Last week, Drs. James Allison and Tasuku Honjo received the 2018 Nobel Prize in Medicine for their seminal work discovering the role of negative immune regulation for cancer therapy. Their work, along with many others in the field, led to the creation of checkpoint blockers, CAR-T therapies, and new modalities still under development, and also sparked a paradigm shift in how physicians and scientists approach the treatment and management of cancer.

One of the largest challenges with adoptive cell therapies has been manufacturing cost and processes. As I wrote in my first CAR-T post, the process of removing T cells from a patient and isolating and reprogramming them in a highly controlled laboratory is time-consuming, costly, and prone to any number of potential manufacturing errors.

Allogene Therapeutics is one of the next-generation of adoptive cell therapy companies development off-the-shelf CAR-T therapeutics. This process involves harvesting and genetically reprogramming immune cells from a healthy donor. The engineered cells can then be used to treat any corresponding cancer patient. An off-the-shelf solution would make production cheaper and turnaround time from lab to patients faster.

Last night, Allogene raised $324 million by offering 18 million shares at a price of $18 per share, the largest biotech IPO of the year. That gives Allogene Therapeutics a fully diluted market value of over $2.1 billion.

I’m not an investor in or otherwise affiliated with Allogene, but I’m interested in the future of cell and gene therapies and their respective public data. I don’t currently hold any positions in any of the companies mentioned below.

Business Overview:

In April, two of Kite Pharma’s former executives, Arie Belldegrun and David Chang, unveiled that they had left Gilead following its $11.9B acquisition of Kite to found Allogene, focused on developing allogeneic CAR therapeutics. The story behind the company’s founding is worth a read.

The company first raised $300 million in Series A financing as part of a strategy to acquire and develop a portfolio of 16 preclinical cell therapy assets from Pfizer and one that has been tested clinically, UCART19, originally developed by Cellectis. This marks Pfizer’s exit from adoptive cell therapy, but they could still benefit from the success of the modality — they gained a 25 percent ownership stake in Allogene with the asset transfer.

In September, Allogene raised an additional $120M in convertible note financing, which both improves their cash position and gives them institutional public company shareholders as strong signaling prior to filing their IPO.

Management:

Allogene benefits from a pair of experienced, successful biotech entrepreneurs who brought one of the first CAR-T products, Yescarta, to market and have remained in the same therapeutic modality.

Arie Belldegrun, Kite Pharma’s former chairman, president, and CEO, is now Allogene’s executive chairman, and David Chang, Kite’s chief medical officer, is Allogene’s president and CEO.

While UCART19 data presented at ASH last year wasn’t great in its pediatric trial and about on par with Kymriah results for its adult trial, the executive’s track record of building meticulous manufacturing processes could improve the fidelity of UCART19, allow them to compete on COGS, and advance the company’s newly ascertained preclinical assets quickly.

Pipeline & Clinical Data

The Allogene portfolio includes 16 pre-clinical T cell therapy assets and UCART19, an allogeneic CAR T therapy targeting CD19 currently in Phase 1 development for the treatment of refractory and relapsed (R/R) acute lymphoblastic leukemia (ALL).

In June, Cellectis introduced the CubiCAR, an an all-in-one CAR architecture that enables CAR T-cell detection, purification, and on-demand depletion via a Rituximab suicide switch, which Allogene will test in the clinic. This could allow faster CAR-T manufacturing, detection, and improvements in its toxicity profile, the latter of which could help swing prior UCART19 data in a positive direction.

The S-1 explains data since ASH2017, painting a rosier picture:

UCART19. In 2016, our collaboration partner, Servier, initiated two clinical trials of UCART19: the CALM trial and the PALL trial. The CALM trial is a Phase 1, open-label, dose-escalation clinical trial in adult patients with R/R ALL. The PALL trial is a Phase 1, open-label, clinical trial in pediatric patients with R/R ALL. In June 2018, interim results from 18 patients in the CALM and PALL clinical trials were presented at the 23rd European Hematology Association Annual Congress. As of April 2018, 13 out of 16 evaluable patients, or 81%, achieved a complete response (CR), defined as the absence of any evidence of cancer, and 12 of those patients, or 92%, achieved a minimum residual disease negative CR (MRD- CR), which occurs when a patient achieves a CR and there is no evidence of ALL cells in the marrow when using sensitive tests such as polymerase chain reaction or flow cytometry. The most common adverse events were related to cytokine release syndrome (CRS) and were generally manageable. Two mild graft-versus-host disease (GvHD) cases in the skin were observed and resolved.

Allogene is using TALENs, a gene editing technology licensed from Cellectis, to remove functional T-cell receptors (TCRs) from its CAR-T products (to reduce the risk of GvHD), and remove CD52 in combination with administering an anti-CD52 antibody to patients, which allows the body to deplete its natural T-cells and allow the infused CAR-T cells to expand and persist in the body.

Allogene and Servier plan to initiate Phase 2 studies for UCART19 in 2019. Allogene has exclusive rights to develop and commercialize UCART19 in the U.S., Servier has exclusive rights for all other countries.

Additionally, Allogene plans to submit an investigational new drug application (IND) in the first half of 2019 for ALLO-501, for the treatment of R/R non-Hodgkin lymphoma (NHL).

Partnerships:

Servier: As mentioned above, Servier holds ex-US rights for UCART19. For countries with state-controlled formularies, pricing will play a large role in how much Servier values the partnership moving forward. The U.K. National Health Service (NHS) recently approved coverage of Novartis’ Kymriah, but funding for Gilead/Kite’s Yescarta has been turned down in draft guidelines by the U.K.’s National Institute for Health and Care Excellence (NICE), an independent body that performs cost-effectiveness analyses for new drugs.

Cellectis: Remains eligible to receive clinical and commercial milestone payments of up to $2.8 billion, or $185 million per target for 15 targets and tiered royalties on net sales of any products that are commercialized by Allogene under the agreement.

Notable in Allogene’s Risks Related to Our Business section is their dependence on Cellectis’ gene editing technology and CAR architecture for manufacturing and Servier’s conduct of clinical trials, tying the operating risks of the latter to Allogene’s quarterly performance.

