The best way to value biotech for deal making and investments

A staggering amount of money is flowing into biotechnology firms. Between 2015 and 2019, over $100 billion in venture capital investment entered the sector. Even in the midst of a pandemic, biotech initial public offerings (IPOs) have remained strong so far in 2020 – Legend Biotech raised $424 million, while Applied Molecular Transport (AMT) raised $154 million and Pliant Therapeutics raised $144 million.

Back in 2017, Gilead Sciences sparked an arms race when it acquired Kite Pharma for almost $12 billion. This was followed by Celgene’s $9 billion acquisition of Juno Therapeutics, which, in turn, was acquired by Bristol-Myers Squibb for $74 billion in 2019. AveXis was also acquired by Novartis for $8.7 billion in 2018, while Spark Therapeutics was acquired by Roche for $4.3 billion and Audentes was acquired by Astellas for $3 billion in 2019.

The value of the companies involved in these deals is largely the result of their IP rights, which begs the question: how is this intellectual property being valued?

There are three widely used approaches for valuing intellectual property:

• calculating design around costs;
• discounted cash-flow analysis (DCF); and
• valuing market comparables.

Calculating costs… does not work

Under this approach, assets are valued based on the cost to create and develop them. The premise is that a buyer involved in an arms-length transaction would be unwilling to pay more for a property than the cost to design around it. To see the cost approach in practice, consider the analogy of a potential buyer looking to purchase a house with an asking price of $1 million. If it costs $200,000 to purchase a similar empty lot and another $500,000 to build a new house, then the buyer would be unwilling to pay much more than $700,000. The constraint on any premium depends on additional costs associated with renting while building the new house or the hassle of building a new house. The cost approach can be useful when evaluating whether to license a patent (or to determine damages for patent infringement) when there is a design around option.

However, in the context of valuing IP rights for a transaction, the backwards-looking nature of this approach may not accurately measure the economic benefit derived from the intellectual property. For instance, two technologies may cost approximately the same to create but may be associated with very different income streams and, therefore, very different valuations. Indeed, in the context of biopharmaceuticals, there is no guarantee that merely matching the R&D spend of a technology will result in an alternative with comparable value. Moreover, given the long development process, it may be too late for the potential buyer to catch up in R&D, regardless of the amount of spending.

Counting cash flows is hard for disruptive technologies

The preferred approach to value intellectual property for transactions is the DCF analysis, which is based on the risk-adjusted present value of the future income stream generated by the underlying technology.

There are three key inputs to the DCF model:

• the expected economic life of the asset;
• expected future cash flows from the asset; and
• the risk associated with realising those future cash flows.

The resulting metric from this analysis is the net present value (NPV).

Both pharmaceutical companies and investors in this space have honed the DCF analysis over several decades, whether for in-house capital budgeting, strategic acquisition or venture capital investment. The economic life of the asset is typically assumed to be the marketing exclusivity provided by the regulatory agency (eg, the Food and Drug Administration in the case of the United States) or loss of patent exclusivity, whichever is later. After that, generics are assumed to enter and displace some portion of the branded share of the compound. While the expected cash flows are estimates, they are based on the anticipated cost of development, projected revenues (assuming the price and number of patients or quantity sold and peak sales), and estimated manufacturing and operational expenses.

Over time, these models have grown more complex, taking into account first mover advantage, payor strategies, reimbursement rates, rebating and promotions and mitigation of generic entry, among other factors. There are numerous databases of past sales by therapeutic class (eg, IMS Health (now IQVIA), Wolters Kluwer and Symphony Health) on which analysts can base these estimates of cash flow. There is also a well-developed approach to incorporating the risk associated with realising those cash flows.

Many sophisticated pharmaceutical companies adjust risk using a measure for the probability of success (known in the pharmaceutical context as the probability of technical and regulatory success), which takes into account the probability of making it through each stage of clinical trials and obtaining regulatory approval.

Venture capital firms use their own discount rates, based on the stage of development. Of course, even in the traditional small-molecule world, these types of forecast can be inaccurate when it comes to assessing the value of innovative products. For example, the projections for Lipitor – Pfizer’s blockbuster drug for managing cholesterol – assumed just a fraction of the approximately $14 billion in peak sales that the drug actually achieved. Between 1992 and 2017, Lipitor generated US sales alone of some $95 billion – $5 billion more than Pfizer paid to acquire Warner-Lambert (which developed Lipitor) in 2000. Perceived as a ‘me too’ late entrant among statins, it was not even expected to reach blockbuster status (defined as $1 billion in annual sales). Instead, Lipitor ended up bursting through that mark in its first year on the market.

