Randomized Trial Comparing Continuous Subcutaneous Insulin Infusion and Conventional Insulin Therapy in Type II Diabetic Patients Poorly Controlled With Sulfonylureas
Objective: To compare the effects of continuous subcutaneous insulin infusion (CSII) and conventional insulin therapy (CIT) in patients with poorly controlled sulfonylurea-treated diabetes mellitus. Research Design and Methods: Twenty-five patients aged 40-65 yr and poorly controlled with sulfonylureas and without severe diabetic complications comprised the study group. Five patients left the study (3 achieved satisfactory glycemic control without insulin, 1 defaulted, 1 developed ketonuria). Ten patients were treated with CSII and 10 with CIT. Outpatient treatment consisted of CIT (twicedaily injections of regular and NPH insulin) or CSII (basal infusion and prandial boluses of regular insulin). Results: Glycosylated hemoglobin improved with both methods of insulin delivery (P < 0.01), but 8 of 10 CSIItreated patients achieved satisfactory glycemic control (HbA, 55 mmol hydroxymethylfurfural (HMF)/mol Hb (equivalent to 11.2% by column chromatography) underwent a full clinical assessment. Blood was drawn after a 12-h fast for blood glucose,
DIABETES CARE, VOL. 14, NO. 8, AUGUST 1991
A.M. JENNINGS AND ASSOCIATES
serum creatinine, serum lipids, HbA,, C-peptide, and islet cell antibody (ICA). Patients without severe diabetic complications who had previously been satisfactorily treated with sulfonylureas for at least 1 yr were approached to take part in this study. Patients were excluded if they had features suggesting type I diabetes (including those with ICA), retinopathy requiring laser therapy, serum creatinine >200 |JLM, severe neuropathy, severe cardiovascular disease, an uncorrected endocrine abnormality, or another life-threatening disease. Of 32 consecutive patients who fulfilled the entry requirements, 25 (78%) agreed to participate. Patients first entered a 3-mo run-in period during which treatment with diet and sulfonylureas was optimized. They were reviewed by a dietitian (S.M.) and a doctor (A.M.J.) every 2 wk for 1 mo and then monthly. They commenced self-monitoring of blood glucose (SMBG) with reagent strips (BM 1 -44, Boehringer Mannheim, Mannheim, Germany). Recommended diets provided - 3 0 % of total energy from fat and 50% from carbohydrates. Patients were encouraged to increase the fiber content of their diet and obese patients were advised to reduce their total energy intake. At the end of the run-in period, patients achieving satisfactory blood glucose control (defined as HbA! 7 mM. The total increase was distributed between the basal and prandial insulins (CSII) or regular and NPH insulins (CIT) according to the pattern of the SMBG results. If blood glucose levels did not improve, a second scale with larger dosage increments was introduced. Insulin dosages were not increased in the event of unexplained hypoglycemia or dietary excess. After the initial period of stabilization, patients were taught to adjust their own insulin therapy. Patients were asked to record hypoglycemic episodes sufficient to interrupt daily activities. Dietary composition was assessed from dietary records, which were kept for 2 days each month (1 weekday and a Saturday or Sunday) and analyzed with a computer program (Microdiet, University of Salford, Salford, UK). Blood was drawn monthly for fasting blood glucose and HbA, and at the beginning and end of insulin therapy for fasting lipids and insulin antibodies. Patients were reviewed by an ophthalmologist before insulin therapy and after 4 mo of treatment. On each occasion, photographs were taken of eight standard fields of each optic fundus, according to the methods of the Early Treatment Diabetic Retinopathy Study (7). At the end of the study, the photographs were analyzed according to the modified Airlie House criteria by one ophthalmologist (J.F.T.) who was unaware of the patient's treatment (8). Patients were asked to complete a booklet of questionnaires on entering the study before commencing insulin therapy and after 4 mo of insulin therapy. The questionnaires included measures of general well-being (depression, anxiety, and positive well-being; 9), satisfaction with treatment (9), and health beliefs (perceived benefits of and barriers to treatment, perceived severity of diabetes and its complications, and perceived vulnerability to complications; 10). Patients were also asked to rate their apprehension about injections and CSII on individual 7-point scales from 0 (not at all apprehensive) to 6 (very apprehensive) at the first visit. Fasting blood glucose and capillary blood glucose were measured by a glucose oxidase method. Total se-
