A scientist once conducted an experiment whereby he taught fleas to jump whenever he said “jump”. His study was prospectively designed to determine the correlation between the number of legs a flea has and hearing. After enrolling 2,500 fleas and teaching them to jump on command, he meticulously removed each flea’s legs one-by-one. After each leg removal, he again commanded the fleas to jump and they did so. Upon removal of the last leg, he again told the fleas to jump. None jumped. His conclusion: Removing all legs from a flea renders a flea deaf. And so goes the hazard of spurious deduction and the topic of this analysis.
Genomic Health (NASDAQ:GHDX) markets a $3,500 test called Oncotype DX that predicts the likelihood of breast cancer recurrence based on the expression levels of 21 tumor-related genes. According to GHDX, oncologists can use the score to determine which patients are most likely to benefit from chemotherapy, thus sparing those that are not likely to benefit from the side effects and insurers the cost, which is estimated at up to $15,000.
Sell-side analysts and the company’s shareholders are electrified by the company since it fits tidily into the standard biotech investing orthodoxy. This orthodoxy demands little more than a large and growing market, hope, hype and a little help from a throng of research analysts willing to surrender their dignity on behalf of management in order to ensure the continued flow of investment banking dollars. In the latter regard, analysts’ efforts have proven valuable as GHDX has burned through over $120 million and will undoubtedly return to the capital markets again and again.
While Oncotype DX may ultimately advance the treatment and health of breast cancer patients there is little evidence to suggest that the company will earn money over the next several years and little evidence to suggest that there is any correlation between enhanced outcomes and the expenditure of $3,500 on Oncotype DX. After marketing Oncotype DX for 2 years, there is no signal of any inherent operating leverage and perhaps this is why the company guided toward a doubling of revenue to $60 million in 2007 while at the same time guiding toward a $28 million loss, which is about the same amount of money the company lost in 2006 on half the revenue. Unfortunately, shorting unprofitable healthcare companies is simply not profitable and burn rate is generally inversely proportional to market capitalization in the short term. However, there are exceptions and GHDX may be one of those. Genomic Health is not a biotech company – it’s not a life sciences company and it’s not a medical device company. It’s a diagnostic company and it deserves a diagnostic company valuation.
While GHDX’s management team perceives themselves to be on the cutting edge of a molecular diagnostics revolution, the company’s most valuable proprietary assets are in essence nothing more than the trademarks associated with the name of their product and the underlying secret algorithms that spew out the results of their test. The company’s patents appear to offer little protection as evidenced by a potential onslaught of competitors that provide or will provide the exact same service with an only marginally-differentiated approach. And while GHDX has established a 3 year lead over their competition, it is but a blink of the eye in the healthcare market. If anything, the benefit of the company’s massive public relations and marketing campaigns will flow primarily to the company’s competitors.
One can argue that this is not an enviable business model and certainly not one that demands an 8x multiple on 2007 revenue. As described below, GHDX’s unusually brief corporate history corroborates the weaknesses. And though GHDX fancies itself a life sciences wonder and a knight in shining armor for breast cancer patients, the fact is that the company exists because of the arrival of RT-PCR technology, inexpensive computers and the resulting ability to analyze massive amounts of data. These resources are widely available to any capitalist paying attention to GHDX’s astounding $475 million+ valuation and there appear to be many who are. AviaraDx, a private company partnered with one of the largest diagnostic service providers (Quest Diagnostics) is likely to be third to market after GHDX and Agendia (whose Mammaprint is the only FDA-approved test for breast cancer recurrence). Aviara’s technology is effectively a reproduction of GHDX’s and is based on equally credible science, but only to the extent that you think either company’s science is credible. The company recently raised a significant amount of capital in a private placement and will likely soon be knocking on GHDX’s door with its first commercial product introduction and will very likely tap the public equity markets to establish a war chest prior to launch.
While both Aviara’s and Agendia’s technology has been dismissed by sell side analysts, who almost seem offended by these companies’ attempts to compete, they have done so without any apparent knowledge of the data supporting Aviara’s product. For example, Lehman Brothers downgraded GHDX on January 17, 2007, specifically citing two factors including the company’s premium valuation and impending competition and yet AviaraDx was not even mentioned as a potential competitor. This may be neglect rather than ignorance but these oversights often blindside investors and lead to the headline shock that can result in a one-day 20% haircut.
