Peter Attia MD409 ‒ Inside modern drug development: the science, economics, and regulatory hurdles
Peter Attia MD
Introducing Lloyd and his path 0:00
Peter Attia opens the conversation with Lloyd, a physician-scientist who trained through an MD-PhD program, practiced rheumatology for about a decade, and ran an NIH-funded lab studying adhesion molecules in the immune system at Brigham and Women's Hospital. He spent his entire academic career there before joining Novartis Institutes over twenty years ago, an organization founded by Mark Fishman at the invitation of then-CEO Dan Vasella to bring the concept of translational medicine, a term coined in the UK, into industry for the first time.
How drug discovery actually starts 3:00
Lloyd explains that drug discovery begins with patients and clinical indications, essentially looking for a needed therapy that does not yet exist. He divides this work into incremental improvements, such as making a drug that requires less frequent dosing or an oral version of an injectable, versus quantum leaps into indications that may not even have official diagnostic codes yet. Incremental work is commercially safer and dominates industry effort, citing statins like Crestor and Lipitor as examples, but Lloyd argues the biggest value to society comes from tackling undescribed needs, even though that requires building new regulatory pathways and educating physicians, patients, and payers. He describes leading Novartis's New Indication Discovery Unit, where his team catalogued roughly 7,000 unmet clinical indications, grouped them into buckets like healthy aging, ENT, renal disease, and fibrotic disease, and eventually developed dozens of projects, including work on frailty in the elderly, a condition he found carried a three-year mortality rate approaching 90 percent for patients forced into nursing homes, worse than most cancers.
Molecule types and patent economics 10:00
Lloyd walks through drug classes: small molecules, which are essentially chemicals made using techniques descended from century-old dye manufacturing, biologicals such as antibodies, peptides, soluble receptors, and gene therapies, and devices, a separate FDA-regulated category covering everything from syringes to implantables. He and Peter then discuss patent law, comparing a drug's development cost and scale to building the Burj Khalifa, but noting that unlike a building, a drug's patent grants only about 10 to 15 years of practical market exclusivity from launch before the formula becomes free for anyone to use. They also touch on trade secrets as an alternative to patents, using Coca-Cola as the classic example, and Lloyd offers two pharmaceutical cases: Armour Thyroid, a desiccated thyroid hormone protected by keeping its manufacturing process secret, and Acthar Gel, a pig-pituitary-derived treatment Lloyd valued for treating acute gout, which was pulled from the market during the BSE scare and later resold by a small company that raised its price by a hundred to a thousandfold after repositioning it for infantile seizures.
Patent Layering as a Strategy 19:00
Drug makers extend protection on a valuable medicine by patenting far more than the core molecule. Using Humira as an example, the discussion describes how a company can patent the composition of matter, then later patent the formulation, the salts, the auto injector, the method of use, the dose, the route of administration, and the manufacturing process, staggering each filing years apart so the effective protection window stretches well beyond a single patent term. This layering, combined with keeping the actual manufacturing process a trade secret, means a competitor might know what a finished pill looks like without ever learning how to reproduce it.
Choosing Small Molecules or Biologics 20:32
Large drug companies tend to be agnostic about format, since they run small molecule, biologic, gene therapy, and even emerging cell therapy programs simultaneously. The choice often comes down to the patient and the disease. A drug for a childhood illness needs to be oral and pleasant tasting, which is why flavored liquid medicines exist, while certain lung conditions call for inhaled formulations because that route makes medical sense.
Sarcopenia as a Working Example 23:01
To illustrate early drug development, the conversation turns to sarcopenia, loosely defined as decreased muscle mass paired with impaired function, measured by grip strength, gait speed, or stair climbing. Because the likely patients are older adults, the drug format has to suit that population, and every added risk in a program, whether about target, format, bioavailability, or toxicity, multiplies against the others, so a string of even 90 percent likely successes can still collapse to near zero overall. This is why failing early is preferable to succeeding through phase three and then failing commercially.
Big Indication versus Orphan Disease 25:30
A strategic choice emerges between chasing a large indication like sarcopenia, which may lack a clearly defined regulatory pathway, or an orphan disease like Duchenne muscular dystrophy, which offers a faster route to approval. One approach is to prove a drug works in the orphan disease first, then pursue broader approval afterward, and both strategies have been used depending on the drug's mechanism and circumstances.
The Difficulty of Measuring Falls 27:01
Falls turn out to have many causes, including weakness, dizziness, poor vision, and lapses in attention, plus factors like slow foot reactivity and reduced type 2A muscle fiber function that let people catch themselves after a stumble. Because clinical records only capture falls serious enough to cause injury, and patients often hide falls out of fear of losing independence, a study was designed using pendant-worn triaxial accelerometers, first tested on ice skaters as a positive control, then deployed on 60 nursing home residents over six months.
A Failed Device Study 33:00
Over six months, 117 confirmed falls occurred among the residents, but the accelerometer device only detected 17 percent of real falls and was wrong 17 percent of the times it claimed one occurred, making it useless for measuring outcomes despite being the best available option at the time. A follow-up attempt using MIT researcher Dina Katabi's WiFi-based motion sensing also failed to pan out, and the fall-prevention drug program had to stop because falls simply could not be measured reliably, leading to a debate about whether just asking participants to self-report would have worked better than relying on sensors.
Myostatin and muscle growth 36:04
The conversation turns to sarcopenia, the loss of muscle mass and strength, which was the starting point for developing bimagrumab, a drug that blocks activin receptors. Researchers already knew about myostatin, a protein discovered by Se-Jin Lee that limits muscle growth, and famous experiments in the 1990s showed that blocking myostatin in mice, chickens, dogs, and cows produced dramatically oversized muscles. In humans the biology is more complex, since myostatin works together with proteins called activins to restrain muscle size, so an effective drug needed to block both by targeting their shared receptor rather than myostatin alone.
How the signaling actually works 41:31
Muscle size depends on nutrition, use, and a biochemical brake system built around a large family of TGF-beta receptors that signal through proteins called SMADs, discovered in fruit fly genetics. This signaling normally suppresses two proteins, Murf1 and atrogin, that break down muscle, so blocking myostatin and activin lets muscle protein accumulate instead of turning over. A related protein called follistatin also inhibits this pathway and boosted muscle in rodent gene therapy experiments, but it turned out impractical as a real treatment: it has a short half-life, would need dosing several times a day, would cost roughly a million dollars a month, and seems to only work during a young developmental window rather than in mature muscle.
Engineering a therapeutic antibody 48:00
Extending a protein's half-life and making it easier to purify can be done by fusing it to the Fc region of an antibody, a technique first developed at Mass General and used in etanercept, a TNF inhibitor that failed in sepsis but succeeded in rheumatoid arthritis. Because myostatin and activin bind their receptors extremely tightly, at very low nanomolar or picomolar concentrations, only a biologic antibody could bind tighter still, so small molecules were ruled out. The Novartis team, working with Morphosys, screened thousands of candidate antibodies using phage display and a glowing luciferase reporter assay, since the target receptors were too scarce to detect by normal methods, eventually landing on an antibody that preferentially blocks the activin receptor type 2B while also hitting type 2A, working across all muscle fiber types.
Years to Find a Candidate 54:31
Identifying just two to ten antibody candidates worth testing took years, not months. The team had to work out the underlying biology, build glowing cell tools that lit up in response to myostatin, and create reagents before they could even begin. The cost at this stage runs into millions of dollars, not yet the tens of millions that come later, and the overall price tag for bringing one approved drug to market today is estimated by a Tufts research group at two to four billion dollars.
