Originally published Oct. 8.
By Turna Ray
A recently published meta-analysis that sheds doubt on the validity of the KIF6 variant as a pharmacogenomic cardiac biomarker has inspired an editorial from two cardiologists cautioning test developers and investigators from rushing to market with gene markers with limited clinical validity.
An international team led by Themistocles Assimes of Stanford University this week published a study in the Journal of the American College of Cardiology looking at the association between the KIF6 Trp719Arg allele and the development of coronary artery disease in more than 17,000 CAD patients and more than 39,000 controls from 19 different studies conducted globally.
The study, which failed to find any association between the KIF6 719Arg allele and CAD, "puts the nail in the coffin” for KIF6 as a marker for gauging risk of coronary disease, Tom Quertermous, William G. Irwin Professor in Cardiovascular Medicine at Stanford and the study’s senior author, said in a statement. “This is such a big study — if there was a significant association between this variant and coronary disease, we would have found it.”
Additionally, the study casts a shadow on Celera's claims that its StatinCheck KIF6 test can predict which patients with heart disease will benefit from statins. “Until very large-scale studies are performed to directly test the marker’s ability to identify statin responders, I would not withhold statins from patients just because their KIF6 test was negative," lead author Assimes said in a statement.
Reviewing the data in an editorial accompanying the study, cardiologists Eric Topol and Samir Damani of the Scripps Research Institute asserted that Celera's StatinCheck test was commercialized without appropriate clinical validation of the marker's impact on CAD and statin response in patients with coronary artery disease.
Urging investigators to vet pharmacogenetic markers "in well-designed, hypothesis-free, genome-wide studies before promoting their use in clinical practice," Topol and Damani wrote in JACC that while there are several cardiovascular-related PGx markers — including apolipoprotein E, LPA, and certain CYP2C19 markers linked to Plavix response — that meet a high evidence threshold for clinical validity, "the KIF6 association has lacked such data from the time of its initial reports.
"Going forward, the KIF6 story should serve as a valuable reminder of the potential pitfalls present in prematurely adopting a genomic test without sufficient evidence," Topol and Damani concluded in the editorial, titled "The KIF6 Collapse."
Celera's Defense
Celera finds the JACC study flawed, however, and accused the Topol/Damani editorial of containing "numerous inaccuracies, misstatements, and omissions." The company's main quibble with the meta-analysis is that in the cohorts analyzed there was no reliable information available on the use of statins. As such, if the majority of patients in these studies were already on statins then this would bias the study's findings on KIF6's link to CAD.
Additionally, Celera noted that the study only looked at patients with nonfatal cases of CAD. Celera thus dismisses the study as irrelevant to the statin claims of its test and claims that its conclusions are old news, since "Celera and others have previously [published] the finding that in case-control studies there was no association between KIF6 and nonfatal MI or CAD."
[ pagebreak ]
Celera is currently in the midst of a strong marketing push for StatinCheck. In response to the meta-analysis and accompanying editorial, it issued a strongly worded press release in defense of its previous claims for the KIF6 test.
If these claims are valid, then the company stands to gain substantial revenue from marketing the test to the approximately 180 million people who are estimated to be carriers of the KIF6 variant worldwide. Earlier this year, Celera CEO Kathy Ordonez said that its subsidiary Berkeley HeartLab has performed nearly 200,000 KIF6 tests since the company launched StatinCheck in the US two years ago.
In an effort to grow the company's cardiovascular test sales, which the company believes will be its future growth driver, the company recently hired 23 sales reps to market the test in the US, and expanded a partnership with Abbott this year to sell StatinCheck in the EU. The company has also inked deals with external labs in Europe and in the US to sell the test.
Celera is working with pharmacy-benefit manager Medco to build the clinical utility evidence in support of KIF6 testing. Last year, Medco and Celera launched a prospective, randomized study to evaluate whether patients' knowledge of their KIF6 gene variant status increases their adherence to statin therapy compared to those not offered the test (PGx Reporter 09/23/09). Positive findings from this Medco study would likely lead to the inclusion of the KIF6 test in Medco's personalized medicine program. Medco did not respond to questions for this article.
With a price tag of approximately $100 per test, the KIF6 test is sold as a prognostic test for gauging coronary heart disease risk and as a predictive test for statin benefit.
Celera claims it has validated the 719Arg allele link to a reduction in risk of coronary heart disease in a combined analysis of seven large prospective studies, and has proven the association between KIF6 and statin benefit through retrospective analysis of four prospective randomized-controlled trials: the secondary prevention Cholesterol and Recurrent Events (CARE) study; the primary prevention West of Scotland Coronary Prevention Study (WOSCOPS); the PROspective Study of Pravastatin in the Elderly at Risk (PROSPER) study; and Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis in Myocardial Infarction 22 (PROVE IT―TIMI 22).
In total, Celera claims that the KIF6 association to coronary heart disease and statin benefit have been validated in approximately 55,000 people, showing that carriers of the KIF6 polymorphism have between a 22 percent and 55 percent greater risk for cardiac events than non-carriers.
Where's the Risk?
