Comparative performance of warfarin pharmacogenetic algorithms in Chinese patients

A novel, rapid and simple UHPLC-MS/MS method for quantification of warfarin in dried blood spots

Warfarin is a common first line anticoagulant with a narrow therapeutic window. Because of the large blood volume needed, previous warfarin determination methods were not applicable to small animals, such as mice. To reduce the number of small animals used needed, we developed and validated a sensitive rapid assay for the simultaneous detection of warfarin enantiomers in mouse dried blood spot (DBS) samples. Analytes were extracted by tert-butyl methyl ether and then separated by a chiral Cellulose-1 column with a mobile phase of 75% acetonitrile (containing 0.1% formic acid). The total chromatographic run time was 3 min. Negative mode electrospray ionization was used for MS/MS detection, where the monitored ion transitions were m/z 307.1 → 161.0 and 341.1 → 284.0 for warfarin and coumachlor (internal standard) respectively. The calibration curves were linear with a correlation coefficient of ≥0.994 for both enantiomers over a concentration range of 10-1000 ng/mL. The satisfactory accuracy and adequate reproducibility of both warfarin enantiomers were validated in terms of intra- and interday precision with mouse DBS cards. The samples were stable at room temperature for at least 14 days. The validated method was applied to a pharmacokinetic study in mice.

A post-hoc interpretable ensemble model to feature effect analysis in warfarin dose prediction for Chinese patients

To interprete the importance of clinical features and genotypes for warfarin daily dose prediction, we developed a post-hoc interpretable framework based on an ensemble predictive model. This framework includes permutation importance for global interpretation and local interpretable model-agnostic explanation (LIME) and shapley additive explanations (SHAP) for local explanation. The permutation importance globally ranks the importance of features on the whole data set. This can guide us to build a predictive model with less variables and the complexity of final predictive model can be reduced. LIME and SHAP together explain how the predictive model give the predicted dosage for specific samples. This help clinicians prescribe accurate doses to patients using more effective clinical variables. Results showed that both the permutation importance and SHAP demonstrated that VKORC1, age, serum creatinine (SCr), left atrium (LA) size, CYP2C9 and weight were the most important features on the whole data set. In specific samples, both SHAP and LIME discovered that in Chinese patients, wild-type VKORC1-AA, mutant-type CYP2C9*3, age over 60, abnormal LA size, SCr within the normal range, and using amiodarone definitely required dosage reduction, whereas mutant-type VKORC1-AG/GG, small age, SCr out of normal range, normal LA size, diabetes and heavy weight required dosage enhancement.

Evaluating warfarin dosing models on multiple datasets with a novel software framework and evolutionary optimisation

Warfarin is an effective preventative treatment for arterial and venous thromboembolism, but requires individualised dosing due to its narrow therapeutic range and high individual variation. Many machine learning techniques have been demonstrated in this domain. This study evaluated the accuracy of the most promising algorithms on the International Warfarin Pharmacogenetics Consortium dataset and a novel clinical dataset of South African patients. Support vectors and linear regression were amongst the top performers in both datasets and performed comparably to recent stacked ensemble approaches, whilst neural networks were one of the worst performers in both datasets. We also introduced genetic programming to automatically optimise model architectures and hyperparameters without human guidance. Remarkably, the generated models were found to match the performance of the best models hand-crafted by human experts. Finally, we present a novel software framework (Warfit-learn) for warfarin dosing research. It leverages the most successful techniques in preprocessing, imputation, and parallel evaluation, with the goal of accelerating research and making results in this domain more reproducible.

Comparison of the prediction performance of different warfarin dosing algorithms based on Chinese patients

Aim: To compare the prediction performance of different warfarin dosing algorithms based on Chinese patients. Materials & methods: A total of 18 algorithms were tested in 325 patients. The predictive efficacy of selected algorithms was evaluated by calculating the percentage of patients whose predicted dose fell within ±20% of their actual stable warfarin dose and the mean absolute error. Results: The percentage within ± 20% and the mean absolute error of the algorithms ranged from 11.9 to 41.2% and -0.20 (-0.29 to -0.11) mg/d to -1.63 (-1.75 to -1.50) mg/d. The algorithms established by Miao et al. and Wei et al. had optimal predictive performance. Conclusion: Algorithms based on geographical populations might be more suitable for the prediction of stable warfarin doses in local patients.