Fully Human Antibodies and Immune Risk 58:01
One of the biggest unknowns at this stage is whether an antibody will trigger an immune response in people. The safest approach is to use a fully human antibody rather than a humanized one, meaning it is built from human genetic material from the start rather than retrofitted from a mouse antibody. Even so, nothing is certain until it is tested in real people, since antibodies can recombine into new sequences that no lab screen fully predicts. Historically the first antibodies ever given to humans were mouse-derived, including OKT3, an anti-CD3 antibody still used today to prevent transplant rejection, along with rabbit anti-thymocyte globulin and horse-serum antivenoms.
Testing in Mice and Primates 1:01:00
Because the human antibody Bimagrumab did not work well in mice, the team engineered a mouse-compatible version called CDDH866 to run early experiments. The results were striking, producing muscle growth in mice that was arguably more dramatic than in pure myostatin knockouts, with mice also becoming stronger and faster despite roughly a thirty percent gain in muscle mass, far more than the four to eight percent later seen in humans, mostly older adults. As soon as the mouse antibody worked, the company began notifying the World Anti-Doping Agency, which has had a screening test ready for over a decade. Moving to non-human primates let researchers confirm expected pharmacology and watch for toxicity, using a weight-of-evidence approach that asks above all whether any harm is monitorable and reversible, since irreversible organ damage typically ends a program. Large companies revisit funding and progress at two to four major checkpoints, a process a startup would experience instead as successive funding rounds.
Funding a small biotech program 1:12:00
For a small company, the key milestone would be showing muscle hypertrophy in the very first clinical study, and the funding needed to go from IND-enabling work through phase one and into a runway for the next round would be roughly 20 to 30 million dollars, comparable to a Series A round. A big company like Novartis can run many such projects in parallel and never has to return to public markets for another 25 million dollars, but internal resources are also stretched, and paradoxically a small company with only one or two shots sometimes has more capital per project than a large company juggling hundreds.
Animal testing and drug attrition 1:16:01
Before a drug reaches humans, it goes through rodent and non-rodent toxicology and DMPK studies, which check whether the drug enters the body, reaches the right place, and behaves as expected, using blood tests, imaging, and autopsies to catch microscopic organ changes. Every drug program carries unknowns, and it's a common industry joke that successful programs have nearly been killed several times before reaching humans. For antibody biologics like bagramab, attrition from initial candidates to actually entering humans is relatively low, with maybe 30 percent making it through, mainly because antibodies are highly specific and rarely cause the off-target toxicity seen with small molecules.
IND filing and GMP manufacturing 1:18:00
Patents are typically filed once a narrowed group of drug candidates has been selected, timed to last as long as possible while still protecting the compound. Filing an IND, the US regulatory permission to test a drug in people, requires proving the manufacturing meets GMP, good manufacturing process, standards, meaning rigorous documentation that what's on the label is truly what's in the bottle, covering purity, activity, and sterility. This is why buying gray-market peptides like retatrutide or BPC-157 online is risky, since there's no such guarantee, and BPC-157 in particular has no known receptor, isn't encoded in the human genome, and rests on data from a single unreproduced investigator.
Toxicology findings and dosing strategy 1:25:00
In rat studies, bagramab caused cardiac hypertrophy, but once heart size was normalized to the animals' dramatically larger bodies, it appeared proportionate, and muscle and heart size both shrank after the drug was withdrawn during a recovery period. Toxicology studies aim to dose animals weekly at levels producing higher blood exposure, both peak and trough, than expected in humans, building in a safety margin, though in some oncology drugs toxic and therapeutic doses overlap and the toxicity is accepted given the medical need.
Risk threshold for healthy volunteers 1:30:00
The guest explains his personal standard for testing drugs in healthy people: he does not want to expose a volunteer to a risk of serious harm greater than the roughly one in one hundred thousand chance of being struck by lightning in a year. If a drug carries a bigger risk of toxicity, testing moves instead to patients who could actually benefit from the therapy, so the risk is balanced against a potential upside. This is not an FDA rule but a personal and industry norm, laid out plainly in consent forms, and he notes that across roughly thirty years in the field, something seriously bad has happened to healthy volunteers only about once every decade.
