Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Therapeutic Dosing of Acenocoumarol: Proposal of a Population Specific Pharmacogenetic Dosing Algorithm and Its Validation in North Indians

  • Saurabh Singh Rathore,

    Affiliation Department of Genetics, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India

  • Surendra Kumar Agarwal,

    Affiliation Cardio-Vascular and Thoracic Surgery, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India

  • Shantanu Pande,

    Affiliation Cardio-Vascular and Thoracic Surgery, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India

  • Sushil Kumar Singh,

    Affiliation Thoracic and Cardio-Vascular Surgery, Chhatrapati Shahuji Maharaj Medical University, Lucknow, India

  • Tulika Mittal,

    Affiliation Department of Genetics, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India

  • Balraj Mittal

    bml_pgi@yahoo.com

    Affiliation Department of Genetics, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, India

Therapeutic Dosing of Acenocoumarol: Proposal of a Population Specific Pharmacogenetic Dosing Algorithm and Its Validation in North Indians

  • Saurabh Singh Rathore, 
  • Surendra Kumar Agarwal, 
  • Shantanu Pande, 
  • Sushil Kumar Singh, 
  • Tulika Mittal, 
  • Balraj Mittal
PLOS
x

Abstract

Objectives

To develop a population specific pharmacogenetic acenocoumarol dosing algorithm for north Indian patients and show its efficiency in dosage prediction.

Methods

Multiple and linear stepwise regression analyses were used to include age, sex, height, weight, body surface area, smoking status, VKORC1 -1639 G>A, CYP4F2 1347 G>A, CYP2C9*2,*3 and GGCX 12970 C>G polymorphisms as variables to generate dosing algorithms. The new dosing models were compared with already reported algorithms and also with the clinical data for various performance measures. Odds ratios for association of genotypes with drug sensitive and resistant groups were calculated.

Results

The pharmacogenetic dosing algorithm generated by multiple regression analysis explains 41.4% (p-value <0.001) of dosage variation. Validation of the new algorithm showed its predictive ability to be better than the already established algorithms based on similar variables. Its validity in our population is reflected by increased sensitivity, specificity, accuracy and decreased rates of over- and under- estimation in comparison to clinical data. The VKORC1-1639 G>A polymorphism was found to be strongly associated with acenocoumarol sensitivity according to recessive model.

Conclusions

We have proposed an efficient north India specific pharmacogenetic acenocoumarol dosing algorithm which might become a baseline for personalised medicine approach for treatment of patients in future.

Introduction

The human genome project completion has opened a new era of pharmacogenetics. Personalized medicine is emerging as a new therapeutic approach in clinical practice in recent times [1], [2]. Coumarinic oral-anticoagulants (COAs) like warfarin, acenocoumarol and phenprocoumon are the most frequently prescribed drugs for managing problems associated with blood coagulation in patients with atrial fibrillation (AF), heart valve replacement (HVR), deep vein thrombosis (DVT), pulmonary embolism and with patients who had undergone orthopaedic surgery [3][6]. More than 2 million patients are given warfarin in USA alone for preventing thromboembolism [7]. In north India, acenocoumarol/acitrom is widely used in place of warfarin. COA therapy is generally given lifelong and its dosing management is very difficult as it has a narrow therapeutic range and there are significant inter-individual as well as interethnic differences in stabilising dosages. Lower doses cause decreased efficacy in anticoagulation and higher doses increase the risk of bleeding events. Therefore, COA usage requires serial monitoring of blood coagulation by prothrombin time and international normalized ratio (INR) measurements. More importantly, the initial phase of COA therapy is very prone for clinical complications associated with over or under dosing. Deranged INR values are often observed in the first weeks of therapy and there is much higher risk of bleeding. To avoid this risk it is advised to predict the initial loading and stabilising doses of COAs [8][11]. At present only clinical parameters are used to predict the drug dose for anticoagulant therapy. For oral anticoagulants, the dose requirement and inter-patient variability are well known to be influenced by age, body weight, dietary vitamin K intake, concomitant disease and interacting medications [4], [12][14]. Genetics is a major role player in variability of drug dosage requirement to achieve therapeutic INR range [15][18]. More than 30 genes were found to be involved in the activity and metabolism of COAs, of which CYP2C9 (gene coding for cytochrome P450 drug metabolizing enzyme) and VKORC1 (gene coding for drug target enzyme) are the most important [19]. Previous studies have shown that ∼30% of the dose variance is explained by single nucleotide polymorphisms (SNPs) in VKORC1 and another ∼12% by two non-synonymous SNPs (*2, *3) in the CYP2C9 region [20], [21]. A genome wide association study has also revealed that VKORC1, CYP2C9 and CYP4F2 are the principal genetic factors responsible for variations in COA dose in white patients and the p values are in concordance with the fact that SNPs in VKORC1 and CYP2C9 regions are most significant [22]. In terms of dose variance, VKORC1 is found to be more influencing than CYP2C9 in some recent studies [23], [24]. However, the therapeutic drug dosage in an individual is determined by complex sets of genetic and environmental factors. Therefore, many attempts have been made in recent years to develop pharmacogenetics guided dosing algorithms based on genetic as well as clinical factors [7], [12], [17], [25]. The algorithm based dose prediction shows the importance of pharmacogenetic testing in patients who are likely to undergo COA therapy. Majority of these studies are based on warfarin dosing and not much has been explored about the pharmacogenetics of acenocoumarol and phenprocoumon. Perez-Andreu et al. have shown the importance of pre-genotyping of VKORC1, CYP4F2, CYP2C9*2 and *3 polymorphisms in patients requiring extreme doses of acenocoumarol [26]. Recently van Schie et al. have reported drug specific algorithms for acenocoumarol and phenprocoumon which are significantly different from warfarin dosing algorithms [27]. In Indian context, there is no information available in terms of genetics responsible for differences in acenocoumarol dosage requirements. Therefore, we have tried to look up in this direction and attempted to develop a pharmacogenetic algorithm to predict the stabilizing drug dosage for a better treatment of Indian patients.