Impact of CYP2C9, VKORC1, ApoE and ABCB1 polymorphisms on stable warfarin dose requirements in elderly Chinese patients

Aim: To analyze the impact of nongenetic factors and gene polymorphisms on warfarin dose requirements in elderly Shanghai Han Chinese patients. Materials & methods: Genotypes of CYP2C9 (rs1799853 and rs1057910), FPGS (rs7856096), ApoE (rs7412 and rs429358), GGCX (rs699664 and rs12714145), EPHX1 (rs4653436, rs1877724, rs1051740 and rs1131873), NQO1 (rs1800566 and rs10517), ABCB1 (rs1045642), VKORC1 (rs9923231) and CYP4F2 (rs2108622) in 214 patients with stable warfarin dose were determined and their demographic characteristics were recorded. Results: Multiple linear regression analysis revealed that VKORC1 rs9923231, CYP2C9*3 rs1057910, ApoE rs7412, age, BMI and concomitant amiodarone could explain 37.0% of the individual variations of daily stable warfarin dose. Conclusion: VKORC1 rs9923231, CYP2C9*3 rs1057910, ApoE rs7412, age, BMI and concomitant amiodarone play an important role in stable dose variation of warfarin in elderly Shanghai Han Chinese patients, whereas ABCB1 rs1045642 is not a significant genetic factor.

Evaluation of the predictive performance of Bayesian dosing for warfarin in Chinese patients

Aim: To evaluate the accuracy and predictive performance of Bayesian dosing for warfarin in Chinese patients. Materials & methods: Six multiple linear regression algorithms (Wei, Lou, Miao, Huang, Gage and IWPC) and a Bayesian method implemented in Warfarin Dose Calculator were compared with each other. Results: Six multiple linear regression warfarin dosing algorithms had similar predictive ability, except Miao and Lou. The mean prediction error of Bayesian priori and posteriori method were 0.01 mg/day (95% CI: -0.18 to 0.19) and 0.17 mg/day (95% CI: -0.05 to 0.29), respectively, and Bayesian posteriori method demonstrated better performance in all dose ranges. Conclusion: The Bayesian method showed a good potential for warfarin maintenance dose prediction in Chinese patients requiring less than 6 mg/day.

Identification of two novel genes SLC15A2 and SLCO1B3 associated with maintenance dose variability of warfarin in a Chinese population

Warfarin is a commonly prescribed and effective oral anticoagulant. Genetic polymorphisms associated with warfarin metabolism and sensitivity have been implicated in the wide inter-individual dose variation that is observed. Several algorithms integrating patients’ clinical characteristics and genetic polymorphism information have been explored to predict warfarin dose. However, most of these algorithms could explain only over half of the variation in a warfarin maintenance dose, suggesting that additional genetic factors may exist and need to be identified. Here, a drug absorption, distribution, metabolism and excretion (ADME) Core Panel Kit-based pharmacogenetic study was performed to screen for warfarin dose-associated SNP sites in Han-Chinese population patients taking warfarin therapy, and the screen was followed by pyrosequencing-based validation. Finally, we confirmed that the common variant rs9923231 in VKORC1 and two novel genes, SLC15A2 (rs1143671 and rs1143672) and SLCO1B3 (rs4149117 and rs7311358), are associated with the warfarin maintenance dose. As has been shown for those carriers with the variant rs9923231 in VKORC1, it was suggested that those subjects with homozygous minor alleles in those four SNPs should take a lower warfarin dose than those carrying the wild type alleles. Together with the established predictor rs9923231 in VKORC1, those four novel variants on SLC15A2 and SLCO1B3 should be considered as useful biomarkers for warfarin dose adjustment in clinical practice in Han-Chinese populations.

