Validating Preclinical Assessment of a Medicine Benefit-Risk Ratio Based on WOE and IV Models

INTRODUCTION. Mathematical models are actively used in biomedical research, including efficacy and safety prediction of medicinal products. Scientific Centre for Expert Evaluation of Medicinal Products has developed and implemented a method for preclinical benefit-risk assessment of medicinal products. The method is based on the binary classification of variables used to calculate WoE (Weight of Evidence) and IV (Information Value) predictors. Calculation algorithms for WoE and IV are based on Bayesian model of prior probability, which allows for risk prediction and informed decision-making regarding pharmacotherapy and its ability to reduce the geno­toxic and embryotoxic effects of environmental teratogens. However, confirmed predictive ability warrants pharmacological validation of a potential corrector and the predictive system computation quality.

AIM. This study aimed to validate preclinical assessment method of the benefit-risk ratio by correcting reprotoxic effects of peat smoke in rats with pharmacotherapy as a case study.

MATERIALS AND METHODS. The study used experimental and statistical analysis. A method developed by Scientific Centre for Expert Evaluation of Medicinal Products was used to confirm predictive significance of preclinical benefit-risk assessment for pharmacotherapeutic correction of peat smoke-induced embryotoxic effects. In order to validate benefit-risk assessment mathematical model and the corrective pharmacological properties of fabomotizole in genotoxicity and embryotoxicity models induced by peat smoke in rats, logistic regression and ROC analysis methods were applied.

RESULTS. AUC analysis (0.701; 0.617–0.786) within pharmacological validation showed that fabomotizole corrective capacity in rats was predicted in the range from “unsatisfactory” to “good” (0.500; 0.239–0.761). This corresponds to WOE/IV estimates, from “low” to “moderate” weight (0.34; –0.99) and “weak” to “strong” information value (0.02; 0.45). Computation quality test for genotoxicity showed an AUC of 0.554 for the predicted probability and 0.432 for the predicted group (random guessing). For embryotoxicity, AUC was 0.701 and 0.782, indicating good predictive ability of the model.

CONCLUSIONS. The validation study has confirmed the predictive value of WoE and IV. The benefit-risk model based on Bayesian prior probability has shown high convergence with ROC analysis in assessing genotoxicity and embryotoxicity of peat smoke, as well as corrective capacity of fabomotizole.

1. Shreder OV, Bunyatyan ND, Goryachev DV, et al. Mathematical prediction of the efficacy of medicinal products in preclinical studies. Bulletin of the Scientific Centre for Expert Evaluation of Medicinal Products. Regulatory Research and Medicine Evaluation. 2022;12(3):315–30 (In Russ.). https://doi.org/10.30895/1991-2919-2022-12-3-315-330

2. Engalycheva GN, Syubaev RD. Development of paediatric medicines: Key risk factors and non-clinical research programmes. Bulletin of the Scientific Centre for Expert Evaluation of Medicinal Products. Regulatory Research and Medicine Evaluation. 2023;13(1):14–26 (In Russ.). https://doi.org/10.30895/1991-2919-2023-500

3. Thomas RS, Paules RS, Simeonov A, et al. The US Federal Tox21 Program: A strategic and operational plan for continued leadership. ALTEX. 2018;35(2):163–8. https://doi.org/10.14573/altex.1803011

4. Vanslette K, Tohme T, Youcef-Toumi K. A general model validation and testing tool. Reliab Eng Syst Safety. 2020;195:106684. https://doi.org/10.1016/j.ress.2019.106684

5. Guzman-Castillo M, Brailsford S, Luke M, Smith H. A tutorial on selecting and interpreting predictive models for ordinal health-related outcomes. Health Serv Outcomes Res Method. 2015;15:223–40. https://doi.org/10.1007/s10742-015-0140

6. 6 Stefaniak F. Prediction of compounds activity in nuclear receptor signaling and stress pathway assays using machine learning algorithms and low-dimensional molecular descriptors. Front Environ Sci. 2015;3:77. https://doi.org/10.3389/fenvs.2015.00077

7. Solomina AS, Schreder OV, Mokrova ED, et al. Fabomotisol as a means of pharmacological correction of pre- and postnatal disorders induced by wood and peat smoke, tobacco smoke, alcohol, hyperglycemia and cyclophosphamide in rat offspring. Molecular Medicine. 2020;18(3):34–48 (In Russ.). https://doi.org/10.24411/2587-7836-2018-10008

8. Shreder OV, Shreder ED, Durnev AD, Seredenin SB. Association of genotoxic and teratogenic effects induced by cyclophosphamide and their modification with afobazole. Hygiene and Sanitation. 2011;(5):64–8 (In Russ.). EDN: OHIKNH

9. Gorbatova DM, Nemova EP, Solomina AS, Durnev AD. Prenatal effects of peat combustion products and their correction with aphobazol in rat offspring. Bulletin of Experimental Biology and Medicine. 2014;158(11):604–5 (In Russ.). EDN: TEWYEB

10. `10. Gorbatova DM, Zhanataev AK, Nemova EP, Durnev AD. DNA damage in placenta cells and rat embryos exposed to peat smoke; antigenotoxic effect of afobazole. Environmental Genetics. 2016;14(2):50–6 (In Russ.). https://doi.org/10.17816/ecogen14250-56

11. Zabelin AV, Pyhalov AA. Finite-element model validation and its execution algorithm. PNRPU Mechanics Bulletin. 2017;(3):216–33 (In Russ.). https://doi.org/10.15593/perm.mech/2017.3.13