Results

Patient characteristics

Patient characteristics are listed and compared in Table 1 between the derivation and validation cohorts. The minor allele frequencies of VKORC1-1639 G>A, CYP2C9*2,*3, CYP4F2 1347 G>A, GGCX 12970 C>G polymorphisms in derivation and validation cohorts were comparable. Table 1 also shows the distribution of genotypes and corresponding mean acenocoumarol doses. The VKORC1 -1639 GG, GA and AA genotypes show daily drug dose requirements of 3.47±1.21 mg/kg,2.60±0.92 and 1.25±0.64 respectively in the derivation cohort. The respective values for the validation cohort were 3.39±1.20 mg/kg, 2.49±0.88 mg/kg and 1.47±0.22 mg/kg. There was only a single patient present each in derivation and validation cohorts with CYP2C9*2*3 genotype and they showed daily maintenance dosage requirements of 1.28 mg/kg and 2.71 mg/kg respectively. Comparatively higher drug dose requirements were observed for CYP2C9*1*1, *1*2 or *1*3 genotypes in both the derivation and validation cohorts. No significant difference was observed between patients with different allelic combinations of CYP4F2 1347 G>A and GGCX 12970 C>G polymorphisms. The trend of dosage requirement was observed for both the cohorts. All other patient characteristic were found to be comparable in both the cohorts.

Pharmacogenetic dosing algorithm by multiple regression

Multiple regression analysis resulted in the following dosing algorithm: dose (mg/day)  = 3.082–0.013(smoking status, 1 for smoker and 0 for non-smoker) –0.433 (gender, 1 for male and 0 for female) –0.004(age in years) + indication(0.327 for DVR and −0.092 for AVR) +0.026(height in centimetres) +0.151(weight in kilograms) –7.660(body surface area in cm2) –0.862(VKORC1 GA) –2.257(VKORC1 AA) –0.049(CYP2C9*2 CT) –0.456(CYP2C9*3 AC) +0.449(CYP4F2 GA) +0.230(CYP4F2 AA) +0.245(GGCX CG) +1.055(GGCX GG) (Table 2). The coefficient of determination (R2) value for this equation is 41.4% (p-value <0.001). It means that 41.4% of variation in acenocoumarol dose is explained by this pharmacogenetic model.

Pharmacogenetic dosing algorithm by linear stepwise regression

Linear stepwise regression produced a more simple equation (including the clinical and genetic factors (Table 2). The R2 value of this algorithm is 37% (p-value <0.001), which is the highest value produced in the stepwise regression modelling for the final step. This second dosing algorithm was found to have similar predictive trend as shown by the multiple regression equation. However, in terms of mean weekly dose and mean absolute error, the multiple regression equation was more accurate. The values of mean weekly dose and mean absolute error by stepwise regression equation were 22.02±4.72 (95% CI 21.08–22.96) and 0.71±7.86 (95% CI −0.76–2.19).

Performance of new dosing algorithm

When compared with clinical data, the new multiple regression algorithm showed improvement in various performance measures like sensitivity (76% vs 51%), specificity (64% vs 49%), rate of overestimation (22% vs 27%), rate of underestimation (15% vs 23%), overall accuracy (63% vs 50%), accuracy in dug sensitive cases (60% vs 51%) and accuracy in drug resistant cases (72% vs 49%). The Cronbach's Alpha constant depicts the relatedness of dosage data (as a group) predicted by the new algorithm or the clinical algorithm with the actual therapeutic dosage data. This value was higher in case of new algorithm in comparison with the clinical data (0.56 vs 0.11) (Table 3).

thumbnail
Table 2. Algorithm development by multiple and linear stepwise regression analyses.

https://doi.org/10.1371/journal.pone.0037844.t002

Comparision of algorithms

The new multiple regression algorithm predicted better in comparison to the therapeutic dose in the aspect of standard deviation from mean weekly dose (Table 4). The mean weekly dose calculated by this algorithm was 21.26±4.82 mg/week (95% CI 20.30–22.21). This value is closest to that obtained by van Schie et al. acenocoumarol dosing algorithm, 23.56±4.67 mg/week (95% CI 22.63–24.49) [27]. The values for the algorithms by Oner Ozgon et al. [28] and Wen et al. [29] were 27.16±1.19 mg/week (95% CI 26.92–27.40) and 27.87±4.94 mg/week (95% CI 26.89–28.85). The mean value for therapeutic data of dose was found to be 21.31±8.35 mg/week (95% CI 19.65–22.97) (Table 4). The mean absolute error (MAE) for the new algorithm was 0.06±7.62 mg/week (95% CI −1.57–1.45). The MAE values were 2.25±7.86 (95% CI 0.69–3.81), 5.84±7.96 (95% CI 4.26–7.42) and 6.56±7.23 (95% CI 5.13–7.99) for van Schie et al. (acenocoumarol algorithm) [27], Oner Ozgon et al. [28] and Wen et al. [29] algorithms (Table 4).

thumbnail
Table 3. Comparision of performance of new algorithms with clinical data.

https://doi.org/10.1371/journal.pone.0037844.t003

Association with acenocoumarol dose

In the study subjects, VKORC1 -1639 G>A polymorphism has the strongest association with acenocoumarol sensitivity according to recessive model (OR 4.42, 95% CI 2.44–7.99, p value<0.05). No other polymorphism was found to be significantly associated with acenocoumarol sensitive, resistant and intermediate dosing groups (Table 5).

thumbnail
Table 4. Mean weekly doses and mean absolute errors according to different algorithms.

https://doi.org/10.1371/journal.pone.0037844.t004

thumbnail
Table 5. Association between acenocoumarol sensitive/resistant/intermediate dose groups and polymorphisms.