Are Genomic Health’s Patents Worth $475 Million?
On August 6, 2006 an article in the New England Journal of Medicine titled “Concordance among Gene-Expression–Based Predictors for Breast Cancer” concluded the following: “Even though different gene sets were used for prognostication in patients with breast cancer, four of the five tested showed significant agreement in the outcome predictions for individual patients and are probably tracking a common set of biologic phenotypes.” One of three derivative conclusions can be reached: either (1) all four concordant tests are accurate with regard to their predictions, (2) all four concordant tests are inaccurate or (3) the one non-concordant test is accurate and all four others are inaccurate. In the end, it doesn’t matter because there appear to be many ways to skin this cat.
Minimal Regulation = Minimal Barriers to Entry
Companies like GHDX and AviaraDx are regulated by the Clinical Laboratory Improvement Amendments, which is overseen by the Centers for Medicare & Medicaid Services. There are currently over 200,000 CLIA certified labs in the United States. While CLIA regulations aren’t toothless and Medicare and Medicaid reimbursement is contingent upon CLIA certification, certification is a cakewalk compared to the requirements for FDA approval for a device or drug and a laboratory only need be approved in one state to reap the benefits of national Medicare/Medicaid coverage. The bottom line is that there is limited governmental oversight of the analytic validity of genetic tests and virtually none over the clinical validity (although the FDA is in the process of determining how to regulate the latter). In many respects, the freedom to operate outside the scope of government interference provides many of the benefits enjoyed by marketers of nutritional supplements who can claim what they want when they want without restriction. This freedom goes to the heart of GHDX’s business model.
Although AviaraDx and Quest have yet to announce CLIA certification, the headline is likely to shave another 15% off GHDX’s stock. And while the timing is uncertain, Aviara’s announcement of its partnership with Quest in December 2006 suggests that it is a near term event. The sell side cognoscenti will undoubtedly articulate their customary objections that “competitive concerns are “overblown” or “premature”, that “weakness in the stock” represents a “buying opportunity” and the equally bizarre assertion that competition will “expand the market”. The fact is that the market is what the market is and the availability of more than one genomic model for the prediction of breast cancer recurrence obviously doesn’t increase the number of women diagnosed with breast cancer.
GHDX was incorporated in August 2000 and launched Oncotype DX in January 2004, representing a little over 3 years of development. R&D expenditures from inception through launch totaled less than $30 million, which is about the cost of getting a drug through phase 1. More important than the dollar value of the expenditures, however, is the disposition of the expenditures. According to GHDX, Oncotype DX has been evaluated in “numerous independent studies involving over 2,600 breast cancer patients.”
The truth is that not a single breast cancer patient had been involved in an Oncotype DX “study” prior to the time the test was marketed. The patients referred to in GHDX’s marketing shtick are archived tumor samples gathered primarily from National Cancer Institute studies that closed years prior to Genomic Health’s existence. This archival tissue is not owned or controlled by GHDX and the “studies” that GHDX relentlessly refers to as “prospective” essentially refer to pre-defined mathematical models. The definitional fraud goes one step further as GHDX refers to the application of their pre-defined models to the tumor samples as “protocols”. And while Genomic Health’s confusion over the definition of the words “patient”, “study” and “protocol” is perplexing and may appear to be immaterial to an investment thesis, it provides valuable insight into the nature of their business and management’s willingness to substitute fact with fiction for marketing purposes.
There is no dispute that Oncotype DX accurately identifies the expression levels of the genes that form the basis of the test but this element of the test is worthless in itself. The value of the test is a function of its ability to establish a valid correlation with the probability of an event (in this case, breast cancer recurrence) and to provide oncologists and patients with actionable information. And clearly, if one can accurately determine the odds of an event occurring and the occurrence of that event can be influenced by taking some action or avoiding some action (in this case, administering or not administering chemotherapy), then we would say that the information is extremely valuable.