Financials:

Cash Position: $410,827,000 in cash and cash equivalents following conversion of the convertible note offering to common stock in the IPO.

Operating Expenses: Estimated at $275M/year based on S-1 financials. R&D makes up 89% of opex. To date, most R&D spend has been on discovery efforts, facility and overhead costs, and clinical trial expenses.

This gives the company a runway of 1.5 years.

As a company in the cell therapy space, and particularly given the company’s off-the-shelf approach, immediate, high-intensity and velocity of manufacturing spend along with larger clinical trials on the horizon means that filing for an IPO while public markets are almost too healthy is a strategic move.

I took a back of the envelope portfolio valuation approach, from the perspective of a potential acquirer, focusing on the two indications of UCART19 (relapsed and refractory pediatric/adult ALL), their single asset currently in clinical testing to determine if the company’s IPO valuation is in line with their current clinical asset potential.

Note that this is an incredibly rough estimation given the early stage of most of the company’s pipeline and market potential, competitive landscape (Novartis, Gilead, Celyad), and rapid scientific pace of adoptive cell therapy technology.

ALL Market Size:

Total cases = 5,960.

• Adults: About one half of adults with ALL are not cured of the disease and ultimately die. One third of patients with standard-risk ALL and two thirds of high-risk patients relapse.
• Children: About 3,000 people younger than age 20 are diagnosed with ALL in the U.S. annually. Between 15 and 20 percent of children who are treated for ALL and achieve an initial complete remission relapse.

Total Market: 1500 adult patients + 600 pediatric patients per year = 2100.

Market penetration = 35% (reasons for estimation below)

• Pro: Growing market — only competition in R/R ALL at the moment is Kymriah, Allogene might gain first-mover advantage in market share if they’re first to bring an allogeneic product to market.
• Pro: COGS drastically lower than Kymriah, allowing Allogene to potentially compete and displace Kymriah on price and speed of treatment deployment to patients.
• Con: Current CAR-T treatments have been limited to large hospitals due to apheresis requirements and management of side effects (cytokine release syndrome, neurotoxicity). The toxicity profile of UCART19 would have to be significantly lower for it to be administered in community settings, where most cancer patients receive care. This last point is what could unleash the market potential of cell therapies and turn them into blockbuster drugs.

Serviceable Market: 735 patients

Estimating Free Cash Flow:

Kymriah for ALL is currently priced at $475,000. If Allogene wants to compete on price but charge what the market will bear, UCART19 could be priced at ~$400,000.

Assuming 35% market share and a $400,000 price per dose, *Peak Annual Revenue: $294,000,000.

*There are some nuances in actual peak revenue given that CMS only reimburses current CAR-T therapy with a base payment of $36,000 and a maximum NTAP (New Technology Add-On payment) of $186,500 under MS-DRG 016, but that’s beyond the scope of this piece. More accurate numbers would need to account for the fraction of future ALL patients receiving public health insurance.

The forecast below assumes a five year ramp to peak sales and a gradual drop over time as competition enters the market and patent exclusivity for both TALENs (2035)and Allogene’s CAR products expires.

Novartis’ Kymriah brought in $12M in Q1 2018 (4x below analyst expectations) and $16M in Q2 2018. The cost and small market size for R/R ALL were likely big issues contributing to those numbers, but Allogene postulates in its S-1 that they’ll achieve higher market penetration because:

As seen in the registrational trials for Kymriah and Yescarta, up to 31% of intended patients ultimately did not receive treatment primarily due to interval complications from the underlying disease during manufacturing or manufacturing failures.

COGS is where Allogene could shine. According to research from Hardman & Co., a British consulting firm, the COGS for a single Kymriah therapy are estimated at $200,000, or 42% of total revenue.

On the potential of allogeneic CAR therapy, Dr. Katy Rezvani, chief of the cellular therapy section at the MD Anderson Cancer Center, said:

“We could target more than one hundred diseases, bringing down the cost of CAR therapy to one per cent of where it is now, which is why we are so very excited about this.”

Generously assuming the above about allogeneic CAR-T unit economics and factoring in advances in cell engineering process automation, I estimate that by 2025, R&D for allogeneic CAR comes down to 10% of the estimated price tag. SG&A is estimated to be 25% based on Gilead numbers, resulting in operating expenses of 35% of sales revenue. I have also not included potential milestone payments to Cellectis for simplicity’s sake.

The tax rate was estimated at 29.84% based on a 21% federal corporate tax rate and an 8.84% California state tax rate.

EBIAT = Revenue * (1–0.35) * (1–0.2984).

Risk-Adjusted NPV:

Weighing revenues for scientific risk, i.e. a 23% chance that a Ph2 drug makes it to market, plus a discount rate of 20% used to evaluate mid-stage biotech companies, gives us a risk-adjusted Net Present Value of $441,104,188, arising from $30,277,188.74 over thirteen years of sales of UCART19 for R/R ALL and $410,827,000 of current cash and cash equivalents. My sales forecast and earnings calculations based on the above are in the Google Sheet below.

Allogene rNPV

Conclusion

Looking at current asset value alone, UCART19 is fundamentally a proof-of-concept asset for allogeneic cell therapy. Allogene’s $2B valuation is, strategically on the part of the company, dependent on additional indications for UCART19 with larger market sizes coming to market alongside their preclinical pipeline, and seeing allogeneic therapies displacing current CAR products among a crowded preclinical atmosphere.

Whether Allogene’s valuation holds up in public markets will be a test of both investor confidence in the biotech sector through Q3 2018 and perhaps a sign of what 2019 might look like for the industry.

Allogene, despite being less than a year old, certainly has the management team and backing to bring multiple novel therapeutics to market, and I hope to see that happen, but the scientific risk of allogeneic cell therapy and sophisticated competitors hot on their heels makes this all but a predictable Kite 2.0.

Thanks to Manjinder Kandola and Chris Devine for reading drafts of this.

As always, hoping to continue the conversation with interested readers. Reach out on twitter @nxpatel. I’d love to hear your thoughts.

On the back of record Q2 2018 profits separate from its core retail business, Amazon can afford to undercut its competition’s prices to expand its moat while simultaneously putting capital to work in new verticals. The company’s acquisition of PillPack isn’t its first major investment into pharmacy, and unlikely to be its last.