https://doi.org/10.1371/journal.pone.0037844.t005

Discussion

It is very difficult for clinicians to predict the accurate COA dose and they depend totally on the traditional physical parameters like sex, weight, height, and age to decide the dose in advancing the therapy. Many evidences of role of genetic markers, influencing warfarin dosage, were shown by different investigators in last few years [13], [14], [30]. Genotyping of patients having VKORC1, CYP2C9 and CYP4F2 variant alleles has been shown to decrease the risk of over-anticoagulation in comparison to a fixed initial dose approach [31], [32]. Distribution of VKORC1-1639 A, CYP2C9*2 and CYP2C9*3 allele frequencies were found to be different for Indians when compared with selected HapMap populations [33]. The ground for population specific dosing regimens, for patients on anticoagulation, is established by such interethnic differences in allele frequencies. The VKORC1 −1639 G>A polymorphism is present in the promoter region of the coding sequence so it decreases the enzyme expression if present in homozygous recessive state. The CYP2C9*2 and *3 polymorphisms are located in exonic region and they decrease the drug clearance as evidenced by Rettie et al. [34] in 1994. We have found only two patients with CYP2C9*2*3 genotype and they required comparatively lower drug doses than those with CYP2C9*1*1, *1*2 or *1*3 genotypes. Since the CYP2C9 enzyme is involved in metabolism of COAs, decrease in their activity will result in lower maintenance drug doses to achieve therapeutic INR.

To prevent the phenotypic extremities of COA treatment, many attempts have been made to develop pharmacogenetics guided dosing regimens for warfarin [12], [17], [21], [28], [29], [35]. The Clinical Pharmacogenetics Implimentation Consortium (CPIC) has published guidelines for the use of pharmacogenomic tests in warfarin dosing (J A Johnson et al.). Many studies have proposed warfarin dosing algorithms based on both genetic and non-genetic factors [21], [35], [36]. The pharmacogenomic algorithms in these studies have different model fitness as evidenced from the R2 values ranging from 34 to 63. There are only a few reports about pharmacogenetic dosing algorithms of acenocoumarol and phenprocoumon. In a recent large European cohort study, acenocoumarol and phenprocoumon algorithms were found to explain 59.4% and 49.0% variations in dose requirements respectively [27]. In the present study, we have generated a new dosing algorithm which is useful in predicting acenocoumarol doses in the Indian population. We have been able to show that 41.4% of the variability in daily maintenance acenocoumarol doses was explained by our pharmacogenetic dosing model based on age, gender, height, weight, BSA, smoking status, indication for cardiac valve replacement surgery, VKORC1 -1639 G>A, CYP4F2 1347 G>A, CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms.

The dosage data generated by our pharmacogenetic algorithm displayed similar skewness as by the therapeutic maintenance dosage data. van Schie et al. [27] acenocoumarol dosing algorithm was most identical to our algorithm in terms of mean weekly dose values. The mean weekly dosage and associated standard deviation for the new algorithm and the van Schie et al. [27] acenocoumarol dosing algorithm is close to that obtained in case of therapeutic data (Table 4). The new algorithm was found to be least erroneous as the MAE was minimum in its case. van Schie et al. [27] acenocoumarol dosing algorithm was very close to our algorithm in this aspect also. Out of other 13 algorithms, 12 were derived from warfarin treated patients and 1 was derived from phenprocoumon treated patients. These algorithms show moderate to higher differences in mean weekly dose and MAE values from the therapeutic dosage data. This observation suggested drug specificity of dosing algorithms and advocated our approach. A single study from south India had used genetic and clinical parameters to develop a pharmacogenetics based dosing algorithm for warfarin and it was found to be more useful in predicting the stabilising drug doses [37]. We have not used their algorithm to predict drug doses for our patient cohort as it uses complex information of vitamin K intake and additional polymorphisms in VKORC1 gene. Our main aim was to target most relevant markers to predict drug dose and keep the algorithm simple. Due to the difference in the type of drug (acenocoumarol in place of warfarin), genetic differences [33], [38] and dietary pattern of north Indians and south Indians, there is a need to derive a dosing regimen for acenocoumarol which is widely used in India.

The new algorithm was compared with the clinical data for different performance measures. It was significantly sensitive (76% vs 51%) than clinical data. The specificity was also increased if new algorithm was used to predict the dose (64% vs 49%). There was decrease in rate of over- and underestimation when new algorithm was compared with clinical data. Moderate increase in accuracy was observed for new algorithm in acenocoumarol sensitive cases and it was significant in acenocoumarol resistant cases (72% vs 49%). The overall accuracy was also higher for new algorithm than clinical data (63% vs 50%). The higher value of the Cronbach's Alpha constant in case of our algorithm as compared with the clinical data (0.56 vs 0.11) proves again the closer proximity of our algorithm with the therapeutic dosing.