The TGN1412 lesson 1:34:02
He recalls the case of a therapeutic antibody targeting CD28, an activating receptor on T cells, which looked safe in animal testing but triggered massive T cell activation and a life-threatening cytokine release when given to six healthy volunteers at once, some of whom died. That disaster, more than twenty years ago, is why trials now use sentinel dosing, giving the drug to one person at a time before expanding a cohort, rather than dosing several volunteers simultaneously. He also mentions a separate small molecule case at a company called Bial, and personal stories from medical school, including radial arterial lines, sleep studies, and inhaled radioactive microspheres, as examples of how far volunteers will go for modest payment.
BEMA trial design and dosing 1:37:01
For the BEMA program, healthy volunteers meant older adults without diagnosable disease, since the drug's target population was older people, and researchers measured muscle mass by MRI and DEXA scans along with blood markers like CK, aldolase, and LDH. Three on-target side effects showed up: muscle cramps, acne, which was rare in older volunteers but more common in younger ones tested later, and first-dose diarrhea. Dosing aimed as high as possible without exceeding animal-tested exposure levels, topping out around fifty to one hundred milligrams per kilogram, informed by earlier cell culture experiments on muscle hypertrophy. The conversation then turns to choosing trial locations, weighing recruitment, investigator quality, regulatory environment, and cost across countries like Germany, the US, Australia, New Zealand, Taiwan, and increasingly China, and clarifies that running a trial under FDA oversight requires it be conducted in the US, while data from trials approved elsewhere can still support a later US application.
Ethnic sensitivity studies 1:47:31
Regulators sometimes require a small phase one trial in a specific population before allowing a larger study in that country. Japan has a formal legal definition of who counts as Japanese, and companies satisfy this requirement by running the study either in Japan itself or in places like Hawaii or California where enough Japanese participants live. China has a similar requirement, driven by documented cases of drug toxicity specific to Han Chinese ethnicity, though its definition of who qualifies is looser, so it is usually simplest to just run the study in China.
Bimagrumab's mixed human results 1:53:00
Novartis ran roughly sixteen phase one and two studies of bimagrumab across different conditions. The drug reliably increased muscle mass in humans, but only by four to eight percent, far below the twenty to thirty percent seen in rodents, and it did not meaningfully improve strength or performance. A nutrition sub-study found that muscle gain scaled with protein intake within the tested range, and cutting protein below normal levels caused muscle loss that the drug still prevented, showing the biology was real even if performance gains were not. A planned nutritional supplement to pair with the drug fell through after Novartis sold its nutrition division to Nestle. A later meta-analysis of sarcopenia trials found the muscle gain translated into only a nine-meter improvement in six-minute walk distance, an effect too small to be clinically meaningful.
From Novartis spinoff to obesity pivot 1:57:30
Novartis's final study tested bimagrumab in type 2 diabetics at a maximal dose of ten milligrams per kilogram monthly for twelve doses, and found meaningful fat loss alongside a seven to eight percent absolute drop in hemoglobin A1C, achieved without any change in diet or exercise. Despite this, Novartis decided the effect size did not justify continued investment and licensed the asset out in 2021 to Versanis Bio, a company built specifically around it, with plans to target obese older adults with low muscle mass. Raising money was difficult since obesity drug development had a long history of commercial failure, and the company managed to raise seventy million dollars after pitching fifty-three investors, short of its hundred million goal. Everything changed when Novo Nordisk's semaglutide data emerged, making incretin drugs the new standard of care for obesity. The company quickly tested bimagrumab alongside semaglutide, tirzepatide, and liraglutide in mice, finding unprecedented additive benefits for fat loss and preservation of lean mass, which led the board to bring in a more experienced CEO and expand the program into what became the Believe, Become, and Behold studies.