In the JACC study, researchers from Stanford, Decode Genetics, Massachusetts General Hospital, and numerous other research centers reported that "none of the 19 studies demonstrated an increased risk of CAD in carriers of the [KIF6] Trp719Arg allele compared with non-carriers." Additionally, the investigators observed that European carriers of the 719Arg allele who had early onset of CAD did not show an increased risk for the disease compared to controls of similar age and compared to non-European subgroups.
The findings led Assimes et al. to conclude that the KIF6 variant was not associated with the risk of clinical CAD. Since the majority of the meta-analysis subjects were of European ancestry, researchers estimated that their findings suggest that for KIF6 mutation carriers in this population, the risk of CAD "is unlikely to be increased by greater than 2 percent compared to non-carriers."
The authors note that the study looked at the impact of KIF6 in early-onset CAD subjects — defined as men younger than 50 years of age and women younger than 60 years of age — but did not find any association between the Trp719Arg variant and CAD, even though previously published studies would suggest that susceptibility alleles would be more prevalent in this subgroup.
[ pagebreak ]
"In this subgroup with early-onset disease, we were able to rule out an increase in risk of 8 percent in Europeans," the study authors wrote in the paper.
Additionally, the researchers also analyzed cohorts of non-European ancestry, and reported no "significant associations" between KIF6 and CAD. However, because the cohorts comprising Hispanic, East Asian, and patients of mixed race or ethnicity were small, the researchers said that their conclusions about KIF6's link to CAD needs further study in these populations "to rule out more modest effects on risk."
The authors also point out three areas of weakness in their analysis, including the inclusion of only non-fatal CAD cases; the inability to measure traditional risk factors of CAD due to the retrospective nature of the study; and uncertainty about the level of statin use among study participants.
With regard to the inclusion of only non-fatal CAD case in the 19 studies analyzed, the authors acknowledge that the exclusion of incident fatal CAD cases could have potentially biased the results. Still, the chances of this are small, the researchers believe, since "the difference in relative risk between these two subgroups of cases would have to be quite large to result in a substantially biased odds ratio in our study, given that the majority of incident CAD events are not fatal, especially among subjects with early-onset disease."
The second potential limitation of the study was that the researchers were unable to measure traditional risk factors "as robustly as they are measured in prospective studies." However, the study authors contend that since prospectively published trials don't suggest that traditional risk factors are correlated with the Trp719Arg polymorphism, they didn't try to adjust the odds ratios for all traditional risk factors of CAD.
The third weakness — and Celera's main contention with the meta-analysis — was that the "study did not allow [Assimes et al.] to explore whether statin use modifies the effect of the 719Arg allele on risk as was done in the WOSCOPS, CARE, and PROVE IT–TIMI 22 trials because reliable information on the use of statins in relation to the incident event was not available for most studies."
However, the authors note that it is unlikely that in the 19 cohorts studied there was a high prevalence of statin use since the study focused recruitment on patients with early onset of CAD, in a younger age bracket.
"We suspect that the overall prevalence of the use of statins in our set of 19 case-control studies is actually lower than the prevalence of use observed at the last follow-up for participants" in previously published studies cited by Celera to support the KIF6 association to CAD, "because a majority of our case-control studies restricted their recruitment and/or genotyping efforts to very early onset cases and young controls," the study authors wrote.
As further evidence of the likelihood of limited statin use in their study cohorts, Assimes et al. cited electronic pharmacy records confirming that around 14 percent of cases and under 5 percent of controls were taking statins "during the appropriate time window of exposure" in the ADVANCE study, which focused on very early-onset CAD in men younger than 45 years and women younger than 55 years.
Finally, the researchers pointed out that the interaction between statin response and KIF6 is questionable because the correlation barely met statistical significant in several studies cited by Celera.
"The interaction term p-value between genotype and statin use was only marginally significant in the WOSCOPS (p = 0.021) and PROVEIT–TIMI 22 (p = 0.018) trials and not significant in the CARE trial (adjusted p = 0.39)," the authors wrote. "Despite these observations, additional high-quality prospective cohort studies of the effect of the KIF6 variant on CAD risk among users and nonusers of statins are needed before any firm conclusions can be made on the validity of this interaction."
A 'Straw Man' Argument?
Celera maintains that Assimes et al. conducted an invalid replication study by looking at cohorts and endpoints that don't match the claims the company is making about the KIF6 test. According to Thomas White, Celera's chief scientific officer, the company is marketing StatinCheck as a test to gauge patients at risk for coronary heart disease, comprising fatal and non-fatal myocardial infarction, cardiovascular death, and "some of the re-vascularization procedures."
[ pagebreak ]
"These authors seem to ignore that's what we claim by stating that what they did [in their study] is what we did, and that's simply not true," Thomas White, Celera's chief scientific officer told PGx Reporter this week.
According to Celera, the claim that KIF6 reduces coronary heart disease risk during treatment with Pravachol was demonstrated in CARE, WOSCOPS, and PROSPER. Additionally, a genetic substudy of PROVE IT―TIMI 22 reported that patients with acute coronary syndrome who were KIF6 carriers experienced a greater reduction in coronary heart disease events when treated with high-dose Lipitor than KIF6 non-carriers.
