The Prediction Model of Warfarin Individual Maintenance Dose for Patients Undergoing Heart Valve Replacement, Based on the Back Propagation Neural Network

  • Qian Li
  • Jing Wang
  • Huan Tao
  • Qin Zhou
  • Jie Chen
  • Bo Fu
  • WenZhe Qin
  • Dong Li
  • JiangLong Hou
  • Jin ChenEmail author
  • Wei-hong Zhang
Original Research Article


Background and Objective

Because of the narrow therapeutic window and huge inter-individual variation, the individual precision on anticoagulant therapy of warfarin is challenging. In our study, we aimed to construct a Back Propagation Neural Network (BPNN) model to predict the individual warfarin maintenance dose among Chinese patients who have undergone heart valve replacement, and validate its prediction accuracy.


In this study, we analyzed 13,639 eligible patients extracted from the Chinese Low Intensity Anticoagulant Therapy after Heart Valve Replacement database, which collected data on patients using warfarin after heart valve replacement from 15 centers all over China. Ten percent of patients who were finally enrolled in the database were used as the external validation, while the remaining were randomly divided into the training and internal validation groups at a ratio of 3:1. Input variables were selected by univariate analysis of the general linear model; 2.0, the mean value of the international normalized ratio (INR) range 1.5–2.5, was used as the mandatory variable. The BPNN model and the multiple linear regression (MLR) model were constructed by the training group and validated through comparisons of the mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), and ideal predicted percentage.


Finally, 10 input variables were selected and a three-layer BPNN model was constructed. In the BPNN model, the value of MAE (0.688 mg/day and 0.740 mg/day in internal and external validation, respectively), MSE (0.580 mg/day and 0.599 mg/day in internal and external validation, respectively), and RMSE (0.761 mg/day and 0.774 mg/day in internal and external validation, respectively) were achieved. Ideal predicted percentages were high in both internal (63.0%) and external validation (59.7%), respectively. Compared with the MLR model, the BPNN model showed a higher ideal prediction percentage in the external validation group (59.7% vs. 56.6%), and showed the best prediction accuracy in the intermediate-dose subgroup (internal validation group: 85.2%; external validation group: 84.7%) and a high predicted percentage in the high-dose subgroup (internal validation group: 36.2%; external validation group: 39.8%), but poor performance in the low-dose subgroup (internal validation group: 0%; external validation group: 0.3%). Meanwhile, the BPNN model showed better ideal prediction percentage in the high-dose group than the MLR model (internal validation: 36.2% vs. 31.6%; external validation: 42.8% vs. 37.8%).


The BPNN model shows promise for predicting the warfarin maintenance dose after heart valve replacement.

Plain Language Summary

Because of the narrow therapeutic window and huge inter-individual variation, the individual precision on anticoagulant therapy of warfarin is still a challenge. According to the rapid development of artificial intelligence, our study was based on a clinical big database and used the advanced algorithm—Back Propagation Neural Network—to construct a prediction model of warfarin maintenance dose. It showed a high prediction accuracy of over 59%, and manifested obvious improvement of the prediction ability in the high-dose group. Hence, the BPNN model shows promise for predicting the precise individual therapy of warfarin.



The authors appreciated the other members who provided their generous contributions during the study.

Author contributions

QL and JW were co-first authors and were responsible for interpretation and analysis of the data, and drafting of the manuscript. HT and QZ were responsible for analysis of the data. JC and WQ were responsible for interpretation of the research. BF, JH and LD performed the conception work, and data acquisition. WHZ was responsible for revising the language for the content and verifying the design of the research. JC, as the corresponding author, was responsible for the conception and design of the research, and revising critically for important intellectual content. All authors approved the final version submitted for publication.

Compliance with Ethical Standards


This project was supported by the National Natural Science Foundation (Project Numbers 71974137 and 81641021) and the National Science and Technology Pillar Program during the Twelfth Five-Year Plan Period (Project Number 2011BAI11B18). The funding sources have no role in the design, implementation, data analysis, and article writing, or in the decision to submit the article for publication.