We have found a strong association of acenocoumarol sensitivity with the heterozygocity and homozygocity for the VKORC1-1639 G>A change. The presence of GA or AA genotype results in very high risk of overdosing. In this condition the drug target enzyme is expressed in lower amounts so lower drug doses can achieve therapeutic INR range in early stage of treatment. Pavani et al. [37] have also reported the high risk of warfarin sensitivity for VKORC1-1639 G>A polymorphism. We have also looked for association of other genotypes in drug sensitive, resistant and intermediate dose groups but could not get any significant one. So we can say that VKORC1−1639 G>A polymorphism is the key player in drug sensitivity. We have observed presence of CYP4F2 1347 G>A polymorphism in the linear stepwise dosing algorithm but it was not found to be associated with any of the dose groups. This may be due to the fact that the linear stepwise regression generates a best fit model for a scalar outcome. This analysis uses both categorical as well as scalar variables as inputs. In contrast, the binary logistic regression used for associating polymorphisms with drug dose groups uses only categorical variables as inputs and it calculates the risk as odds ratios. In other words, we can say that VKORC1, CYP4F2, weight and gender explain 37% of variation in acenocoumarol dosage requirements. The R2 values of 41.4% and 37% in respective cases of linear stepwise regression and multiple regression algorithms suggest that VKORC1 and CYP4F2 are major contributors in acenocoumarol dosage variability. The contributions of VKORC1and CYP4F2 genotypes in overall predictive power of dosing algorithm are 21% and 3.7% respectively. There is only ∼4% increase in R2 value by including the other genetic factors like CYP2C9 and GGCX and other non-genetic factors in multiple regression modelling. Cardiac valve replacement surgery phenotypes (double/mitral/aortic valve replacement) have ∼1.5% contribution in explaining the variability of dose requirements. This is reflected by the fact that excluding these variables from multiple regression modelling and including CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms gives an algorithm with R2 value of 40%. This also reflects that CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms explain about 3% of variability in the acenocoumarol dosage requirements.

The low frequencies of CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms in north Indian population explain the lower contribution of these markers in variability in dosage requirements. Also, the CYP2C9*2 and *3 alleles are less frequent in the derivation cohort so their contribution is small in the algorithm. This is a limitation of our study as there may be some bias in patient selection for the two cohorts. Moreover, Loebstein et al. [39] reported that an Arg36Tyr polymorphism VKORC1 was responsible for higher dose requirement in warfarin resistant patients and was significantly associated with higher drug dose requirement. As we have not analysed this polymorphism in our patient group, it is possible that this or some other polymorphisms might also contribute to some extent for acenocoumarol resistance.

Conclusion

The multiple regression algorithm can be used for a more accurate prediction of acenocoumarol doses. It explains 41.4% variability in acenocoumarol dosage requirements. The linear stepwise algorithm uses genotyping of only two polymorphisms and it can be used for better cost to benefit ratio as its results are comparable with the more complex multiple regression algorithms. Polymorphisms in VKORC1 and CYP4F2 came out to be the principal genetic determinants explaining 37% acenocoumarol dosage variability. The less frequent CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms in Indian population do determine the dosage variability but to an extent of ∼3% only. In comparison to other dosing models, which are mostly based on warfarin, our drug specific algorithms show more accurate acenocoumarol dosage prediction.

Future perspective

Development of new oral anticoagulants like Dabigatran (direct thrombin inhibitor) and Rivaroxaban (direct factor Xa inhibitor) have sparked a new hope of overcoming the risk factors associated with vitamin K antagonists. These new drugs may replace the traditional oral anticoagulants in future.

However, till these new drugs are widely used, there is a need to replicate this work in a larger sample size as well as in various regions and ethnic groups of India before introduction in routine clinical practice. Improved therapy outcomes based on such studies will allow successful application of individualised dosing in clinical setting.

Methods

Ethics Statement

The study protocol was approved by the institutional ethical committee of Sanjay Gandhi Post Graduate Institute of Medical Sciences (SGPGIMS). The authors followed the norms of World's Association Declaration of Helsinki. All the participants gave written informed consent to participate in the study.

Patients

The present study was carried out on the DNA samples of patients who had undergone surgery for aortic/mitral/double valve replacement and were being followed-up for regular PT-INR (Prothrombin Time-International Normalized Ratio) testing. The PT-INR testing was done every 2 weeks for each patient. A total of 7586 patients on COA therapy were screened and of which 225 patients of northern Indian origin were recruited from the outpatient department of Department of Cardiovascular and Thoracic Surgery, Sanjay Gandhi Post Graduate Institute of Medical Sciences and Department of Thoracic & Cardio-Vascular Surgery, Chhatrapati Shahuji Maharaj Medical University, Lucknow, India. The patient sample collection was done during a period of about one and half year from March 2010 to August 2011. We have divided the patients into two cohorts, namely derivation cohort of 125 patients and validation cohort of 100 patients. The selection of patients for derivation and validation cohorts was done on a random basis. All these subjects were taking maintenance dosage of acenocoumarol to achieve the INR in a therapeutic range (between 2.0 to 3.5) for at least 3 consecutive months. Patients aging less than 18 years or those suffering from diabetes, liver disease, chronic diarrheal conditions or malabsorption were excluded from the study. Patients were given dietary advice and food charts were prepared in order to avoid any interference with acenocoumarol pharmacokinetics and pharmacodynamics. Patients with interfering drug administration and showing non-compliance and were excluded from the study. Clinical data like age, body weight, height, gender, smoking habits, indication for acenocoumarol therapy, average maintenance dose and ethnicity were recorded. The average maintenance dose was calculated as mean of the dose values during the 3 month period when last two consecutive stable INR values were documented. The north Indian ethnicity of patients was decided according to their place of residence in the last three generations, food habits and mother tongue (Hindi or related languages).

Genotyping

Blood samples were collected once in vacutainer vials coated with ethylenediaminetetraacetic acid along with the samples taken up for regular PT-INR tests. Standard salting-out method was used to isolate genomic DNA from blood samples. DNA was checked both for quantity as well as quality by gel electrophoresis and spectrophotometry using the NanoDrop Analyzer (ND-1000) spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). The ratio of absorbance of DNA at 260 and 280°nm were between 1.7 and 1.9. The checked DNA was stored at −40°C. The genotyping for VKORC1 -1639 G>A, CYP4F2 1347 G>A, CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms was done by polymerase chain reaction and restriction fragment length polymorphism (PCR–RFLP) [17], [40][42]. Negative (genomic DNA absent) and positive (known heterozygote) controls were used in each PCR reaction. Genotyping was repeated in ∼10% samples by different lab members to check for genotyping errors and 100% concordance was obtained. For the GGCX 12970 C>G polymorphism, Taqman probe-based genotyping was also performed by using the ABI prism 7900 using SNP discrimination assays designed by Applied Biosystems (ABI) (Applied Biosystems, Foster City, CA, USA).

