses were similar in the basal and prandial groups, our findings suggest that the initiation of insulin with a basal formulation, as compared with a prandial formulation, is of benefit before intensification to a basal–prandial regimen. It is likely that the greater ratio of prandial to total insulin in the prandial group explains the greater reduction in postprandial glucose levels and the higher rate of hypoglycemia than in the basal group. The lower weight gain in the basal group may be persistence of the difference observed at 1 year or may reflect a continuing need to correct the higher rates of hypoglycemia, as seen in the biphasic and prandial groups, with a higher carbohydrate intake. The higher rate of hypoglycemia and weight gain in the prandial group than in the basal group was consistent with the findings in other trials.^3,25

Recent trials of intensive glycemic control have shown risks of severe hypoglycemia.^26,27 In our trial, we found that reasonable levels of glycemic control could be achieved with a low rate of major hypoglycemia, particularly when therapy was initiated with basal insulin. However, the possible association between hypoglycemia and death from cardiovascular causes supports continuing research in this area. Reassuringly, the rate of hypoglycemia in our study was no greater in patients reaching the 6.5% target than in those who did not reach this target.

Future research will need to explore the relative inability of a minority of patients to achieve an adequate reduction in glycated hemoglobin levels, regardless of the insulin regimen used, and to examine whether this outcome is associated with physiological or behavioral factors in either patients or health care professionals.²⁸

In conclusion, our findings comparing three different insulin strategies provide an evidence base to guide the addition of insulin to oral antidiabetic therapy and its intensification in clinical practice. The results support the initial addition of basal insulin to oral therapy, with subsequent intensification to a basal–prandial regimen, consistent with consensus recommendations.²¹ Using this approach, a majority of patients were able to achieve glycemic targets safely, with rates of hypoglycemia and weight gain that were lower than those in either the biphasic group or the prandial group.

The following investigators participated in the 4-T trial (asterisks indicate employees of Novo Nordisk): Steering Committee — R. Holman (chair), J. Darbyshire, M. Davies, A. Dyer Toft,* A. Farmer, M. Gall,* J. Levy, G. Nelson, H. Schou.* Data and Safety Monitoring Board — S. del Prato (chair), I. Campbell, R. Gray, R. Hills, S. Marshall, J. Scarpello. Policy Advisory Group — M. Davies (chair), A. Adler, P. Harvey, G. Neary, M. Bilous, J. Farmer, S. Gray, R. Jones, C. Kelly, D. Mellor, A. Millward, D. Moore, D. Russell-Jones, J. Nolan. Coordinating Center — B. Barrow, J. Darbyshire, Z. Doran, C. Dudley, A. Gilligan, V. Gregory, J. Hartweg, E. Harris, R. Holman, J. Keenan, I. Kennedy, K. Macdonald, D. McLeod, S. Motupally, R. Roberts, S. Paul, I. Stratton, K. Thorne, A. Tse, L. Tucker, M. Usman. Central Laboratory — J. Carpenter, R. Carter, K. Fisher, K. Islam, R. Klyne, L. Mansfield, A. Platt, B. Shine, A. Tse, T. Waknell. Novo Nordisk Study Team* — A. Aggarwal, J. Blakemore, A. Brown, N. Bryant, M. Budwal, C. Carrington, F. Chambers, B. Chubb, L. Clemente, L. Crawshaw, N. Dunmore, M. du Preez, J. Fenton, M. Fitch, J. Fox, C. Gordon, J. Hauff, K. Hoppen, D. Hutchins, R. Lewis, D. Lighter, L. Lowe, S. McQuade, S. Mansfield, I. Minns, A. Monk, N. Griffiths, M. Patterson, S. Peck, C. Pike, R. Meakin, A. Rosenfalck, J. Shadbolt, S. Shamash, T. Sorensen, M. Stack, P. Stella, R. Taylor, E. Townshend, P. Wilkinson, S. Williams, R. Young. Investigators — England: A. Adler, A. Akintewe, S. Atkin, T. Barnett, S. Bennett, R. Bilous, C. Bodmer, L. Borthwick, R. Davies, C. Fox, N. Furlong, I. Gallen, G. Gill, R. Gregory, P. Harvey, A. Hassey, S. Heller, P. Home, D. Hopkins, A. Johnson, E. Jude, D. Kerr, S. Kumar, J. Litchfield, J. Lorains, K. McLeod, P. McNally, M. Mansfield, D. Matthews, A. Millward, P. O'Hare, S. Page, A. Panahloo, D. Robertson, M. Rossi, R. Rowe, D. Russell-Jones, M. Sampson, I. Scobie, W. Stephens, C. Strang, J. Vora, T. Wheatley, D. Whitelaw, P. Wiles, P. Winocour; Ireland — F. Dunne, B. Kinsley, J. Nolan, S. Sreenan; Northern Ireland — J. Andrews, W. Henry, S. Hunter; Scotland — P. Abraham, A. Collier, S. Gray, A. Jaap, G. Leese, I. Malik, D. Matthews, J. Walker.