Conflict of interest

Qian Li, Jing Wang, Huan Tao, Qin Zhou, Jie Chen, Bo Fu, WenZhe Qin, Dong Li, JiangLong Hou, Jin Chen, and Wei-hong Zhang declare that they have no conflicts of interests.

Ethics approval

All methods and study protocols have been approved by the Ethics Committee of West China Hospital of Sichuan University (ChiECRCT-201792). As this was a retrospective study, as per the ethical approval documents.

Informed consent

Informed consent has been exempted.


  1. 1.
    Lim WY, Lloyd G, Bhattacharyya S. Mechanical and surgical bioprosthetic valve thrombosis. Heart. 2017;103(24):1934–41.PubMedGoogle Scholar
  2. 2.
    Egbe AC, Pislaru SV, Pellikka PA, et al. Bioprosthetic valve thrombosis versus structural failure: clinical and echocardiographic predictors. J Am Coll Cardiol. 2015;66(21):2285–94.PubMedCrossRefGoogle Scholar
  3. 3.
    Puri R, Auffret V, Rodes-Cabau J. Bioprosthetic valve thrombosis. J Am Coll Cardiol. 2017;69(17):2193–211.PubMedCrossRefGoogle Scholar
  4. 4.
    Osnabrugge RL, Mylotte D, Head SJ, et al. Aortic stenosis in the elderly: disease prevalence and number of candidates for transcatheter aortic valve replacement: a meta-analysis and modeling study. J Am Coll Cardiol. 2013;62(11):1002–12.PubMedCrossRefGoogle Scholar
  5. 5.
    Bove T, Van Belleghem Y, Francois K, et al. Low target-INR anticoagulation is safe in selected aortic valve patients with the Medtronic Open Pivot mechanical prosthesis: long-term results of a propensity-matched comparison with standard anticoagulation. Interact Cardiovasc Thorac Surg. 2017;24(6):862–8.PubMedCrossRefGoogle Scholar
  6. 6.
    Di Minno G, Russolillo A, Gambacorta C, Di Minno A, Prisco D. Improving the use of direct oral anticoagulants in atrial fibrillation. Eur J Intern Med. 2013;24(4):288–94.PubMedCrossRefGoogle Scholar
  7. 7.
    Eikelboom JW, Connolly SJ, Brueckmann M, et al. Dabigatran versus warfarin in patients with mechanical heart valves. N Engl J Med. 2013;369(13):1206–14.PubMedCrossRefGoogle Scholar
  8. 8.
    Marcy TR, Truong T, Rai A. Comparing direct oral anticoagulants and warfarin for atrial fibrillation, venous thromboembolism, and mechanical heart valves. Consult Pharm. 2015;30(11):644–56.PubMedCrossRefGoogle Scholar
  9. 9.
    Nishimura RA, Otto CM, Bonow RO, et al. 2017 AHA/ACC Focused Update of the 2014 AHA/ACC Guideline for the Management of Patients With Valvular Heart Disease: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. J Am Coll Cardiol. 2017;70(2):252–89.PubMedCrossRefGoogle Scholar
  10. 10.
    Anderson J, Horne B, Stevens S, et al. Randomized trial of genotype-guided versus standard warfarin dosing in patients initiating oral anticoagulation. Circulation. 2007;116(22):2563–70.PubMedCrossRefGoogle Scholar
  11. 11.
    Klein TE, Altman RB, Eriksson N, et al. Estimation of the warfarin dose with clinical and pharmacogenetic data. N Engl J Med. 2009;360(8):753–64.PubMedCrossRefGoogle Scholar
  12. 12.
    Liu R, Li X, Zhang W, Zhou H-H. Comparison of nine statistical model based warfarin pharmacogenetic dosing algorithms using the racially diverse international warfarin pharmacogenetic consortium cohort database. PLoS One. 2015;10(8):e0135784.PubMedPubMedCentralCrossRefGoogle Scholar
  13. 13.
    Ugrinowitsch C, Fellingham GW, Ricard MD. Limitations of ordinary least squares models in analyzing repeated measures data. Med Sci Sports Exerc. 2004;36(12):2144–8.PubMedCrossRefGoogle Scholar
  14. 14.
    Tao Y, Chen YJ, Fu X, Jiang B, Zhang Y. Evolutionary ensemble learning algorithm to modeling of warfarin dose prediction for Chinese. IEEE J Biomed Health Inform. 2019;23(1):395–406.PubMedCrossRefGoogle Scholar
  15. 15.
    Dong J, Shi GH, Lu M, et al. Evaluation of the predictive performance of Bayesian dosing for warfarin in Chinese patients. Pharmacogenomics. 2019;20(3):167–77.PubMedCrossRefGoogle Scholar
  16. 16.
    Tao Y, Chen YJ, Xue L, Xie C, Jiang B, Zhang Y. An ensemble model with cluster assumption for warfarin dose prediction in Chinese patients. IEEE J Biomed Health Inform. 2019. (Epub 7 Jan 2019).PubMedCrossRefGoogle Scholar