739
CSII IN TYPE II DIABETES
rum cholesterol, serum triglyceride, and high-density liproprotein (HDL) cholesterol were measured with Boehringer reagents on a Cobas biocentrifugal analyzer (Roche, Welwyn Garden City, UK); HDL cholesterol was measured after precipitation of other lipoproteins with dextran sulfate and magnesium. Insulin antibodies were measured by radioimmunoassay (11) and ICA by indirect immunofluorescence (12). C-peptide was measured by modification of the method described by Heding with the interassay coefficient of variation being 8% (13). HbA, was measured by a colorimetric method and reported in rnillimole HMF per mole Hb (14). The reference range was 29-39 mmol HMF/mol Hb, and the interassay coefficient of variation was 6%. This method was compared with a column chromatography method (Bio-Rad, Richmond, CA) in the Department of Clinical Chemistry at this hospital, and over the range of 4-13% the following relationship was observed: HbA! (%) = 0.21 (HbA! in mmol HMF/mol Hb) - 0.35 (unpublished observations). Statistical methods. Data were analyzed nonparametrically with Wilcoxon's rank-sum test and Wilcoxon's signed-rank test. Fisher's exact test and x2-analysis were used to measure association, and correlation was assessed with Spearman's rank method. Statistical analyses were two tailed unless stated. Analyses were performed with the computer program Microstat (Ecosoft, IN) on an Olivetti M24 personal computer. Analyses involving psychological variables were conducted with an Algol statistics package on an RML 380Z personal computer, and SPSSX on an IBM 3083 computer.
TABLE 1 Characteristics of patients Treatment group
Age (yr) n (M/F) Diabetes duration (yr) Percentage ideal body weight Fasting C-peptide (nM)
Conventional insulin therapy
Continuous subcutaneous insulin infusion
61 (52-63) 10(6/4) 6(2-10)
58 (42-65) 10(6/4) 6(2-18)
122 (82-164) 0.66 (0.38-2.09)
118(98-158) 0.76(0.27-2.13)
Results expressed as medians with ranges in parentheses.
Hb, whereas only 3 of 10 patients treated with CIT were able to do so (Fisher's exact test = 0.035 [1-tailed], P = 0.069 [2-tailed], x2-analysis, P < 0.05). The mean decrement in HbA, was 26% (range 3-47%) in CSIItreated patients and 20% (-4-37%) in CIT-treated patients. There was no significant difference in the level of glycemic control achieved by patients above or below 120% ideal body weight. Median daily insulin dose was 0.58 U/kg (range 0.39-0.81 U/kg) for patients treated with CSII, a mean of 58% that was given as a basal infusion. Median daily insulin dose for CIT-treated patients was 0.65 U/kg (range 0.35-1.9 U/kg, NS vs. CSII), 67% of which was given as NPH insulin. Weight gain during insulin therapy was comparable in patients treated with CSII (median 4.5 kg, range - 4 . 5 - 9 . 0 kg) and those treated with
RESULTS Five patients did not progress to insulin therapy. One patient developed features suggesting type I diabetes (ketonuria and weight loss), one defaulted during the run-in period, and three achieved satisfactory glycemic control with sulfonylurea therapy and diet. Twenty patients were treated with insulin, 10 were randomized to CSII, and 10 to CIT. Oral therapy consisted of gliclazide 320 mg/day (14 patients), glyburide 20 mg/day (4 patients), or chlorpropamide 500 mg/day (2 patients). Patients randomized to CSII did not differ significantly from those randomized to CIT in baseline characteristics (Table 1). All 20 patients completed 4 mo of insulin therapy, although one CSII-treated patient died of a myocardial infarction before his final retinal assessment. Before commencing insulin therapy, fasting blood glucose, mean blood glucose, and HbA, were comparable in the two groups of patients (Table 2). After 4 mo of insulin therapy, fasting blood glucose, mean blood glucose, and HbA, had significantly improved in patients treated with CSII and in those treated with CIT but all three measures of glycemic control were significantly lower in the group of patients treated with CSII than in those treated with CIT (Table 2). Eight of 10 patients treated with CSII achieved HbA, 4 kg, despite frequent dietary advice. Rapid improvement in glycemic control has been associated with early deterioration in retinopathy in type I diabetes (19,20,28), although with longer duration of treatment, retinopathy may improve (29,30). In this study, only one patient developed the soft exudates characteristically associated with intensified insulin therapy (30). The infrequency of marked retinal deterioration may reflect several factors. First, 50% of patients attending both-retinal assessments had no evidence of retinopathy at the first assessment, a feature associated with a low risk of deterioration in type I diabetec patients undergoing rapid improvement in glycemic control (20,28). Second, HbA! concentrations were not as close
