A total of 198 patients switched to AHCL therapy, 99 (50%) patients to Minimed 780G, 16 (8%) to Diabeloop and 83 (42%) to Tandem CIQ.

Among these 99 using Minimed 780G, 6 came from traditional CSII, 10 from MDI, 41 from SAP and 42 from HCL. Of the 16 using Diabeloop, 15 came from CSII and 1 from HCL. Of the 83 using Tandem CIQ, 38 came from CSII, 10 from MDI, 34 from SAP and 1 from HCL. These data are shown in Table 1.

Table 2 summarizes the observed values of HbA1c and CGM metrics at different time points for the entire study cohort, along with the percentage of patients simultaneously achieving the targets of HbA1c < 7%, TIR > 70%, and TBR < 4%.

A month after the switch (T1 timepoint), there was a significant improvement in CGM metrics and HbA1c for all patients, even after adjusting for age, gender, disease duration, and previous insulin strategies: GRI −10.7, GMI −0.27%, CV −2.1%, TAR>250 −3.7%, TAR180-250 −5.6%, TIR + 9.7%, HbA1c −0.54% (all p < 0.001, Table 3).

The improvements observed after the switch were maintained throughout the observational period. At 3 months following the switch, significant improvements were observed: GRI −10.7, GMI −0.25%, CV −2.3%, TAR>250 −3.2%, TAR180-250 −5.4%, TIR + 9.2%, HbA1c −0.55% (all p < 0.001). At T3, the variation in TBR54-69 also became significant (−0.43%, p = 0.007). These improvements persisted at 6 months: GRI −9.5, GMI −0.22%, CV −1.6%, TAR>250 −3.2%, TAR180-250 −4.9%, TIR + 8.5%, HbA1c −0.57% (all p < 0.001), and TBR54-69 −0.37% (p = 0.02). At 12 months, the favourable trends continued: GRI −9, GMI −0.18%, CV −1.6%, TAR>250 −2.6%, TAR180-250 −4.6%, TIR + 7.9%, HbA1c −0.52%, and TBR54-69 −0.62% (all p < 0.001). For TBR<54, no significant improvements were detected at any timepoint.

No significant gender differences have been evidenced. BMI did not significantly change after the switch. At the T0 timepoint, the percentage of overweight and obese females was 29.5% and 15.2%, respectively; these percentages did not change significantly at T12 (26.1% and 16.2%, respectively). For males, the percentages of overweight and obese individuals were 44.4% and 17.3% at T0, and 46% and 13.5% at T12 (no statistically significant change from T0 to T12). At T0 and T12, the weight of females was 66.6 ± 14.5 kg and 67.3 ± 15.4 kg, respectively (p = 0.12), while for males it was 81.9 ± 13.9 kg and 81.1 ± 13.3 kg (p = 0.54).

An analysis of the association between GRI and CGM metrics revealed a strong and statistically significant association at T3 and T6 for: GRI and TAR>250 (R = 0.86, p < 0.001), GRI and TAR180-250 (R = 0.77, p < 0.001), and GRI and TIR (R = -0.94, p < 0.001). However, no significant association emerged between GRI and TBR54-69 or TBR<54 (with R values of 0.09, p = 0.22, and 0.31, p < 0.001, respectively).

Figure 1 shows the statistical distribution of HbA1c, GRI, GMI, and CV at the study time points for the entire cohort, using boxplots. Figure 2 illustrates the percentages for TAR, TIR, and TBR with the barplot. These visualizations provide a comprehensive overview of the key metrics and their variations over the study period.

Boxplot showing the distribution of HBA1c, GRI, GMI, CV in all patients, over the entire study period

TAR, TIR, TBR in all patients over the entire study period

Table 3 summarizes the variations in CGM metrics and HbA1c after the switch, comparing the Minimed 780G, Diabeloop, and Tandem CIQ devices at the T1, T3, T6, and T12 time points (1, 3, 6, and 12 months).

The comparison showed that at T1, the Diabeloop device significantly outperformed the others in terms of GRI, GMI, CV, TAR>250. Specifically, the Diabeloop group exhibited a mean GRI reduction of 20.6, compared to a reduction of 9 and 11.1 in the 780 and Tandem groups, respectively (P = 0.032). The GMI in the Diabeloop group decreased by 0.47%, versus 0.36% in the Minimed 780G group and 0.17% in the Tandem group (p = 0.008). Similarly, CV decreased by 7.3% with Diabeloop, in contrast to decreases of 0.7% and 2.7% in the 780 and Tandem groups, respectively (p < 0.001). The mean reduction in TAR>250 was 9.5% for Diabeloop, compared to 2.5% for 780 and 4.2% for Tandem (p = 0.017). Furthermore, at T1, the Minimed 780G device recorded the highest average decrease in TAR180-250 at 7.2% (compared to 4.7% for Diabeloop and 4% for Tandem, p = 0.020), while Tandem achieved the most significant reduction in TBR54-69 at −0.79 (+ 0.24 for 780 and −0.34 for Diabeloop, p = 0.004).

Beyond T1, the previously observed differences in performance improvements for GRI, GMI, TAR>25 and TBR54-69 across the three devices were no longer evident. Diabeloop continued to demonstrate statistically significant better CV reductions at the later assessments of T3 and T6 (p = 0.002 and p = 0.005, respectively) and a significantly better TAR>250 at T3 (p = 0.018). The superiority of the Minimed 780G device in reducing TAR180-250 continued to be evident at T3 and T6 (p = 0.021 and p = 0.037, respectively).

The HbA1c reductions achieved by the three devices evidenced no statistical differences at 1, 3, 6, and 12 months.

Figure 3 shows the distribution of HbA1c, GRI, GMI, and CV at post-switch timepoints, stratified by AHCL therapy. Figure 4 shows the distribution of these metrics over the entire study period, stratified by prior insulin therapy. Supplementary Tables S1 and S2 present mean values and standard deviations of HbA1c and CGM metrics at all time points, stratified by post-switch AHCL devices and by the four pre-switch insulin therapies, respectively.

Boxplots showing the distribution of HbA1c, GRI, GMI, and CV at post-switch timepoints, stratified by AHCL therapy

Boxplots showing the distribution of HbA1c, GRI, GMI, and CV, stratified by prior insulin therapy, over the entire study period

The evaluation of acute and chronic complications one year before and one year after the intervention was conducted. Acute complications included severe hypoglycemia and ketoacidosis, whereas chronic complications included arterial disease, nephropathy, neuropathy, retinopathy, and cardiopathy.

Acute complications were assessed both during the study period and throughout the entire available clinical history recorded at our hospital before the switch.

When comparing the 12 months pre- and 12 months post-change of therapy, no differences in acute complications (severe hypoglycemia 1 vs 1, ketoacidosis 1 vs 1) were noted. When the entire clinical history before the switch was compared with the one year following the switch, a notable decrease was observed in cases of severe hypoglycemia, dropping from 54 to 1 (p < 0.001), and in cases of ketoacidosis, reducing from 11 to 1 (p = 0.011). The singular case of severe hypoglycemia after the switch was linked to a sensor error, and the ketoacidosis incident to a pump cannula blockage, indicating no post-AHCL system activation complications were due to algorithm faults.

In terms of chronic complications, there was no significant change in the prevalence of arterial disease, nephropathy, neuropathy, retinopathy, or cardiopathy following the switch.