  17. 17.
    Hamberg AK, Hellman J, Dahlberg J, Jonsson EN, Wadelius M. A Bayesian decision support tool for efficient dose individualization of warfarin in adults and children. BMC Med Inform Decis Mak. 2015;15:7.PubMedPubMedCentralCrossRefGoogle Scholar
  18. 18.
    Tao H, Li Q, Zhou Q, et al. A prediction study of warfarin individual stable dose after mechanical heart valve replacement: adaptive neural-fuzzy inference system prediction. BMC Surg. 2018;18(1):10.PubMedPubMedCentralCrossRefGoogle Scholar
  19. 19.
    Grossi E, Podda GM, Pugliano M, et al. Prediction of optimal warfarin maintenance dose using advanced artificial neural networks. Pharmacogenomics. 2014;15(1):29–37.PubMedCrossRefGoogle Scholar
  20. 20.
    Solomon I, Maharshak N, Chechik G, et al. Applying an artificial neural network to warfarin maintenance dose prediction. Isr Med Assoc J. 2004;6(12):732–5.PubMedGoogle Scholar
  21. 21.
    Dong L, Shi YK, Xu JP, et al. The multicenter study on the registration and follow-up of low anticoagulation therapy for the heart valve operation in China [in Chinese]. Zhonghua yi xue za zhi. 2016;96(19):1489–94.PubMedGoogle Scholar
  22. 22.
    Lenzini P, Wadelius M, Kimmel S, et al. Integration of genetic, clinical, and INR data to refine warfarin dosing. Clin Pharmacol Ther. 2010;87(5):572–8.PubMedPubMedCentralCrossRefGoogle Scholar
  23. 23.
    Steyerberg EW, Harrell FE Jr. Prediction models need appropriate internal, internal-external, and external validation. J Clin Epidemiol. 2016;69:245–7.PubMedCrossRefGoogle Scholar
  24. 24.
    Zahedi F. An introduction to neural networks and a comparison with artificial intelligence and expert systems. Interfaces. 1991;21(2):25–38.CrossRefGoogle Scholar
  25. 25.
    Carlucci D, Renna P, Schiuma G. Evaluating service quality dimensions as antecedents to outpatient satisfaction using back propagation neural network. Health Care Manag Sci. 2013;16(1):37–44.PubMedCrossRefGoogle Scholar
  26. 26.
    Sridevi K, Sivaraman E, Mullai P. Back propagation neural network modelling of biodegradation and fermentative biohydrogen production using distillery wastewater in a hybrid upflow anaerobic sludge blanket reactor. Bioresour Technol. 2014;165:233–40.PubMedCrossRefGoogle Scholar
  27. 27.
    Rumelhart DE, Hinton GE, Williams RJ. Learning internal representations by error propagation. Amsterdam: Elsevier Inc.; 1988.Google Scholar
  28. 28.
    Xiang LC, Xiao LH, Li M, et al. Diagnosis values of back propagation neural network integrating age, transrectal ultrasound characteristics and serum PSA for prostate cancer [in Chinese]. Sichuan da xue xue bao Yi xue ban. 2016;47(1):77–80 (84).PubMedGoogle Scholar
  29. 29.
    Sheiner LB, Beal SL. Some suggestions for measuring predictive performance. J Pharmacokinet Biopharm. 1981;9(4):503–12.PubMedCrossRefGoogle Scholar
  30. 30.
    Justice AC, Covinsky KE, Berlin JA. Assessing the generalizability of prognostic information. Ann Intern Med. 1999;130(6):515–24.PubMedCrossRefGoogle Scholar
  31. 31.
    Zhao W, Berger V. Imbalance control in clinical trial subject randomization-from philosophy to strategy. J Clin Epidemiol. 2018;101:116–8.PubMedCrossRefGoogle Scholar
  32. 32.
    Moons KG, Kengne AP, Grobbee DE, et al. Risk prediction models: II. External validation, model updating, and impact assessment. Heart. 2012;98(9):691–8.PubMedCrossRefGoogle Scholar
  33. 33.
    Dong L, Shi Y, Xu J, et al. The multicenter study on the registration and follow-up of low anticoagulation therapy for the heart valve operation in China. Natl Med J China. 2016;96(19):1489–94.Google Scholar
  34. 34.
    Gu Q, Kong Y, Schneede J, et al. VKORC1-1639G > A, CYP2C9, EPHX1691A > G genotype, body weight, and age are important predictors for warfarin maintenance doses in patients with mechanical heart valve prostheses in southwest China. Eur J Clin Pharmacol. 2010;66(12):1217–27.PubMedCrossRefGoogle Scholar
  35. 35.