When compounded with Amazon’s existing logistics infrastructure, business units, online retail market share, and most notably, new modalities of enabling purchases (Echo, Go), Amazon’s jump into the pharmacy business resembles the following:

1. A front door to health care even closer to home than retail clinics: Amazon owns 62% of the smart speaker market, nearly 34 million devices across the United States.
2. Market density and leverage unmatched by any health insurer, including the federal government: 310M active customers on amazon.com, 100M Prime members (51% of U.S. households).
3. Distribution nodes suited for drug storage and delivery: 34% of U.S. households live within 5 miles of a Whole Foods store.
4. A seamless online pharmacy with home medication delivery funneling into a robust e-commerce offering: I previously used this to describe Walmart’s potential acquisition of PillPack, but Amazon beat them to the punch.

It’s worth restating Amazon’s mission statement to realize what their goals are:

We seek to be Earth’s most customer-centric company.

And what that means for industries outside retail:

“Amazon has no particular desire to be a grocer, and contrary to conventional wisdom the company is not even a retailer. At its core Amazon is a services provider enabled — and protected — by scale.”

— Ben Thompson

Does Amazon have the pieces to make this work? Will the mass market embrace at-home delivery of prescription drugs? Is their customer base large enough to compete, or at least play friendly, with hesitant PBMs? I dove into Amazon’s history in pharmacy, recent acquisitions, and the avenues for growing their market and customer value through their non-core business units.

TABLE OF CONTENTS:

• The Entrypoint
• Whole Foods Rx
• Alexa Meets PharmacyOS
• The PBM Problem

The Entrypoint

First, some history. In 1999, Amazon bought a 46% stake in Drugstore.com, which, after never recording an annual profit, sold its mail-order pharmacy business for $5M in 2010 and was shut down in 2016. Despite Amazon’s efforts to market Drugstore.com products on its own website, challenges from incumbents (notably PBMs), regulations, and logistical hurdles sunk the venture.

Drugstore.com’s pricing power and convenience, however, did have an advantage in cash-pay pharmaceuticals, such as Viagra and Propecia, which usually aren’t covered by insurance. In other words, when patients — rather than insurers — were making purchasing decisions, Drugstore.com had an edge.

The parallel in today’s health care market are patients who are either uninsured or the rising number of Americans on high-deductible health plans, which usually don’t offer first-dollar coverage for medications. As a result, pharmacies hold significant leverage, and will often charge sky-high prices to these groups. Cash-pay alone makes up 8% of the U.S. drug market, or 400M 30-day-equivalent prescriptions.

After struggling to get past regulators, the company’s near $1B acquisition of PillPack is a signal that it’s willing to spend significant capital to obtain strategic relationships with both suppliers and pharmacy benefits managers along with statewide pharmacy licenses to sell prescription drugs.

Amazon has been selling over-the-counter medicine and other personal health care products for the past year, a notable characteristic being that these products are 1) made by many manufacturers that primarily compete on price (functionally commodities) and 2) not usually covered by insurance. This positions the generic drugs market, which currently make up one-third of global pharmaceutical sales and are growing at 3x the rate of branded drugs, an entrypoint into the industry that is conducive to Amazon’s proven marketplace model.

With news that GoodRx is looking for a buyer, Amazon’s pharmacy strategy could mirror their Echo and retail strategy. First, control the supply (Amazon Basics : PillPack), then use ease-of-purchasing (Echo : GoodRx) to selectively funnel privately-labeled Amazon products or bundled goods to consumers.

Yet, many of the same logistical, regulatory, and competitive challenges that sunk Drugstore.com and hampered prior Amazon efforts in pharmacy remain, which brings us to the role of Whole Foods.

Whole Foods Rx

Logistics

The logistical challenges for drugs (cold storage, perishability, degradation during transit, limited SKUs) parallel many of the same issues that kept Amazon out of the grocery business until its Whole Foods acquisition, and why AmazonFresh has burned cash for a decade without much success. Amazon needs immediate scale in its pharmacy business to avoid wasting stock of perishable drugs. Additionally, the relatively limited number of outpatient drugs doesn’t give Amazon the same selection advantage it has in CPG over brick-and-mortar retailers.

But thinking about pharmaceutical storage and distribution in the context of traditional fulfillment centers misses the mark. This is where Amazon’s 13.7B acquisition of Whole Foods becomes important.

Whereas fullfillment centers represent high fixed costs that benefit from economies of scale as Amazon reaches more customers and third-party merchants, the Whole Foods acquisition represents a large capital expenditure in an existing, decentralized network of “storage” units with the physical infrastructure (basic/cold storage) and proximity to major population centers to support the supply and distribution of pharmaceuticals. 34% of U.S. households live within 5 miles of a Whole Foods store.

Playing with TAM & LTV

Considering the framework that Ben Thompson has used to describe Amazon’s Whole Foods acquisition, drugs are simply another “primitive” in the ‘Amazon Grocery Services’ architecture, akin to meat, fruits, and vegetables.

Amazon’s partnership with Xealth, which connects patients to health care software and products, signals their awareness that “pharma of the future” will go beyond pills — including digital therapeutics that influence neuroplasticity and behavioral health, biosensors as early warning systems for the exacerbation of medical conditions, and specialized diets to combat chronic disease.

Whole Foods improves Amazon’s last mile delivery challenge, and piggybacks off a service and brand synonymous with health. There’s an approximately 50% overlap between existing Whole Foods stores and Prime Now (2-hour delivery) zip codes. Amazon can use existing, often recurring, deliveries of prescription drugs via Prime Now as a springboard to up-sell customers on Whole Foods products (i.e. produce, meals), delivered alongside their prescriptions — increasing each customer’s lifetime value.

There’s growing evidence that condition-specific meal plans can be as effective, if not moreso, than a specialty drug and reduce long-term health care spending. Amazon’s logistics backbone could deliver medically-necessary Whole Foods meals to consumers, akin to their AmazonFresh program, as a supplement to the medications that they know their consumers are buying to treat conditions such as Type II diabetes or congestive heart failure.