It all sounds great and not very complicated. In fact, the exact same process is undertaken by drug companies except it takes drug companies an estimated $800 million and ten years to get from point A to B in contrast to the $30 million and three years that it took GHDX to get their $3,500 test on the market. The difference is that a drug company’s expenditures are rewarded with an average of 10 years of typically unassailable marketing exclusivity. The brief development history of Oncotype DX suggests that either (a) the good people at Genomic Health are remarkable scientists or (b) they took some shortcuts and are doing something entirely different than the drug companies. In fact, they have. But some may object and say that this is an inaccurate comparison; after all Genomic Health manufacturers and markets diagnostics whereas drug companies design and market drugs. It’s therefore a different business. But it’s actually a perfectly valid comparison inasmuch as Genomic Health claims to have conducted numerous prospective and independent clinical trials involving over 2,500 women with breast cancer and Oncotype DX is being marketed on the basis of the ability to maximize treatment benefits.
In that light, GHDX sounds a lot like a drug development company and its valuation suggests that investors feel the same way. Management clearly appreciates the value to their stock price of having conducted “multiple clinical trials” inasmuch as the correlation between absurd valuations and the words “protocol” and “clinical trial” is well established in the bizarre world of biotech. If this isn’t evident to some investors then they need only get a couple of friends together, buy some white lab coats, incorporate a business named “Bio-Something-Or-Other” and knock on one of the many investment banking firm’s doors to present their business plan.
To understand what GHDX has done let’s consider the development of a drug to treat high blood pressure. Aside from the actual clinical trials, there are several other important issues that need to be considered. First, we need a tool to measure the thing we are trying to measure. In this case, its blood pressure and the tool used to measure it is a blood pressure cuff – this tool allows us to calculate a blood pressure “score”. Genomic Health’s blood pressure cuff is the black box computer program that spits out a recurrence score.
Second, we need to be confident that the blood pressure score is correlated with some outcome, the odds of some outcome or even the recurrence of some outcome. This is a more difficult task and often takes decades to establish. For example, it was decades before the relationship between high blood pressure and the long term effects of the condition became evident. This is because it takes a long time for high blood pressure to wreak its havoc.
Third, we need to quantify what constitutes a clinically meaningful change in blood pressure. This too takes many years to establish. Genomic Health has done all of these things in only three years and they have done it without the burden of enrolling a single patient in a trial or conclusively demonstrating a benefit to an actual human being.
The benefits of Oncotype DX instead relate to either women who have already died of breast cancer or women who were fortunate enough not to have died. In neither case, however, did Oncotype DX have anything to do with the outcomes.
Since Genomic Health is all about the validity and utility of its test it is vital to understand the methodological flaws inherent in their “clinical trials”. And while these flaws do not indicate that Oncotype DX is entirely without value to oncologists and patients, they do highlight the limitations of the test and indirectly the weaknesses of the business model. A brief overview of the development process is enlightening.
Step one in developing Oncotype DX involved the analysis of 250 cancer-related genes from publicly available gene databases and publicly available literature. The relationship of the expression of these genes to recurrence was determined by analyzing three historical studies of breast cancer where the outcomes of the patients were known. In total, 447 patient outcomes were studied. Having done this, GHDX identified 21 relevant genes from which it developed an algorithm that generates a Recurrence Score that stratifies patients into risk categories. Step two involved GHDX’s validation “studies” to determine the ability of the score to predict the odds of recurrence of breast cancer.
To do this, GHDX analyzed tumor samples from a 2,617 patient study conducted by the National Cancer Institute from 1982 to 1988. Usable tumor samples were available for only 668 patients in the relevant treatment group and all outcomes of this subgroup were known in advance. So here we are at step two and we’ve already introduced several horrors from the drug development realm – subgroup analysis, non-random populations and the introduction of a multitude of systematic errors. It is well established that human beings cannot choose samples in a random fashion and it is clear that study authors, even in the case of more rigorous randomized, controlled trials, often reject the null hypothesis from the get go. This tendency is exacerbated when greed is substituted for science and the de novo goal of research is to raise money in an IPO.
The bottom line is that investors need to determine what they are investing in and whether they are willing to pay a biotech multiple for a commodity service provider.
GHDX 1-yr chart
Disclosure: Author has no position in GHDX.