Health & Medical Diabetes

Intensive Glycemic Control in Noninsulin-Treated T2DM

Intensive Glycemic Control in Noninsulin-Treated T2DM

Research Design and Methods


This was a 12-month, prospective, multicenter, open-label, parallel-group, randomized, controlled clinical trial; the full protocol was previously reported. The study was approved by the Ethics Committee of each site and complies with the Helsinki Declaration. All patients provided written informed consent before enrollment.

Setting and Participants


The trial was conducted at 39 diabetes clinics in Italy. Patients with type 2 diabetes not treated with insulin (disease duration 1–10 years), aged 35–75 years, and with HbA1c 7.0–9.0% were eligible. Patients were ineligible if they had insulin treatment for >7 days, previous use of structured SMBG, impending complications of diabetes, or limited life expectancy or if they were pregnant, breastfeeding, or intended to become pregnant.

Randomization


Allocation ratio was 1:1. A computerized random number generator was used to select random permuted blocks of four. Details on randomization restriction and block size were not disclosed to investigators. Randomization was stratified by the diabetes treatment at enrollment (diet only or diet plus diabetes medications). Allocation information was sealed in sequentially numbered opaque envelopes prepared by the clinical research organization managing the trial.

Interventions


A commercially available educational program (Accu-Chek EduCare; Roche Diagnostics, Monza, Italy) was used to provide standardized diabetes information to patients in both groups. The program is organized into subject-specific modules and includes charts and other materials to facilitate patient engagement. Sessions on nutrition, physical activity, SMBG, and diabetes medications were provided at baseline and additional modules were completed throughout the study.

Patients in the intensive structured monitoring (ISM) group were required to perform 4-point capillary glucose measures before breakfast and lunch, 2 h after lunch, and 5 h after lunch but before dinner 3 days/week, every week (2 working days [Monday–Friday] and 1 weekend day [Saturday or Sunday]), for 12 months. ISM patients were trained to interpret SMBG data and were given a diary listing glycemic targets as follows: <110 mg/dL for fasting glucose levels and glucose levels before lunch; <50 mg/dL as difference between postprandial and preprandial glucose levels; and suggestions for reaching treatment goals. Patients in the active control (AC) group were required to complete a 3-day, 4-point profile before their visits at months 6 and 12 to obtain data for comparison with the ISM group. These data were not available for use by clinicians for glycemic evaluation or medication adjustments.

At each follow-up visit (months 3, 6, 9, and 12), investigators performed physical examinations; recorded BMI, blood pressure, and heart rate; and collected blood samples for HbA1c measurements. HbA1c for statistical analysis was measured by the central laboratory (Laboraf Diagnostica e Ricerca, Milan, Italy) using the Variant II testing systems (Bio-Rad, Segrate, Italy).

At each visit, investigators prescribed diabetes medication aiming at an HbA1c target <7.0% in both groups. With ISM patients, investigators reviewed and discussed the SMBG and diary and reviewed and recommended changes in diet and physical activity. SMBG data from ISM patients were downloaded to a computer through a wireless device (Accu-Chek Smart-Pix system; Roche Diagnostics, Monza, Italy) and analyzed using ad hoc software that provided easy-to-read summary statistics (Supplementary Fig. 1). For adjusting diabetes medications, investigators had the option to use a treatment algorithm based on guidelines from international and national scientific societies (Supplementary Fig. 2). Incretin mimetics and DPP-4 inhibitors were not included in the algorithm because they were unavailable in Italy when the protocol was written. Once they became available, investigators were notified that they could be used for study patients. The algorithm guided changes in diabetes medications (type or dosage) based on mean fasting or preprandial glucose levels, differences between postprandial and preprandial glucose, and hypoglycemic events. In the AC group, SMBG data were not available for viewing in patient meters and data were not downloaded until the end of the study; therefore, adjustments of diabetes medications were based exclusively on HbA1c and hypoglycemic events.



(Enlarge Image)



Figure 1.



Flow of PRISMA study participants.







(Enlarge Image)



Figure 2.



Least-square mean difference in HbA1c (%) during the study by treatment group in the ITT population (A) and PP population (B).




Outcomes


Two primary end points were tested in hierarchical order: change in HbA1c levels from baseline to month 12 and percentage of patients reaching/maintaining the risk target (low blood glucose index [LBGI] ≤2.5 together with high blood glucose index [HBGI] ≤5) from baseline to month 12. LBGI and HBGI are summary statistics computed from SMBG data shown to predict the risk of hypoglycemia and hyperglycemia, respectively. LBGI increases when the number or extent (or both) of low SMBG measurements increases, whereas HBGI increases when the frequency or the extent (or both) of high SMBG measurements increases (see Supplementary Table 1 for computation of LBGI and HBGI). The hierarchical approach of primary end points avoids multiplicity issues with adjustment of type I error because the second co-primary end point is tested only if the first is statistically significant at 0.05.

