Risk factors and prediction of hypoglycaemia using the Hypo-RESOLVE cohort: a secondary analysis of pooled data from insulin clinical trials

Aims/hypothesis The objective of the Hypoglycaemia REdefining SOLutions for better liVES (Hypo-RESOLVE) project is to use a dataset of pooled clinical trials across pharmaceutical and device companies in people with type 1 or type 2 diabetes to examine factors associated with incident hypoglycaemia events and to quantify the prediction of these events. Methods Data from 90 trials with 46,254 participants were pooled. Analyses were done for type 1 and type 2 diabetes separately. Poisson mixed models, adjusted for age, sex, diabetes duration and trial identifier were fitted to assess the association of clinical variables with hypoglycaemia event counts. Tree-based gradient-boosting algorithms (XGBoost) were fitted using training data and their predictive performance in terms of area under the receiver operating characteristic curve (AUC) evaluated on test data. Baseline models including age, sex and diabetes duration were compared with models that further included a score of hypoglycaemia in the first 6 weeks from study entry, and full models that included further clinical variables. The relative predictive importance of each covariate was assessed using XGBoost’s importance procedure. Prediction across the entire trial duration for each trial (mean of 34.8 weeks for type 1 diabetes and 25.3 weeks for type 2 diabetes) was assessed. Results For both type 1 and type 2 diabetes, variables associated with more frequent hypoglycaemia included female sex, white ethnicity, longer diabetes duration, treatment with human as opposed to analogue-only insulin, higher glucose variability, higher score for hypoglycaemia across the 6 week baseline period, lower BP, lower lipid levels and treatment with psychoactive drugs. Prediction of any hypoglycaemia event of any severity was greater than prediction of hypoglycaemia requiring assistance (level 3 hypoglycaemia), for which events were sparser. For prediction of level 1 or worse hypoglycaemia during the whole follow-up period, the AUC was 0.835 (95% CI 0.826, 0.844) in type 1 diabetes and 0.840 (95% CI 0.831, 0.848) in type 2 diabetes. For level 3 hypoglycaemia, the AUC was lower at 0.689 (95% CI 0.667, 0.712) for type 1 diabetes and 0.705 (95% CI 0.662, 0.748) for type 2 diabetes. Compared with the baseline models, almost all the improvement in prediction could be captured by the individual’s hypoglycaemia history, glucose variability and blood glucose over a 6 week baseline period. Conclusions/interpretation Although hypoglycaemia rates show large variation according to sociodemographic and clinical characteristics and treatment history, looking at a 6 week period of hypoglycaemia events and glucose measurements predicts future hypoglycaemia risk. Graphical Abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-024-06177-6.


Medical history definitions
Medical history events of interest were defined as shown in ESM Table 1.

Concomitant Medications definitions
Concomitant medications were defined as shown in ESM Table 2.

XGBoost
A XGBoost model forms its prediction as the sum of a number of predictive trees where each tree provides a given prediction via a series of sequential decision rules with each decision rule based on a single covariate.The result of a decision rule in a tree determines which decision rule is applied next to the data.The process continues until the last rule in the given sequence is reach and a predictive value conditioned on the decisions is given.When fitting a decision tree, a new decision rule is only added to the tree if it provides sufficient predictive gain.Given a covariate can appear in multiple decision rules within a single tree and can appear in multiple trees, the average predictive gain of all decision rules using the covariate is used as a measure of the importance of that covariate.XGBoost easily fits non-linear effects of covariates and interaction effects between covariates without requiring explicit user-intervention in modelling.Due to differences in study populations it is reasonable to expect that effect sizes of common risk factors may be different within different trials for example.However, XGBoost models each 6 week time slice for an individual as independent.
Grid search was performed of the following parameters of XGBoost: eta (values: 1.0, 0.1, 0.01, and 0.001), gamma (values: 1.0, 0.1, and 0.001), and max_depth (values: 2, 4, 8, and 16).Association with Antibiotics, Anti-thyroid, and Blood glucose lowering medications were excluded from this analysis as numbers of observations were low.Only a single type 1 diabetes study contained Human only insulin, and this study recorded no severe hypoglycaemia events and so no association was estimated in this case.Blood glucose is self-monitoring blood glucose For continuous covariates this is the increase in hypoglycaemia rate for every standard deviate change in covariate for the first 6 weeks of the study and for categorical covariates it is the increase in hypoglycaemia rate with respect to the reference category.In both cases adjusted for age, sex, diabetes duration and study identifier as fixed effects, and individual identifier as random effect.* Associations where the confidence interval does not cross 1, and are therefore significant (P < 0.05), are highlighted in bold.

ESM
ESM Table 10: Type 2 diabetes minimally-adjusted associations of baseline covariates without imputation with hypoglycaemia events across the trial duration.Association with Antibiotics and Anti-thyroid medications were excluded from this analysis as number of observations were low.Blood glucose is self-monitoring blood glucose Level 3 events were low in Human+Analogue insulin leading to wide confidence intervals when estimating associations with insulin origin.For continuous covariates this is the increase in hypoglycaemia rate for every standard deviate change in covariate for the first 6 weeks of the study and for categorical covariates it is the increase in hypoglycaemia rate with respect to the reference category.In both cases adjusted for age, sex, diabetes duration and study identifier as fixed effects, and individual identifier as random effect.* Associations where the confidence interval does not cross 1, and are therefore significant (P < 0.05), are highlighted in bold.

Table 4 :
Candidate covariates for type 1 diabetes Level 2 or worse hypoglycaemic event analysis .

Table 9 :
Type 1 diabetes minimally-adjusted associations of baseline covariates without imputation with hypoglycaemia events across the trial duration.