Drug Development Failure How GLP-1 Development Was Abandoned in 1990

Abstract

Many factors determine whether and when a class of therapeutic agents will be successfully developed and brought to market, and historians of science, entrepreneurs, drug developers, and clinicians should be interested in accounts of both successes and failures. Successes induce many participants and observers to document them, whereas failed efforts are often lost to history, in part because involved parties are typically unmotivated to document their failures. The GLP-1 class of drugs for diabetes and obesity have emerged over the past decade as clinical and financial blockbusters, perhaps soon becoming the highest single source of revenue for the pharmaceutical industry (Berk 2023). In that context, it is instructive to tell the story of the first commercial effort to develop this class of drugs for metabolic disease, and how, despite remarkable early success, the work was abandoned in 1990. Told by a key participant in the effort, this story documents history that would otherwise be lost and suggests a number of lessons about drug development that remain relevant today.

The well-worn trope that "history is written by the winners" applies as much to the history of biotechnology companies as it does to the wars and other geopolitical events to which the phrase is more commonly applied. In the 48 years since the biotech industry began with the launch of Genentech in 1976, followed soon thereafter by Biogen, Chiron, Amgen, and many others, much has been written to celebrate an industry that invigorated and reshaped the pharmaceutical business (Hughes 2011; Maraganore 2009). Of course, not all early entries lived to celebrate their success. In these early years (as well as today), companies failed for many reasons, including flawed science or therapeutic hypotheses, ineffective leadership and staffing, defective business models, inadequate funding, and simply bad luck. Both the successes and failures should interest historians of science and entrepreneurship, as both provide important lessons, and it's important to document some of the more interesting "failures" before relevant details are irretrievably lost.

The enormous medical benefits and financial rewards of the GLP-1 class of therapeutics have incentivized me to document the brief history of a company I cofounded in 1987, Metabolic Biosystems (MetaBio). My goal is to record the circumstances under which MetaBio "failed" as a company before those involved are no longer capable of doing so. The most important success of MetaBio was its demonstration of the anti-diabetic efficacy of GLP-1 in humans, a finding cut short by decisions of the startup biotech and Pfizer, the major pharma company that obtained exclusive rights to the work and rapidly funded it. Nearly 40 years after MetaBio's founding, as GLP-1–related therapeutics are now established as an immense success, how and why these early efforts to develop the class were prematurely terminated should be of interest to historians, entrepreneurs, drug hunters, and clinicians.

The story begins with John Baxter, an endocrinologist and pioneer in the cloning of hormones, including growth hormone (Seeburg et al. 1977). Baxter was a highly regarded professor of medicine at UCSF, and after the establishment of the first cohort of biotech companies by several of his friends and colleagues, investors encouraged Baxter to start one as well. California Biotechnology (Cal-Bio) launched in 1982. While remaining a full-time faculty member at UCSF, Baxter recruited an outstanding cohort of scientists to CalBio's Mountain View, California, facility, where they pursued a diverse group of projects related to cardiovascular biology, pulmonary function, growth factors, inflammation, and other areas. Over time, CalBio's major asset became a formulation of B-type atrial natriuretic peptide, and Auriculin—subsequently named Natrecor—was eventually approved by the Food and Drug Administration (FDA) in 2001 as a treatment for acute congestive heart failure, an approach that eventually was seen as controversial (Kesselheim, Fischer, and Avorn 2006). Along the way, CalBio changed its name to Scios, and the company was acquired by Johnson and Johnson in 2003 for $2.4 billion.

This story is not about CalBio/Scios per se, however, but about MetaBio, the small company that I cofounded with two Harvard colleagues in 1987 as a wholly owned subsidiary of CalBio, a corporate structure that may have contributed to its ultimate demise. This origin story begins with a project I developed with two other colleagues at Harvard Medical School (HMS),—Alan Moses and Martin Carey.—The project was unrelated to our individual core research interests, which in my case was the mechanisms of insulin resistance. In 1983, Moses and I became interested in mechanisms to enhance nasal absorption of insulin as a possible approach to diabetes therapeutics (Moses et al. 1983). Working with Carey, a bile salt authority, we identified and patented an adjuvant class (fusidic acid derivatives) that enabled control of blood glucose excursions when administered at mealtime as a nasal spray formulation with insulin (Gordon et al. 1985). We sought to identify a company interested in developing this with us. After several unproductive meetings with potential investors, I had the opportunity to discuss this idea with fellow endocrinologist Baxter, who immediately said, "You should do this with CalBio!" Soon thereafter, Beth Israel Hospital—where I worked, and which controlled the patent—licensed the discovery to CalBio. CalBio helped us extend our early work, and soon struck a deal with Eli Lilly to further advance the insulin delivery project (Longenecker et al. 1987). The 1985 Calbio annual report featured this "Nazdel" drug delivery platform prominently as a major company asset, referencing having licensed it from Beth Israel and Brigham and Women's Hospitals.