However, if you continue to think about Whole Foods as simply another customer for ‘Amazon Grocery Services’, pre-packaged meal kits (already available at Amazon Go) tailored to specific conditions could also be sold to other large ACV customers— notably health systems and local governments embracing the ‘food as medicine’ philosophy, increasing Amazon’s addressable market in the space.

The Geisinger Health System in Pennsylvania piloted a program to ‘prescribe food as a specialty drug’ for patients with Type 2 diabetes. Costs for those patients dropped by a staggering average of 80%, from $240,000 per member per year to $48,000 per member per year. In nearby Philadelphia, the Metropolitan Area Neighborhood Nutrition Alliance has been delivering free meals to low-income patients with serious illnesses for nearly thirty years, showing annual reductions in health care spending of 55% of patients who receive meals compared to those that don’t. Fundamentally, the types of foods that patients ate had a significant and clinically observable impact on their health outcomes.

Rising costs for Medi-Cal, the state’s Medicaid system that covers over a third of its population, and a high burden of chronic disease and food insecurity among Californians made Geisinger’s pilot the perfect experiment to bring out west.

The California pilot program, which officially launched this month with support from state senator Ben Allen along with California assembly members Blanca Rubio and Richard Bloom, provides $6 million in funding over three years to six nonprofit organizations to offer diet guidance and deliver free meals to Medi-Cal patients that need special diets to manage certain medical conditions. The ‘Food is Medicine’ pilot program will deliver meticulously formulated meals to 1,000 people with congestive heart failure over a 12-week span.

Additionally, as of early 2018, Medicare Advantage plans can receive federal reimbursement for groceries and home-delivered meals for covered patients. This functionally hands Amazon the federal government as an indirect customer for a patient population that is expected to nearly double by 2025.

Alexa Meets PharmacyOS

Although companies like Express Scripts made their name in mail order pharmacy, the mail order market itself has been shrinking as pharmacies began to vertically integrate with other elements of the health care system. Retail pharmacies in particular have been cutting in-store prices to move customers into their physical retail stores, where they also stock and sell those same customers personal health care products, groceries, convenience goods, and home supplies.

This is where Amazon has a convenience advantage. Why ask customers to go to a physical location to increase basket size when you own the world’s best online marketplace?

That being said, when it comes to shifting the demand curve, Amazon has taken a different approach than its retail pharmacy competitors. Often times, buying the same item via an Echo is cheaper than searching for the same item on amazon.com. For traditional CPG purchases (e.g. batteries, hangers, electrical cables) Amazon will automatically purchase and deliver its own AmazonBasics labelled products via Echo, which improves their margins.

How frequently this service is used, however, is uncertain. According to The Information, only 2% of customers with Alexa-enabled devices have made purchases via Alexa in 2018, and 90% of those that did make a purchase over voice didn’t try it again. Other surveys (below) have claimed that up to 34% of customers with Alexa-enabled devices have made a voice purchase. People were more likely to buy commodity items that didn’t merit additional research and reading reviews. Amazon is betting that offering low prices via voice vs. online shopping will increase that percentage.

However, a more interesting case in the context of pharmacy is repeat purchases, that is, prescription refills and subscription models for personal health products. Look at the way many Echo users have set up their Alexa reminders to date — primed to both help users complete their medication regimen and internally track when they need support or a refill, automatically fulfilled by Amazon’s logistics backbone.

In fact, Americans are in the pharmacy a lot, filling an average of 12.6 prescriptions per year, making refill fulfillment and adherence a sustainable, and growing, market.

This infrastructure complements PillPack’s PharmacyOS system, which was built to both streamline internal operations and strengthen the customer relationship, which CEO TJ Parker explained:

PharmacyOS is allowing us to continuously improve our back-end operations, which are invisible to customers, but result in an increasingly simple and frictionless service. We’re executing on the promise we make as a company — which we’ll do everything we can to make it as easy as possible for people to take their medications as prescribed. Today, we’re working on making the ecosystem as useful as possible. While a customer might only go to their doctor once every few months, they interact with medications every day. We want to make the most of that daily opportunity to support our customers and helping them take control of their health. In the future that will likely include new digital services, offline programs, and additional tools for caregivers.

With 34 million Echo devices already distributed nationally, and options like the $50 Echo Dot inexpensive enough to make its way into most American homes, Alexa, bolstered by PharmacyOS, could quickly and reasonably become a staple for home health guidance by increasing both the frequency of touchpoints and the size of the network. Similar to the Whole Foods scenario, an online drug marketplace, augmented by voice when available, will also help increase the LTV of both PillPack and Amazon customers by recommending and bundling complementary health care goods, e.g. insulin with a blood glucose monitor, test strips, lancets, and diabetes-friendly shoes.

The PBM Problem

The monopolistic elephants in the room, Pharmacy Benefits Managers see Amazon as an obvious threat; their primary interest is in protecting their market share, part of which comes from the mail order pharmacy units (e.g. Express Scripts) that are clearly Amazon’s first target in the prescription drug market.

Amazon has two options moving forward, either:

1. Convince an increasingly smaller number of PBMs, i.e. Express Scripts, OptumRx, CVS Caremark, that have consolidated 80% of the market to partner with them.
2. Build their own vertically-integrated PBM/Pharmacy that gives them complete control over the supply chain, distribution, and customer experience.

While these large PBMs have previously worked with PillPack, its acquisition and Amazon’s cash on hand will likely leave them hesitant to continue the relationship. Option #2, on the other hand, would be time consuming, throw Amazon into a whole mess of regulatory compliance, probably require acquiring an existing PBM to build on top of, and then somehow compete with nearly monopolistic PBMs that currently process nearly 6 billion 30-day-equivalent prescriptions each year.

However, if successful, this strategy gives Amazon complete control over the customer relationship that parallels the way they’ve achieved success in their core marketplace business. Amazon’s business strategy, growing profits, and capital efficiency make it likely that, while they may try to play nice with PBMs in the near-term, I wouldn’t count them out on leveraging their loyal customer base to take on the PBM giants within the next five years. The bigger question is, will their competition do anything about it?

