NEW YORK (GenomeWeb) – In the late 1990s, when genetics researcher Scott Diehl first came across data on Interleukin Genetics' test for assessing the risk of gum disease, he immediately noted weaknesses in the analyses.
"I was very troubled that the [genetic] test didn't have scientific validity," said Diehl, who from 1993 to 2002 was the chief of the molecular epidemiology and disease indicators branch at the National Institutes of Health's National Institute of Dental and Craniofacial Research.
Now as a professor of oral biology and health informatics at the Rutgers School of Dental Medicine, Diehl's concerns about Interleukin's periodontal disease risk tests – an earlier version called PST and a newer iteration called PerioPredict – have only grown as the company has commercially launched the tests.
A research team led by Diehl recently reanalyzed a study that Interleukin says supports the use of its tests, alongside conventional risk factors, to determine whether a person is at low risk of periodontal disease and can do well with just one annual dental cleaning as opposed to the standard two cleanings most people get.
In its reanalysis, published in the March issue of the Journal of the American Dental Association, Diehl's group was unable to find statistical support for any association between PST and tooth extraction — the patient outcome that, according to Diehl, is ultimately of interest in evaluating whether or not to use the test — or any interaction between the test and preventive visits. Such evidence, Diehl said, "is an essential requirement for changing the standard of dental care based on the use of this test."
However, Interleukin and its University of Michigan collaborators on the reanalyzed study argued in a counterpoint in JADA that the study was never designed to find an association with tooth extractions.
"Reanalyzing study data is always reasonable, but the approach used by Diehl and [colleagues] was fundamentally flawed," Interleukin Founder and CEO Kenneth Kornman told GenomeWeb in a statement. "It was less a reanalysis than a misuse of the University of Michigan study database, asking a question of the study that it was neither designed, nor powered, to answer."
Interleukin's PST gauges two interleukin-1 (IL-1) gene variations. In late 2013, the company transitioned to the newer PerioPredict, which gauges four different IL-1 variations. The company acknowledges that PerioPredict gauges different IL-1 markers than PST, but in the original study on the PST test, researchers from Interleukin and elsewhere noted that the newer test yields comparable results in terms of which patients need one or two annual teeth cleanings. Diehl and colleagues also challenged this assertion with their reworking of the data.
"The IL-1 PST and PerioPredict genetic tests are not equivalent, as they classify patients' risks differently, and neither test has any evidence supporting their use for patient stratification or any other aspect of dental care," his group concluded in the JADA paper.
Given what Diehl believes to be lackluster evidence backing the ability of the test to gauge patients' outcomes (e.g. whether they're at risk for tooth extractions), he is particularly concerned that Interleukin has inked deals with dental insurers who are already using the test to determine whether they should cover only one cleaning for those the test deems to be at low risk for severe gum disease, while continuing to reimburse for two cleanings for high-risk patients.
According to Interleukin, around 500 million yearly dental visits in this country cost more than $100 billion and preventive visits comprise 76 percent of dental services. However, a preventive dental care strategy that includes its gum disease genetic risk test may help lower healthcare expenses, suggests Interleukin.
Interleukin published a financial analysis in its original paper that showed a preventive dental care strategy for patients deemed to be low risk for tooth extraction by their PST test has the potential of saving approximately $4.8 billion annually in a population of 175 million covered adult lives. However, Diehl raised the concern that since his analysis shows the test has no significant association whatsoever with this patient outcome, such a cost saving strategy would harm many patients' dental health.
Ultimately it boils down to a disagreement over the extent to which IL-1 markers underlying Interleukin's test contribute to the risk of gum disease and periodontitis progression. The firm claims on a website for PerioPredict that IL-1 genetic variations gauged by the test are the "most prevalent known risk factor associated with development of severe periodontitis, found in approximately 30 percent of the population." However, Diehl points out that the published data from genome-wide association studies to date has shown that IL-1 variants do not contribute significantly, if at all, to periodontitis risk.
Interleukin states on the website for PerioPredict that people who have a negative test can still get gum disease and tells doctors to advise their low-risk patients that it's important to care for their teeth at home and come in for regular checkups. For high-risk patients, the company advises doctors to tell patients they won't necessarily get periodontitis, but they are at heightened risk for the condition and their insurance plan may cover more frequent annual preventative visits.
Interleukin has financial ties with the dental insurance industry. In 2012, the Delta Dental Plan of Michigan purchased 500,000 shares of Interleukin for $3 million, giving a Delta Dental representative a seat on the company's board (that board member has since been replaced). As a large provider of dental benefits, Delta Dental also covers PerioPredict in certain employer groups.
Interleukin is also in the midst of a three-year deal with Renaissance Health Service Corporation, which in 2013 agreed to pay a fixed reimbursement price for every PST genetic test processed by Interleukin for a Renaissance customer. RHSC has also funded a study involving Interleukin's periodontal disease risk test.