    Lee VW, You JH, Lee KK, Chau TS, Waye MM, Cheng G. Factors affecting the maintenance stable warfarin dosage in Hong Kong Chinese patients. J Thromb Thrombolysis. 2005;20(1):33–8.PubMedCrossRefGoogle Scholar
  36. 36.
    Jie-qiong S, Shao-wen LIU, Song-wen C, et al. The relationships between anticoagulation therapy during operation and thromboembolic events in patients with persistent/permanent atrial fibrillation who underwent radiofrequency catheter ablation [in Chinese]. Chin J Cardiac Arrhythm. 2009;13(3):209–12.Google Scholar
  37. 37.
    Wang JT, Dong MF, Song GM, Ma ZS, Ma SJ. Combined low-dose aspirin and warfarin anticoagulant therapy of postoperative atrial fibrillation following mechanical heart valve replacement. J Huazhong Univ Sci Technol Med Sci. 2014;34(6):902–6.PubMedCrossRefGoogle Scholar
  38. 38.
    Chen L, Xiao Y, Ma R, et al. Bipolar radiofrequency ablation is useful for treating atrial fibrillation combined with heart valve diseases. BMC Surg. 2014;14:32.PubMedPubMedCentralCrossRefGoogle Scholar
  39. 39.
    Dong L, Shi Y, Tian Z. The follow-up of 12 pregnant women with anticoagulation therapy after mechanical heart valve replacement. Chin J Obstet Gynecol. 2001;36(8):465–7.Google Scholar
  40. 40.
    Li D, Xin-hui L, Ying-kang S, Er-yong Z, Teng D, Li XR. Low-intensity anticoagulation therapy in the pregnant women with mechanical heart valves:a report with 56 cases [in Chinese]. Chin J Thorac Cardiovasc Surg. 2011;27(1):8–10.Google Scholar
  41. 41.
    Lip GYH, Al-Saady N, Jin J, et al. Anticoagulation control in warfarin-treated patients undergoing cardioversion of atrial fibrillation (from the Edoxaban Versus Enoxaparin-Warfarin in Patients Undergoing Cardioversion of Atrial Fibrillation Trial). Am J Cardiol. 2017;120(5):792–6.PubMedCrossRefGoogle Scholar
  42. 42.
    Limdi NA, Beasley TM, Baird MF, et al. Kidney function influences warfarin responsiveness and hemorrhagic complications. J Am Soc Nephrol. 2009;20(4):912–21.PubMedPubMedCentralCrossRefGoogle Scholar
  43. 43.
    Hijazi Z, Hohnloser SH, Oldgren J, et al. Efficacy and safety of dabigatran compared with warfarin in relation to baseline renal function in patients with atrial fibrillation: a RE-LY (Randomized Evaluation of Long-term Anticoagulation Therapy) trial analysis. Circulation. 2014;129(9):961–70.PubMedPubMedCentralCrossRefGoogle Scholar
  44. 44.
    Kucuk M, Ozdemir R, Karacelik M, et al. Risk factors for thrombosis, overshunting and death in infants after modified blalock-taussig shunt. Acta Cardiol Sin. 2016;32(3):337–42.PubMedPubMedCentralGoogle Scholar
  45. 45.
    Shao-hui W, Liang Z, Wei-feng J, et al. Risk factors of atrial fibrillation recurrence in patients who obtained ablation endpoints with longstanding persistent atrial fibrillation [in Chinese]. Int J Cardiovasc Dis. 2014;41(04):268–70.Google Scholar
  46. 46.
    Sharabiani A, Bress A, Douzali E, Darabi H. Revisiting warfarin dosing using machine learning techniques. Comput Math Methods Med. 2015;2015:560108.PubMedPubMedCentralCrossRefGoogle Scholar
  47. 47.
    Ma Z, Wang P, Gao Z, Wang R, Khalighi K. Ensemble of machine learning algorithms using the stacked generalization approach to estimate the warfarin dose. PLoS One. 2018;13(10):e0205872.PubMedPubMedCentralCrossRefGoogle Scholar
  48. 48.
    Li Q, Tao H, Wang J, et al. Warfarin maintenance dose prediction for patients undergoing heart valve replacement—a hybrid model with genetic algorithm and back-propagation neural network. Sci Rep. 2018;8(1):9712.PubMedPubMedCentralCrossRefGoogle Scholar
  49. 49.
    Saffian SM, Duffull SB, Wright DFB. Warfarin dosing algorithms underpredict dose requirements in patients requiring ≥ 7 mg daily: a systematic review and meta-analysis. Clin Pharmacol Ther. 2017;102(2):297–304.PubMedCrossRefGoogle Scholar
  50. 50.
    Peng Q, Huang S, Chen X, et al. Validation of warfarin pharmacogenetic algorithms in 586 Han Chinese patients. Pharmacogenomics. 2015;16(13):1465–74.PubMedCrossRefGoogle Scholar
  51. 51.
    Centers for Medicare & Medicaid Services. Medicare coverage database—potential NCD topics.