Though never adequately explained, Lilly abandoned the collaboration over the next couple of years, to the extreme consternation of the team, citing "issues with irritation in animal toxicology work." As nasal drug delivery was not our primary area of scientific interest, we regretfully accepted CalBio's decision to end the program. But I remained in touch with the irrepressible Dr. Baxter, and over lunch during one of his visits to Boston in 1987, he asked me to consider starting a new company focused specifically on diabetes and obesity. No such start-up existed at the time. Perhaps unsurprisingly, I answered, "Wow, of course." He asked me to think seriously about what projects such a company might undertake, and who else might become involved, and I agreed to do so.

My close friend and former National Institutes of Health (NIH) collaborator Ron Kahn had become the director of research at the Joslin Diabetes Center and was also an HMS professor. Stimulated by the Baxter challenge, we spent several memorable weekend sessions in my living room discussing potential project areas, and we eventually brought in Bruce Spiegelman, another HMS professor with whom I had been collaborating. Within a month, we developed a list of actionable ideas that excited us. Upon hearing these, Baxter was also excited and proposed a business model that was rapidly implemented. The "Newco" of which we would be founders would be a wholly owned subsidiary of CalBio, operating out of its facility and with CalBio's employees. "Newco"—rapidly named Metabolic Biosystems—would have a scientific advisory board whose members we would pick and that we would chair. After a minimum of five years, CalBio could buy out our "founders' interest" at a "market valuation," or we could alternatively force sale of our interest to CalBio. Lawyers got involved, and MetaBio was soon a reality. This was not the optimal approach to maximize founder value, but it enabled the company to have a quick and painless start.

CalBio leadership rapidly decided to seek a major pharma company to co-invest in the effort, both to provide a needed infusion of capital and expertise, and to further validate the effort. We spiffed up our internal plan, and with the help of Baxter, presented it to Pfizer, a major pharma company with an active interest in metabolic disease, as well as several other companies. Remarkably, within a few weeks, we received a positive response from Pfizer, proposing a deal by which they would obtain exclusive rights to our entire program. The joint venture would represent all activity of both CalBio and Pfizer to develop treatments or diagnostics addressing diabetes or obesity using the tools of biotechnology that CalBio brought to the table and that Pfizer apparently sought. Pfizer would provide $30 million over five years and would lead any clinical development programs. Legal documents were created, edited, and signed in July 1988, officially launching the MetaBio/Pfizer alliance. My colleagues and I received equity in MetaBio/CalBio and consulting fees, which seemed like an appropriate arrangement, and we began making regular visits to both the Mountain View offices of CalBio and the Groton, Connecticut, headquarters of Pfizer Research.

The alliance was initially overseen by a joint Research Committee, with two scientists from CalBio—John Lewicki and Karen Talmadge, the latter serving as Head of Diabetes and Obesity Research at CalBio/MetaBio until 1990—and, perhaps surprisingly, two very senior scientists from Pfizer. These were Nancy Hutson, Pfizer Senior Vice President of Global R&D and director of Pfizer's Research and Development site in Groton, and Gregory Gardiner, then Group Director of R&D Operations of Pfizer Central Research, from which position he led Pfizer's external biotechnology strategy. The seniority of the Pfizer participants signaled the seriousness of their interest. They had certainly decided that the joint effort we proposed offered considerable value to Pfizer.

So, what was the MetaBio research portfolio that garnered such rapid interest in 1987/88? MetaBio projects fell into four categories: insulin analogues, enhancers of insulin action, adipocyte secreted factors, and gut factors regulating metabolism. A brief discussion of how we pitched these in our initial proposal follows.