In an earnings call last week, Interleukin's Kornman said he was surprised that JADA would publish the reanalysis by Diehl's group, calling the work intentionally misleading. While acknowledging he himself is an academic, Kornman characterized the paper as the "perspective of academics … that like to take pot shots occasionally." Kornman was a department chair and professor at the University of Texas Health Science Center at San Antonio and also holds academic appointments at Harvard University.
Characterizing Diehl et al.'s analysis as a "flawed approach," Kornman said the JADA paper "contradicts even long-standing evidence that more preventive care and monitoring is needed to manage patients who smoke or have diabetes."
Different conclusions
In the large observational study that Interleukin often cites as evidence to support its tests and that Diehl's group reanalyzed, researchers from the company and elsewhere genotyped some 5,000 patients who were in Delta Dental of Michigan's insurance claims database and had 16 years of claims data.
In this study, published in 2013 in the Journal of Dental Research and funded by the NIH and Renaissance Health Services, researchers led by the University of Michigan's William Giannobile stratified people into high- and low-risk groups for development and progression of periodontal disease based on three risk factors: diabetes, smoking, and results of the PST genetic test. Patients were deemed low risk if they were non-smokers, not diabetics, and didn't have the IL-1 risk variants gauged by the PST test; and high risk if they had any one of the three risk factors.
Giannobile et al. wanted to find out whether in the high- and low-risk groups the number of dental visits was associated with the level of tooth loss. They reported that in the low-risk population, tooth loss wasn't associated with the number of yearly dental cleanings. But for high-risk patients who had one of the three risk factors, those who visited the dentist twice a year had lower rate of tooth loss than those with one visit. And the more risk factors they had, the more tooth loss events they experienced, the study authors reported, suggesting that people with more than one risk factor may need more than two preventative visits annually.
It is worth noting the Giannobile, chair of the University of Michigan School of Dentistry's department of periodontics and oral medicine, is also the editor-in-chief of the Journal of Dental Research, the journal in which the aforementioned analysis was published.
In their recent reanalysis in JADA, Diehl and colleagues came to very different conclusions. Diehl was able to get additional data from Giannobile et al. to garner information about the number of study participants who got one or two dental visits and the percentage of those that had one or more tooth extractions over a period of 16 years. They then conducted univariate statistical analysis to determine if smoking, diabetes, the PST test, and the number of preventative visits were independently associated with tooth loss; multivariate analysis to assess if these factors act independently on the risk of tooth loss; and interaction tests to gauge if genetic test results interacted with the number of preventative visits.
Diehl's team concluded that diabetes and smoking are risk factors associated with increased risk of tooth extraction, and two annual preventative dental care visits are associated with lower risk. But upon statistical analysis Diehl and colleagues were unable to find that the PST genetic test was associated with the number of tooth extractions. Moreover, the number of preventive visits viewed in light of patients' different risk factors, including their genetic test results, didn't seem to make a difference in terms of outcomes.
Most of the primary analysis in the Giannobile paper used the PST test; some secondary analysis involved the more recently launched PerioPredict. Giannobile et al. noted in their paper that the two versions of the genetic tests – PerioPredict and PST – yielded "comparable" results in discerning which patients needed one or two preventative visits. Moreover, a study that was partially funded by Interleukin, co-authored by company scientists, and published in the Journal of Periodontal Research earlier this year concluded that different IL-1 genetic markers assessed by PerioPredict influence the severity of chronic periodontitis in different ethnicities.
But Diehl's group found fault with the assertion that the old and new test yield comparable results. They used data from the 1,000 Genomes database, comparing the "high risk" and "low risk" classifications by PST and PerioPredict in four ethnic groups and found that in the same ethnic population, the two tests classified different percentages as "high risk."
For example, the PST test deemed 11 percent of African Americans from the Southwest US as being at high risk of periodontal disease. But when this population was tested by PerioPredict, there was a fivefold increase in the number of people deemed to be high risk. Furthermore, Diehl and colleagues reported that 56 percent of African Americans and 28 percent of Caucasians are risk classified differently under PST versus PerioPredict.
No GWAS support
In Diehl's view, the reason that his group came to such divergent conclusions arises from the fact that IL-1 genetic markers are only weakly associated the risk of periodontal disease. Moreover, tooth extractions can happen in middle aged adults due to a number of factors, with periodontitis being only one such cause.
The IL-1 markers that Interleukin's tests analyze were discovered by candidate gene approaches, which investigate the association between a genetic variation and a phenotype based on prior knowledge of the gene's impact on a particular disease or trait.
"These interleukin genes aren't within 100 miles of being the most significant findings in the genome-wide association studies that have been performed on periodontitis," Diehl said. "They're not even close." GWAS approaches, in contrast, scan the genome for common SNPs in people with and without the phenotype of interest.
