Roche's $2.1 billion retrograde bet on Scott Gottlieb's FDA

Can we please put US health tech back in the oven for two more years?

It's tough to admit through the lens of my American Exceptionalism that the US health tech space is a generation behind the curve. I don't blame the current crop of health tech companies, the reason it's so bad is because of highly fragmented & onerous regulatory burdens, and an almost decade-old healthcare bill that's done a lot more harm than good.

If you follow health tech globally, you know that nowhere comes close to what's happening in India and China. Consumer and healthcare technology cycles are synchronized, which has meant their health tech foundations are based on platform plays. It's no wonder their highest valued health tech companies started as doctor booking services, India's Practo and China's Ping An Good Doctor -- about to go IPO last valued ~ $5.39b.

In both those markets, there has been a rapid decline in barriers to entry. For any health tech company that exists in the US, I'll bet a couple software engineers, right out of an Indian Institute of Technology, got together and tried it, failed, pivoted a few times, failed some more, and are probably now looking at their fintech friends with envy. The iteration cycles are rapid and India and China's massive populations allow companies to reach commercially viable network effects and economies of scale. Couple this with a low regulatory burden and that explains why home health aid platforms in India are emerging at an accelerated rate, whereas the US's HomeHero raised $23m and collapsed in less than four years.

The HITECH Act, a well-meaning blunder.

A fundamental flaw in the US is the lack of synchronicity between technology cycles of inpatient care, outpatient services, and consumer goods. The reason for this was a $25.9b premature injection of financing for expanding health information technology through the HITECH Act, passed along with the 2009 American Recovery and Reinvestment Act. Hospitals were the first to take advantage of this program, seeing what appeared to be the obvious benefits of unifying administrative, billing, and patient records. Before passage, ~10% of hospitals and clinics were using computers. Within a few years of passing, 70% of hospitals were filled with hulking desktops and mobile kiosks powered by gigantic batteries.

Physician practices lagged behind, they didn't seem to think the benefits were clear. When hospitals reached 70%, they were stuck at 30%. Their concerns were repeatedly validated in study after study showing little to no benefit from technology adoption, and at a macro-level, the productivity of US healthcare hadn't budged. It wasn't until the threat of penalty provisions in the HITECH act that would decrease their Medicare payments that adoption trends picked up. At this moment, we're above 90% on both.

In the backdrop of these changes was the exponential explosion of computing power and efficiency. By the time clinics started to adopt health information technology en mass, an iPad had the same processing power at a fraction of the cost hospitals had bought their IT infrastructure for. To date, medicine is still searching for the efficiencies technology was supposed to bring, and even now those clinic systems are outdated in comparison to the consumer technology market.

IT overhauls are expensive capital investments that companies generally undertake only when necessary. This has created siloed healthcare systems and inhibited the creation of true platform based health tech entities that can link multiple levels of care delivery. The encouraging note here is hospital systems and clinics are currently updating their legacy infrastructure, resulting in healthcare IT's adoption cycle to sync up with consumer technology for the first time.

This means the technology ecosystem in a few years from now will look nothing like what it is today. If you're a health tech company, you need to brace yourselves for some brutal pivots. In addition, data agnostic technology companies will also be entering into the space by applying general purpose machine learning and AI techniques. You'll need to have answers to some challenging questions: Is there some benefit to being a health tech specialist versus a data generalist? Is there something special about healthcare data that requires separate analytic treatment?

So, what's the point of spending capital right now creating software for a market fragmented into rotten walled gardens?

Roche's health tech investments are defying all market and policy trends.

Coming back to Roche's one-step-behind health investment strategy, I would have thought their less than a year old experience in dropping $100m for the diabetes management platform mySugar would have cooled them off to health tech. Almost simultaneously with their purchase, Medtronic introduced an artificial pancreas diabetes model in the US. While Roche is struggling to figure out how to incorporate mySugar into their own artificial pancreas platform, Medtronic is absolutely crushing the diabetes market. Surprise! It turns out patients don't want a diabetes management app, they want their diabetes managed.

This makes the Flatiron investment even more perplexing, it defies all US healthcare market and policy trends. Roche has a lot of very capable policy analysts, I have to assume they are not involved in Roche's deal teams. There are two sides to the Flatiron business model, both with unfavorable medium-term outlooks. The front facing side is a SaaS model that provides oncology-specific electronic health record software to 265 community cancer centers.

There's not a more endangered species in the great consolidation wars than community cancer centers.

Hospital groups have been purchasing these up at a rapid rate because their high-level services sit near the top of the payments food chain. There's no "uncompensated care" in cancer care, an individual can't show up at an emergency department and get chemotherapy. Community centers were recently granted a small reprieve by reforms to the 340b payments program that had heavily incentivized hospitals to provide expensive pharmacotherapies in the outpatient setting, but Medicare part B payment increases will more than cancel out this deduction for 85% of hospitals.

How is it that stand-alone cancer community centers still exist? The reason is that physicians are paid for cancer medications. Unlike any other field of medicine, doctors providing these therapeutics receive reimbursement for the average sales price of the medication plus an additional 6%. The more expensive medication you prescribe, the more money you make, with some treatment regimens costing over $100,000, it can be a quite lucrative business.

"Buy and bill" is a historical artifact in the payments space, a policy that was adopted during a time when chemotherapies were basically poisons that killed cancer cells at a slightly higher rate than normal ones. These toxic medications justified payments as compensation for having to store and administer them, the payment wasn't supposed to be a revenue driver.

