Endocrine Abstracts

Introduction: Lipohypertrophy (LH) is a chronic complication of diabetes mellitus that caused by frequent subcutaneous injections of insulin. Nowadays, LH were modified due to good quality modern insulin and expansion their concentration. As a result, some difficulties of diagnosis were appeared.

Design: The aim of this study has been to develop the estimation risk model of insulin induced LH in diabetic patients. This study was done on 140 diabetic patients (89 females and 51 males) who had been under the treatment with insulin a mean 8 years. Observation and palpation techniques, as well as ultrasonography of subcutaneous fat were used in assessing LH in these diabetics. All patients were divided into two groups. First group included 117 patients with LH, second – 23 diabetics without LH. Further, all LH risk factors were statistically processed using Spearman’s, Kendall tau, Gamma rank correlation coefficients and binary logistic regression. Results were statistically significant when P<0.05.

Results: All risk factors were analyzed using rank correlation coefficients on first stage. Statistically insignificant parameters were eliminated (P>0.05). Ten factors from 23 were remained after first stage. Further, 10 parameters were subjected to ROC-analysis. Measure AUC was determined. All risk factors had high predictive value (AUC>0.5). So, they were used to development the estimation risk model. On the basis of binary logistic regression the estimation risk model was created. Predictive value of model was 86% taking into account threshold cut-off 0.3 and confidence interval 95%. Efficacy of estimation risk model were tested on 34 diabetic patients.

Conclusions: Nowadays, LH remains severe insulinotherapy complication. Primary prevention is necessary for diabetic patients with pathologic areas of subcutaneous fat. Therefore, we developed the estimation risk model with good quality and high predictive value (86%) for diabetic patients who are under the treatment with insulin.