Predicting the efficacy of anti-inflammatory therapy in COVID-19

Predicting the efficacy of anti-inflammatory therapy in COVID-19

I.M. Sukhomlinova^1,2, I.G. Bakulin¹, M.Yu. Kabanov^1,2 

¹North-Western State Medical University named after I.I. Mechnikov, St. Petersburg

²St. Petersburg Hospital for War Veterans, St. Petersburg

Background: in the 20^th century, the development of prognosis methods became an integral part of science and the result of the combined experience in the world scientific community. Proactive administration of anti-inflammatory drugs (dexamethasone, olokizumab, baricitinib and tocilizumab) has become the leading direction of pathogenetic therapy. Nevertheless, the issue of predicting the therapy outcomes with these drugs is poorly understood.

Aim: to form the methods for predicting the efficacy of anti-inflammatory therapy (AIT) with dexamethasone, baricitinib, olokizumab and tocilizumab by multiple regression analysis, decision tree (neural network) and their clinical testing.

Patients and Methods: a retrospective analysis of 229 cases concerning severe and moderate course of COVID-19 requiring AIT was conducted. 2 methods of predicting the therapy outcome have been formed (multiple regression analysis and neural network in python 3.8.10). In order to test the created methods, a prospective comparative study was conducted with the inclusion of 108 patients.

Results: the dependence of AIT outcomes on age-sex characteristics, comorbidity and disease severity, data from instrumental and laboratory studies, anamnesis and objective examination of patients by the time of the AIT initiation was established. A computer program has been developed based on a decision tree (neural network) "Index for predicting the efficacy of anti-inflammatory therapy in COVID-19» and «Method for predicting the risk of fatal outcome in patients with severe and moderate COVID-19 during proactive AIT" (multiple regression analysis), with a computer program developed on its basis «Assessment prognosis of the efficacy of anti-inflammatory therapy in COVID-19» (multiple regression analysis). During retrospective and prospective studies, it was found that the use of dexamethasone as monotherapy for severe COVID-19 is impractical due to its low efficacy. In the absence of positive trend of the patient’s condition during therapy with tocilizumab, olokizumab, it is necessary to consider the possibility of repeated administration of anticytokine drugs for vital indications due to their significant effect on reducing the severity progression of the disease (p<0.00001). In the course of a prospective study, it was revealed that in conditions of changes in the disease course or methods to pathogenetic treatment, a program based on a decision tree seems more promising.

Conclusion: the development of artificial intelligence and neural networks allows to significantly improve the quality of predictive models, especially in clinical practice in the absence of standards or clinical recommendations.

Keywords: COVID-19, efficacy prediction, anti-inflammatory therapy, dexamethasone, olokizumab, baricitinib, tocilizumab.

For citation: Sukhomlinova I.M., Bakulin I.G., Kabanov M.Yu. Predicting the efficacy of anti-inflammatory therapy in COVID-19. RMJ. 2023;2:72–76.