Statistical analyses

SPSS version 17.0 software (SPSS Japan, Tokyo, Japan) was used to perform all the statistical analyses. Descriptive statistics was used to calculate the patient characteristics. The minor allele frequencies were calculated by counting the total number of alleles for a particular SNP in all the patients and then dividing the allele with low prevalence by the allele with higher prevalence and expressed as percentage. The derivation cohort was used to generate the pharmacogenetic model which was used to predict the drug dosage in the validation cohort. The parameters used in multiple regression for deriving the pharmacogenetic model were age, sex, height, weight, body surface area (BSA), smoking status, indication for cardiac valve replacement surgery, VKORC1 -1639 G>A, CYP4F2 1347 G>A, CYP2C9*2, *3 and GGCX 12970 C>G polymorphisms. A simpler pharmacogenetic model was also developed by linear stepwise regression using the above said variables. The genotype data was entered as 0 for absence and 1 for presence. We have not used complex dietary information of vitamin K uptake because it was too difficult to obtain that in clinical setting. Further, the patients were given dietary advice in form of charts and any noncompliance in this regard was treated as exclusion criteria. Apart from these pharmacogenetic algorithms comprising of both genetic as well as clinical factors, we have also generated an equation based only on clinical factors. This was done by putting all except the genetic factors in linear stepwise regression. We have extracted eleven algorithms through literature search which use similar parameters for dosage prediction. These algorithms were used to predict the doses based on the patient profile in the validation cohort of our study. The dosage predicted by our pharmacogenetic model was compared with the therapeutic dosage and with those predicted by the twelve algorithms. Of these 12 algorithms, 11 have been derived from patient cohorts using warfarin for anticoagulation therapy. One algorithm used patients on acenocoumarol treatment for deriving pharmacogenetic dosing algorithm. For these comparisons, we have computed the mean values of the dosage along with standard deviation (SD) and 95% confidence interval (CI) values for each algorithm. Comparison of the mean absolute error (MAE) of each algorithm was used to evaluate the performance of each algorithm. The absolute error values were calculated by deducting the actual dose from the predicted dose and their means were defined as MAE. The MAE values were calculated along with the SD and 95% CI values.

We have used SPSS software to make 3 quartiles of therapeutic drug doses of all the patients. Patients on a maintenance dose of ≤17.0 and ≥24.5 mg/week were considered acenocoumarol sensitive and resistant respectively while the patients in the range of 17.0-24.5 mg/week were assumed to be in intermediate range. As a measure of association, odds ratio values for different genotypes were calculated in both drug sensitive and drug resistant groups. The performance of our pharmacogenetic algorithms was compared with that of clinical algorithm in terms of sensitivity, specificity, rate of over and underestimation, accuracy and relatedness as a group. All these parameters were obtained by calculations by Fisher's exact test.

Author Contributions

Conceived and designed the experiments: BM SKA SSR. Performed the experiments: SSR. Analyzed the data: SSR BM. Contributed reagents/materials/analysis tools: SSR SKA SP TM SKS. Wrote the paper: SSR BM.

