Key result
A clinical characteristic-based regression model accurately predicted individual basal insulin demand, with no significant difference from actual rates (MD -0.5 IU/d; 95% CI -1.5 to 0.6; P=0.46).
Why the study?
Basal rate profiles in type 1 diabetes on insulin pump therapy show enormous inter-individual heterogeneity, and predictive tools based on clinical characteristics may facilitate therapy.
Can clinical characteristics accurately predict individual basal rate profiles in adult patients with type 1 diabetes on insulin pump therapy?
Cohort (n=339)
Can clinical characteristics accurately predict individual basal rate profiles in adult patients with type 1 diabetes on insulin pump therapy?
A multivariate regression model based on clinical characteristics can accurately predict individual basal insulin rate profiles in adults with type 1 diabetes, potentially facilitating insulin pump therapy optimization.
May support model-guided basal insulin dosing; leaves open prospective validation before practice change.
Background: Basal rate profiles in patients with type 1 diabetes on insulin pump therapy are subject to enormous inter-individual heterogeneity. Tools to predict basal rates based on clinical characteristics may facilitate insulin pump therapy. Methods: Data from 339 consecutive in-patients with adult type 1 diabetes on insulin pump therapy were collected. Basal rate tests were performed over 24 hours. A mathematical algorithm to predict individual basal rate profiles was generated by relating the individual insulin demand to selected clinical characteristics in an exploratory cohort of 170 patients. The predicted insulin pump profiles were validated in a confirmatory cohort of 169 patients. Findings: Basal rates (0.27 ± 0.01 IU.d −1. kg −1 ) showed circadian variations with peaks corresponding to the “dawn” and “dusk” phenomena. Age, gender, duration of pump treatment, body-mass-index, HbA 1c , and triacylglycerol concentrations largely predicted the individual basal insulin demand per day (IU/d; exploratory vs prospective cohorts: r 2 = 0.518, P < .0001). Model-predicted and actual basal insulin rates were not different (exploratory cohort: Δ 0.1 (95% CI −0.9; 1.0 U/d; P = .95; prospective cohort: Δ −0.5 (95% CI −1.5; 0.6 IU/d; P = .46). Similarly, precise predictions were possible for each hour of the day. Actual and predicted “dawn” index correlated significantly in the exploratory but not in the confirmatory cohort. Interpretation: Clinical characteristics predict 52% of the variation in individual basal rate profiles, including their diurnal fluctuations. The multivariate regression model can be used to initiate or optimize insulin pump treatment in patients with type 1 diabetes.
No takes yet. Share an insight, caveat, or question.
Nauck et al. (2020) conducted a cohort in Type 1 diabetes (n=339). Clinical characteristics-based prediction model vs. Actual basal insulin rates was evaluated on Difference between model-predicted and actual basal insulin rates in the prospective cohort (MD -0.5 IU/d, 95% CI -1.5 to 0.6, p=0.46). A clinical characteristic-based regression model accurately predicted individual basal insulin demand, with no significant difference from actual rates (MD -0.5 IU/d; 95% CI -1.5 to 0.6; P=0.46).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: