Lloyd Klickstein on How New Drugs Get Made

At a cocktail party three decades ago, a guest asked oncologist Steve Rosenberg how new cancer drugs get made. Do scientists just walk into a kitchen and grab things? Peter Attia retells that story near the start of this conversation with Lloyd Klickstein (05:48) because it captures something true: even smart people have almost no picture of how a molecule becomes a medicine. Klickstein has spent more than two decades filling in that picture, first at the drugmaker Novartis, then across a string of biotech startups, and this conversation uses one drug, bimagrumab, as a guide through the whole process, from a spreadsheet of unmet medical needs to a combination trial that beat bariatric surgery.

The starting point was not a molecule. It was a list. At Novartis, Klickstein and colleagues built a catalogue of roughly 7,000 unmet clinical indications, diseases nobody yet had a good drug for (09:04). One bucket was frailty in older adults. Klickstein, who saw these patients as a practicing rheumatologist, cites a statistic that still stops him: once a frail older adult is forced into a nursing home, the three-year mortality rate approaches 90 percent (10:59).

"Being a frail elderly person who has to go to a nursing home was worse than most cancers." (Lloyd Klickstein, [10:59])

That number set off a research program aimed at sarcopenia, the medical term for muscle loss paired with weakness. But before Klickstein's team could treat it, they had to measure it, and the first attempt failed outright. They fitted nursing-home residents with pendant accelerometers meant to detect falls automatically. Over six months, 60 residents produced 117 confirmed falls, and the device caught only 17 percent of them (34:11). Attia suggests an obvious workaround, just ask residents to self-report, and Klickstein's answer is a small lesson in humility: recall bias, device unfamiliarity, and residents' fear of losing their independence all made self-report unreliable too. The team eventually shelved the whole falls-prevention program because they could not measure the outcome they wanted to change.

What a Patent Actually Buys

Before any biology, there is arithmetic. A drug patent lasts 20 years from filing, but because years disappear into research and trials before a drug ever reaches a pharmacy, the real payoff window, called market exclusivity, is usually just 10 to 15 years from launch (17:01). That clock is why large drug companies stagger separate patents on formulation, dose, delivery device, and manufacturing process, stretching protection well past the original patent on the drug itself, a practice Klickstein says can be abused (21:01). Not every drug takes the patent route. Klickstein describes two older medicines protected by trade secret instead of patent: Armour Thyroid, a desiccated animal-thyroid hormone whose exact composition was never disclosed, and Acthar Gel, an ACTH extract purified from pig pituitaries that Klickstein once used in the emergency room for gout attacks, until it was pulled from the market during the mad cow scare and later relaunched at up to a thousand times the price (18:39).

This economic logic explains why Klickstein is blunt about gray market research peptides, the compounds sold online labeled not for human use that mimic drugs like semaglutide or the experimental peptide BPC-157. What a buyer is actually paying for when they choose an approved drug, he says, is GMP, short for good manufacturing process, the documented, inspected chain of custody that guarantees what is on the label is what is in the bottle (76:48). BPC-157, he notes, has never been reproduced outside the lab of the single scientist who first described it, has no known human receptor, and is not encoded in the human genome. He calls it the platform for the greatest grift in the wellness industry (80:28).

From Mouse Miracle to Modest Human Gain

The middle of the conversation turns to the biology behind bimagrumab. In the 1990s, geneticist Se-Jin Lee discovered myostatin, a protein that restrains muscle growth; block it in mice and the animal builds like a bodybuilder. But myostatin alone does not explain human muscle regulation. Klickstein's team at Novartis, working with researchers David Glass and Chris Liu, found that a second molecule, activin A, shares the job, and blocking both together, by targeting their shared receptor, is necessary for maximal hypertrophy in humans (39:22). Because the receptor needed binding at extremely high affinity, in the low picomolar range, small-molecule chemistry could not do the job precisely enough, so the team built an antibody instead, screening thousands of candidates for the stickiest binder (47:52).

Before any of this reached patients, Klickstein had to decide how much risk was acceptable in a healthy volunteer. His personal rule, borrowed from a mentor named Bob Schmauder: never expose a healthy volunteer to a serious risk greater than the annual odds of being struck by lightning, about one in 100,000 (85:58). If a drug's risk exceeds that bar, move testing into patients who stand to benefit, so risk is balanced by potential reward. It is not an FDA regulation, just Klickstein's own standard, and he notes that across three decades in the industry, something seriously wrong happens to healthy volunteers roughly once every ten years, most notoriously the Tegenero antibody trial in London, where six volunteers were dosed simultaneously with a T-cell-activating antibody and several nearly died (87:49).

The mouse results for bimagrumab were dramatic, producing more hypertrophy than even myostatin knockout animals. The human results were not. Bimagrumab increases human muscle mass by only 4 to 8 percent, far short of the 20 to 30 percent gains seen in rodents (61:11), and Klickstein is candid that biology, not dosing, explains the gap. Worse, bigger muscles did not translate into meaningfully better function: a meta-analysis of Novartis's sarcopenia trials found only a nine-meter improvement in six-minute walk distance, not enough, in Klickstein's own judgment, to prevent falls. And in a pattern he says is common to anabolic agents generally, muscles got larger but not stronger without resistance training layered on top of the drug (109:08).

Novartis eventually licensed bimagrumab out. Klickstein helped found Versanis Bio to redevelop it, initially for sarcopenic obesity, a hard sell in 2021: of 53 investors approached, almost none wanted anything to do with obesity drugs, a field littered with commercial failures (114:55). Then Novo Nordisk's semaglutide data landed, and the calculation changed overnight. Versanis pivoted to testing bimagrumab as a partner drug for GLP-1 therapies. The resulting BELIEVE trial ran nine treatment arms combining two doses of bimagrumab with two doses of semaglutide, plus placebo, over 72 weeks of treatment. In the highest-dose combination group, patients lost 22 to 23 percent of starting body weight, and 45.7 percent of starting body fat (125:28), a figure Klickstein compares directly to bariatric surgery.

"This, to me, this is the first medical therapy that gives fat loss equivalent to or superior than bariatric surgery." (Lloyd Klickstein, [125:28])

The trial was not without cost. LDL cholesterol rose about 20 percent in treated patients, an effect Klickstein attributes to activin receptors acting directly in the liver rather than an off-target reaction (127:53). Eli Lilly acquired the company before Versanis could run its planned phase three trial, and the drug's future, according to public trial registries, is now Lilly's to write.

Klickstein closes on two forward bets. One is selective inhibition of mTOR, short for mechanistic target of rapamycin, a nutrient-sensing pathway that extends life in yeast, worms, flies, and mice when dialed down; the trouble is that current rapamycin-class drugs also suppress a second complex, mTORC2, with side effects nobody wants over decades (133:29). The other is his newest company, built on an odd clue: certain cancer drugs that inhibit a stress-signaling pathway cause skin cancer as a side effect in about 10 percent of older patients (138:06). Klickstein's bet is that gently activating that same pathway, rather than blocking it, could prevent cancer instead of causing it, starting with patients who have already had five or more skin cancers and face a 50 percent chance of another within a year (139:59). It is a strange, backward way to find a drug, reading a side effect for what it implies about the biology it interrupted. It is also, in miniature, the whole method this conversation describes: start with the patients nobody is treating well, and let the biology tell you what is missing.