The KIF6 gene variant has also been reported to predict risk of CHD in prospective population cohort studies, including the Atherosclerosis Risk in Communities, or ARIC, study of 13,907 middle aged Americans; the Cardiovascular Health Study in 4,522 patients aged 65 or older; and the Women's Health study of more than 25,000 patients who are older than 45 years and without a previous history of coronary heart disease.
According to White, the non-fatal CAD cohort is a major flaw that limits the JACC meta-analysis from being a valid replication of the validation studies conducted by Celera to develop and market StatinCheck.
"The key points of the Assimes article are that the KIF6 variant is not really associated with atherosclerosis-related disease. It's a coronary artery study, and some of those people have a non-fatal myocardial infarction in the sub-study, so many of them just had atherosclerosis," White said. "Well, we and others published studies six years ago that KIF6 wasn't associated with atherosclerosis. So, this is absolutely nothing new. They are using a straw man to knock down something we've never claimed before."
Certainly, Celera's marketing materials for StatinCheck promote it as a test that predicts increased coronary heart disease risk and reduction of cardiac events during statin therapy. However, in an April press release, Celera reported data from a mouse-model study that suggests that the company was looking into KIF6's link to atherosclerosis.
In the press release, the company reported that early findings from a study with researchers at the University of Washington and Indiana University suggested that KIF6 was specifically expressed in macrophages and chondrocytes in mouse atherosclerotic lesions, but not in normal arteries. "Multiple independent studies have replicated the association of KIF6 with risk for coronary heart disease and event reduction during treatment, but this is the first biological function study that links the KIF6 protein to atherosclerosis," White said at the time in a statement.
Celera's other objection with the Assimes et al. analysis is that it only includes non-fatal cases. Celera's validation studies include fatal cases, which is "a big difference because you would want genetic markers to predict people who will die from myocardial infarction as opposed to people who just have myocardial infarction. Worldwide over half of heart attacks are fatal," White added.
Approximately 70 percent of the samples studied in the JACC meta-analysis are from patients with coronary artery disease, which is atherosclerosis, and a subset is myocardial infarction. "The fact that the early myocardial infarction cases may not have been taking statins, that's a reasonable point," White acknowledged. "But we didn't claim this was a marker for early myocardial infarction."
Finally, Celera takes issue with the lack of certainty in the study's analysis with regard to the prevalence of statin use.
"They don't know what the statin use in the populations they studied," White said. "And since this is a test for both risk and the reduction of heart events like death in the people who take statins, if any statins are being used in the populations they studied, then it would have prevented the people from being cases. So, they are selecting against people who may be using statins, because those people are protected."
Winner's Curse?
As independent reviewers of Assimes et al., Topol and Damani note in their JACC editorial that Celera's validation studies to identify KIF6's link to CAD "used antiquated candidate gene-based methods," which are often plagued by false positives and the so-called "winner's curse:" the inability of variants identified by candidate-gene methods to be validated in subsequent studies.
[ pagebreak ]
Furthermore, they charged that there is a dearth of mechanistic investigations confirming KIF6's link to atherosclerosis, lipid metabolism, CAD, or MI, and noted that in the study, the researchers found that "no consistent expression of [the KIF6 variant] has been demonstrated in relevant tissues such as vasculature."
Holding up as a cautionary tale the too-rapid commercialization of StatinCheck in the face of questionable clinical data, Topol and Damani urge investigators to validate markers in genome-wide association studies.
"No variant in KIF6 has yet emerged as a statistically significant marker in more than eight GWASs conducted to date on dyslipidemia, CAD, and MI," they pointed out in the editorial. "Such a common variant (present in greater than 35 percent of Europeans) should have clearly surfaced in these GWASs if the association was truly valid."
Celera challenged Topol and Damani's assertion about the robustness of its KIF6 studies, however, noting that the association between KIF6 and coronary heart disease was discovered in a "large gene-centric study of thousands of gene-based SNPs, called putative functional SNPs.
"It's wasn't a picket-fence approach like he means [with] GWAS," White said. "Nonetheless, it's not a candidate-gene approach at all." He added that the replication of the KIF6 association in the seven large randomized-controlled studies "is proof that the marker is real."
Topol and Damani anticipate in their editorial that "proponents of KIF6 testing" will consider Assimes et al. to be biased due to the possibility of high statin use in the cohorts studied. "However, that argument holds little merit," they wrote, pointing out that "several of the initial reports noting the KIF6 association with CAD did not indicate that statin users were excluded or censored before their analysis."
Topol and Damani further agreed with Assimes et al. that the prevalence of statin use was likely to be lower in the early-onset disease patients.
Finally, Topol and Damani noted that there is little mechanistic data in GWAS analysis on KIF6-mediated statin response. "Until such studies are conducted, pharmacogenetic tests with claims of identifying statin hyper-responders or hyporesponders should be viewed with considerable skepticism," they wrote.
Meanwhile, Celera has said it is on track to file for premarket approval from the FDA for StatinCheck before the end of the year.