Estimation of the warfarin dose with a pharmacogenetic refinement algorithm in Chinese patients mainly under low-intensity warfarin anticoagulation

Pharmacogenetic (PG) dosing algorithms have been confirmed to predict warfarin therapeutic dose more accurately;however, most of them are based on standard intensity of warfarin anticoagulation, and their utility outside this range is limited. This study was designed to develop and validate a PG refinement algorithm in Chinese patients mainly under low-intensity warfarin anticoagulation. Consented Chinese-Han patients (n=310) under stable warfarin treatment were randomly divided into a derivation (n=207) and a validation cohort (n=103), with 83% and 80% of the patients under low-intensity anticoagulation, respectively. In the derivation cohort, a PG algorithm was constructed on the basis of genotypes (CYP2C9*3 and VKORC1–1639A/G) and clinical data. After integrating additional covariates of international normalised ratio (INR) values (INR on day 4 of therapy and target INR) and genotype of CYP4F2 (rs2108622), a PG refinement algorithm was established and explained 54% of warfarin dose variability. In the validation cohort, warfarin dose prediction was more accurate (p <0.01) with the PG refinement algorithm than with the PG algorithm and the fixed dose approach (3 mg/day). In the entire cohort, the PG refinement algorithm could accurately identify larger proportions of patients with lower dose requirement (≤2 mg/day) and higher dose requirement (≥4 mg/day) than did the PG algorithm. In conclusion, PG refinement algorithm integrating early INR response and three genotypes CYP2C9*3, VKORC1–1639A/G, CYP4F2 rs2108622) improves the accuracy of warfarin dose prediction in Chinese patients mainly under low-intensity anticoagulation.

Association Between the Lower Extremity Deep Venous Thrombosis, the Warfarin Maintenance Dose, and CYP2C9*3, CYP2D6*10, and CYP3A5*3 Genetic Polymorphisms: A Case-Control Study

OBJECTIVE

This study explored the association between the CYP2C9*3/CYP2D6*10/CYP3A5*3 genetic polymorphisms with lower extremity deep venous thrombosis (LEDVT) and the warfarin maintenance dose.

METHODS

Five hundred thirty-six patients who were pathologically diagnosed with LEDVT after surgery were included in the LEDVT group. At the same time, 540 patients without LEDVT who underwent surgery were recruited as the control group. Patients were given warfarin at an initial dose of 2.5-3.0 mg. Blood samples were collected to detect the initial and stable international normalized ratio (INR) values. The warfarin maintenance dose was obtained if the INR remained within a range of 2.0-3.0 for 3 consecutive days. The genotype distribution and haplotype analysis of the CYP2C9*3/CYP2D6*10/CYP3A5*3 alleles were analyzed using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) testing and SHEsis software, respectively. Logistic regression analysis was used to analyze the risk and protective factors for LEDVT.

RESULTS

The A/G genotypes, G/G genotypes, and G allele of CYP3A5*3 in the LEDVT group were observed with increased frequency compared with the control group. The LEDVT group displayed a higher ACG haplotype frequency, and lower ACA and ATA haplotype frequencies than the control group. Age, diabetes, low-density lipoprotein, CYP3A5*3 and the ACG haplotype were independent risk factors for LEDVT. High-density lipoprotein and the ACA haplotype were independent protective factors for LEDVT. The genotype distributions of the CYP2C9*3, CYP2D6*10, and CYP3A5*3 genetic polymorphisms were associated with the warfarin maintenance dose.

CONCLUSION

The CYP3A5*3 genetic polymorphism may be an important risk factor for LEDVT. Moreover, CYP2C9*3, CYP2D6*10, and CYP3A5*3 are associated with the warfarin maintenance dose.

Verification of five pharmacogenomics-based warfarin administration models

Objective:

This study aims to screen and validate five individual warfarin dosing models (four Asian model algorithms, namely, Ohno, Wen, Miao, Huang, and the algorithm of International Warfarin Pharmacogenetic Consortium, namely IWPC algorithm) with the aim of evaluating their accuracy, practicality, and safety.