DIABETES CARE, VOL. 14, NO. 8, AUGUST 1991
A.M. JENNINGS AND ASSOCIATES
to the nondiabetic range as those reported for intensive insulin therapy in type I diabetes (4,19,20), and the decrement in HbA, was not as pronounced as that reported elsewhere (31). Third, the number of patients studied was small. These observations need to be extended to a larger group of patients before firm conclusions can be drawn regarding retinal changes during rapid improvement in glycemic control in type II diabetes. During this study, questionnaires were used to evaluate several psychological parameters. Within-group analyses revealed that CIT-treated patients reported significant improvements in treatment satisfaction after 4 mo of insulin treatment. CSII-treated patients reported significantly fewer barriers to treatment and greater perceived treatment cost-effectiveness (benefits less barriers) after 4 mo of insulin treatment. They also perceived the severity of their diabetes to be significantly reduced. Although one patient in this group reported a marked deterioration in treatment satisfaction, most CSII-treated patients reported improved treatment satisfaction. Indeed, treatment satisfaction increased significantly when scores for all insulin-treated patients were analyzed together. In the CSII-treated patients, there was a positive correlation between the degree of apprehension about CSII at recruitment and the HbA, after 4 mo of insulin therapy. Although it would not be wise to generalize from one isolated correlation in a small number of patients, a previous study has also shown that patients' concerns may be predictive of treatment outcome (32). Patients included in this study were representative of patients
Objective: To compare the effects of continuous subcutaneous insulin infusion (CSII) and conventional insulin therapy (CIT) in patients with poorly controlled sulfonylurea-treated diabetes mellitus. Research Design and Methods: Twenty-five patients aged 40-65 yr and poorly controlled with sulfonylureas and without severe diabetic complications comprised the study group. Five patients left the study (3 achieved satisfactory glycemic control without insulin, 1 defaulted, 1 developed ketonuria). Ten patients were treated with CSII and 10 with CIT. Outpatient treatment consisted of CIT (twicedaily injections of regular and NPH insulin) or CSII (basal infusion and prandial boluses of regular insulin). Results: Glycosylated hemoglobin improved with both methods of insulin delivery (P < 0.01), but 8 of 10 CSIItreated patients achieved satisfactory glycemic control (HbA, 55 mmol hydroxymethylfurfural (HMF)/mol Hb (equivalent to 11.2% by column chromatography) underwent a full clinical assessment. Blood was drawn after a 12-h fast for blood glucose,
DIABETES CARE, VOL. 14, NO. 8, AUGUST 1991
A.M. JENNINGS AND ASSOCIATES
serum creatinine, serum lipids, HbA,, C-peptide, and islet cell antibody (ICA). Patients without severe diabetic complications who had previously been satisfactorily treated with sulfonylureas for at least 1 yr were approached to take part in this study. Patients were excluded if they had features suggesting type I diabetes (including those with ICA), retinopathy requiring laser therapy, serum creatinine >200 |JLM, severe neuropathy, severe cardiovascular disease, an uncorrected endocrine abnormality, or another life-threatening disease. Of 32 consecutive patients who fulfilled the entry requirements, 25 (78%) agreed to participate. Patients first entered a 3-mo run-in period during which treatment with diet and sulfonylureas was optimized. They were reviewed by a dietitian (S.M.) and a doctor (A.M.J.) every 2 wk for 1 mo and then monthly. They commenced self-monitoring of blood glucose (SMBG) with reagent strips (BM 1 -44, Boehringer