Secondary end points included the following: changes in HBGI and LBGI; changes in SMBG frequency; changes in diabetes therapy (type of medication or dosage); frequency and severity of hypoglycemic episodes; changes in blood pressure, estimated glomerular filtration rate calculated according to the creatinine-based Modification of Diet in Renal Disease equation, lipid profile, and BMI; changes in diabetes-specific quality of life questionnaire scores and diabetes-specific locus of control questionnaire scores; and study-related and diabetes-related adverse events. The diabetes-specific quality of life questionnaire used in the Diabetes Control and Complications Trial, translated into Italian, modified for patients with type 2 diabetes, and validated, includes the following three domains: satisfaction (score 14–70); impact (score 28–92); and worry (score 5–25); higher score indicates poor quality of life. The diabetes-specific locus of control questionnaire, translated into Italian, includes the following three domains: internal; powerful others; and chance (scores 6–36); the domain with the highest score indicates locus of control.

Mild hypoglycemia and moderate hypoglycemia were defined as symptoms consistent with hypoglycemia or glucose levels ≤60 mg/dL or ≤50 mg/dL, respectively, without loss of consciousness; given some degree of overlapping between mild and moderate hypoglycemia, we present them combined as nonsevere hypoglycemia. Severe hypoglycemia was defined as an event with symptoms consistent with hypoglycemia during which the person required the assistance of another person or intravenous glucose or glucagon administration. The event might be confirmed by the finding of a glucose level <50 mg/dL. Participants were informed of their risk of hypoglycemia and instructed to record any hypoglycemic event in a diary and to contact the clinic if severe or repeated nonsevere hypoglycemia occurred.

In a post hoc analysis, we tested the difference in the proportion of participants in the ISM group or AC group who reached clinically meaningful HbA1c reductions of <0.3, <0.4, or <0.5%.

Statistical Analysis


Five hundred patients in each group were needed to achieve 90% power to detect a significant (at the two-sided 5% level) 0.3% difference between the ISM and AC groups in the mean HbA1c change at month 12 compared with baseline, assuming a 1.25% SD and 25% attrition. Primary and secondary end points were analyzed on an intent-to-treat (ITT) basis including all randomized patients. Primary and secondary end points also were analyzed in the per protocol (PP) population, consisting of all randomized patients who completed the study without major protocol violations and were compliant with the SMBG regimen (i.e., ≥80% of the required SMBG measurements in the ISM group and ≤200 unstructured discretionary SMBG measurements in the AC group, the maximum measurements recommended for these patients by the Italian Standards of diabetes care). Results for the PP population are reported for primary end points and selected secondary end points.

Statistical analyses were performed using SAS (version 9.02, TS level 02M0). The first co-primary end point was analyzed using a mixed linear model with randomized group, center, visit, and randomized group-by-visit interaction as fixed effects and baseline HbA1c as covariate. An unstructured variance–covariance matrix was used to model the correlation between repeated measurements within each patient. Restricted maximum likelihood estimates and two-sided 95% CI of the mean difference between randomized groups at month 12 were calculated using the Newton-Raphson algorithm. Cross-sectional comparisons were performed using time-by-time contrasts programmed using the SAS Mixed Procedure. Missing data were filled in by multiple imputation assuming a missing at-random mechanism of dropout. The Monte Carlo Markov Chain technique implemented in SAS Proc MI was used to obtain 50 imputed datasets. Rubin rules implemented in SAS Proc MIANALYZE were used to combine effect estimates and to estimate 95% CIs to allow for uncertainty attributable to missing data. For the post hoc analysis, the last observation carried forward technique was used to complete missing values for patients who did not complete the study. The incidence rate ratio of hypoglycemia was estimated using Poisson regression. A two-sided test with P ≤ 0.05 was considered statistically significant. The interaction between randomized group and center was assessed, with a two-sided P ≤ 0.10 considered statistically significant for the test of interaction. The analysis of the second co-primary end point was based on the Cochran-Mantel-Haenszel test controlling for clinical site effects. Secondary end points were analyzed according to the type of variable. Summary statistics and two-sided 95% CIs were computed for mean changes (continuous variables) and risk differences (categorical variables).

Leave a reply