Materials and Methods:

Patients’ CYP2C9*3 and VKORC1–1639G >A genes were genotyped, and patient-related information and steady warfarin doses were recorded. The difference between the predicted dose and actual maintenance dose of each model was compared.

Results:

The prediction accuracies of the Huang and Wen models were the highest. In terms of clinical practicality, the Huang model rated the highest for the low-dose group, whereas the Ohno and IWPC models rated the highest for the middle-dose group. The models tended to markedly overpredict the doses in the low-dose group, especially the IWPC model. The Miao model tended to severely underpredict the doses in the middle-dose group, whereas no model exhibited severe overprediction.

Conclusions:

Since none of the models ranked high for all the three criteria considered, the impact of various factors should be thoroughly considered before selecting the most appropriate model for the region's population.

Development of a novel individualized warfarin dose algorithm based on a population pharmacokinetic model with improved prediction accuracy for Chinese patients after heart valve replacement

The gene-guided dosing strategy of warfarin generally leads to over-dose in patients at doses lower than 2 mg/kg, and only 50% of individual variability in daily stable doses can be explained. In this study, we developed a novel population pharmacokinetic (PK) model based on a warfarin dose algorithm for Han Chinese patients with valve replacement for improving the dose prediction accuracy, especially in patients with low doses. The individual pharmacokinetic (PK) parameter - apparent clearance of S- and R-warfarin (CLs) was obtained after establishing and validating the population PK model from 296 recruited patients with valve replacement. Then, the individual estimation of CLs, VKORC1 genotypes, the steady-state international normalized ratio (INR) values and age were used to describe the maintenance doses by multiple linear regression for 144 steady-state patients. The newly established dosing algorithm was then validated in an independent group of 42 patients and was compared with other dosing algorithms for the accuracy and precision of prediction. The final regression model developed was as follows: Dose=-0.023×AGE+1.834×VKORC1+0.952×INR+2.156×CLs (the target INR value ranges from 1.8 to 2.5). The validation of the algorithm in another group of 42 patients showed that the individual variation rate (71.6%) was higher than in the gene-guided dosing models. The over-estimation rate in patients with low doses (<2 mg/kg) was lower than the other dosing methods. This novel dosing algorithm based on a population PK model improves the predictive performance of the maintenance dose of warfarin, especially for low dose (<2 mg/d) patients.

Frequency of Common CYP2C9 Polymorphisms and Their Impact on Warfarin Dose Requirement in Pakistani Population

Polymorphisms in cytochrome P450 (CYP) 2C9 (CYP2C9) gene result in interindividual variability in warfarin dose requirement. There is a need for characterization of genotype frequency distribution in different populations for construction of customized dosing algorithms to enhance the efficacy and reduce the toxicity of warfarin therapy. This study was carried out in Pakistani population to evaluate the contribution of common CYP2C9 polymorphisms to warfarin therapy. A total of 550 stable patients taking warfarin were enrolled after medical history, physical examination, and laboratory investigations. Single blood sample was collected after informed consent. Genomic DNA was extracted, and genotype analysis for CYP2C9*2 and CYP2C9*3 polymorphisms was done by polymerase chain reaction-restriction fragment length polymorphism assay. A number of samples were also analyzed by direct DNA sequencing for validation of the results. Data were analyzed using SPSS version 20. Genotype frequency distribution of CYP2C9*2 and CYP2C9*3 was found to be different from other populations. Of these 2 polymorphisms, CYP2C9*2 did not demonstrate significant effect on warfarin dose requirement, whereas CYP2C9*3 did show significant effect ( P value = .012). It is concluded that there is a need to study genotype frequency distribution and their effect on warfarin dose variability among different populations due to diversity in outcome.

Warfarin Pharmacogenetics Reevaluated: Subgroup Analysis Reveals a Likely Underestimation of the Maximum Pharmacogenetic Benefit by Clinical Trials

OBJECTIVES

Various patient subgroups were examined to determine which ones obtain the largest pharmacogenetic improvements in warfarin dose accuracy. Subgrouping schemes of recent clinical trials were analyzed for comparison.

