Abstract
Background
Pump basal-rate reduction during Ramadan may prevent hypoglycemia but can increase hyperglycemia. Its dose-response relation and phenotype-dependent trade-off remain uncertain.
Methods
Thirty UVA/Padova virtual patients (10 children, 10 adolescents, 10 adults) underwent paired 30-day Ramadan conditions delivering 100%, 90%, 80%, 70%, or 60% of the pump steady-state basal rate. The primary complete-case analysis included 29 virtual patients. Outcomes were time in range (TIR), time below range (TBR), time above range (TAR, including TAR
>
250 and TAR
>
300), and rescue corrections. Paired permutation tests and bootstrap intervals quantified effects. Robustness and revision analyses used three sensor seeds, three seeded meal/timing scenarios with a held-out evaluation, a reactive-correction sensitivity analysis, a time-of-day decomposition, terminal-event reporting, and a negative-control (sham) test of the pre-Ramadan marker.
Results
Mean TIR was 81.7%, 79.9%, 73.9%, 65.3%, and 56.0% at 100%, 90%, 80%, 70%, and 60%, respectively (
P
0.001). Compared with 100%, 90% basal reduced TBR by 1.6 percentage points (95% CI −2.1 to −1.2) while changing TIR by −1.8 points (95% CI −3.7 to −0.3; Holm-adjusted
P
= 0.031). Severe hyperglycemia rose steeply with deeper reduction (TAR
>
250: 2.9% at 100% to 14.8% at 60%). Enabling realistic reactive corrections raised TIR at the deepest policies but preserved the same policy ordering. Pre-Ramadan TBR stratified the TIR response across reduced policies (omnibus patient-blocked permutation
P
= 0.023), with the strongest gradients at 70% and 60% basal; a negative-control analysis showed this signal was not specific to TBR but reflected a broader baseline-glycemia pattern. In a held-out scenario, a revised safety-first selection rule improved hypoglycemia-target attainment (39% to 79%) and reduced TBR by 2.9 points but incurred a 5.2-point TIR cost, demonstrating that prioritizing fasting safety involves a glycemic trade-off.
Conclusion
Ramadan pump basal-rate reduction should be treated not as a uniform percentage change but as a phenotype- and context-dependent trade-off. Among the reduced policies, 90% basal produced the smallest TIR loss while lowering hypoglycemia; pre-Ramadan TBR was an exploratory candidate stratifier; and meal-size associations differed by policy. These mechanistic findings apply to simulated pump basal-rate modulation and require prospective, CGM-anchored validation before any clinical use.