References

  1. 1. Guttmacher AE, Collins FS (2002) Genomic medicine – a primer. N Engl J Med 347: 1512–1520.AE GuttmacherFS Collins2002Genomic medicine – a primer.N Engl J Med34715121520
  2. 2. Evans WE, Relling MV (1999) Pharmacogenomics: translating functional genomics into rational therapeutics. Science 286: 487–491.WE EvansMV Relling1999Pharmacogenomics: translating functional genomics into rational therapeutics.Science286487491
  3. 3. Hirsh J (1992) Antithrombotic therapy in deep vein thrombosis and pulmonary embolism. Am Heart J 123: 1115–1122.J. Hirsh1992Antithrombotic therapy in deep vein thrombosis and pulmonary embolism.Am Heart J12311151122
  4. 4. Hirsh J, Dalen J, Anderson DR, Poller L, Bussey H, et al. (2001) Oral anticoagulants: mechanism of action, clinical effectiveness, and optimal therapeutic range. Chest 119: 8S–21S.J. HirshJ. DalenDR AndersonL. PollerH. Bussey2001Oral anticoagulants: mechanism of action, clinical effectiveness, and optimal therapeutic range.Chest1198S21S
  5. 5. Laupacis A, Albers G, Dalen J, Dunn M, Feinberg W, et al. (1995) Antithrombotic therapy in atrial fibrillation. Chest 108: 352S–359S.A. LaupacisG. AlbersJ. DalenM. DunnW. Feinberg1995Antithrombotic therapy in atrial fibrillation.Chest108352S359S
  6. 6. Stein PD, Dalen JE, Goldman S, Schwartz L, Theroux P, et al. (1995) Antithrombotic therapy in patients with saphenous vein and internal mammary artery bypass grafts. Chest 108: 424S–430S.PD SteinJE DalenS. GoldmanL. SchwartzP. Theroux1995Antithrombotic therapy in patients with saphenous vein and internal mammary artery bypass grafts.Chest108424S430S
  7. 7. Anderson JL, Horne BD, Stevens SM, Grove AS, Barton S, et al. (2007) Randomized trial of genotype-guided versus standard warfarin dosing in patients initiating oral anticoagulation. Circulation 116: 2563–2570.JL AndersonBD HorneSM StevensAS GroveS. Barton2007Randomized trial of genotype-guided versus standard warfarin dosing in patients initiating oral anticoagulation.Circulation11625632570
  8. 8. Hirsh J, Fuster V, Ansell J, Halperin JL (2003) American Heart Association/American College of Cardiology Foundation guide to warfarin therapy. Circulation 107: 1692–1711.J. HirshV. FusterJ. AnsellJL Halperin2003American Heart Association/American College of Cardiology Foundation guide to warfarin therapy.Circulation10716921711
  9. 9. Ezekowitz MD, James KE, Radford MJ, Rickles FR, Redmond N (1999) Initiating and Maintaining Patients on Warfarin Anticoagulation: The Importance of Monitoring. J Cardiovasc Pharmacol Ther 4: 3–8.MD EzekowitzKE JamesMJ RadfordFR RicklesN. Redmond1999Initiating and Maintaining Patients on Warfarin Anticoagulation: The Importance of Monitoring.J Cardiovasc Pharmacol Ther438
  10. 10. Ansell J, Hirsh J, Dalen J, Bussey H, Anderson D, et al. (2001) Managing oral anticoagulant therapy. Chest 119: 22S–38S.J. AnsellJ. HirshJ. DalenH. BusseyD. Anderson2001Managing oral anticoagulant therapy.Chest11922S38S
  11. 11. Ageno W, Squizzato A, Dentali F, Crowther M (2005) Tailoring warfarin induction doses to reflect individual and disease-specific factors. Am J Med 118: 143–144.W. AgenoA. SquizzatoF. DentaliM. Crowther2005Tailoring warfarin induction doses to reflect individual and disease-specific factors.Am J Med118143144
  12. 12. Gage BF, Eby CS (2003) Pharmacogenetics and anticoagulant therapy. J Thromb Thrombolysis 16: 73–78.BF GageCS Eby2003Pharmacogenetics and anticoagulant therapy.J Thromb Thrombolysis167378
  13. 13. Kamali F, Khan TI, King BP, Frearson R, Kesteven P, et al. (2004) Contribution of age, body size, and CYP2C9 genotype to anticoagulant response to warfarin. Clin Pharmacol Ther 75: 204–212.F. KamaliTI KhanBP KingR. FrearsonP. Kesteven2004Contribution of age, body size, and CYP2C9 genotype to anticoagulant response to warfarin.Clin Pharmacol Ther75204212
  14. 14. Aquilante CL, Langaee TY, Lopez LM, Yarandi HN, Tromberg JS, et al. (2006) Influence of coagulation factor, vitamin K epoxide reductase complex subunit 1, and cytochrome P450 2C9 gene polymorphisms on warfarin dose requirements. Clin Pharmacol Ther 79: 291–302.CL AquilanteTY LangaeeLM LopezHN YarandiJS Tromberg2006Influence of coagulation factor, vitamin K epoxide reductase complex subunit 1, and cytochrome P450 2C9 gene polymorphisms on warfarin dose requirements.Clin Pharmacol Ther79291302
  15. 15. Wadelius M, Pirmohamed M (2007) Pharmacogenetics of warfarin: current status and future challenges. Pharmacogenomics J 7: 99–111.M. WadeliusM. Pirmohamed2007Pharmacogenetics of warfarin: current status and future challenges.Pharmacogenomics J799111
  16. 16. Rieder MJ, Reiner AP, Gage BF, Nickerson DA, Eby CS, et al. (2005) Effect of VKORC1 haplotypes on transcriptional regulation and warfarin dose. N Engl J Med 352: 2285–2293.MJ RiederAP ReinerBF GageDA NickersonCS Eby2005Effect of VKORC1 haplotypes on transcriptional regulation and warfarin dose.N Engl J Med35222852293
  17. 17. Sconce EA, Khan TI, Wynne HA, Avery P, Monkhouse L, et al. (2005) The impact of CYP2C9 and VKORC1 genetic polymorphism and patient characteristics upon warfarin dose requirements: proposal for a new dosing regimen. Blood 106: 2329–2333.EA SconceTI KhanHA WynneP. AveryL. Monkhouse2005The impact of CYP2C9 and VKORC1 genetic polymorphism and patient characteristics upon warfarin dose requirements: proposal for a new dosing regimen.Blood10623292333