To apply the DCF model to the financial valuation of a gene therapy candidate, we first need its economic life expectancy. The experience of the first approved gene therapy, Glybera, is a useful example here. uniQure received approval for Glybera, indicated for a rare disease called lipoprotein lipase deficiency, in 2012 in Europe. Glybera was priced at €1.1 million (approximately $1.4 million). Only one commercial patient in Germany received the drug and despite an apparently positive outcome for the patient (who was hospitalised some 40 times before the treatment but not once after), uniQure abandoned the drug in Europe and never completed the approval process in the United States. Sticker shock clearly played a role in Glybera’s demise, although rapidly evolving technology may render existing products obsolete well before the expiration of underlying patents.

For instance, there are two approved gene therapies for cancer in the United States: Kymriah and Yescarta. Both are chimeric antigen receptor T-cell (CAR-T) therapies and are based on using the patient’s own white blood cells. Both are meant to be one-shot therapies. Therefore, like Glybera but unlike Lipitor, there are no refills of prescriptions to treat a chronic disease. As a result, a newer therapy that is better on efficacy or safety, or one that is allogeneic (as opposed to being autologous) can easily wipe out the market potential for existing therapies.

In other words, there may not be a first-mover advantage of the kind that is often apparent for drugs that treat chronic conditions such as diabetes or hypertension, where existing products continue to maintain sales (especially from refills for existing patients) despite the entry of newer therapies. We have heard from the chief financial officers of some prominent companies developing gene therapies that these are some of the most difficult financial projections that they have undertaken in their professional careers.

The pricing of gene therapy products is another key input that is also evolving. Recently, BioMarin attracted the ire of Senator Bernie Sanders when its CEO indicated that the company was looking to price its haemophilia A gene therapy – Valrox – at between $2 million and $3 million. Glybera, by comparison, was ‘just’ $1.4 million. From an economic perspective, this pricing is not without merit. The annual cost for treating severe haemophilia A using replacement factor VIII is more than $250,000 in the United States. Even if we assume, conservatively, a lifespan of 60 years and no cost increases, that totals about $15 million per patient. With a relatively high discount rate, the present value is still several million dollars. Moreover, the current treatment requires several injections each week over a lifetime. By contrast, Valrox is expected to be a one-time therapy, which would thus justify a price premium for convenience and improved quality of life.

However, sensitivity to pricing is not driven solely by the total cost of the therapy but also by lack of evidence over the durability of the response (eg, will factor VIII levels fall off over time and, if so, after how many years?). Indeed, BioMarin recently published data that seems to indicate continuing declines in factor VIII levels over time. Equally important is the changed cash-flow streams for payors. Instead of paying, say, $250,000 per year over the lifetime of the patient, payors may need to fork out an order of magnitude higher upfront lump sum. Biopharmaceutical companies have floated a number of novel ways to structure these payments, ranging from payment for demonstrated performance to payment as an annuity over a fixed number of years.

Of course, a standard approach to dealing with uncertainty in financial modelling is to carry out a Monte Carlo simulation. At its simplest, this could be a base-case scenario bookended with a worst and best-case analysis. For disruptive technologies in their infancy, it is hard to know where these bookends lie. Equally tough is the distribution of uncertainty around the base case (which itself can be difficult to estimate). Therefore, even a full-blown sophisticated Monte Carlo analysis could be limited in its ability to evaluate disruptive technologies. However, it could provide value compared to a DCF model based simply on point estimates in the face of great uncertainty.

Finally, the third significant input into the DCF model is an estimate of risk associated with realising these future cash flows. The pharmaceutical sector has a lot of historical data on which to base these estimates of probability of success. However, it is unclear how well experience from small molecules or even monoclonal antibodies transfers to novel gene therapies. In any case, all of these risks would translate into a higher discount rate in the DCF model, which, in turn, would produce a low NPV. By contrast, as illustrated earlier, the actual valuations for gene therapy investments and acquisitions can be substantial.

In summary, a DCF model can be difficult to put together for disruptive technologies, including the current state of development of gene therapies.

Looking to market comparables

When asking how these valuations are being determined, Audentes provides a useful example. Its sole product under development was for an ultra-rare neuromuscular disease known as X-linked myotubular myopathy (XLMTM), with an estimated target patient population of about 40 boys in the United States and potential revenues of $80 million. Yet, as indicated earlier, Astellas paid more than 37 times this amount ($3 billion) for the company in 2019. How can such an eye-popping multiple be justified? As one equity analyst from Piper Jaffray noted: “We’ve long viewed the scarcity value of Audentes’ fully integrated manufacturing and broad product portfolio as highly attractive.”