Carl Quintanilla on Twitter

Walgreens CEO says "not particularly worried" when asked about $AMZN's acquisition of Pillpack: conference call "Pharmacy world is much more complex than just delivering certain pills" (via @reuters)

If you found any of this helpful or interesting (or disagree completely), please share and/or reach out to me on twitter @nxpatel. I’d love to hear your thoughts.

Related health care strategy pieces:

• The Amazon-Berkshire-JPM Health Care Playbook
• Walmart Health: A Deep Dive into the $WMT Corporate Strategy in Health Care

Amazon, Berkshire Hathaway, and JP Morgan Chase (ABC) announced last Wednesday that Dr. Atul Gawande will serve as CEO for their unnamed health care venture first publicized in January.

“I’m thrilled to be named CEO of this healthcare initiative, I have devoted my public health career to building scalable solutions for better healthcare delivery that are saving lives, reducing suffering, and eliminating wasteful spending both in the US and across the world. Now I have the backing of these remarkable organizations to pursue this mission with even greater impact for more than a million people, and in doing so incubate better models of care for all. This work will take time but must be done. The system is broken, and better is possible.”

— Atul Gawande

This choice was met with split reactions, but the larger question is how Atul will deploy the resources of these three giants of industry to affect a much smaller, yet more distributed, network of employees than most health systems and insurers.

In this essay, I dive into the trends that compelled ABC to take on the health care industry, how each company’s strengths contribute to tackling some of health care’s biggest challenges, and how Gawande, who’s made a career of moving ideas from the written page into the clinic, could combine the best of health services research, technology, and strategy to deliver a transformative health care experience.

Table of Contents:

• The Challenge
• The Goals
• The Play

The Challenge

Over the last decade, the total cost of annual health insurance coverage has increased by over 50%, squeezing both employer margins and hampering wage growth.

One of the most troubling trends in particular is the shift of insurance coverage from more comprehensive benefits packages to high-deductible health plans (HDHPs), which offer cheaper premiums, but require enrollees to pay a large out-of-pocket cost (ranging anywhere from $1350-$6650 for an individual) before they receive any help from their insurer. Given that 40% of Americans don’t have enough cash in the bank to cover even a $400 emergency expense, requiring them to pay multiples of that for health care renders their coverage all but useless.

Now, put this into the context of the largest employee pools and their characteristics for each of these three companies:

Amazon | warehouse fulfillment workers: usually short-term (1-year) jobs, resulting in unstable health insurance coverage, in a physically demanding role with high occupational hazard.

Berkshire Hathaway | manufacturing/construction workers: unionized blue collar jobs, distributed across midwest and southern states with fewer health care resources and high occupational hazard.

JP Morgan Chase| bank tellers: middle class, distributed nationally, largely suburban.

None of these groups are well served in the current U.S. health care system. Most earn enough to be disqualified for Medicaid, but don’t make enough for a private sector plan with a $2000 deductible to provide any useful coverage without bankrupting families’ savings.

Hence, the initiative by ABC to take on rising health care costs for employers and poor coverage for their employees.

The Goals

Because the new organization’s structure and solution(s) are yet to be decided or announced, it makes sense to examine the stated goals of the venture from the words of Bezos, Dimon, and Buffett.

The new company will be independent and free from profit-making incentives and constraints. The new venture’s initial focus would be on technology that provides simplified, high-quality and transparent care.

All three executives made it clear that the effort would first and foremost be focused on their approximately 1.2 million employees, bolstered by their data and bargaining power to lower health care costs. If successful, the solutions that the venture develops may be gradually rolled out to all Americans.

At the Aspen Ideas Festival, Gawande outlined his objective as CEO to eliminate health care waste along three axes:

1. High administrative costs — taking middlemen out of the system.
2. Pricing failures — unit prices per product or service are too high.
3. Misutilization — wrong care pushing out the right care.

It’s important to note that these are not independent challenges, but inextricably linked. Pricing failures are partly said to arise from high administrative costs and middlemen. As I wrote in STAT:

“The cost of labor makes up more than half of the total cost of delivering health care. So as health care employment rises, it pulls up health care prices. Expenses are passed down to employers and patients in health insurance premiums, which are escalating at a rate outpacing inflation.

More jobs and higher prices haven’t translated to better health care. The rapid growth in employment has largely come from adding administrative and management jobs, rather than from adding clinicians responsible for direct patient care. Administrative costs in U.S. health care are the highest in the developed world, accounting for more than 25 percent of spending in the sector.”

However, supplemental labor is often what helps reduce misutilization of health care services. Health coaches help change patient behavior for chronic disease management, back office staff handle claims and scheduling so clinicians can focus on patient care, and medical scribes give physicians more quality time with patients rather than their keyboards.

Misutilization is often exacerbated by consumer price sensitivity as a result of HDHPs. Out-of-pocket spending on health care isn’t a problem in countries like Singapore where the government has price controls on medical services, but research shows that, due to high U.S. prices, HDHPs are coarse and ineffective, and don’t make consumers better shoppers for high-value care. They simply tamp down utilization of all health care services.

Before going through mechanisms to tackle each of these challenges, let’s examine the strengths of each firm involved in the venture, and what they, at a high level, could contribute to each axis.

Amazon

The world’s most consumer-centric company has strong bargaining power with suppliers due to market reach of its e-commerce business, scale moat in its unit economics and logistics, affluent neighborhood presence via Whole Foods, internal incubation of advanced logistics startups like Prime Now and Air, and most notably, success in public markets despite — until recently — negative to light profits.

Berkshire Hathaway

Given that Berkshire has fewer than 30 employees but oversees a basket of businesses with a collective 360,000 employees, Buffett has an eye for and emphasis on strong management teams when making acquisitions. The firm’s collection of companies also lends itself to strategic capital allocation — redeploying profits from a cash heavy subsidiary into others that are capital intensive (like a health insurance risk pool). Finally, some management aspects of Berkshire’s GEICO insurance unit, notable for its low operating costs and high market share relative to competitors, might be translatable to a health insurance play.

JP Morgan Chase

The world’s leading investment bank by revenue offers sheer size, diversity of assets, and access to capital markets.

