The Prediction Model of Warfarin Individual Maintenance Dose for Patients Undergoing Heart Valve Replacement, Based on the Back Propagation Neural Network
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.
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.
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 has been exempted.
- 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
- 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
- 27.Rumelhart DE, Hinton GE, Williams RJ. Learning internal representations by error propagation. Amsterdam: Elsevier Inc.; 1988.Google Scholar
- 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.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
- 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
- 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.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.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
- 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
- 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
- 51.Centers for Medicare & Medicaid Services. Medicare coverage database—potential NCD topics. http://www.cms.hhs.gov/mcd/ncpc_view_document.asp?id=1.