This was a single-center, retrospective, observational study. Participants in this study were individuals with type 1 diabetes treated at the Division of Diabetology and Metabolism, Tokyo Women’s Medical University Hospital aged 18 years or above, who switched from MDI to SAP between February 2015 and December 2019 (SAP group). Individuals who had used SAP before were excluded. As controls, individuals who changed from MDI to CSII (except patch pumps) between April 2014 and December 2019 (CSII group) were included. Type 1 diabetes diagnosis was based on the Japan Diabetes Society diagnostic criteria [9,10,11]. The decision to choose between SAP or CSII was made after consulting the individual and their physician. The SAP devices available at our institution were: Medtronic Minimed 640G and 620G (Medtronic plc, Dublin, Ireland). The available CSII devices were: Medtronic Paradigms 722 and 712 (Medtronic plc, Dublin, Ireland), TOP Syringe Pump TOP-8200 (TOP Corporation, Tokyo, Japan), and Medtronic Minimed 640G and 620G without CGM. Although the Medtronic Minimed 770G became available during the observation period, participants who started using the hybrid closed-loop function were excluded. Also, LGS and PLGS in MiniMed 640G were also excluded. Participants who started using LGS or PLGS feature were discontinued from observation. For glucose monitoring, individuals in the CSII group underwent self-monitoring of blood glucose (SMBG), and individuals in the SAP group underwent real-time CGM (Enlite sensor). However, because intermittently scanned CGM (isCGM) (FreeStyle Libre, Abbott Laboratories, Chicago, IL, USA) was launched in 2017, some participants in the CSII group used both SMBG and isCGM. Participants were followed up for up to 3 years.

Individuals aged < 18 years were excluded. The observation was terminated when: SGLT2 inhibitors were used; a participant became pregnant; real-time CGM was started in the CSII group; LGS, PLGS, or a hybrid closed-loop insulin pump was initiated in the SAP group; or a participant on CSII was switched to SAP.

In our department, insulin pumps are introduced either on an inpatient or outpatient basis. For the basal insulin dose, the total daily dose (TDD) of insulin for the past week was determined; approximately 80% of the TDD was assumed to be the initial TDD of the insulin pump, and 30–40% of the TDD divided by 24 h was set as the initial value (units/hour) of the basal rate [12]. The amount of bolus insulin was determined based on the amount of rapid insulin at the time of MDI. When carbohydrate counting was performed, the initial insulin sensitivity factor was defined as 1700/TDD (mg/dL/unit), and the initial carbohydrate-to-insulin ratio was defined as 450/TDD (g/unit) [12]. Both values were set with appropriate fine-tuning based on the blood glucose measurements and CGM data. The insulin used was selected in discussion between the physician and the participant. Either aspart, lispro, or glulisine was used.

SAP, CSII, CGM, and SMBG were covered by health insurance. The insulin pump was introduced under the supervision of doctors and nurses at our institution, and follow-up was also performed at our institution. Clinical examinations were performed as part of routine medical care. A typical evaluation schedule at our institution is as follows: blood glucose, HbA1c, and urine analysis are routinely measured every 2 months, and blood count, liver function, and renal function are routinely measured every 2–6 months. For micro vascular complications, retinopathy was routinely evaluated at least annually by an ophthalmologist. Retinopathy was classified according to the Fukuda Classification or Modified Davis Classification. Nephropathy was assessed by measuring urinary albumin/creatinine ratio (uACR), usually once a year, on awakening. An uACR of 30-299 was classified as early nephropathy, and an uACR of 300 or more as overt nephropathy.

Patient education on how to operate the insulin pump, what to do in case of pump issues, and interpretation of CGM results were provided mainly by nurses holding CDEj (an expert in lifestyle guidance in the clinical treatment of diabetes) in the SAP group. The setting of treatment goals and configuration about low glucose suspend were determined individually through consultation between the physician and pump users. For most individuals, the treatment goal was HbA1c < 7% and no severe hypoglycemia.