METHODS

The accuracy of a pharmacogenetic dose algorithm was determined retrospectively in comparison to that of a clinical algorithm in subgroups of the International Warfarin Pharmacogenetics Consortium (IWPC) patient database (n = 2,274) and of newly studied clinic patients (n = 146).

RESULTS

White patients with low-dose genotypes (*1*3/AA, *2*2/AA, *2*3/GA, *2*3/AA, *3*3/GG, *3*3/GA, and *3*3/AA) achieved the largest pharmacogenetic improvements in warfarin dose accuracy. Mean absolute dosing error (MAE) in this subgroup of IWPC and newly studied patients was reduced 75.7% and 89.7%, respectively. White IWPC patients with >2 variants or ≥2 mg/day absolute difference between pharmacogenetic and clinical dose predictions obtained MAE reductions of 71.1% and 65.3%, respectively. By comparison, unstratified populations and subgroups of a major clinical trial, when replicated in IWPC patients, obtained smaller MAE reductions of 31.8% to 48.2%. Blacks and Asians obtained substantially smaller dose accuracy improvements overall than whites.

CONCLUSIONS

Patient subgroups were identified that obtained the largest pharmacogenetic improvements in warfarin dose accuracy. These subgroups have not been analyzed in clinical trials to date, likely resulting in underestimation of the pharmacogenetic benefit.

Validation of warfarin pharmacogenetic algorithms in 586 Han Chinese patients

AIM

To compare the accuracy of the nine selected algorithms for predicting warfarin dose with 586 Han Chinese patients.

MATERIALS & METHODS

Genotyping of VKORC1 1639G>A, CYP2C9*2 and CYP2C9*3 variants was performed. Both the mean absolute error and ideal estimation value were used for comparison.

RESULTS

The top three performers were from East Asians. The algorithms from Caucasians generally performed better in the medium-dose subgroup (>3 and <7 mg/day), while the algorithms from East Asians generally performed better in the low-dose subgroup (≤ 3 mg/day). None of the algorithms performed well in the high-dose subgroup (≥ 7 mg/day).

CONCLUSION

Algorithms built for specific ethnic groups and preassigned-dose groups are suggested for better prediction.

Methods for Predicting Warfarin Dose Requirements

BACKGROUND

The aim of this study was to compare the predictive performance of different warfarin dosing methods.

METHODS

Data from 46 patients who were initiating warfarin therapy were available for analysis. Nine recently published dosing tools including 8 dose prediction algorithms and a Bayesian forecasting method were compared with each other in terms of their ability to predict the actual maintenance dose. The dosing tools included 4 algorithms that were based on patient characteristics (2 clinical and 2 genotype-driven algorithms), 4 algorithms based on international normalized ratio (INR) response feedback and patient characteristics (2 clinical and 2 genotype-driven algorithms), and a Bayesian forecasting method. Comparisons were conducted using measures of bias (mean prediction error) and imprecision [root mean square error (RMSE)].

RESULTS

The 2 genotype-driven INR feedback algorithms by Horne et al and Lenzini et al produced more precise maintenance dose predictions (RMSE, 1.16 and 1.19 mg/d, respectively; P < 0.05) than the genotype-driven algorithms by Gage et al and Klein et al and the Bayesian method (RMSE, 1.60, 1.62, and 1.81 mg/d respectively). The dose predictions from clinical and genotype-driven algorithms by Gage et al, Klein et al, and Horne et al were all negatively biased. Only the INR feedback algorithms (clinical and genotype) by Lenzini et al produced unbiased dose predictions. The Bayesian method produced unbiased dose predictions overall (mean prediction error, +0.37 mg/d; 95% confidence interval, 0.89 to -0.15) but overpredicted doses in patients requiring >8 mg/d.

CONCLUSIONS

Overall, warfarin dosing methods that included some measure of INR response (INR feedback algorithms and Bayesian methods) produced unbiased and more precise dose predictions. The Bayesian forecasting method produced positively biased dose predictions in patients who required doses >8 mg/d. Further research to assess differences in clinical endpoints when warfarin doses are predicted using Bayesian or INR-driven algorithms is warranted.

