Hospitalized patients with hyperglycemia require a longer hospital stay and face several complications [1, 2]. The glycemic level at fasting (fasting blood glucose, FBG) and 2-h postprandial blood glucose (2hBG) targeting 7.8 and 10.0 mmol/L, respectively, in non-critically ill patients with hyperglycemia, could prevent adverse outcomes [2, 3].

Insulin is superior to other medications for the rapid control of hyperglycemia in hospitalized patients. The normal physiological pattern of insulin secretion by the pancreas consists of basal release and burst of bolus insulin. American Association of Clinical Endocrinologists (AACE) and American Diabetes Association (ADA) recommend basal-bolus insulin instead of slide-scale insulin for glycemic control due to its effectiveness and safety [4]. However, the latest survey revealed that more than 30% of physicians still prefer to use slide-scale insulin in clinical practice because the administration of bolus insulin is complex [5,6,7]. Many studies simplified the insulin administration regimen, for example, by using basal plus [7], reducing correction insulin administration [8, 9], using premixed insulin [10], or only using basal insulin [11].

In healthy individuals, the ratio of basal and bolus insulin levels is 50:50 [12]. Therefore, when initiating insulin therapy in hospitalized patients with type 2 diabetes (T2D) one-half of the total daily dose (TDD) was given as a basal dose (glargine or determir) once daily at bedtime, and the other half was given as a bolus (aspart, lispo, or glulisine) in equally divided doses before breakfast, lunch, and dinner regardless of TDD calculation based on body mass index (BMI) or random glucose level [2, 13, 14]. At the end of the corresponding clinical trials, Umpierrez found that the ratio of basal-bolus was about 1:1 in their two clinical trials (glargine to glulisine, 22:20 units/day and 43:42 units/day) and Meyer et al. found the ratio as about 1:1 (glargine to glulisine 33:36 units/day) [13,14,15].

Currently, the initial and final ratio of basal-bolus are kept at 1:1. We also designed a study that had a titrating ratio of 1:1 during the middle period. Surprisingly, when FBG and three 2hBG reached the target levels, the ratio was 1:1.5 (glargine to aspart, 25:36 units/day) [16]. Liu et al. used a continuous subcutaneous infusion of titrated insulin and the study found a ratio of 1:1.5 when the target glucose level was attained [17]. A study with the Latin American non-intensive care unit patients with T2D used basal-bolus and had a ratio of about 1:1.5 (glargine to glulisine, 22:31 units/day) [18].

This multicenter, randomized, controlled clinical study aimed to investigate whether the weight-based, 1:1.5 basal-bolus insulin initiation and titration using an algorithm is superior to 1:1.

This multicenter, randomized prospective study was performed in the department of endocrinology of four medical centers (The Second Affiliated Hospital of Guangzhou Medical University, The Third Affiliated Hospital of Southern Medical University, the Fifth Affiliated Hospital of Guangzhou Medical University, and the First Affiliated Hospital of Shenzhen University) between October 2021 and June 2022. Informed consent was obtained from each participant. This study was approved by the ethics Committee of our hospital.

Patients who were aged between 18 and 75 years and diagnosed previously or newly with T2D, with a blood glucose (BG) level of > 10.0 mmol/L on admission were randomly selected for the current study. The patients were excluded if they were in one of the following criteria: (i) patients who received insulin therapy at a daily dosage of > 0.4 U/kg before admission. This is because one of our group initiated insulin 0.4 U/kg, if patient had used a daily dosage of > 0.4 U/kg before admission, their glucose maybe worsen ; (ii) patients who were unable to eat; (iii) patients who received corticosteroid therapy; (iv) patients who had renal insufficiency with the plasma creatinine concentration of ≥ 130 µmol/L or liver insufficiency (aspartate aminotransferase or alanine aminotransferase concentration of ≥ two-fold normal range), this is because insulin was metabolized in liver and kidney; (v) patients who were pregnant; (vi) patients with a previous or current history of malignant tumors.

Randomization codes were generated using a computer program (SPSS V.25.0). Patients were randomly assigned at a ratio of 1:1 to the two treatment groups on the first day of admission at the four medical centers. Neither patients nor investigators were masked to the treatment group.

Basal insulin bolus consisting of a subcutaneous injection of glargine (Sanofi Aventis Deutschland GmbH, Frankfurt, Germany) at bedtime and aspart (Novo Nordisk, Bagsværd, Denmark) before each of three meals. The blood glucose levels of the patients were checked during the round at 09:00 h and insulin titration dose was determined by physicians based on FBG in the morning and 2hBG after breakfast from the same day and 2hBG after lunch and dinner from the previous day.

All other antidiabetic agents were discontinued on the day of admission. Insulin was initiated at a TDD of 0.5 units/kg in the 1:1.5 group and 0.4 units/kg in the 1:1 group. The 1:1.5 group received 40% (0.2 units/kg) of the TDD as glargine and 60% (0.3 units/kg) as aspart. The 1:1 group received 50% (0.2 units/kg) of the TDD as glargine and 50% as aspart (0.2 units/kg). Glargine was administered as a single daily dose while aspart was divided into three equal parts.

Both glargine and aspart were titrated using a weight-based algorithm. In the 1:1.5 and 1:1 groups, glargine was titrated at 0.1 units/kg/day. In the 1:1.5 group, aspart was titrated at 0.05 units/kg/day, and in the 1:1 group, aspart was titrated at 1/3 of 0.1 units/kg/day before each meal. When one 2hBG level reached the target, the aspart was not titrated further (Table 1). The FBG and 2hBG target levels were set at 7.8 mmol/L and 10.0 mmol/L, respectively, as recommended by the American Endocrine Society, respectively (Table 1) [2]. If hypoglycemia was seen, the corresponding insulin titration was held.

Blood glucose levels at five points were measured, including FBG, 2hBG after three meals, and BG at 03:00 h using a glucose meter (Accu-Chek Advantage; Roche Diagnostics, Basel, Switzerland). Additionally, the glucose levels were measured when patients reported symptoms of hypoglycemia. Hypoglycemia was classified into three categories (level 1:3.0 ≤ BG < 3.9 mmol/L; level 2: BG < 3.0 mmol/L; level 3: a severe event that requires assistance from another person for treatment of hypoglycemia). Hemoglobin A1c (HbA1c) was tested in all patients on day 2 of hospitalization.

All centers ordered dietary profile according to our textbook. First, we calculated ideal weight (kg) which is height (cm) minus 105. Second, we calculated total energy between 25 (overweight or obese) and 30 (normal weight or lean) kcal/kg/day. Third, in our diet, protein was given 1.0 g/kg ideal weight, fat was given 0.8 g/kg ideal weight, and left energy was given as carbohydrate which account for about 50–55% of total energy. Total calories were divided in a ratio of 1:2:2 across the three daily meals.

The primary outcome of the study was the time for achieving FBG and 2hBG target levels. The secondary outcome was the incidence of hypoglycemia during hospitalization.

The primary endpoint was considered on the day when 2hBG reached the target level after three meals. Based on our previous study [9], a significant difference in the time was considered when 2hBG reached the target level on any 1 day between the two groups. Assuming significant differences of three points with α = 0.05% and 90% power, the required number of patients for each group was 86. To allow a 20% dropout rate, it needed 215 patients and we recruited 220 patients. The basic characteristics of subjects and outcome variables were compared using an independent t-test or χ2 test as appropriate. Statistical analyses were performed using the SPSS version 25.0 (SPSS, Chicago, IL). The p-value of < 0.05 was considered statistically significant. Data were provided as means ± SD or median (range).