Assessing the relationship between markers of glycemic control through flexible copula regression models

Glycated haemoglobin (HbA1c) is a sensitive marker of blood glucose in patientswith diabetes. However, levels can vary considerably, even among individualswith similar mean blood glucose concentrations. Other glycated proteins, such asfructosamine, can also act as blood sugar markers, but estimating HbA1c and fruc-tosamine via independent models may lead to errors of interpretation regardingdisease severity. From a clinical standpoint, it would be of great interest to knowthe factors that affect the mean concentration of both HbA1c and fructosamine,that influence the variability in the concentrations of these glycated markers, andthat cause HbA1c/fructosamine discordance. Flexible models are required that illus-trate the behaviour of these variables as well as the association between them. Thepresent work reviews existing models that might serve in this regard. Flexible cop-ula regression models using P-splines, were used to provide a better understandingof the behaviour of both glycated proteins, and the relationship between them underthe possible influence of different covariates. This work shows the usefulness ofthis type of models in practice, and provides a basis for its clinical interpretation bymeans of an understandable case study. Ultimately, to better understand the effectsof each continuous covariate, they were represented at the true scale of the response variables