Verification of pharmacogenetics-based warfarin dosing algorithms in Han-Chinese patients undertaking mechanic heart valve replacement

OBJECTIVE

To study the performance of pharmacogenetics-based warfarin dosing algorithms in the initial and the stable warfarin treatment phases in a cohort of Han-Chinese patients undertaking mechanic heart valve replacement.

METHODS

We searched PubMed, Chinese National Knowledge Infrastructure and Wanfang databases for selecting pharmacogenetics-based warfarin dosing models. Patients with mechanic heart valve replacement were consecutively recruited between March 2012 and July 2012. The predicted warfarin dose of each patient was calculated and compared with the observed initial and stable warfarin doses. The percentage of patients whose predicted dose fell within 20% of their actual therapeutic dose (percentage within 20%), and the mean absolute error (MAE) were utilized to evaluate the predictive accuracy of all the selected algorithms.

RESULTS

A total of 8 algorithms including Du, Huang, Miao, Wei, Zhang, Lou, Gage, and International Warfarin Pharmacogenetics Consortium (IWPC) model, were tested in 181 patients. The MAE of the Gage, IWPC and 6 Han-Chinese pharmacogenetics-based warfarin dosing algorithms was less than 0.6 mg/day in accuracy and the percentage within 20% exceeded 45% in all of the selected models in both the initial and the stable treatment stages. When patients were stratified according to the warfarin dose range, all of the equations demonstrated better performance in the ideal-dose range (1.88-4.38 mg/day) than the low-dose range (<1.88 mg/day). Among the 8 algorithms compared, the algorithms of Wei, Huang, and Miao showed a lower MAE and higher percentage within 20% in both the initial and the stable warfarin dose prediction and in the low-dose and the ideal-dose ranges.

CONCLUSIONS

All of the selected pharmacogenetics-based warfarin dosing regimens performed similarly in our cohort. However, the algorithms of Wei, Huang, and Miao showed a better potential for warfarin prediction in the initial and the stable treatment phases in Han-Chinese patients undertaking mechanic heart valve replacement.

Influence of warfarin dose-associated genotypes on the risk of hemorrhagic complications in Chinese patients on warfarin

This study was designed to evaluate the effect of the warfarin dose-associated genotypes, CYP2C9*3 (rs1057910), VKORC1 -1639 G/A (rs9923231), and CYP4F2 1347 C/T (rs2108622), on hemorrhagic complications in Han Chinese patients. Consecutively recruited patients requiring more than 1 year of warfarin treatment were followed from the initiation of warfarin anticoagulation for at least 3 months. CYP2C9*3, VKORC1 -1639 G/A, and CYP4F2 1347 C/T were genotyped by sequencing. The association between genotypes and warfarin hemorrhagic complications was evaluated using Cox proportional hazard regression, adjusted for demographic and clinical factors. Of 312 eligible patients obtaining stable warfarin anticoagulation in 3 months, 11 major and 69 minor hemorrhages occurred over 147 person-years. The CYP2C9*3 genotype conferred an increased risk of all [hazard ratio (HR) 3.07, 95 % confidence interval (CI) 1.57-6.01] and minor hemorrhage (HR 3.28, 95 % CI 1.62-6.65), but not major hemorrhage (HR 0.44, 95 % CI 0.04-4.72). CYP2C9*3 also conferred an increased risk of over-anticoagulation with international normalization ratio (INR) ≥4 (HR 2.92, 95 % CI 1.08-7.85). VKORC1 -1639 G/A, and CYP4F2 rs2108622 did not confer significant increase in risk for hemorrhage or over-anticoagulation. Kaplan-Meier curves showed that time to all hemorrhagic events was significantly shorter for patients with CYP2C9*3 genotype than non-carriers (P = 0.001), but not for patients with VKORC1 -1639 G/A or CYP4F2 rs2108622 genotype (P = 0.3 and 0.2). CYP2C9*3 may be the main genetic factor in hemorrhagic complications in Chinese patients under warfarin anticoagulation.