COAG, EU-PACT, and a tale of three trials: pharmacogenomic-guided warfarin dosing

This entry is the third part of a series on late-breaking clinical trials from the American Heart Association Scientific Sessions 2013. For a list of all reviewed trials, click here.

Summaries:
COAG: In the Clarification of Optimal Anticoagulation through Genetics (COAG) study [1], 1015 patients at sites throughout the US were randomized to an initial warfarin dosing strategy that incorporated both pharmacogenomic and clinical factors or clinical factors only. After 5 days, both arms were managed according to a standardized adjustment algorithm. At 4 weeks of follow-up, there was no difference in the primary outcome of INR time-in-therapeutic range (TTR) between the two groups (45.2% vs. 45.5% in the pharmacogenomic and clinically-guided group, respectively, p = 0.91). A difference was found only in the subgroup of self-identified black patients, where TTR was lower with the pharmacogenomic-guided strategy (35.2% vs. 43.5% with control, p = 0.01). No differences in thrombotic or bleeding events were observed.

EU-PACT trials: In a similarly designed study conducted by the European Pharmacogenetics of Anticoagulant Therapy (EU-PACT) group [2], 548 patients were randomized to a warfarin dosing strategy that used both pharmacogenomic and clinical factors or clinical factors only. After the first 5-7 days, patients were managed according to local standards of practice. At a follow-up of at least 10 weeks, no difference in the primary endpoint of TTR was observed between the two groups (61.6% vs. 60.2% in the pharmacogenomic and clinically-guided groups, respectively, p = 0.52). An improvement in TTR was observed with pharmacogenomic-guided dosing at 4 weeks (52.8% vs. 47.5% with control, p = 0.02), a pre-specified secondary endpoint of the study. No differences were observed in specific subgroups, nor were there any differences in clinical outcomes.

In a second trial conducted by the EU-PACT group [3], 455 patients were randomized to a dosing strategy comprised of both pharmacogenomic and clinical factors or usual care, which was defined as an initial dosing strategy of 10 mg, 5 mg, and 5 mg for days 1-3 among patients < 75 years of age and 5 mg daily for days 1-3 in patients > 75 years. After 5 days, patients were managed according to local standards of care. After a follow-up period of 12 weeks, the pharmacogenomic-guided strategy improved TTR by a mean 7% (67.4% vs. 60.3% with usual care, 95% CI 3.3-10.6, p < 0.001). The median time to reach therapeutic INR was 8 days shorter in the pharmacogenomic-guided group (21 days vs. 29 days with usual care, p < 0.001). No differences in clinical endpoints were observed.

Commentary:
Let me be first to admit that I missed the fact that there were three studies of pharmacogenomic-guided warfarin dosing strategies published simultaneously. The major headline from AHA13 was that this strategy did not confer significant improvements in the management of therapy, and I walked away thinking that was the end of the story. I even tweeted so. However, based on the findings of the third study, I think further discussion of these trials is warranted.

Note: For the sake of simplicity, I will refer to the EU-PACT study that compared pharmacogenomic plus clinical factors versus clinical factors alone as EU-PACT-1 and the study comparing pharmacogenomic plus clinical factors to usual care as EU-PACT-2.

First, the COAG and EU-PACT-1 trials were almost identical in design, with the main difference being their length of follow-up (4 weeks in COAG vs. a minimum of 10 weeks and goal of 12 weeks in EU-PACT-1) [1,2]. Both obtained genotypes for the major polymorphisms expected to influence warfarin management (i.e., CYP2C9 an VKORC1) and compared 5 days of an algorithm that combined this information with clinical features already known to impact warfarin dosing (intervention) versus an algorithm that used only clinical features (control). While this design was essential for determining whether the addition of genetic information would influence outcomes, one should be cautious not to interpret the control arm as representing the standard of care for most practices. As an example of how complex the control algorithm was, the following equation was used to determine the initial dose of warfarin in COAG:

Dose (in mg/day) = exp[0.613 – (0.0075 x age) + (0.156 if black race) + (0.108 if smokes) + (0.425 x body surface area) – (0.257 if on amiodarone) + (0.216 x target INR) + (0.0784 if indication for warfarin is DVT/PE)]

While these and other dosing calculations have been validated in smaller studies, I am not aware of any large robust clinical trials comparing them to usual care. Even so, they are not included in current practice guidelines and are therefore unlikely to be widely used.  Whether or not they have been validated, they certainly incorporate many of the features one should consider when managing warfarin therapy.  Given their inclusion in the dose determinations for these two studies, I am not surprised that the addition of genetic characteristics failed to have an incremental impact on TTR control in a study of only a few hundred patients.  I imagine many thousands of patients would be necessary to distinguish the impact of genetic characteristics on top of all of the other adjustments already included in the equation.

Although no differences in TTR were observed at 4 weeks in COAG, an improvement with the pharmacogenomic-guided dosing strategy was observed at this time interval in EU-PACT-1 (although the primary analysis was conducted at 12 weeks). The cause of this discrepancy is unclear but the poor performance of the pharmacogenomic-guided algorithm among blacks in COAG (40% of the study population) may have been a contributor.

The design of EU-PACT-2 represents a more accurate comparison of a pharmacogenomic-guided dosing strategy and usual care [3].  While some ambulatory care practices and inpatient consultation services may utilize equations similar to the ones highlighted above, the vast majority of patients receiving a diagnosis that warrants anticoagulation therapy are initiated on a fixed dose with subsequent adjustments for changes in INR. The specific dose selected often incorporates many of the features included in these dosing equations, but not in a formal sense. When compared to this usual care approach, EU-PACT-2 demonstrated that a strategy incorporating both pharmacogenomic and clinical factors improves TTR control.  Unfortunately, because an arm comprised of patients being managed using only clinical factors was not included in this trial, it is impossible to know the true incremental impact of genetic information on warfarin dosing.

Finally, all three trials were conducted mostly at large academic medical centers with access to specialist providers, including pharmacists specifically trained in the management of anticoagulation therapy.  Although the intervention was largely blinded for the first 5-7 days, the specialized care provided at these centers may have influenced the potential impact of the pharmacogenomic-guided strategy by the end of study follow-up at 4-12 weeks. What would have been more interesting to know is whether this strategy would have conferred improvements among patients managed by general care practitioners or those discharged from the hospital who were unable to follow up for INR management until 1-2 weeks later. Perhaps my practice environment has biased me, but I feel these latter scenarios are far more common.

Bottom line:
The addition of genetic information to an algorithm already incorporating clinical features known to influence warfarin dosing does not improve management at 12 weeks, although differences may be observed in certain subgroups or at earlier time points. On the other hand, an algorithm incorporating both genetic and clinical information significantly improves warfarin management when compared to usual care.


References

1. Kimmel SE, French B, Kasner SE, Johnson JA, Anderson JL, Gage BF, et al. A Pharmacogenetic versus a Clinical Algorithm for Warfarin Dosing. N Engl J Med. 2013 Nov 19;
2. Verhoef TI, Ragia G, de Boer A, Barallon R, Kolovou G, Kolovou V, et al. A Randomized Trial of Genotype-Guided Dosing of Acenocoumarol and Phenprocoumon. N Engl J Med. 2013 Nov 19;
3. Pirmohamed M, Burnside G, Eriksson N, Jorgensen AL, Toh CH, Nicholson T, et al. A Randomized Trial of Genotype-Guided Dosing of Warfarin. N Engl J Med. 2013 Nov 19;