Mannheim, Mannheim, Germany). Recommended diets provided - 3 0 % of total energy from fat and 50% from carbohydrates. Patients were encouraged to increase the fiber content of their diet and obese patients were advised to reduce their total energy intake. At the end of the run-in period, patients achieving satisfactory blood glucose control (defined as HbA! 7 mM. The total increase was distributed between the basal and prandial insulins (CSII) or regular and NPH insulins (CIT) according to the pattern of the SMBG results. If blood glucose levels did not improve, a second scale with larger dosage increments was introduced. Insulin dosages were not increased in the event of unexplained hypoglycemia or dietary excess. After the initial period of stabilization, patients were taught to adjust their own insulin therapy. Patients were asked to record hypoglycemic episodes sufficient to interrupt daily activities. Dietary composition was assessed from dietary records, which were kept for 2 days each month (1 weekday and a Saturday or Sunday) and analyzed with a computer program (Microdiet, University of Salford, Salford, UK). Blood was drawn monthly for fasting blood glucose and HbA, and at the beginning and end of insulin therapy for fasting lipids and insulin antibodies. Patients were reviewed by an ophthalmologist before insulin therapy and after 4 mo of treatment. On each occasion, photographs were taken of eight standard fields of each optic fundus, according to the methods of the Early Treatment Diabetic Retinopathy Study (7). At the end of the study, the photographs were analyzed according to the modified Airlie House criteria by one ophthalmologist (J.F.T.) who was unaware of the patient's treatment (8). Patients were asked to complete a booklet of questionnaires on entering the study before commencing insulin therapy and after 4 mo of insulin therapy. The questionnaires included measures of general well-being (depression, anxiety, and positive well-being; 9), satisfaction with treatment (9), and health beliefs (perceived benefits of and barriers to treatment, perceived severity of diabetes and its complications, and perceived vulnerability to complications; 10). Patients were also asked to rate their apprehension about injections and CSII on individual 7-point scales from 0 (not at all apprehensive) to 6 (very apprehensive) at the first visit. Fasting blood glucose and capillary blood glucose were measured by a glucose oxidase method. Total se-
739
CSII IN TYPE II DIABETES
rum cholesterol, serum triglyceride, and high-density liproprotein (HDL) cholesterol were measured with Boehringer reagents on a Cobas biocentrifugal analyzer (Roche, Welwyn Garden City, UK); HDL cholesterol was measured after precipitation of other lipoproteins with dextran sulfate and magnesium. Insulin antibodies were measured by radioimmunoassay (11) and ICA by indirect immunofluorescence (12). C-peptide was measured by modification of the method described by Heding with the interassay coefficient of variation being 8% (13). HbA, was measured by a colorimetric method and reported in rnillimole HMF per mole Hb (14). The reference range was 29-39 mmol HMF/mol Hb, and the interassay coefficient of variation was 6%. This method was compared with a column chromatography method (Bio-Rad, Richmond, CA) in the Department of Clinical Chemistry at this hospital, and over the range of 4-13% the following relationship was observed: HbA! (%) = 0.21 (HbA! in mmol HMF/mol Hb) - 0.35 (unpublished observations). Statistical methods. Data were analyzed nonparametrically with Wilcoxon's rank-sum test and Wilcoxon's signed-rank test. Fisher's exact test and x2-analysis were used to measure association, and correlation was assessed with Spearman's rank method. Statistical analyses were two tailed unless stated. Analyses were performed with the computer program Microstat (Ecosoft, IN) on an Olivetti M24 personal computer. Analyses involving psychological variables were conducted with an Algol statistics package on an RML 380Z personal computer, and SPSSX on an IBM 3083 computer.
TABLE 1 Characteristics of patients Treatment group
Age (yr) n (M/F) Diabetes duration (yr) Percentage ideal body weight Fasting C-peptide (nM)
Conventional insulin therapy
Continuous subcutaneous insulin infusion
61 (52-63) 10(6/4) 6(2-10)
58 (42-65) 10(6/4) 6(2-18)
122 (82-164) 0.66 (0.38-2.09)
118(98-158) 0.76(0.27-2.13)
Results expressed as medians with ranges in parentheses.