  18. 18. Lal S, Jada SR, Xiang X, Lim WT, Lee EJ, et al. (2006) Pharmacogenetics of target genes across the warfarin pharmacological pathway. Clin Pharmacokinet 45: 1189–1200.S. LalSR JadaX. XiangWT LimEJ Lee2006Pharmacogenetics of target genes across the warfarin pharmacological pathway.Clin Pharmacokinet4511891200
  19. 19. Wadelius M, Chen LY, Eriksson N, Bumpstead S, Ghori J, et al. (2007) Association of warfarin dose with genes involved in its action and metabolism. Hum Genet 121: 23–34.M. WadeliusLY ChenN. ErikssonS. BumpsteadJ. Ghori2007Association of warfarin dose with genes involved in its action and metabolism.Hum Genet1212334
  20. 20. D'Andrea G, D'Ambrosio RL, Di Perna P, Chetta M, Santacroce R, et al. (2005) A polymorphism in the VKORC1 gene is associated with an interindividual variability in the dose-anticoagulant effect of warfarin. Blood 105: 645–649.G. D'AndreaRL D'AmbrosioP. Di PernaM. ChettaR. Santacroce2005A polymorphism in the VKORC1 gene is associated with an interindividual variability in the dose-anticoagulant effect of warfarin.Blood105645649
  21. 21. Wadelius M, Chen LY, Lindh JD, Eriksson N, Ghori MJ, et al. (2009) The largest prospective warfarin-treated cohort supports genetic forecasting. Blood 113: 784–792.M. WadeliusLY ChenJD LindhN. ErikssonMJ Ghori2009The largest prospective warfarin-treated cohort supports genetic forecasting.Blood113784792
  22. 22. Takeuchi F, McGinnis R, Bourgeois S, Barnes C, Eriksson N, et al. (2009) A genome-wide association study confirms VKORC1, CYP2C9, and CYP4F2 as principal genetic determinants of warfarin dose. PLoS Genet 5: e1000433.F. TakeuchiR. McGinnisS. BourgeoisC. BarnesN. Eriksson2009A genome-wide association study confirms VKORC1, CYP2C9, and CYP4F2 as principal genetic determinants of warfarin dose.PLoS Genet5e1000433
  23. 23. Lindh JD, Lundgren S, Holm L, Alfredsson L, Rane A (2005) Several-fold increase in risk of overanticoagulation by CYP2C9 mutations. Clin Pharmacol Ther 78: 540–550.JD LindhS. LundgrenL. HolmL. AlfredssonA. Rane2005Several-fold increase in risk of overanticoagulation by CYP2C9 mutations.Clin Pharmacol Ther78540550
  24. 24. Caldwell MD, Awad T, Johnson JA, Gage BF, Falkowski M, et al. (2008) CYP4F2 genetic variant alters required warfarin dose. Blood 111: 4106–4112.MD CaldwellT. AwadJA JohnsonBF GageM. Falkowski2008CYP4F2 genetic variant alters required warfarin dose.Blood11141064112
  25. 25. Wadelius M, Chen LY, Downes K, Ghori J, Hunt S, et al. (2005) Common VKORC1 and GGCX polymorphisms associated with warfarin dose. Pharmacogenomics J 5: 262–270.M. WadeliusLY ChenK. DownesJ. GhoriS. Hunt2005Common VKORC1 and GGCX polymorphisms associated with warfarin dose.Pharmacogenomics J5262270
  26. 26. Perez-Andreu V, Roldan V, Lopez-Fernandez MF, Anton AI, Alberca I, et al. (2010) Pharmacogenetics of acenocoumarol in patients with extreme dose requirements. J Thromb Haemost 8: 1012–1017.V. Perez-AndreuV. RoldanMF Lopez-FernandezAI AntonI. Alberca2010Pharmacogenetics of acenocoumarol in patients with extreme dose requirements.J Thromb Haemost810121017
  27. 27. van Schie RM, Wessels JA, le Cessie S, de Boer A, Schalekamp T, et al. (2011) Loading and maintenance dose algorithms for phenprocoumon and acenocoumarol using patient characteristics and pharmacogenetic data. Eur Heart J 32: 1909–1917.RM van SchieJA WesselsS. le CessieA. de BoerT. Schalekamp2011Loading and maintenance dose algorithms for phenprocoumon and acenocoumarol using patient characteristics and pharmacogenetic data.Eur Heart J3219091917
  28. 28. Oner Ozgon G, Langaee TY, Feng H, Buyru N, Ulutin T, et al. (2008) VKORC1 and CYP2C9 polymorphisms are associated with warfarin dose requirements in Turkish patients. Eur J Clin Pharmacol 64: 889–894.G. Oner OzgonTY LangaeeH. FengN. BuyruT. Ulutin2008VKORC1 and CYP2C9 polymorphisms are associated with warfarin dose requirements in Turkish patients.Eur J Clin Pharmacol64889894
  29. 29. Wen MS, Lee M, Chen JJ, Chuang HP, Lu LS, et al. (2008) Prospective study of warfarin dosage requirements based on CYP2C9 and VKORC1 genotypes. Clin Pharmacol Ther 84: 83–89.MS WenM. LeeJJ ChenHP ChuangLS Lu2008Prospective study of warfarin dosage requirements based on CYP2C9 and VKORC1 genotypes.Clin Pharmacol Ther848389
  30. 30. Higashi MK, Veenstra DL, Kondo LM, Wittkowsky AK, Srinouanprachanh SL, et al. (2002) Association between CYP2C9 genetic variants and anticoagulation-related outcomes during warfarin therapy. JAMA 287: 1690–1698.MK HigashiDL VeenstraLM KondoAK WittkowskySL Srinouanprachanh2002Association between CYP2C9 genetic variants and anticoagulation-related outcomes during warfarin therapy.JAMA28716901698
  31. 31. Teichert M, van Schaik RH, Hofman A, Uitterlinden AG, de Smet PA, et al. (2009) Genotypes associated with reduced activity of VKORC1 and CYP2C9 and their modification of acenocoumarol anticoagulation during the initial treatment period. Clin Pharmacol Ther 85: 379–386.M. TeichertRH van SchaikA. HofmanAG UitterlindenPA de Smet2009Genotypes associated with reduced activity of VKORC1 and CYP2C9 and their modification of acenocoumarol anticoagulation during the initial treatment period.Clin Pharmacol Ther85379386
  32. 32. Perez-Andreu V, Roldan V, Anton AI, Garcia-Barbera N, Corral J, et al. (2009) Pharmacogenetic relevance of CYP4F2 V433M polymorphism on acenocoumarol therapy. Blood 113: 4977–4979.V. Perez-AndreuV. RoldanAI AntonN. Garcia-BarberaJ. Corral2009Pharmacogenetic relevance of CYP4F2 V433M polymorphism on acenocoumarol therapy.Blood11349774979