Note: It’s worth mentioning the goals Jamie Dimon outlined in his annual letter to JPM shareholders. Some of what he hopes for has been previously attempted in employer-sponsored insurance plans to date, but with little success. Among his goals (italics mine):

• Aligning incentives among doctors, insurers, and patients (e.g. outcomes-based contracts).
• Reducing fraud and waste (e.g.workflow automation, transparent pricing).
• Giving employees more access to telemedicine and better wellness programs (e.g.Omada Health, Lemonaid Health).
• Figuring out why so much money is spent on end-of-life care (Also why Humana, a large Medicare Advantage insurer, is on a hospice acquisition spree).

The Play

“But what they’re saying to me is that resources won’t be the problem. Human behavior will be. And achieving scale will be.”

— Atul Gawande

The ABC Play: A virtual primary care model, with direct contracting for downstream utilization of health care services, paid for under a value-based insurance design system.

Gawande’s advocacy for the merits of incremental care lends itself to building a model of primary care that reduces long-term costs while also harnessing the competitive advantages of Amazon, Berkshire, and JPM to deliver a distributed, scalable service.

Kevin O’Leary broke down the three methods in which incremental care can reduce overall health care spending (his table below). Each arm of this potential structure for the ABC health care venture operates on one or more of the three methods, or buckets.

Virtual Primary Care:

Harvard economist Amitabh Chandra outlined the employee distribution and diversity challenge of providing health care to 1.2M ABC employees best:

Amitabh Chandra on Twitter

The challenges of improving health care for Amazon employees in Lexington, KY are quite different than improving them for JP Morgan bankers in NYC. How does one manage this? If one can, why couldn't existing payers manage it? These whys are at the heart of business strategy.

New primary care models like Iora Health and One Medical run brick-and-mortar clinics across the country, each able to spread due to their focus on serving niche demographics (e.g. MA and high-income respectively) rather than commercial populations that directly compete with one another. However, with just one million widely distributed employees, adopting a copycat model of existing primary care startups won’t be effective at managing diverse population needs or reaching the per-patient membership benchmark per-clinic to remain financially sustainable.

Instead, harnessing Amazon’s software engineering and product resources to adopt the Virtual Primary Care model (below flowchart), pioneered by Sherpaa Health, increases panel size per physician (e.g. 2500–3000 patients) and reduces fixed costs associated with physical infrastructure, which makes the unit economics of primary care more appealing. Reduced operating expenses allows this model to offer consumers a lower per-patient-per-month price point (Goal #2). For a more thorough explanation of the unit economics of primary care models, read Part 1 of Kevin’s series.

This approach uses Bucket Three — decreasing overall demand for downstream health care services by keeping people healthier. Ideally a network of health coaches to encourage personalized health and well-being, asynchronous messaging for seamless communication and guidance with care team members, and a streamlined workflow process for care teams will see improved patient health outcomes and reduced clinical burnout. By no mistake, Gawande’s research has been studying workflow implementation and changing the behavior of health systems for nearly a decade, and his team will likely have the methods to put a system like this into place nationally.

This particular bucket could take years, if not decades, to prove to be cost-effective at scale, but because this venture is a non-profit funded by three companies with some of the largest cash reserves on the planet, they can afford to be patient.

Value-based Insurance Design:

Nisarg Patel on Twitter

@chrissyfarr @CNBC Amazon enters the insurance market, provides a great UX, gains a ESI monopoly on zero profit margins, turns around and squeezes suppliers, and then becomes America's public option. This works solely b/c the market has ignored them being in the red for their entire existence.

Unlike HDHPs, the value-based insurance design (VBID) approach sets copayment rates on the value of clinical services — the benefits and costs — rather than the costs alone. This tunable approach to benefits design nudges consumers towards high-value care and away from low-value, expensive care (Goal #3). This approach uses Bucket Two — reducing misutilization of health care services.

In this instance, VBID plans could be modified to target individuals with specific diagnoses (e.g. reduced copayments for beta-blockers for CHF patients) or those living in particular geographic regions. In theory, this could also function to reduce long-term costs (Bucket Three) by incentivizing people to use the care most likely to improve their long-term health.

These systems will be more costly to implement due to the need for cooperation and transfer of eligibility data between payers and the point of care. This is why existing VBID initiatives have been limited to conditions like diabetes, where it’s easier to ascertain condition status simply from medications taken. Obtaining integrated claims data would be ideal here and help create in-house VBID plans across different disease areas.

Given those needs, Oscar Health’s 1) customer-facing concierge support and analytics that utilizes digital interventions to nudge users towards higher value care along with 2) an in-house claims processing unit powered by software (Goal #1) offers the perfect partnership (or very expensive acquisition) opportunity here.

By refining the team’s operational approach, using data to learn and evolve, automating claims holds and routing, developing sophisticated matching algorithms for doctor and member information, and flagging risky charges early, his team has been able to reduce accuracy issues, pay claims faster (on average, within 5 days of receipt), and increase the number of claims that move through the system without needing a person to look at them.

Direct Contracting

Walmart has bypassed insurers (Goal #1) and is instead buying health care for its 1.4M U.S. employees directly from health care organizations, using its status as the largest private employer in the U.S. as bargaining leverage to negotiate lower rates (Goal #2).

Along with Boeing, GE, and others, Walmart has negotiated bundled payment contracts with health systems via the Employers Centers of Excellence Network (ECEN), a collection of employers directly purchasing health care for their employees. This approach uses Bucket One — influencing utilization of downstream health care services. As the Harvard Business Review writes:

ECEN began first with total joint replacement, then added spinal surgery, and recently has begun providing bariatric (weight loss) surgery at selected centers. ECEN provides employees of participating companies with 100% coverage for all travel and medical expenses at carefully selected healthcare systems; patients pay no out-of-pocket costs. Participating employers benefit from the quality assurance of the ECEN’s rigorous center selection process and the financial savings from paying competitive, pre-set rates for bundled care negotiated between participating hospitals [and ECEN].

ABC has a collective 1.2M employees and could follow the same methodology outlined by ECEN (below, bolding mine) to identify and contract with health systems that provide high-quality specialty care under bundled payments. Charges per episode-of-care average 10–15% less than traditional fee-for-service medicine and the selection process offers the same transparency that ABC has been chomping at the bit to achieve.