This is a new phenomenon due to the ballooning cost of cancer medications, and it's broadly acknowledged the policy has outlived its usefulness. President Trump's recent budget proposal provides guidance on how increases of buy and bill payment will be linked to consumer price inflation going forward, and it shouldn't be surprising if HHS Secretary Alex Azar attempts to phase out these payments in the next few years.

So the front-end of Flatiron's business model will collapse in the medium term because it's currently propped up by an antiquated payment policy, what about the back?

The back-end of Flatiron's model is to clean and aggregate cancer clinic data to sell to pharmaceutical companies. The company earned itself a year or two of being ahead of other health data entities from its brute force strategy of hiring large teams of medical coders to clean up the unstructured data. Getting slightly in front of the curve was an incredibly costly endeavor, and AI-based data mining techniques are making breakthroughs in unstructured data analysis nearly every day. When Google announced last month that it had created a more accurate adverse event prediction model from seven years worth of uncleaned health records data, one headline read "Google paper stirs interest, but not seen as transformative." We've come a long way.

The next trend they're betting against is patient data ownership and portability. I could careless how iron-clad Flatiron may believe its data use agreements to be, their ability to throw up a barrier to entry by preventing rivals access to their EMR information will not survive the collective onslaught of patient advocacy groups, other (health) tech companies, consolidated healthcare systems, and Alex Azar's HHS. His combative remarks made on Monday on the department's crusade for data portability highlight the administration's stance, which was immediately reemphasized a day later at the HIMSS18 by CMS administrator Seema Verma and Jared Kushner.

Another misjudged trend is what the relative value real-world evidence (RWE) studies will have on accelerating FDA approval for cancer treatments. The presumption here being they can mine clinic data to both use the data to supplement clinical trials for new medications and detect new use cases for old therapeutics. The issue with clinical trials supplementation is you must assume future cancer therapeutics will behave like previous ones. They will not.

Next generation cancer drug discovery investments fall into a few categories: immunomodulators, CAR T Cell, and precision drugs. The underlying biology of these treatment modalities means their benefits will be obvious and pronounced, or they won't work at all. With massive effect sizes, the cohorts necessary for drug approval should decline. We should expect the "unprecedented" 73% response rate to treatment seen in last month's study of untreated renal cell carcinoma to become very "precedented." The phase 1b study had 52 patients and has received breakthrough designation status by the FDA. With results that huge, you don't need anything beyond traditional reporting methods to find new uses and to move it quickly through the approval process.

Roche's own PD-1 inhibitor Tencentriq is already in phase III trials for 18 indications, it seems like the benefit of having Flatiron in this context would be to prevent further drug testing for cancers Tencentriq doesn't work in to prevent large and expensive setbacks like the one last May.

So what's left exactly?

The only unique avenue for Roche to make returns on their investment is if they're able to discover novel combinations of current on the market drugs that could solely be discovered through digitally aggregating large cohorts. This implies the effect sizes of these combination treatments will be reminiscent of the traditionally small response rates of the past. Consider, a 2014 study found that the cancer drugs approved from 2002 to 2014 increased survival by an average of 2.1 months.

One element this strategy has going for it is the FDA commissioner's retrograde stance to the role of his administration in weighing the relative benefits of a drug into the approval process. In an interview with Kaiser Health News, "Gottlieb said it’s not his job to help insurance companies or government programs decide which drugs to cover. Health systems and insurers 'have a difficult time saying no,' Gottlieb said, 'so they want to put the regulator in the position of saying no.' ” Gottlieb's views run counter to the slow evolution of the FDA's stance on the topic as there had been instances in recent history where new applications were rejected for lack of superiority.

Based on these trends Roche has a couple years to squeeze out some high-cost low-benefit treatment regimens, and this may be their calculus. Even if it was only a few hits, the cost of cancer care is so high that could be enough to justify the price tag.

But the macros for US healthcare are so good, I still want investment exposure.

For the time being, I wouldn't be looking at the US healthcare space for outsized returns. There are opportunities -- the market is still acting confused by pharmaceutical-insurer mergers which are ultimately going to be profitable. People are warming up to the model though, when CVS-Aetna announced their merger both share prices declined whereas when the Cigna-Express Scripts merger was announced today only Cigna saw a significant slump. Getting in on a merger bond sale could be of interest for those bored with health stocks and index funds.

In a few years, US health tech may have some viable investment opportunities depending on how much more consolidation occurs. At some stage, there will come a point where it will make sense for incumbents to transition to internal health tech development. Healthcare will always fundamentally be about geography, if I'm massive hospital system and you're a small health tech start-up working in my market, the valuation I'd give you for a buyout is going to be more reflective of the hassle it would be for me to acquire your customers independently, rather than hockey stick growth projections.

Sadly, US health tech will never claim the crown of being the most innovative or disruptive. To know what's on the frontier of possibilities, look to India and China.

Enjoyed this? Check out my previous series of posts on health care, development, and technology. Series 1:

Consumer genomics – a poor value proposition

Ule for retail pharmacy – hurry up and exist

K-Clinics – a morally challenging solution to a global problem

Venture debt – an ecosystem designed to discourage

Blockchain for health and development – an actual silver bullet

Series 2:

Techno-paternalistic development – blockchain use case

2018 - rise of the “brownfield enablers” in emerging markets healthcare

Filipino-Germanic relations – a global healthcare workforce solution

2017 – the year impact investing jumped the shark

Reforming impact investing – the need for an ecosystem approach