Hb, whereas only 3 of 10 patients treated with CIT were able to do so (Fisher's exact test = 0.035 [1-tailed], P = 0.069 [2-tailed], x2-analysis, P < 0.05). The mean decrement in HbA, was 26% (range 3-47%) in CSIItreated patients and 20% (-4-37%) in CIT-treated patients. There was no significant difference in the level of glycemic control achieved by patients above or below 120% ideal body weight. Median daily insulin dose was 0.58 U/kg (range 0.39-0.81 U/kg) for patients treated with CSII, a mean of 58% that was given as a basal infusion. Median daily insulin dose for CIT-treated patients was 0.65 U/kg (range 0.35-1.9 U/kg, NS vs. CSII), 67% of which was given as NPH insulin. Weight gain during insulin therapy was comparable in patients treated with CSII (median 4.5 kg, range - 4 . 5 - 9 . 0 kg) and those treated with
RESULTS Five patients did not progress to insulin therapy. One patient developed features suggesting type I diabetes (ketonuria and weight loss), one defaulted during the run-in period, and three achieved satisfactory glycemic control with sulfonylurea therapy and diet. Twenty patients were treated with insulin, 10 were randomized to CSII, and 10 to CIT. Oral therapy consisted of gliclazide 320 mg/day (14 patients), glyburide 20 mg/day (4 patients), or chlorpropamide 500 mg/day (2 patients). Patients randomized to CSII did not differ significantly from those randomized to CIT in baseline characteristics (Table 1). All 20 patients completed 4 mo of insulin therapy, although one CSII-treated patient died of a myocardial infarction before his final retinal assessment. Before commencing insulin therapy, fasting blood glucose, mean blood glucose, and HbA, were comparable in the two groups of patients (Table 2). After 4 mo of insulin therapy, fasting blood glucose, mean blood glucose, and HbA, had significantly improved in patients treated with CSII and in those treated with CIT but all three measures of glycemic control were significantly lower in the group of patients treated with CSII than in those treated with CIT (Table 2). Eight of 10 patients treated with CSII achieved HbA, 4 kg, despite frequent dietary advice. Rapid improvement in glycemic control has been associated with early deterioration in retinopathy in type I diabetes (19,20,28), although with longer duration of treatment, retinopathy may improve (29,30). In this study, only one patient developed the soft exudates characteristically associated with intensified insulin therapy (30). The infrequency of marked retinal deterioration may reflect several factors. First, 50% of patients attending both-retinal assessments had no evidence of retinopathy at the first assessment, a feature associated with a low risk of deterioration in type I diabetec patients undergoing rapid improvement in glycemic control (20,28). Second, HbA! concentrations were not as close
DIABETES CARE, VOL. 14, NO. 8, AUGUST 1991
A.M. JENNINGS AND ASSOCIATES
to the nondiabetic range as those reported for intensive insulin therapy in type I diabetes (4,19,20), and the decrement in HbA, was not as pronounced as that reported elsewhere (31). Third, the number of patients studied was small. These observations need to be extended to a larger group of patients before firm conclusions can be drawn regarding retinal changes during rapid improvement in glycemic control in type II diabetes. During this study, questionnaires were used to evaluate several psychological parameters. Within-group analyses revealed that CIT-treated patients reported significant improvements in treatment satisfaction after 4 mo of insulin treatment. CSII-treated patients reported significantly fewer barriers to treatment and greater perceived treatment cost-effectiveness (benefits less barriers) after 4 mo of insulin treatment. They also perceived the severity of their diabetes to be significantly reduced. Although one patient in this group reported a marked deterioration in treatment satisfaction, most CSII-treated patients reported improved treatment satisfaction. Indeed, treatment satisfaction increased significantly when scores for all insulin-treated patients were analyzed together. In the CSII-treated patients, there was a positive correlation between the degree of apprehension about CSII at recruitment and the HbA, after 4 mo of insulin therapy. Although it would not be wise to generalize from one isolated correlation in a small number of patients, a previous study has also shown that patients' concerns may be predictive of treatment outcome (32). Patients included in this study were representative of patients
