The Pharmagellan Guide to Biotech Forecasting and Valuation

Frank S. David, Seth Robey and Andrew Matthews · 2017

Thesis

The value of an early-stage drug depends on a chain of uncertain inputs, from how many patients it could treat to its odds of approval, and the way to forecast it well is to set out each input explicitly and anchor it to published evidence on how drugs have actually performed, rather than to a single optimistic guess.

The main ideas

A forecast's main value is forcing its builders to state and challenge their assumptions.
Revenue is built from separate inputs: patient population, adherence, price, market share, launch curve and loss of exclusivity.
Each input should be anchored to benchmarks from research, filings and the history of comparable drugs.
New drugs take years to reach peak sales, and the shape of the launch curve changes an asset's value a great deal.
Revenue falls sharply when exclusivity ends and generic competitors arrive.
Costs, from manufacturing to sales force to trials, must be modeled as carefully as revenue.
Valuation adjusts future cash flows for the probability that the drug fails and for the time value of money.

Why build the model

The Pharmagellan Guide to Biotech Forecasting and Valuation, published in 2017, is a short reference book of about 140 pages by Frank S. David, Seth Robey and Andrew Matthews. David, a physician and scientist trained at Columbia, worked in oncology strategy at AstraZeneca and in transactional consulting at the investment bank Leerink Partners before founding Pharmagellan, an advisory firm on drug research strategy. He later became a professor of the practice at Tufts.

The book is built for executives, investors, dealmakers and entrepreneurs who build or use financial models of drugs still in development. Rather than offering a proprietary method, it works through the inputs to a model one at a time, giving benchmarks for each drawn from more than 150 references: peer-reviewed studies, industry reports and company filings.

David has argued that forecasts made years before a drug launches are bound to be imprecise. Their value lies in forcing a team to set out its key assumptions where others can see and challenge them.

Implication: when you read a drug forecast, ask for the assumption behind each input and the evidence for it, not just the final number.

The size of the market

The book divides revenue into its parts. It starts with the addressable patient population: how many people have the condition, how many are diagnosed, and how many would be eligible for the drug.

It then turns to adherence and compliance. Patients who are prescribed a drug often fail to fill the prescription or stop taking it, so the number of people who actually take the drug is smaller than the number who could.

Pricing and market share complete the picture. A forecast must estimate what payers will pay and what share of eligible patients the drug will win against existing treatments and other new entrants.

The mechanism is that each step narrows the market, and an error at any step multiplies through the rest.

Implication: build revenue forecasts from the patient up, step by step, so that each narrowing is visible and can be checked.

Launch curves

A drug does not reach its peak sales at launch. Its uptake follows a curve that rises over several years, and the shape of that curve has a large effect on value, because sales earned later are worth less today and patent protection runs out on a fixed date.

Robey and David published research on launch curves in Nature Reviews Drug Discovery in 2016. Studying drugs approved from 2000 to 2002, they found that first-in-class drugs and later entrants took about the same time, roughly six years, to reach peak market penetration. This contradicted an older study of cardiovascular drugs from the 1980s, which had found that first-in-class drugs ramped up much more slowly than later ones. They argued that small changes in the assumed time to peak can sharply change how attractive an asset appears.

Implication: test how much an asset's value depends on the speed of its launch, and do not assume a novel drug will ramp up faster or slower without evidence.

Loss of exclusivity

The other end of a drug's life matters as much as its launch. When patents and regulatory exclusivity expire, generic or biosimilar competitors can enter, and prices and sales typically fall steeply.

The book treats loss of exclusivity as a distinct input. A forecast must estimate when protection ends and how fast revenue declines afterward, which depends on how many competitors enter and on the type of drug. Data from the U.S. Department of Health and Human Services show that prices of generic drugs commonly fall by 70 to 80 percent or more within a few years when many competitors enter.

The mechanism is that most of a drug's value is earned in the years between peak sales and the end of exclusivity, so the date of that end sets a ceiling on its value.

Implication: check the patent and exclusivity timeline before trusting any long-range revenue forecast.

Costs

The second part of the book covers expenses. A model must estimate the cost of goods sold, sales, general and administrative costs, the size and cost of a sales force, marketing, research and development, and regulatory costs.

These inputs are often treated more casually than revenue, but they determine how much of the revenue becomes profit. Development costs come early, before any revenue, and so weigh heavily in a valuation. A sales force must be built before launch, and its size depends on how many physicians need to be reached.

Implication: give costs the same scrutiny as revenue. A plausible sales forecast with unrealistically low costs overstates value as much as an inflated sales forecast does.

Valuation

The final part concerns valuation, through two inputs: the probability of technical and regulatory success, and the discount rate.

Most drugs in development fail. A valuation must therefore weigh each future cash flow by the probability that the drug will reach that point, using benchmarks for the odds of passing each stage of clinical trials and regulatory review. This approach is usually called risk-adjusted net present value.

The discount rate then reduces future cash flows to their value today, reflecting the cost of capital and the time before the drug earns money.

The mechanism is that an early-stage drug's value is a probability-weighted average of success and failure, and a forecast that assumes success overstates it many times over.

Implication: when an early-stage asset is valued, find out what probability of success and what discount rate were assumed, since small changes in either move the answer a long way.

Limits

Its benchmarks date from the mid-2010s. Drug pricing, the effects of the Inflation Reduction Act and the market for biosimilars have changed since, so its figures need updating against newer evidence.
It lists the parts of a model rather than building one. Readers have noted that it explains each input without walking through a complete worked model, and some wanted more on valuation.
Its valuation approach has known weaknesses. Average success rates by trial phase ignore what is specific to a given drug, and practitioners disagree on what discount rate belongs in a model that already adjusts for risk.

The Interpretation of Financial Statements — reading the financial statements that the book's benchmarks draw on.
The Emperor of All Maladies — the history of cancer research and the long path from discovery to treatment.