  33. 33. Rathore SS, Agarwal SK, Pande S, Mittal T, Mittal B (2010) Frequencies of VKORC1 -1639 G>A, CYP2C9*2 and CYP2C9*3 genetic variants in the Northern Indian population. Biosci Trends 4: 333–337.SS RathoreSK AgarwalS. PandeT. MittalB. Mittal2010Frequencies of VKORC1 -1639 G>A, CYP2C9*2 and CYP2C9*3 genetic variants in the Northern Indian population.Biosci Trends4333337
  34. 34. Rettie AE, Wienkers LC, Gonzalez FJ, Trager WF, Korzekwa KR (1994) Impaired (S)-warfarin metabolism catalysed by the R144C allelic variant of CYP2C9. Pharmacogenetics 4: 39–42.AE RettieLC WienkersFJ GonzalezWF TragerKR Korzekwa1994Impaired (S)-warfarin metabolism catalysed by the R144C allelic variant of CYP2C9.Pharmacogenetics43942
  35. 35. Gage BF, Eby C, Johnson JA, Deych E, Rieder MJ, et al. (2008) Use of pharmacogenetic and clinical factors to predict the therapeutic dose of warfarin. Clin Pharmacol Ther 84: 326–331.BF GageC. EbyJA JohnsonE. DeychMJ Rieder2008Use of pharmacogenetic and clinical factors to predict the therapeutic dose of warfarin.Clin Pharmacol Ther84326331
  36. 36. Klein TE, Altman RB, Eriksson N, Gage BF, Kimmel SE, et al. (2009) Estimation of the warfarin dose with clinical and pharmacogenetic data. N Engl J Med 360: 753–764.TE KleinRB AltmanN. ErikssonBF GageSE Kimmel2009Estimation of the warfarin dose with clinical and pharmacogenetic data.N Engl J Med360753764
  37. 37. Pavani A, Naushad SM, Rupasree Y, Kumar TR, Malempati AR, et al. (2011) Optimization of warfarin dose by population-specific pharmacogenomic algorithm. Pharmacogenomics J. A. PavaniSM NaushadY. RupasreeTR KumarAR Malempati2011Optimization of warfarin dose by population-specific pharmacogenomic algorithm.Pharmacogenomics J
  38. 38. Rosemary J, Adithan C (2007) The pharmacogenetics of CYP2C9 and CYP2C19: ethnic variation and clinical significance. Curr Clin Pharmacol 2: 93–109.J. RosemaryC. Adithan2007The pharmacogenetics of CYP2C9 and CYP2C19: ethnic variation and clinical significance.Curr Clin Pharmacol293109
  39. 39. Loebstein R, Dvoskin I, Halkin H, Vecsler M, Lubetsky A, et al. (2007) A coding VKORC1 Asp36Tyr polymorphism predisposes to warfarin resistance. Blood 109: 2477–2480.R. LoebsteinI. DvoskinH. HalkinM. VecslerA. Lubetsky2007A coding VKORC1 Asp36Tyr polymorphism predisposes to warfarin resistance.Blood10924772480
  40. 40. Deng S, Zhu G, Liu F, Zhang H, Qin X, et al. (2010) CYP4F2 gene V433M polymorphism is associated with ischemic stroke in the male Northern Chinese Han population. Prog Neuropsychopharmacol Biol Psychiatry 34: 664–668.S. DengG. ZhuF. LiuH. ZhangX. Qin2010CYP4F2 gene V433M polymorphism is associated with ischemic stroke in the male Northern Chinese Han population.Prog Neuropsychopharmacol Biol Psychiatry34664668
  41. 41. Aynacioglu AS, Brockmoller J, Bauer S, Sachse C, Guzelbey P, et al. (1999) Frequency of cytochrome P450 CYP2C9 variants in a Turkish population and functional relevance for phenytoin. Br J Clin Pharmacol 48: 409–415.AS AynaciogluJ. BrockmollerS. BauerC. SachseP. Guzelbey1999Frequency of cytochrome P450 CYP2C9 variants in a Turkish population and functional relevance for phenytoin.Br J Clin Pharmacol48409415
  42. 42. Rieder MJ, Reiner AP, Rettie AE (2007) Gamma-glutamyl carboxylase (GGCX) tagSNPs have limited utility for predicting warfarin maintenance dose. J Thromb Haemost 5: 2227–2234.MJ RiederAP ReinerAE Rettie2007Gamma-glutamyl carboxylase (GGCX) tagSNPs have limited utility for predicting warfarin maintenance dose.J Thromb Haemost522272234
  43. 43. Carlquist JF, Horne BD, Muhlestein JB, Lappe DL, Whiting BM, et al. (2006) Genotypes of the cytochrome p450 isoform, CYP2C9, and the vitamin K epoxide reductase complex subunit 1 conjointly determine stable warfarin dose: a prospective study. J Thromb Thrombolysis 22: 191–197.JF CarlquistBD HorneJB MuhlesteinDL LappeBM Whiting2006Genotypes of the cytochrome p450 isoform, CYP2C9, and the vitamin K epoxide reductase complex subunit 1 conjointly determine stable warfarin dose: a prospective study.J Thromb Thrombolysis22191197
  44. 44. Zhu Y, Shennan M, Reynolds KK, Johnson NA, Herrnberger MR, et al. (2007) Estimation of warfarin maintenance dose based on VKORC1 (-1639 G>A) and CYP2C9 genotypes. Clin Chem 53: 1199–1205.Y. ZhuM. ShennanKK ReynoldsNA JohnsonMR Herrnberger2007Estimation of warfarin maintenance dose based on VKORC1 (-1639 G>A) and CYP2C9 genotypes.Clin Chem5311991205
  45. 45. Miao L, Yang J, Huang C, Shen Z (2007) Contribution of age, body weight, and CYP2C9 and VKORC1 genotype to the anticoagulant response to warfarin: proposal for a new dosing regimen in Chinese patients. Eur J Clin Pharmacol 63: 1135–1141.L. MiaoJ. YangC. HuangZ. Shen2007Contribution of age, body weight, and CYP2C9 and VKORC1 genotype to the anticoagulant response to warfarin: proposal for a new dosing regimen in Chinese patients.Eur J Clin Pharmacol6311351141
  46. 46. Ohno M, Yamamoto A, Ono A, Miura G, Funamoto M, et al. (2009) Influence of clinical and genetic factors on warfarin dose requirements among Japanese patients. Eur J Clin Pharmacol 65: 1097–1103.M. OhnoA. YamamotoA. OnoG. MiuraM. Funamoto2009Influence of clinical and genetic factors on warfarin dose requirements among Japanese patients.Eur J Clin Pharmacol6510971103