Hospitals and individual participating physicians undergo a thorough and iterative evaluation process. Fewer than 5% of healthcare systems initially identified for participation in ECEN meet all of the quality requirements for consideration. An invitation-only request for information from ECEN leads to an extensive review of the system’s quality, outcomes, and patient satisfaction data.

After an assessment phone call that includes PBGH, HDP, as well as physicians and administrators from the health system under consideration, an extensive request for proposal is sent for completion. Candidate centers must provide information including detailed clinical protocols, surgical-patient selection criteria, clinical registry participation, information on multidisciplinary shared decision-making, as well as institutional and physician-level performance metrics. These metrics include length of stay, return to surgery, infection rates, and procedure-specific outcomes such as joint dislocation after hip replacement and nerve covering tears occurring during spinal surgery.

For centers remaining in consideration after these initial steps, PBGH and HDP perform an in-depth site visit utilizing a patient tracer methodology. The patient experience is reviewed from arrival at the airport, throughout the care cycle, and to departing for home. A successful visit moves the process along to negotiation and contracting.

Conclusion

On aggregate, I have a feeling that, operationally, this venture will be an extension of the process that Gawande has already cultivated at his health systems innovation group Ariadne Labs (Atul is also a proponent of #TrustTheProcess).

As a final note, two things not covered in the above essay, but important to mention:

• Drug pricing — An ongoing lawsuit between Express Scripts and CareZone could pave the way for Amazon to enter the prescription pharmacy business (and why they should buy PillPack if Walmart doesn’t).
• Nutrition — there’s growing evidence that condition-specific meal plans can be as effective, if not moreso, than a specialty drug and reduce long-term health care spending. Amazon’s logistics backbone could deliver medically-necessary meals to employees akin to their AmazonFresh program as a component of a VBID plan specific to conditions such as diabetes or congestive heart failure.

Thanks to Franklin Yang and Abhishek Dharan for reading drafts of this.

If you found any of this helpful or interesting (or disagree completely), please share and/or reach out to me on twitter @nxpatel. I’d love to hear your thoughts.

The staccato, ephemeral blips of an EKG monitor are no match for the deafening crunch of potato starch meeting tooth enamel. So naturally, the first thing I noticed about Cindy when I walked into her hospital room was the family-sized bag of Lays in her hands.

Cindy (not her real name) was suffering from congestive heart failure (CHF); the muscles in her heart were too weak to effectively pump blood through her body. The bag of chips was just one of the snacks that her husband bought from a nearby convenience store minutes before I arrived to take her medical history. As a Medicaid family, the two of them were living paycheck-to-paycheck and stacked multiple part-time jobs to make ends meet. Because of the uncertain duration of Cindy’s stay in the hospital, her husband wanted to give her the most caloric food he could with the little money they received each month from federal food stamps.

Cindy, like most patients with CHF, was likely told by her physicians to eat a low-sodium diet. Because her heart was weak, blood was prone to backing up into the vessels of her lungs and limbs. High levels of salt pull fluid into the bloodstream and increase pressure inside the vessels. As blood pressure rises, the excess fluid is forced into the lungs, displacing oxygen and making it difficult–if not impossible­–to breathe.

Cindy excitedly told me that she was leaving the hospital the next day. My preceptor and I mentioned the chips to one of her nurses, and hoped that she’d receive some diet counseling from a nurse or her physician before she was discharged. Cindy must have known better. What could have compelled her husband to buy a bunch of junk food over a few apples? Whatever benefits they were receiving from the federal food stamp program, SNAP, clearly weren’t enough for Cindy’s health.

One week later, while moving from room-to-room to find a patient to interview, I was surprised to see Cindy back in the hospital. I asked the floor nurse why she had been readmitted. Cindy had experienced severe shortness of breath from a buildup of fluid in her lungs, symptoms likely due to her salt intake at home. I found her–anxious and unsettled–lying in a hospital bed; once again hooked up to an EKG monitor. There weren’t any chips by her bedside today.

Along with Cindy and her husband, more than 42 million people–one in eight Americans–receive food stamps. However, the impact of budgetary squabbles over food stamp spending is most often manifested inside the walls of America’s academic medical centers. Food insecurity, the lack of access to nutritionally adequate food, has been linked to higher risks of developing hypertension, heart disease, stroke, cancer, diabetes, arthritis, kidney disease, and hepatitis, among others. While helpful, the limited purchasing power of food stamp recipients means they often have to turn to the cheapest calories, usually processed foods made from refined sugars and oils. It’s no wonder, then, that America’s poorest families have much higher obesity rates relative to the rest of the U.S. population.

Cindy was put on furosemide, a drug that helps the body remove excess water and salt by increasing urine output. Her legs shrunk and her breathing slowly returned to normal. If the rest of her tests came back normal, she would be set to go home within a few days. I wondered if she’d ever return. Doctors can manage the worst manifestations of chronic disease in the hospital, but the environment that these patients return to once they leave–block corner convenience stores, cheap junk food, and limited options for nutrition–only exacerbates their conditions. The furosemide was treating a symptom of Cindy’s disease, but left the underlying cause unaddressed. Could smarter food policy be a better medicine?

Hospitals took the lessons learned from patients like Cindy into their own hands. The Geisinger Health System in Pennsylvania piloted a program to ‘prescribe food as a specialty drug’ for patients with Type 2 diabetes. The program provided fresh fruits and vegetables, whole grains, and lean proteins to feed each program participants’ entire household with two healthy meals for five days per week. After 18 months, this approach of ‘meals as medicine’ led to a drop in HbA1c levels from an average of 9.6% initially to 7.5% after the program. To put this in context, diabetes patients taking multiple medications for blood sugar management can expect their HbA1c to drop somewhere between 0.5 to 1.2%. Geisinger also tracked how total health care spending for 37 of their patients on the program changed over the course of 18 months. Costs for those patients dropped by a staggering average of 80%, from $240,000 per member per year to $48,000 per member per year. In nearby Philadelphia, the Metropolitan Area Neighborhood Nutrition Alliance has been delivering free meals to low-income patients with serious illnesses for nearly thirty years, showing annual reductions in health care spending of 55% of patients who receive meals compared to those that don’t. Fundamentally, the types of foods that patients ate had a significant and clinically observable impact on their health outcomes.