The discovery and rapid deployment of insulin into clinical practice at the beginning of the 20th century was one of the great scientific and medical breakthroughs of modern times (Flier and Kahn 2021). The FDA approval of recombinant human insulin in 1982, the first recombinant human therapeutic, was another insulin-related first of great importance. Despite the remarkable benefits of these breakthroughs, it was apparent to both physicians and patients that insulin therapeutics had much room for further improvement; both more sustained and more rapid kinetics after injection were desired. We were in contact with several acknowledged leaders in the synthesis and analysis of insulin analogues with favorable biologic and pharmacokinetic properties and proposed to engage them to identify novel analogues. We were aware of similar efforts by major insulin producers (such as Lilly and Novo Nordisk) but believed we could compete. Clinical trials with human proinsulin were proceeding at Lilly, based on data that proinsulin had preferential effects on the liver that might produce therapeutic benefits (Revers et al. 1984). We proposed to design a better "hepato-specific" insulin based on emerging structural insights.

Insulin resistance is an important contributor to the pathophysiology of type 2 diabetes, and the research programs of Kahn and Flier both addressed this problem (Flier, Kahn, and Roth 1979). Kahn's lab had discovered that the insulin receptor was an insulin-stimulated tyrosine kinase and was heavily focused on identifying factors that antagonized this activity that might be inhibited (Kasuga et al. 1983). The initial MetaBio program would seek to identify inhibitors of tyrosine phosphatases that might enhance insulin signaling.

Spiegelman's lab had discovered several adipocyte secreted proteins, and he and I collaborated to show that tissue expression and blood levels of one of these, named adipsin, was dramatically reduced in ob/ob and db/db mice (Flier et al. 1987). Our plan with MetaBio was to determine whether restoring adipsin levels to normal with recombinant protein would have metabolic benefits in obese mice; we would also look for novel adipocyte secreted proteins with biologically important roles. This included a major effort to discover the product of the as yet undiscovered ob gene, using both genetic and biochemical approaches (identifying a secreted protein missing from media bathing adipocytes from ob/ob vs control adipocytes). This work began in 1988, six years before the positional cloning of the ob gene in 1994 and identification of the secreted protein leptin (Zhang et al. 1994).

It was known in 1988 that two gut-derived hormones—GIP and GLP-1—could importantly enhance insulin secretion, but which of these "incretin" hormones was the most biologically critical, and how they contributed to physiology and disease, was unclear. We declared our intention to further explore this important unsolved question.

The projects described above were the basis for MetaBio's formation, but within weeks an unexpected event supercharged the program, and likely produced the rapid interest of Pfizer by bringing GLP-1 into the fold. In May 1987, I attended the annual meeting of the American Society of Clinical Investigation in Atlantic City, at the time the key meeting for scientific interchange across broad areas of academic medicine. My practice over previous years was to carefully examine the thick book of abstracts arriving by mail a couple of weeks before the meeting to identify key talks I wished to attend. One that caught my attention was from the lab of Harvard endocrine scientist Joel Habener, then a leader of genetic approaches to physiology and endocrine disease (Potts et al. 1980; Zingg, Goodman, and Habener 1984). Presented in a metabolism session at 8am on Saturday morning, this abstract reported discovery through cloning of the glucagon gene locus of a putative new gastrointestinal peptide, which they named glucagon-like peptide-1 (GLP-1). The 10-minute talk by a fellow from the lab showed that GLP-1 (also called insulinotropin) enhanced insulin secretion in a glucose-dependent manner—that is, enhancement occurred only when glucose was elevated, and the more elevated, the greater the enhancement. There was no effect when glucose was low. Although the presented work involved islet cells and the perfused rat pancreas, I recognized the potential importance of this discovery, and when the talk was over, I approached Habener, whom I had previously met.

I congratulated him for the beautiful study and asked if he was working with a company to further develop it. He told me he had patented the discovery, but it was not yet licensed to a company. I told him of the recently forming startup biotech under the aegis of CalBio. Joel knew and admired Kahn, Spiegelman, and Baxter, and he asked me to contact him as soon as the meeting was over to discuss this further. It's hard to believe, but within a few weeks, CalBio/MetaBio had negotiated an exclusive worldwide license to the Howard Hughes Medical Institute and Massachusetts General Hospital (HHMI/MGH) patents for GLP-1 as a therapy for diabetes, in exchange for royalties on future sales. In retrospect, this was quite a remarkable outcome. At the time of this agreement, the Habener group had not yet reported, and I don't believe had carried out, any human studies.