Similarly, in 2019 Roche paid $4.8 billion for Spark Therapeutics even though, at that time, Spark’s only product was Luxterna, a gene therapy for a rare form of blindness, with net product sales of only $27 million in 2018. Roche only accounted for about $195 million in net identifiable assets in a $4.8 billion acquisition.

A year earlier, Juno Therapeutics was acquired by Celgene for a whopping $9 billion. At the time, Juno did not have a single approved product. Indeed, even by mid-2020, Juno still has no approved products, with its CAR-T therapy for certain types of cancer still awaiting approval from the FDA. At the time of the acquisition, Celgene explained that what drove the valuation was “a novel scientific platform and scalable manufacturing capabilities to position Celgene at the forefront of future advances in the science of cellular immunotherapy” (source: Celgene). Celgene was acquired in 2019 by Bristol-Myers Squibb for $74 billion.

The catalyst for these large valuations of gene therapy companies was the 2017 acquisition of Kite Pharma by Gilead Sciences for almost $12 billion. At the time, Kite’s CAR-T therapy – Yescarta – was under priority review by the FDA (and was subsequently approved). However, the initial indication for which Yescarta received approval – relapsed or refractory diffuse large B-cell lymphoma (DLBCL) – estimated only 7,500 patients in the United States each year (and a similar number in Europe). Given the low patient population, higher cost of manufacturing and lower margins, the investment seemed to focus on entry into a novel approach to treating cancer that might pay off in several successful future products, rather than on Yescarta itself.

A leading player making a big bet is often a hallmark of how valuations get set in disruptive industries. Consider, for instance, Facebook’s acquisition of WhatsApp for $19 billion in 2014 on then revenues of just $20 million (or 1% of its acquisition cost). What Facebook actually acquired was a platform with more than 600 million active monthly users. Similarly, Gilead’s acquisition led to significant additional deals in the biotech space, including the large ones discussed earlier. Demand also increased for biotech IPOs. For instance, pre-money median valuations of biotech firms increased from about $200 million in 2016 to about $350 million by 2018 (source: Cowen). Unsurprisingly, there has also been a significant flow of venture capital money into biotech firms, increasing from about $19 billion in 2017 to about $28.7 billion in 2018.

Valuing disruptive technologies

So, how should intellectual property be valued during the early stages of disruptive technologies?

Perhaps the best-suited approach is to think of valuation in terms of real options. An option gives the owner the right – but not the obligation – to buy or sell an asset at a future date at an agreed-upon price. Real options involve real rather than financial assets. A key feature of options is that decisions are made after some uncertainties have been resolved. The option price is analogous to the initial cost of a project (or initial funding in a start-up). When venture capital firms invest in a Series A financing of a start-up, they are essentially purchasing an option in that company. A typical venture capitalist may invest between $5 million and $20 million in a start-up and bring on other investors to form a syndicate in case the funding need is greater. Then, depending on the performance of the start-up, the investors can abandon or expand their investment in the company.

The real options approach to valuation differs from the DCF model significantly. First, when we perform a DCF analysis to calculate NPV, the premise is that we can estimate the inputs discussed earlier with some high degree of economic certainty. However, as we have seen, estimating future cash flows, the duration of those cash flows and the likelihood of realising those cash flows can be difficult for disruptive technologies. By contrast, the real options approach requires only that we invest a small amount (the option price) upfront and wait and see how the company does before deciding on the next investment. This approach more closely mimics how venture capital firms generally make investment decisions.

When faced with greater uncertainty and risk, it is also useful to diversify. If we bet on a single product (or technology), we face a binary outcome of 0 (failure) or 1 (success) – what mathematicians call a Bernoulli distribution. Now suppose we invest in a company developing four drug candidates. Each one by itself still has a binary outcome of failure or success in terms of commercialisation. However, if the fate of the four products is independent of one another, then the outcome of all those individual drug candidates follows a binomial distribution. The higher the number of drug candidates a company has in its pipeline, the more the shape of the binomial distribution resembles the familiar bell-shaped curve. In other words, the risk from investing goes down. Biotech companies that garner high valuations often have several drug candidates in the pipeline.

Eventually, of course, the technology stabilises, the development process becomes more certain and the regulatory environment more transparent. We can also gather historical data on market size, performance of previous products and financial terms of previous transactions. As a result, we can eventually rely more on the traditional tried-and-true methods of valuation, such as the DCF analysis. However, even then, valuation based on a real options framework is still the better approach. It better reflects how decisions are made in the real world and allows for initial investments to be expanded, abandoned or delayed (the options) as uncertainty is resolved with time. As a result, a real option-based valuation more accurately reflects the underlying value being assessed, whether it is intellectual property such as a patent or the expected value of a product such as a gene therapy.