NEW YORK (GenomeWeb) — A study by researchers from the University of Florida and the University of Illinois has demonstrated why dosing algorithms for warfarin have performed poorly in patients with African ancestry, especially in recent randomized clinical trials, showing that such algorithms can mis-dose specific genotypes important for the African American population.
The study, published recently in the journal Pharmacogenetic and Genomics, both confirms and provides more direct explanation for troubling findings from the COAG trial last year, in which African American patients fared worse on a genotype-guided warfarin dosing strategy than when given the drug using a clinical algorithm alone.
According to Larisa Cavallari, the senior author of the new study, researchers speculated at the time of the COAG publication that the reason African Americans did worse with genotype-guided dosing was because the variants that are prominent and important in this population were not part of the PGx algorithm used in the trial. However, no one had shown directly that this was the case.
In her study, Cavallari set out to measure directly how well the algorithm used in COAG — as well as another warfarin dosing algorithm developed by the International Warfarin Pharmacogenetics Consortium and used in a separate clinical trial, EU-PACT — would predict the dose for African Americans carrying several specific genotypes known to be important in this population in regard to differences in warfarin response.
The researchers examined the association of four CYP2C9 alleles, and the rs12777823 G>A genotype in a cohort of 274 warfarin-treated African Americans, and compared their actual stable warfarin doses with what would be predicted by either the warfarindosing.org algorithm used in the COAG trial, or the IWPC algorithm used in EU-PACT.
Overall, the COAG algorithm tended to predict higher doses than the IWPC algorithm across the different genotypes studied, and both algorithms overestimated the dose for patients with CYP2C9 variants.
The authors wrote that the COAG algorithm overdosed warfarin by more than 1 mg/day in 64 percent of the study cohort with a CYP2C9 variant, and 70 percent of those homozygous for the rs12777823 variant allele. The IWPC method, meanwhile, overdosed more than 1 mg/day in 47 percent of CYP2C9 variant allele carriers and underestimated doses in 42 percent of rs12777823 wild-type homozygotes.
The group also calculated how the COAG algorithm would perform with the genotypes studied taken into consideration.
The algorithm allows, through warfarindosing.org, an option for inputting CYP2C9*5 and CYP2C9*6, although this was not done in the COAG trial. Cavallari and her colleagues made this change, and then estimated a 20 percent dose reduction for patients with CYP2C9*8 and *11 alleles, which cannot be added to the algorithm, and by 7 mg/week for rs12777823 GA, and 9 mg/week for rs12777823 AA, which also cannot be added.
The dosage predicted by this combined strategy was very similar to the actual observed average dose in the cohort, the authors wrote.
The team also looked briefly at two other existing algorithms, one published by a team led by Wenndy Hernandez of the University of Chicago, and another by a group led by Vanderbilt University's Andrea Ramirez.
The Hernandez group's algorithm, which includes CYP2C9*5, CYP2C9*8, CYP2C9*11, and rs12888823, performed well in variant allele carriers, but under-dosed non carriers, Cavallari and her coauthors reported.
The other method, by the Ramirez group, which includes CYP2C9*6, CYP2C9*8, and CYP2C9*11, but not CYP2C9*5 or rs12888823, performed well in those without a variant, but overdosed variant allele carriers, mainly due to overdosing those with rs12777823 variants.
Overall, Cavallari said, the findings provide important direct evidence of why African Americans, who comprised about one third of the total COAG trial cohort, did worse with PGx dosing versus non-PGx dosing.
Though the IWPC algorithm also showed problems in dosing African Americans in Cavallari's study, its use in the EU-PACT trial was in a cohort of 99 percent Caucasian European subjects. Fittingly, in that trial, PGx-guided dosing showed a significant improvement over clinical dosing in terms of patients'' time spent in therapeutic INR range during the initial 12-weeks of treatment.
In COAG, meanwhile, investigators found that in the first four weeks of treatment, genotype-guided warfarin administration was no better than the clinical algorithm in terms of the mean time that the two groups remained in therapeutic INR range.
According to Cavallari and her coauthors, their data help explain these divergent results, as well as illustrate how important accounting for population-specific variants is and will be in genotype-guided dosing for warfarin, and potentially for other drugs as well.
One more large clinical trial of genotype-guided warfarin dosing is currently still underway. Iverson Genetics has said it is hoping to report final data from its Warfarin Adverse Event Reduction for Adults Receiving Genetic Testing at Therapy Initiation, or WARFARIN, study by the end of 2015. However, according to Cavallari and her coauthors, genotyping in this trial is also limited to variants important mainly to Europeans.
Cavallari said that she's not sure whether there would be funding support for a larger prospective trial in non-European populations, but added that she's not sure one is necessary.
"Clinicians have asked for randomized controlled trial data showing genoytping improves outcomes over not genotyping, but I'm not sure that we need that."
"If we have strong data showing a connection between genotype and pharmacodynamic and pharmacokinetic data, and we know those patients need a lower dose, why shouldn't we use that to help dose patients?" she asked.
"The problem with any of these large trials is that [PGx dosing] is not going to benefit most people in the trial. Most patients have genotypes associated with a normal dose, and it's only the outliers where you see a benefit, so to have enough power to detect that is very difficult," she explained.
What will be a more pressing question, Cavallari said, is establishing the best ways for this information to be used clinically, in anticipation of growing use of preemptive genotyping.
"With preemptive PGx, you are taking the cost question out of it, so the need for randomized data on improved outcomes is less important. If it benefits some and doesn't harm others, then why not use genotype data," she said. "The problem is that COAG showed that doing this wrong could actually harm African Americans … so I think it will be important that we do have algorithms that account for variants across race groups."
Moving forward, Cavallari said that it will also be important to continue to look at how additional newly discovered variants affect predicted dose, as well as focus research on other ethnic and racial groups, like Hispanics, in which different variants may play a role in warfarin dose requirements.