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Qian Li
    • 1
  • Jing Wang
    • 2
  • Huan Tao
    • 3
  • Qin Zhou
    • 4
  • Jie Chen
    • 5
  • Bo Fu
    • 6
  • WenZhe Qin
    • 7
  • Dong Li
    • 8
  • JiangLong Hou
    • 8
  • Jin Chen
    • 1
    Email author
  • Wei-hong Zhang
    • 9
    • 10
  1. 1.Department of Evidence-Based Medicine and Clinical Epidemiology, West China HospitalSichuan UniversityChengduChina
  2. 2.Department of Career Development DivisionThe Fourth Affiliated Hospital of Anhui Medical UniversityHefeiChina
  3. 3.Department of Hematology, West China HospitalSichuan UniversityChengduChina
  4. 4.Department of NutritionThe Second Affiliated Hospital of Chongqing Medical UniversityChongqingChina
  5. 5.Department of AnesthesiologyChina Mianyang Central HospitalMianyangChina
  6. 6.Department of Cardiovascular SurgeryTianjin Central HospitalTianjinChina
  7. 7.Department of Social Medicine and Health ManagementShandong UniversityJinanChina
  8. 8.Department of Cardiovascular SurgeryWest China Hospital, Sichuan UniversityChengduChina
  9. 9.Department of Research Laboratory for Human Reproduction, Faculty of Medicine and School of Public HealthUniversité Libre de Bruxelles (ULB)BrusselsBelgium
  10. 10.International Centre for Reproductive Health (ICRH), Department of Public Health and Primary CareGhent UniversityGhentBelgium

Personalised recommendations