In Diehl's view, Interleukin has latched on to the IL-1 markers because the firm has IP around them. "They've got a hammer in their hand, so everything in the world looks like a nail," he said.
In an editorial published in the same issue of JADA as the Diehl study, John Ioannidis, professor of health research and policy at the Stanford School of Medicine, lauded Giannobile et al. for sharing their data with Diehl's group, but seemed to put more weight on Diehl's reanalysis.
Noting the reanalysis by Diehl's team "is more appropriate than the original analysis," Ioannidis said that the IL-1 genetic variants included in the PST and PerioPredict tests do not discriminate between patients who experience tooth loss or those who would benefit from two or one dental cleanings.
"Diehl and colleagues correctly pointed out how the weak evidence on the genetic risk factors and the artificial dichotomy of 'low-' and 'high-risk' patients in Giannobile and colleagues study led to misleading conclusions," he wrote. "With no discriminating ability, the PST and PerioPredict genetic tests lack rationale for clinical use." Ioannidis further characterized the IL-1 markers as lacking genome-wide significance for their association to periodontitis risk.
In its counterpoint to Diehl's analysis in JADA, Giannobile's group noted that its study was never intended to assess whether the risk factors predicted tooth extraction outcomes or to gauge whether the individual risk factors had any associations with the number of dental cleanings that Diehl's group explored. They said their aim was to determine whether tooth loss differed for patients who got one or two cleanings in high- and low-risk groups.
"We recognized that the data set lacked the statistical power either to assess each risk factor individually or the interaction of each risk factor with the number of cleanings in this study," Giannobile and colleagues wrote, accusing Diehl's group of misinterpreting their study and misunderstanding the high- and low-risk patient groupings.
Interleukin's Kornman further told GenomeWeb in an email that the individual influence of IL-1 markers on tooth loss has been addressed by many studies. "Elevated IL-1 levels are well established clinically and scientifically as a critical determinant of periodontitis severity, progression, tooth loss, and disease recurrence," he said. "We believe, and the evidence supports, that testing for IL-1 gene variants provides the dental professional with important information to assess prevention and treatment strategies for their patients."
Giannobile and colleagues acknowledged in their counterpoint that GWAS studies haven't born out the significance of IL-1 markers to periodontitis risk, but noted that this doesn't necessarily discount the markers.
"For example, in the Framingham risk score for predicting 10-year risk for a cardiovascular event, high total cholesterol alone in a 50-year-old man accounted for only approximately 12 percent of the heart attacks," they wrote in their counterpoint. "Yet we know that lowering cholesterol levels within the context of a few other risk factors is still regarded as valuable in the clinical prevention and management of cardiovascular disease."
In the end, Diehl is concerned that Interleukin is cherry picking data from studies that support the use of its test for gauging gum disease risk. Diehl is willing to acknowledge that IL-1 gene variants may have a small, weak association with periodontitis risk, but probably account for as little as 1 percent of the risk. "But to jump from there and say you have a clinically valid test for periodontitis [risk] … there is absolutely no evidence for that," he said. Diehl added that his comments to GenomeWeb are his own views and not those of Rutgers University.
In Interleukin's earnings call last week, Kornman noted that this year the company will focus on expanding the market for PerioPredict through discussions with payors and employer groups. Specifically, the company plans to reach out to entities in charge of designing reimbursement coverage policies so the test might be paid for under preventative dental care schemes.
The company is also planning to use the IL-1 markers in cardiovascular disease and osteoarthritis, which Diehl is further troubled by, having also noted data analysis issues in publications related to these disorders. Interleukin is providing IL-1 and lipoprotein(a) SNP testing for studies investigating Isis Pharmaceuticals' ISIS-APO(a)-LRx, and is involved in developing osteoarthritis genetic tests. The company introduced its Inherent Health brand of wellness-oriented genetic tests in 2009 for weight management, heart health, bone health, and nutrition.
The plan to bolster adoption and reimbursement this year for PerioPredict, currently the company's flagship test, comes after the company recorded a 2014 net loss of $6.3 million compared to $7.1 million in 2013. Interleukin also reported $1.8 million in revenues in 2014 compared to $2.4 million in 2013. Interleukin said the 2013 revenues were greater due to higher volumes of pre-paid test kits due to an Amway promotion and additional royalties. In a recent 10-K filing with the Securities and Exchange Commission, Interleukin said it doesn't "expect to receive any significant revenues from PerioPredict until 2016, at the earliest, and the timing of any such revenues may be substantially later."
This article has been updated to clarify that 2013 revenues were greater than in 2014 because of more revenue from test kits sales, not necessarily attributable to PerioPredict, and provide additional information about when the company expects to realize significant revenues from the test.