The primary outcome was a comparison of HbA1c over time and between the SAP and CSII groups over a 3-year period. The secondary outcome consisted of four analyses: (1) a comparison of body weight and insulin use over time and between the SAP and CSII groups over a 3-year period; (2) changes in the proportion of participants in each group who experienced severe hypoglycemia in each year to assess safety; (3) a comparison of the percentage of time that the CGM was active and the CGM metrics among SAP users; and (4) factor analysis of participants who achieved HbA1c < 7%.

HbA1c, body weight, and insulin dosage at outpatient visits were obtained from medical records, and changes in pre-introduction values at 12, 24, and 36 months were determined. The SAP and CSII groups were compared in terms of the percentage of participants achieving HbA1c < 7%, which was the target for most individuals; comparisons were performed annually. Daily insulin dose was divided into total insulin, basal insulin, and bolus insulin, divided by body weight (kg). The insulin change at a given time was calculated as the rate of change calculated from: [(insulin dose at that time − insulin dose before pump introduction)/insulin dose before pump introduction]. Four diabetologists oversaw the outpatient clinic under almost the same policy.

For comparison of severe hypoglycemia frequency, records of severe hypoglycemia were extracted from medical records. The presence or absence of severe hypoglycemia was tabulated 12 months before the start of pumps and every 12 months after the start of pumps, and the presence or absence of severe hypoglycemia during the 12 months before the start of pumps and 25–36 months after treatment. Severe hypoglycemia was defined as “an event requiring another person’s assistance” [13]. Participants introduced to the pump less than 1 year after their first visit to our institution were excluded from this analysis.

For the third analysis, CGM data were obtained for the SAP group only. The CGM metrics were defined as the percentage of glucose data falling within the target range of 70–180 mg/dL as time in range (TIR), the percentage of data above the target range (> 180 mg/dL) as time above range (TAR), and the percentage of data below the target range (< 70 mg/dL) as time below range (TBR) [14]. The percentage of active CGM time was calculated by dividing the number of CGM data points by 288 × number of days. The first month of SAP was excluded from the analysis because CGM may have started a few days after the start of insulin pumps in the outpatient introduction protocol at our institution. The first year was considered 1 month from 1 month after the start of SAP, the second year was considered 1 month from 13 months after the start of SAP, and the third year was considered 1 month from 25 months after the start of SAP to test whether years 2–3 changed compared to the first year. For this analysis, participants whose highest percentage of active CGM time was less than 50% in each year were excluded, because they were considered to have little CGM use.

For the factor analysis of participants who achieved HbA1c < 7%, we investigated factors associated with achieving HbA1c < 7% after 1 or 3 years of treatment for participants whose HbA1c was ≥ 7% before starting the insulin pumps. The following items were tested: treatment group (SAP or CSII), sex, estimated glomerular filtration rate (eGFR), body weight, HbA1c at initiation, total/basal/bolus insulin dose at MDI, age, age of onset, duration of diabetes, presence of severe hypoglycemia in the previous year, and percentage of active CGM time in the first and third years.

Baseline characteristics are presented as mean ± standard deviation or median (interquartile range) for continuous variables. Comparisons between the SAP and CSII groups were made by t-test, and categorical variables were compared by Fisher’s exact test. Changes in continuous variables over time and between-group comparisons were performed using mixed-effects models for repeated measures (MMRM). The model was adjusted for age, eGFR, insulin dose before pump introduction, and retinopathy as fixed covariates. Multiplicity was corrected using the Dunnett–Hsu method with pre-start as a control. Results are shown as least square mean ± SE. The percentage of active CGM time and CGM metrics were calculated using MMRM with the value in the first year as the covariate, corrected by the Dunnett–Hsu method with the value in the first year as the control. McNemar’s test was performed to test for changes in the proportion of individuals experiencing severe hypoglycemia and those who achieved HbA1c < 7%. Statistical analyses were performed using JMP Student Edition 18.2.2 (JMP Statistical Discovery LLC, Cary, NC, USA) and SAS version 9.4 (SAS Institute Inc.).

The study was conducted in accordance with the Declaration of Helsinki. This study was a retrospective observational study, and the requirement for informed consent was waived. Instead, the institution offered participants the opportunity to opt-out. This study was approved by the Ethics Committee of Tokyo Women's Medical University (Date of approval: January 10th, 2023; Approval No. 5535).