Research by a rapidly assembled and excellent MetaBio team proceeded in several of the areas mentioned. We had identified a relatively hepato-specific insulin analogue and were well on the way to identifying the product of the ob gene. But my focus here is on GLP-1, since today analogues of this molecule are exciting therapies for both type 2 diabetes and obesity, earning a yearly combined $20 billion in revenues for several companies (including Novo and Lilly), a number expected to rise several fold over the next decade (Berk 2023; Kreymann et al. 1987). Given that—in 1988—we had a dominant GLP-1 patent position, excellent scientific staff, sufficient funding, and an alliance with a pharma giant with expertise in diabetes, how is it possible that CalBio/MetaBio didn't become another Genentech on the back of GLP-1 alone?

The failure wasn't for lack of major progress in key areas of research. Working together with Pfizer at Groton, the joint team began to infuse GLP-1 into normal and then type 2 diabetes subjects. The work clearly showed that GLP-1 infusions enhanced insulin secretion in normal and diabetic subjects, and that it lowered glucose levels when administered over a period of days in subjects with diabetes. At the time, published data on GLP-1 in humans was scant; Steven Bloom's group in London had published a paper at the end of 1987 showing an acute insulinotropic action of GLP-1 in normal subjects (Kreymann et al. 1987). The first data on GLP-1 action in diabetic human subjects was published in 1992 by Habener and colleagues at MGH (Nathan et al. 1992).

We also quickly discovered that the half-life of GLP-1 was only several minutes. On that basis, the MetaBio team launched a focused effort to identify the responsible protease, an independent approach to boosting the GLP-1 pathway in diabetes by inhibiting the protease that inactivated it. Though substantial progress was being made, Pfizer decided to pull support for this program before final identification of what years later was identified as the enzyme DPP-4, oral inhibitors of which were approved by the FDA in 2006 for diabetes treatment (Drucker 2007; Holst and Deacon 1998). The "gliptin" class of drugs peaked at annual revenues of around $10 billion and remains a major oral therapy for type 2 diabetes.

The demonstration in 1990 by the Pfizer/MetaBio program that several days of GLP-1 infusion lowered blood glucose in diabetic subjects was critical to positioning GLP-1 as a therapy for diabetes. Remarkably, the team also presented data to our internal group, showing that GLP-1 slowed gastric emptying and reduced hunger. These observations were extensively discussed by our joint team, and both were indicators of the eventual efficacy of these agents against diabetes and obesity. Sadly, none of this definitive work was ever presented publicly or published, so its existence, until now, has been known only to those involved. The Pfizer approach, enshrined in the founding documents of the alliance, was to keep the findings confidential, and no one thought twice about this at the time. The role of publishing data was dramatically different in pharma than in the academic world in which the founders had lived.

Despite our excitement about the work, we noticed a problem with how it was being evaluated by Pfizer, most specifically by Gardiner and Hutson, who had important input to the final decision-makers for Pfizer on the ongoing collaboration. As we were enjoying the positive human infusion data and anticipating a major effort to design both protease resistant GLP-1 analogues and an inhibitor of the responsible protease as new diabetes therapies, I was shocked when told that senior Pfizer leadership had concluded that there would never be another injectable therapy for diabetes other than insulin. What led them to this conclusion was never explained (and of course it proved totally false in the real world over subsequent decades), but the implication for our program was unambiguous. Pfizer stated that continuation of the project would require us to identify a route for GLP-1 administration other than injection, such as trans-nasal or transcutaneous. The MetaBio team and the founders vigorously disagreed with Pfizer's view that an injectable GLP-1 analogue would never be worth developing, but our arguments had no effect. Based on knowledge at the time about trans-nasal and transcutaneous delivery of peptides like GLP-1, and the short half-life of the protein, we considered it extremely unlikely that either approach could be successfully developed. We nevertheless undertook efforts to develop these modalities, and they were predictably unsuccessful.

