Financial Engineering for Biotech Success

In 2016, JAMA¹ published a paper whose results had been apparent for a while: More than half of all interventional drug and biologics trials conducted between 1998 and 2008 failed. The majority of the failures were due to inefficiency. Investment in biotech began to soften.²

Andrew W. Lo, PhD

One person well aware of pharma’s failures was Andrew W. Lo, PhD, a finance professor at the MIT Sloan School of Management and the director of the MIT Laboratory for Financial Engineering (LFE). Several years before the JAMA publication, Lo and his LFE colleagues launched a research agenda to identify the reasons for the multiple failures. He needed to find those reasons to answer his ultimate question: why biotech investing had become so risky. It wouldn’t be the first time his group focused on mega amounts of money: The Lo lab is known for its work on hedge funds.³

MDT: Before we get into the pharma world, can you discuss a bit of the work on hedge funds?

AL: This was in the early 2000s, a few years before regulators established rules requiring the largest hedge funds to disclose the amounts of risk they were taking on. We developed statistical measures of those risks and discovered a strong upward trend starting in 2005, a clear warning sign for a major financial crisis. At the time, the economy was going strong so when we raised these concerns in our publications and conference presentations, no one paid much attention. Until 2008.

MDT: Is that what financial engineering is? A way to solve problems?

AL: Exactly. Financial engineering uses a combination of economic and mathematical principles and data science to solve financial problems in a world fraught with risk and uncertainty.⁴

MDT: Isn’t risk the same thing as uncertainty?

AL: No. Risk is the kind of randomness that you can quantify, like the odds of winning blackjack in Las Vegas. Uncertainty, on the other hand, is the kind of randomness that you can’t quantify, the “unknown unknowns”. Financial markets are driven by both risk and uncertainty; investors are willing to bear risk for the right price, but are much less tolerant of uncertainty, at any price.

MDT: Do you find risk interesting?

AL: Very much so! Financial markets—and a significant segment of these markets includes the pharmaceutical and biotechnology industries—are critical for the well-being of every aspect of society. Risk is a central feature of these markets.

MDT: Please explain what piqued your interest in clinical trials.

AL: I became interested largely for personal reasons. About 15 years ago, several friends and a family member all died of cancer within a few years of each other, and that was a big shock to me. In trying to be helpful to them, I discovered that finance plays a huge role in drug development, and a key input into those financial calculations is the probability of success of clinical trials. That’s what motivated me to understand how to estimate these quantities more accurately.

MDT: What were your initial observations about the biopharma community?

AL: The pharma industry is focused mainly on the science and medicine of disease and the development of therapeutics, which makes total sense. But my collaborators and I realized that a common bottleneck to biomedical innovation is how drug development is financed, and this part of the business has received much less attention.

MDT: Any other observations?

AL: Because there was a lack of familiarity in the biopharma industry with financial engineering concepts like portfolio theory, diversification, hedging, and risk management, this suggested tremendous opportunities by applying these concepts systematically, more so than any other industry I’ve been involved in. The main reason is cultural differences between the scientific/medical side of biopharma and the financial side.

MDT: Please expand on your ideas.

AL: One example involves how probabilities of success (PoS) of clinical trials are estimated. The most common method is to use historical average success rates, but this approach doesn’t take into account the entire clinical path of a drug candidate from pre-clinical to Phases I through III, and then a new drug or application. Investors care more about the likelihood of success path by path, not phase by phase which is typically how historical estimates are computed. So my collaborators and I applied machine-learning techniques to forecast clinical trial outcomes using over 200 different characters or “features” such as whether the drug candidate is a small or large molecule, does it qualify for the FDA’s fast track designation or breakthrough therapy, does the sponsor have a track record of achieving FDA approvals, and so on.⁵

MDT: How did the biotech industry receive your path-by-path formulation?

AL: We received a lot of positive feedback and have published several peer-reviewed journal articles about our approach. We’ve even collaborated on an article with Novartis in which we summarized the results of a contest we held to see whether their internal teams could beat our forecasts. Something like 300 teams submitted about 3,000 models.⁶

MDT: Your group has published other ideas regarding risk as well.

AL: Yes. One of our earliest publications in this field made the point that portfolio diversification was extremely important to reduce the financial risks to biotech venture capitalists.⁷ Venture capitalism is vital to biomedical innovation because of the high cost of therapeutic development and the outsized and binary risks associated with clinical testing.⁸

MDT: Was that idea controversial?

AL: At first, there was a lot of resistance. Actually, that’s a bit presumptuous of me. At first there was complete indifference. I was a total outsider to the biopharma industry, so I was largely ignored. In 2010, I gave a talk at a Milken conference and suggested that to really move the needle in oncology, we would need a megafund of $30 billion. I didn’t pull that amount out of thin air—I read the literature of the cost of drug development, asked how many drug candidates are needed to achieve portfolio diversification, and then calculated the risk and reward of the portfolio for various levels of funding, which is my stock in trade. But when I mentioned the $30 billion figure in my talk, there was this collective hush in the audience because the number was so much larger than any biotech fund in existence at the time. This was well before private equity funds started investing in the biopharma sectors, and a number of industry insiders argued that no one could manage a fund of that size.

MDT: What did you do?

AL: I drew on my prior experience in starting my own asset management company, AlphaSimplex Group, where I learned that new financial products often take time for investors to appreciate, and that launching smaller “proofs of concept” can be a good way to begin. So I collaborated with industry experts to launch a several biotech portfolio companies including BridgeBio Pharma and Roivant Sciences. Eventually, this portfolio approach gained more traction and now there are several multi-billion-dollar biotech private equity biotech funds, including, Blackstone and KKR. I guess these ideas weren’t so crazy after all!

Christine Bahls is a freelance writer for medical, clinical trials, and pharma information.

References

1. Hwang TJ, Carpenter D, Lauffenburger JC, Wang B, Franklin JM, Kesselheim AS. Failure of Investigational Drugs in Late-Stage Clinical Development and Publication of Trial Results. JAMA Intern Med. 2016;176(12):1826–1833. doi:10.1001/jamainternmed.2016.6008
2. Mullen P. Where VC Fears To Tread. Biotechnol Healthc. 2007 Oct;4(5):29-35. PMID: 22478674; PMCID: PMC2651715.
3. Hasanhodzic, Jasmina and Lo, Andrew W., Can Hedge-Fund Returns Be Replicated?: The Linear Case. Journal of Investment Management, Vol. 5, No. 2, Second Quarter 2007, Available at SSRN: https://ssrn.com/abstract=989605
4. Andrew W. Lo. Infinite MIT. https://infinite.mit.edu/video/andrew-w-lo
5. Chi Heem Wong, Kien Wei Siah, Andrew W Lo, Estimation of clinical trial success rates and related parameters, Biostatistics, Volume 20, Issue 2, April 2019, Pages 273–286, https://doi.org/10.1093/biostatistics/kxx069
6. Siah KW, Kelley NW, Ballerstedt S, et al. Predicting drug approvals: The Novartis data science and artificial intelligence challenge. Patterns (NY). 2021 Jul 21;2(8):100312. doi: 10.1016/j.patter.2021.100312. PMID: 34430930; PMCID: PMC8369231.
7. Fernandez, Jose-Maria, Roger M. Stein, and Andrew W. Lo (2012), Commercializing Biomedical Research through Securitization Techniques, Nature Biotechnology 30 (10), 964–975.
8. Lo A, Thakor RT. Financial intermediation and the funding of biomedical innovation: A review. Working Paper 30594. National Bureau of Economic Research. October 2022.