Technological advancements in diabetes has improved treatment possibilities. The big amount of data obtained from continuous glucose measurement (CGM) devices gave new investigation perspectives of blood glucose variation and other parameters connected to diabetic treatment. The aim of the paper is to investigate the relationship between age, duration of diabetes and sex of the patient with blood glucose levels measured with continuous glucose monitor. Blood glucose level was characterized with the relative time spent in hyper- and hypoglycaemia, central tendency (mean and median) and dispersion (standard deviation and interquartile range). These parameters were measured on n=58 Hungarian children suffering from type 1 diabetes mellitus. The children were poorly controlled, with non-blinded continuous glucose monitor performed to optimize therapy. The univariate association between the aforementioned parameters was investigated with scatterplots (with non-parametric smoothing), linear and Spearman-p correlation coefficients. Multivariate structure was investigated with canonical correlation analysis. It was found that age and duration of diabetes was not significantly associated with any descriptor of blood glucose, but hemoglobin Alc (HbAlc) was positively associated with time spent in hyperglycaemia (but not with hypoglycaemia), with central tendency and with dispersion. The association with dispersion is likely a spurious relationship induced by the central tendency (as a confounder). We observed that the associations of the descriptors with HbAlc are non-monotone, with a break in the positive correlation at 10%. We hypothesize that this is caused by the poor compliance of the patients, i.e. those in worse state increasingly “cheat” during CGM measurements, that is, they pay more attention to their blood glucose levels than what is usually done when no CGM measurement is present; therefore breaking the expected positive, monotone association.