Endocrine Abstracts

Background: Post-COVID-19 condition (long COVID-19, post-acute COVID-19, long-term effects of COVID-19) is an emerging health problem in people recovering from COVID-19 infection within the past 4-6 months. Patients with type 2 diabetes (T2D) are in the risk group for a more severe course of COVID-19 and the development of its complications.

Aim: to define the prevalence and prediction of post-COVID-syndrome development in patients with T2D according to retrospective analysis based on national survey in Ukraine.

Method: The retrospective analysis include data from 403 patients who suffered from COVID-19 infection in different regions on Ukraine. Among these patients, 168 (41.7%) developed post-COVID syndrome and 235 (58.3%) reported it absence. Patients were asked to fill out specially developed questionnaires for the purpose of retrospective assessment of the main parameters. The questionnaires included the following information: anthropometric indicators, year of diagnosis ofT2D, existing T2D complications history of COVID-19, COVID-19 severity and treatment, post-COVID phenotype and symptoms, duration of post-COVID syndrome, hypoglycemic therapy, levels of HbA1C, lipids and basic biochemical indicators. The stepwise multivariate logistic regression and PNN (Probabilistic Neural Network) models were used to select independent risk factors. The ROC curve analysis was used to assess the accuracy of models.

Results: As a result of the selection, 8 independent factor associated with the risk of post-COVID development in T2D patients were selected: treatment of COVID-19 with steroids (OR 1.73, 95% CI 1.06 – 2.84; P=0.029), remdesmevir (all patient presented), mechanical ventilation (OR 36.9, 95% CI 4.2 – 322; P=0.001), myocardial infarction (OR 2.68, 95% CI 1.3 – 5.5; P=0.007) and stroke (OR 4.18, 95% CI 1.79 – 9.75; P=0.001) in anamnesis, T2D duration (OR 0.94, 95% CI 0.9 – 0.97; P=0.001), commbination of anti-diabetic drugs with insulin (OR 2.98, 95% CI 1.35 – 6.58; P=0.007) and used of insulin analogues (OR 2.94, 95% CI 1.34 – 6.48; P=0.007). It should be noted that the indicators. In ROC analysis 8-factorial model constructed with PNN AUROC 0.831; 95% CІ 0.791–0.866) significantly better predict post-COVID syndrome development as compared multiple regression model (AUROC 0.759; 95% CІ 0.714–0.800; P=0.004)

Conclusion: Treatment of COVID-19 with steroids, remdesmevir, mechanical ventilation, myocardial infarction and stroke in anamnesis, T2D duration, commbination of anti-diabetic drugs with insulin and use of insulin analogues are the main predictors of post-COVID development. The prediction of post-COVID with PNN are clearly accurate as compared to step-wise logistic regression model.