Naming drugs and the tide suffix 2:06:00
Generic drug names follow international nomenclature rules, where the suffix is fixed by drug class and the company can propose a prefix or infix, which is why the antibody Bimagrumab ends in mab for monoclonal antibody and draws its prefix from Bhima, the Indian god said to be as strong as ten thousand elephants. GLP1 drugs similarly share a class suffix, tide, seen in exenatide, liraglutide, semaglutide, and tirzepatide.
The Believe trial design 2:10:30
The Believe study began as a planned 24-week trial in sarcopenic obesity but was expanded once Novo Nordisk's obesity data emerged, turning into a nine-arm, full factorial trial combining low and high doses of bimagrumab with low and high doses of semaglutide plus placebo, run partly during the pandemic with supply chain problems and an open-label semaglutide arm because a matching placebo injector could not be made. The study grew to 507 patients over 104 weeks, with 48 weeks as the primary endpoint, and required raising additional funding beyond the original 70 million dollars.
Results, side effects, and what comes next 2:13:30
In the high-dose combination arm, patients lost 22 to 23 percent of starting body weight and 45.7 percent of starting body fat, a fat loss comparable to bariatric surgery, though strength gains measured by grip strength were modest. One unexpected finding was a roughly 20 percent rise in LDL cholesterol, described as an on-target effect since activin receptors are present in the liver, though the exact mechanism remains unstudied. Eli Lilly later acquired the company and the drug, and combination studies with tirzepatide continue, while the guest expects future obesity treatment to follow a pattern of injectable induction therapy followed by oral maintenance drugs. The conversation closes with reflections on mTOR inhibition, where selective TORC1 inhibition without suppressing TORC2 is seen as the key unsolved challenge for developing safer, more effective longevity-oriented drugs.
Questioning intermittent fasting in older age 2:24:31
The discussion turns to a finding that young rodents downregulate a growth pathway called mTor in response to fasting, but old rodents do not show the same response. This raises doubts about whether intermittent fasting works the same way in older people as it does in the young, though testing this directly would require liver biopsies, which are invasive and not something people readily volunteer for, and current imaging tools cannot substitute for that data.
Shifting toward real prevention in medicine 2:25:31
Looking ahead a decade, the biggest source of optimism is a shift from a sick care system toward genuine preventive medicine. Making that shift requires stronger primary care and, practically, billing codes for preventive visits, since right now a doctor cannot bill for seeing a patient specifically to prevent cancer. Despite these institutional hurdles, people seem to be waking up to wanting to stay healthy rather than wait to get sick and treated.
A new drug aimed at preventing cancer 2:26:30
The idea behind the new company is that drugs known to cause cancer as a side effect are likely blocking a cancer protective pathway, so switching that pathway back on gently might prevent cancer. The model drug is sorafenib, a kinase inhibitor used for kidney and liver cancer, which causes skin cancers in about ten percent of older patients by blocking a pathway called ribotoxic stress, the same pathway targeted by potent toxins like diphtheria toxin, sarin, and ricin. The approach is to activate this pathway mildly and reversibly, with the hope of preventing perhaps half of cancers, starting with skin cancer since it is nearly as common as all other cancers combined. A planned phase two trial would enroll adults fifty and older who have had at least five prior skin cancers, a group with a fifty percent chance of another within a year, testing low dose, high dose, and placebo to measure actual cancer prevention.
Melanoma sensitivity and future testing 2:29:31
A parallel effort with the Broad Institute ran the compound through a panel of a thousand cancer cell lines to see which tumors respond, and melanoma emerged as by far the most sensitive, for reasons not fully explained by mutational burden alone. This suggests the drug could work as a treatment, not just a prevention, for melanoma, and prevention itself has precedent, since an Australian study showed intensive sunscreen use reduced melanoma compared with usual practice. Any broader claim that the mechanism prevents cancers beyond skin will be hard to test before approval because cancer incidence is rare, though larger phase three trials could eventually explore it. The conversation closes by noting other companies have seen this data at conferences and are now developing similar pathway inhibitors, while Eli Lilly's own related drug work is still ongoing, apparently being tested alongside newer GLP-1 drugs.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