About three years into the collaboration, Pfizer informed CalBio it would end its commitment after four rather than five years, permitted by the agreement with due notice. And indeed, they did just that. The other scientific founders and I tried to convince CalBio leadership to seek alternative external funding to continue the promising work, or to fund it internally, but this didn't happen. Why? CalBio's primary focus and most of its funding was to advance development of Natrecor, a project led by the company's scientific director, John Lewicki, work that eventually led to the company's sale to Johnson and Johnson. Via a process that wasn't transparent, our founder's interest in MetaBio was assessed to be of a certain (we considered low) value, and on that basis, our ownership share was "bought out." We pushed back on the terms and obtained legal advice, but in the end we accepted the terms. With that, our nascent efforts to be the first company to develop a GLP-1 based diabetes therapy, either an analogue resistant to degradation, or an inhibitor of the protease we expected would soon be discovered ended, both emerging more than a decade later as blockbuster therapies.

The lens of history is powerful. In this case, retrospective analysis makes it clear that our effort in the GLP-1 therapeutic space was far ahead of its time. If the full force of CalBio and Pfizer had pressed ahead to develop either a more stable GLP-1 analogue, or an inhibitor of the protease that would soon have been identified, it's extremely likely we would have succeeded. Patients would have benefited, and the company's success would have been enormous. What accounted for the decision to precipitously end the work?

A clear answer wasn't evident to me in 1991/92 when MetaBio ceased to function, but it's worth considering some possible explanations. Ceasing the work on GLP-1 and other promising projects was opposed by the MetaBio founders and the MetaBio scientific team, so answering this question has two parts: why did Pfizer make this decision, and why, in response to that decision, did CalBio not continue the work?

A definitive answer with regard to Pfizer would require review of internal records or interviews with key decision-makers, neither of which is possible now, more than 30 years later. What is clear is that the decision by Hutson, Gardiner, and others was mistaken, based on faulty judgments about the potential role of GLP-1 as a therapy for diabetes, the potential for an injectable formulation of GLP-1 analogues to be successfully adopted, and the potential that clinically obvious effects on appetite and gastric emptying might enable future use in obesity therapy, among others. We interacted with the key figures at Pfizer leading metabolic research and those higher up in the corporate pharmaceutical chain of command. I had been deeply impressed by their rapid decision to invest in our company, and I was equally dumbfounded by their decision to end their investment despite convincing early evidence of the program's success. On several occasions in future years, when recounting this history to others in the biopharmaceutical industry—including some at Pfizer—I was met with looks of surprise and doubt, suggesting that perhaps I was telling a tall tale, or exaggerating some of the history. I hope to counter that by writing this piece.

Of course, companies in all industries make strategic errors all the time. But given the success of GLP-1–related approaches to therapy of two extremely common diseases, this mistaken call by both Pfizer and a hungry startup is especially noteworthy. Calling attention to this failure doesn't serve the interests of present companies or previous participants in the work. On the other hand, it took several years after the publication of the beneficial actions of GLP-1 in humans by several academic groups for other pharmaceutical companies to see the value and enter the fray, eventually producing the therapies we have today. Or stated another way, Pfizer was not alone in failing to appreciate the potential therapeutic value of intervening in this pathway.

What about CalBio's decision, in the wake of Pfizer terminating the collaboration, to shut down the ongoing MetaBio effort? I had great respect for the CalBio scientists and leaders, and they were in an outstanding position to assess the value of what had been rapidly developed with advice from the founders and funded entirely by Pfizer. In the light of history, their decision to terminate the program can also be judged as a colossal error. What caused this to happen? Once again, many of the key players have passed from the scene (John Baxter passed away in 2012) or moved on to other phases of their lives. I will raise several hypotheses.

In recent discussions, Rich Casey, CalBio CEO at the time, made several points. First, his greatest focus, and that of his shareholders and external collaborators, was on Natrecor, and without Pfizer support, they judged there to be insufficient funds to continue the GLP-1 project, despite its early success. He recalls pitching the program to large companies in the diabetes space, like Lilly and Novo, and recalls them not expressing interest. He surmised the Pfizer decision led others to be wary, reflecting a "herd mentality," and he was told "the word on the street" was that future therapies for type 2 diabetes would all be orally administered drugs, rather than injectables, the view stated by Pfizer brass to CalBio.

