Subjects.
This retrospective cohort study was conducted at the Changhua Christian Hospital (CCH), Taiwan. A total of 23629 patients with type 2 diabetes were screened for eligibility using registry data from the Diabetes Case Management Program (DCMP) at the CCH Diabetes Care Center between January 2007 and December 2017. The DCMP provides standardized comprehensive diabetes care including lifestyle assessment, physical examination, laboratory evaluation, and diabetes self-management (DSM) education (such as instruction on nutrition, diet, exercise, medication, self-monitoring of blood glucose (SMBG), and problem-solving skills aimed at reducing related complications). All participants in the program received education during scheduled teaching sessions. Care is delivered by a coordinated multidisciplinary team, including physicians, and certified diabetes educators (registered nurses and dietitians). A detailed description of the program has been reported elsewhere (24). The diagnosis of T2DM was based on the criteria established by the American Diabetes Association, by satisfying one of the following: a fasting plasma glucose value ≥ 126 mg/dL, a 2-h plasma glucose ≥ 200 mg/dL during a 75-g oral glucose tolerance test, a random plasma glucose ≥ 200 mg/dL in a patient with classic symptoms of hyperglycemia, or a glycated hemoglobin (HbA1c) level of ≥ 6.5%. Excluded from the study were individuals with type 1 and other types of diabetes (n = 49), those with less than 3 years of analyzable data (n = 3926), diabetes duration exceeding 12 months at enrollment (n = 15990), pregnancy (n = 1), baseline HbA1c < 7 (n = 1010), no anti-diabetic medication treatment (n = 270), DKA/HHS during 3 years (n = 35), age below 18 or above 70 (n = 1020), and an estimated glomerular filtration rate (eGFR) < 30 mL/min/1.73 m2 (n = 72). The analysis encompassed 1256 eligible patients, with 1142 in the oral antidiabetic therapy group and 114 in the early insulinization group, following 1:1 propensity score matching to mitigate bias (Fig. 1). Methods were performed in accordance with the relevant guidelines and regulations. The Institutional Review Board of CCH granted the waiver for informed consent and approved the study (IRB No: 210318).
Data collection. Data were collected from the hospital's electronic medical records, including the DCMP diabetes registry, prescriptions, laboratory data, and CCH research database. Patients with type 2 diabetes were referred to the Diabetes Care Center by diabetes specialists, typically 2 to 6 weeks after their first outpatient clinic visit. Upon enrollment in the DCMP, patients underwent comprehensive assessments, including surveys, physical examinations, and laboratory tests. They received standardized one-on-one diabetes self-management education sessions. A certified diabetes educator evaluated patients' knowledge on glycemic control, willingness towards DSM, frequency of SMBG, and medication adherence through face-to-face interviews post-course (24).
Outcome variables.
Glycemic control, the primary outcome, was assessed using HbA1c levels measured at baseline and at 3, 6, 9, 12, 24, and 36 months post-enrollment, measured through ion-exchange high-performance liquid chromatography using the VARIANTTM II Turbo system .
Major exposure variables. Participants who received early insulinization for 3 to 6 months were in the insulin group, while those receiving OADs alone were in the OAD group.
Variables for subgroup analysis. Participants were categorized based on HbA1c levels (< 9.0% or ≥ 9.0%), sex (male or female), BMI (< 30 kg/m2 or ≥ 30 kg/m2), and age at onset of diabetes (< 40 or ≥ 40).
Other control variables. Basic data included age at onset of diabetes, gender, education level, and family history of diabetes. Health-related behaviors included smoking (within the preceding year), alcohol consumption (more than once weekly within the preceding year), and physical activity [regular (≥ 30 min/day, ≥ 3 days/week), occasional (less rigorous than regular exercise), or no exercise]. Knowledge regarding glycemic control was defined as an understanding of the need for and methods of controlling blood glucose. Willingness toward DSM was defined as the motivation to learn self-management techniques. Medication adherence was defined as taking medication regularly at the dose recommended by the physician over the past week. Four-point scales were used to assess the three aforementioned variables. Data were merged into simple dichotomies (i.e., top-two-box vs. bottom-two-box) and categorized as adequate (yes) or inadequate (no) for analysis (24).
Physical examination included measurement of blood pressure (BP), height, and body weight. Systolic and diastolic BP were measured with the patients in a seated position after a 10-min rest. Body mass index was calculated as body weight (kg)/height (m2). Baseline laboratory data included total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), creatinine, and glutamic pyruvic transaminase (GPT) levels measured using a UniCel DxC 800 Synchron Clinical System (Beckman Coulter, Brea, CA, USA). eGFR was calculated using the equation recommended by the National Kidney Foundation. Data on the major non-psychiatric comorbidities described in the Charlson Comorbidity Index (CCI) during the year preceding enrolment were collected from the CCH research database. Major comorbidities, including congestive heart failure (CHF), coronary artery disease (CAD), cerebrovascular accident (CVA), hypertension (HTN), hyperlipidemia, chronic kidney disease (CKD), cancer, cirrhosis, rheumatoid arthritis (RA), myocardial infarction (MI), peripheral vascular disease (PVD) and dysrhythmia were analyzed as independent variables (24).
Statistical analysis. Frequency counts with percentages and mean ± standard deviation (SD) were used to describe categorical and continuous variables, respectively. Differences between the two groups were assessed using the chi-square test for categorical variables and the Student’s t-test for continuous variables. To mitigate the influence of confounding variables, we employed 1:1 propensity score matching. Kaplan-Meier method was utilized to generate cumulative incidence curves for well-controlled events, and discrepancies between these curves were assessed using log-rank tests. Cox proportional hazard models were employed to evaluate the relationship between insulin use status and well-controlled events. The multivariate Cox proportional regression model was adjusted for the propensity score. Hazard ratios (HR) with 95% confidence intervals (CI) were used to express the results. Subgroup analyses were conducted to evaluate the association between insulin and glycemic control in subgroups stratified by various HbA1c levels (≥ 9.0% or < 9.0%), sex (male or female), BMI (≥ 30 kg/m2 or < 30 kg/m2), and age at onset of diabetes (≥ 40 or < 40). All analyses were two-tailed and performed using IBM SPSS Statistics version 22 (IBM Corp., Armonk, NY, USA) and SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA). The significance level was set at 0.05.