Among 42 Japanese patients treated with the insulin pump, 35 were consecutively selected who met the inclusion and exclusion criteria and who signed an informed consent. We used a Paradigm 712 pump (Medtronic, Northridge, CA) with which we program basal insulin infusion hourly every 0.05 units. All patients used rapid-acting insulin in their pumps and were asked to monitor blood glucose four times daily. To exclude factors that could modify insulin dose requirements (²), the patients were hospitalized in Osaka University Medical Hospital from July 2007 to July 2010 to adjust the insulin pump. We excluded individuals with eating disorders, concomitant dietary restrictions, unstable retinopathy, renal failure, or pregnancy or who were using other antidiabetic agents or steroids. During the 2–3 weeks of hospitalization, an eight-point glucose testing (before and 2 h after each meal [2200 and 0300 h]) was performed to allow optimization of basal insulin rates according to the previous report (⁴). The diabetic meals in all studies were 25–30 kcal/kg ideal body weight, consisting of 50–60% carbohydrate, 15–20% protein, and 20–25% fat and were prepared by dietitians. Overnight basal insulin rates were evaluated with the blood glucose readings at 2200, 0300, and 0700 h and set using the same method (⁴). Upon verification of the basal rates, bolus insulin was determined by physicians using carbohydrate counting. All meals were consumed within 20 min; no additional food or drink was consumed unless required to treat hypoglycemia. The target fasting glucose and 2-h postprandial glucose values were set at 100 and 150 mg/dL in each participant. TDD, TBD, and percentage of TBD to TDD (%TBD) were collected after achieving the target glucose level at least for 3 days. The average of the eight-point glucose testing for the final 3 days of the study period was collected. HbA_1c, TDD, and %TBD at 6 months after the study period in 26 patients were collected. Demographic data are presented as means ± SD.

Characteristics of the study subjects were as follows: sex, 11 males/24 females; age, 39.7 ± 10.9 years (20–69) [mean ± SD (range)]; body weight, 56.5 ± 10.0 kg (36–80); BMI, 22.0 ± 3.0 kg/m² (14.6–28.7); duration of diabetes, 22.1 ± 11.4 years (2–44); HbA_1c, 7.26 ± 1.40% (5.0–11.5).

The average of the eight-point blood glucose was shown in Fig. 1A (upper panel). According to the meal omission, basal insulin dose was adjusted in each patient. Basal insulin rate was shown in Fig. 1A (lower panel). TBD was 8.7 ± 2.9 units (3.3–15.3), TDD was 31.6 ± 8.5 units (18–50), and TDD/kg was 0.56 ± 0.13 units/kg (0.32–0.84). As shown in Fig. 1B, %TBD was 27.7 ± 6.9% (16.6–43.8). No patients required 50% TDD. In addition, there was no relationship between BMI, age, HbA_1c, and %TBD (data not shown).

HbA_1c changed from 7.42 ± 1.32 to 6.95 ± 1.46% (P < 0.01), TDD changed from 32.0 ± 7.8 to 33.9 ± 12.0 units/day (P = 0.21), and %TBD changed from 28.9 ± 5.1 to 29.7 ± 6.2% (P = 0.54) at 6 months after the study period.

We investigated here the basal insulin requirement in C-peptide–negative type 1 diabetic patients. Our results show that basal insulin requirement is ∼30% TDD in the inpatients on diabetic diets prepared by a dietitian (Fig. 1B). In addition, the maximal basal insulin requirement in all patients was 43.8% TDD, and no patients required 50% TDD. These results indicate the currently widely published formula (TBD = 0.5 × TDD) (¹–³) overestimates the TBD. King (⁵) also recently suggested that this should be revised to TBD = 0.4 × TDD to prevent excess basal insulin treatment. In our results, TBD was ∼30% TDD, which was less than King’s estimation. We assume that this is the result of the strict diabetic diet in the hospital setting. Diabetic meals are served to all patients. These meals are close to the macronutrient content as the meals recommended by the American Diabetes Association (⁶).

We consider that previous estimation (TBD = 0.5 × TDD) might have been influenced by several factors as follows. One is diet failure, i.e., increased fat intake. Fat ingestion may cause delay of gastric emptying, leading to late glucose excursion (⁷). When eating fat-rich pizza, it was recommended to give dual-wave bolus insulin for 8 h, which means giving bolus insulin for 8 h after normal bolus insulin to suppress late glucose excursion induced by fat (⁸). The basal insulin requirement in the daily life after discharge might be more than in the hospital if patients eat more fat than recommended. However, %TBD stayed at the same level without deteriorating glucose control 6 months after the study period. The other factor is the misunderstanding of blood glucose monitoring data in well controlled patients. There are some patients who set their basal insulin higher than required to lower their blood glucose. In these patients, although the peak glucose value is high, HbA_1c is seemingly well controlled, which leads to the misunderstanding about their basal requirement.

We did not use a continuous glucose monitoring system (CGMS) in our study. Although there is a report showing that eight-point glucose testing a day is comparable to a continuous glucose monitoring system, eight-point glucose testing has the possibility of missing midnight hypoglycemia (⁹). In such a case, it might be better to further reduce the TBD. Furthermore, this study is of small size and is not designed to be a randomized study comparing different percentages of basal and bolus.

In conclusion, basal insulin requirement is ∼30% TDD in Japanese type 1 diabetic patients who use the insulin pump.

No potential conflicts of interest relevant to this article were reported.

A.K. and H.K. researched data and wrote the manuscript. T.Y., M.M., and K.M. reviewed and edited the manuscript. N.F., K.F., and T.Y. researched data. M.T., F.S., and T.M. edited the manuscript. I.S. researched data and contributed to discussion.