California policymakers took notice. Rising costs for Medi-Cal, the state’s Medicaid system that covers over a third of its population, and a high burden of chronic disease and food insecurity among Californians made Geisinger’s pilot the perfect experiment to bring out west.

The California pilot program, which officially launched this month with support from state senator Ben Allen along with California assembly members Blanca Rubio and Richard Bloom, provides $6 million in funding over three years to six nonprofit organizations to offer diet guidance and deliver free meals to Medi-Cal patients that need special diets to manage certain medical conditions. The ‘Food is Medicine’ pilot program will deliver meticulously formulated meals to 1,000 people with congestive heart failure over a 12-week span.

As the first state-sponsored prescription of meals as medicine, the spillover effects of success in California could be tremendous, and potentially entwine food policy with health care debates on Capitol Hill. With both sides of the aisle harping on deficits and health care prices, why not follow the evidence? Including meals as a part of federal Medicaid coverage could both give our most vulnerable population more financial freedom to make healthier choices and reduce long-term spending on expensive treatments for chronic disease. For patients like Cindy, an apple a day really could keep the doctor away.

Nisarg Patel is a student at Harvard Medical School and the Harvard School of Dental Medicine, and cofounder of Memora Health.

Could We Prescribe Meals as Medicine? was originally published in Op-Med on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Nisarg Patel

The first pediatric patient I saw was in 2010, while I was still a college student scurrying after attendings within the pale eggshell walls of a greater Phoenix hospital. The kid wasn’t much younger than I was, either 14 or 15, but he was just moody enough that I felt like 18-year-old me was old and gray in comparison.

The boy was suffering from bipolar disorder, but, despite instruction from his physicians and nudging from his parents, refused to take his medications. I couldn’t help but notice that throughout the visit, where I stood quietly in the corner of the outpatient clinic room, his eyes were glued to the screens of his Nintendo DS. The fleeting sounds springing from the console were just loud enough to fill the room and make the rest of us mildly uncomfortable.

I could relate. As a college freshman, my weekends were bookended with rounds of Super Smash Bros. and FIFA with friends across the hall. If only Mario Kart could treat his condition, I thought.

Eight years later, I couldn’t have imagined that the nascent field of “digital therapeutics” might just make that offhand thought a reality.

Pear Therapeutics’ CEO Corey McCann defines digital therapeutics as “patient-facing software intended to treat disease.” Digital therapeutics founders argue that their programs take advantage of our brain’s neuroplasticity — the ability for neurons in the central nervous system to ‘rewire’ themselves and result in neurochemical changes — to drive a clinical effect. This mechanism of action lends itself well to treat health problems that have an evidence-based behavioral or psychological mode of management, such as cognitive-behavioral therapy. The ideation, development, and testing of digital therapeutics mirrors that of traditional pharmaceuticals, with the distinction that, instead of small molecules, algorithms and user experience are the assets. Software could thus be taken down the typical path of a drug pipeline and receive a drug-like reimbursement.

What does this look like in practice? An interesting example comes from Akili Interactive Labs, a company building action video games to treat pediatric ADHD. Their premise is that sensory stimuli from action and adventure video games activates attentional networks in the brain at random, and tweaking algorithms and the user experience of those games ever so finely combines various sensory stimuli in a way that activates the specific region of the brain needed to treat ADHD. These pediatric patients forget that the game is medicine; they’re instead drawn into the imagery and adventure of steering down a river with aliens. Although hundreds of startups are building digital therapeutics, only a handful, such as Akili, have gone through the rigorous, large-scale clinical testing and validation to bring them up to pharma’s playing field.

In 2017, the Food and Drug Administration approved the first digital therapeutic, Reset, from Pear Therapeutics, that could be commercialized and reimbursed as a prescription drug. The product itself is a digital packaging of cognitive behavioral therapy, and acts as a monotherapy to treat addiction to alcohol, cannabis, cocaine, and stimulants. A second digital product from Pear, ResetO, is currently under priority review by the FDA and hopes to be approved as a combination therapy with buprenorphine for opioid use disorder. Other companies with rigorous, promising clinical data include Virta Health for patients with diabetes, Omada Health for patients with prediabetes, and Propeller Health for patients with asthma and COPD.

While digital therapeutics may look just like any iPhone app at first glance, that’s not quite the case behind the scenes.

Many digital therapeutics companies have been building teams and workflows modeled off the traditional drug development process, including having products pass through a specialized pipeline: research and development teams, quality groups, clinical study groups, and regulatory filing groups. Naturally, this process requires larger investments and more rigorous development cycles than traditional consumer software. Notably missing among most is the large sales force that Big Pharma is known for, which may explain why the majority of digital therapeutics products are still unfamiliar to most physician’s ears, even in Boston and San Francisco. These startups have begun partnering with large pharmaceuticals to bridge that gap, but would like to eventually grow to the point where those partnerships are a bonus rather than a requirement for their products to spread through the market.

But will FDA approvals and peer-reviewed publications be enough to convince doctors to prescribe these? Will patients give digital programs their health information in an era where privacy scandals among U.S. technology giants has eroded the public’s trust?

Getting to that stage itself is hard enough. Moving products through regulatory approval, not to mention persuading health insurance companies to reimburse them, is incredibly long and laborious. Not only do insurers have to trust fledgling startups to take care of patients, but also ensure that patient data will remain private and secure. Digital therapeutics companies are hoping that clinical validation and regulatory approval will be enough to gain the trust of physicians and help them transition from seeing these products as apps to digital medicines that will truly help their patients.

The biggest challenge, however, is not physician trust or the fidelity of the technology itself, but rather the habits of patients. Patients at-large don’t yet regard software as medicine, meaning products must not only show efficacy, but also meaningfully engage their users. If 50% of Americans don’t take their medication as prescribed, what will make them want to engage with these digital programs even half as often? Only time, technology, and aliens will tell.

Nisarg Patel is a student at Harvard Medical School and the Harvard School of Dental Medicine, and cofounder of Memora Health.

Prescribing software as a specialty drug was originally published in Op-Med on Medium, where people are continuing the conversation by highlighting and responding to this story.