Although a small company, I recall some internal tensions between the original CalBio group and the newly launched and smaller MetaBio program, which was operating within the same facility and business enterprise. When Pfizer decided to withdraw, CalBio was making progress on their core Natrecor program for treatment of acute heart failure, a program rarely mentioned at our MetaBio meetings. I can imagine internal discussions about a need to focus company resources on this most advanced asset once Pfizer support for MetaBio unexpectedly evaporated.

It has not escaped my attention that the outcome might also have been different if one of the founders had decided to leave academia to push these projects from within. It often takes a strong internal champion to achieve continued investment in the face of competing priorities. I remember some discussion of this idea, but neither Kahn, Spiegelman, nor I were ever interested in making such a transition. The assets of MetaBio might also have been spun out into an independent company, freeing it from the need to compete for funding with the clinical stage Auriculin program, and providing a cash infusion and metabolic focus. For unclear reasons, perhaps related to the state of capital markets at the time, that approach wasn't taken. So, confused and disturbed as we were by the outcome, we accepted the "buyout," which was less than we viewed as fair but nevertheless quite tangible, and moved on.

It's hardly newsworthy when a biotech startup fails to develop into a successful company, whether on its own or via partnership or acquisition by another company that recognizes its value. That's the nature of the venture capital business, whereby savvy investors make bets on an array of early ideas and promising people, knowing full well that most will fail. They do so in the hope that a minority of their investments will hugely succeed and justify this investment strategy. The skills and strategies that produce both successes and failures deserve analysis, for the sake of history and to inform future efforts to achieve higher performance.

Of course, even the biggest pharmaceutical companies make decisions to advance or kill programs on a regular basis, integrating information from many sources to put their large but still limited resources to work. In this process, given the many uncertainties that exist, mistakes will surely be made. To shed light on Pfizer's mistaken decision with regard to GLP-1, it may be useful to review the process by which Novo Nordisk, the Danish pharmaceutical company that now has the greatest share of the GLP-1 market, came to develop this class of drugs. This was a long and complex process with many setbacks, recently reviewed in a paper by Lotte Knudsen (2019), who ran the program at Novo from 1995. Novo began studying GLP-1 in 1992 and obtained rights to the Habener patents after MetaBio relinquished them. Their initial efforts involved attempts to create protease resistant analogues to produce a once daily treatment for testing, but these efforts were unsuccessful. A breakthrough came in 1995, when Knudsen decided to produce an analogue with an acylated fatty acid that, based on prior research, might be expected to retain activity and prolong half-life by binding to albumen (Spector 1975). This approach proved successful. Phase three trial results for liraglutide in diabetes were published in 2008/9, and the drug was registered for use in 2009 for diabetes, and in 2014 for obesity. Research on its effects to reduce appetite and cause weight loss went back to 1995 at Novo, with uncertainty about whether the mechanism to reduce appetite involved the vagus nerve, the brain, or other sites of action. Clinical success in obesity required the insight that doses required for obesity (as opposed to diabetes) required gradual upward titration to reach effective doses without causing excessive nausea or other gastrointestinal complaints. New molecular entities combining GLP-1 agonism with other hormones, like GIP, are showing even greater effectiveness, and oral versions are on the horizon (Abassi 2023).

I tell this story at a moment in time when articles about the tremendous therapeutic and financial success of GLP-1–related drugs—in both diabetes and obesity—are appearing daily in the scientific and popular press. While these GLP-1 success stories are of course important, so too is the previously untold story of Pfizer and MetaBio. Pfizer found other routes to remain successful (some, such as their mRNA COVID19 vaccine, also in-licensed from smaller companies). I suspect current leaders and employees are unaware of Pfizer's early ownership and dismissal of the potential for GLP-1. Sadly, MetaBio, the little company that could—and perhaps should have—been hugely successful with GLP-1, was a decade or more ahead of its time. Despite every possibility of success, MetaBio went down because there were mistaken ideas about what was possible and what was not in the realm of metabolic therapeutics, and because proper corporate structure and adequate capital are always issues when attempting to survive predictable setbacks. Even the biggest blockbusters can be dismissed as unworthy by smart and ambitious people, either because the underlying biologic mechanisms are inadequately understood, there are mistaken ideas of how the market will react to new approaches, drug development confronts difficult roadblocks, or bad luck intervenes. These same dynamics are undoubtedly playing out today in the complex world of pharmaceutical development, and it will likely take decades to determine just how costly some of today's mistakes will prove to be.