[@PeterAttiaMD] 409 ‒ Inside modern drug development: the science, economics, and regulatory hurdles

Link: https://youtu.be/GoNjdK-I0Bc

Duration: 155 min

Transcript: Download plain text

Short Summary

An interview on the Drive Podcast with Lloyd, an MD-PhD trained at Brigham and Women's Hospital who practiced rheumatology before 20+ years at Novartis Institutes, traces drug-development case studies including the failed fall-detection trial, the myostatin/activin-pathway antibody bimagrumab, antibody engineering with MorphoSys, toxicology, healthy-volunteer risk, and a new cancer-prevention venture targeting the ribotoxic stress pathway.

Key Quotes

1. "And there are a lot of serious diseases in medicine that are worse than cancer in terms of clinical outcomes, but we don't treat them the same way and and I think we should." (00:09:05)
2. "Essentially, we need if the if the lian myostatin and actin for the receptor is sticky, you need an inhibitor that's even stickier." (00:50:59)
3. "So we use our best clinical judgment to say that I don't want to expose people to a risk greater than that of a lightning strike in a year." (01:50:02)
4. "So if a drug causes cancer, it must be most likely inhibiting a cancer protective pathway. Most drugs are inhibitors of things." (04:32:46)

Detailed Summary

Guest Background

Lloyd is an MD-PhD trained at Brigham and Women's Hospital in Boston who practiced rheumatology for about 10 years while running an NIH-funded adhesion-molecule lab. He left academia ~20 years ago to join Novartis Institutes, founded by Mark Fishman under CEO Dan Vasella, where Fishman tasked him with founding the New Indications Discovery Unit.

• After Novartis out-licensed bimagrumab, Lloyd became founding CEO of Versanis Bio with Elon Zipkin, raising $70M from Atlas Venture (Michael Gladstone) and Medicxi (Nick Williams, Vonnie Marie) before being acquired by Eli Lilly.
• Lloyd is now starting a new cancer-prevention company targeting the ribotoxic stress pathway, beginning with melanoma and skin cancer prevention.
• He has executed both regulatory strategies for muscle drugs: pursuing large indications (sarcopenia) and using orphan diseases like Duchenne as faster regulatory on-ramps.

Translating Drug Discovery Into Industry

Novartis Institutes brought "translational medicine" into industry; the term was coined at Oxford in the UK, and Novartis was its first industrial manifestation. Lloyd frames drug discovery as patient- and indication-first, contrasting incremental improvements (e.g., rosuvastatin vs. lovastatin) with quantum-step new therapies.

• Drug-development infrastructure is heavily biased toward incremental improvements because they are lower risk; entirely new indications often require building regulatory pathways and physician/patient/payer awareness from scratch.
• Lloyd frames successful drug development as patient-centered, with the unmet need driving modality, target, and clinical-trial design rather than the other way around.

The New Indications Discovery Unit

Mark Fishman tasked Lloyd with founding the unit and gave it a budget to find the most needed medicines Novartis was not pursuing. The unit compiled ~7,000 unmet clinical indications, grouped them into roughly seven or eight therapeutic buckets (healthy aging, ENT, renal, fibrotic), and grew to dozens of projects before being trimmed.

• Lloyd cited a frail-elderly three-year mortality approaching 90%—worse than most cancers—as a motivating data point that seeded the muscle-disease effort.
• The unit's broad portfolio approach allowed systematic triage of indications by unmet need before prioritization.

Modalities, Patents, and Formulation Constraints

Modalities split into small molecules (historically from dye companies ~100 years ago), biologics (antibodies, peptides, soluble receptors), gene therapies, and devices. Pharma patents cover composition of matter, formulation, salts, auto-injector, indication, dose, route, and manufacturing method, with companies staggering filings to extend effective life (Humira as the textbook example).

• Disease constraints dictate format: childhood diseases need palatable oral formulations (grape/cherry), and some lung conditions require inhaled drugs rather than oral or injectable options.
• Manufacturing is split across multiple parties (substance, formulation, packaging, distribution), accounting for a large share of branded-drug cost, and GMP mandates high documentation standards on purity, activity, contamination, and sterility.

Sarcopenia and the Failed Fall-Detection Study

Sarcopenia is being defined as decreased muscle mass with impaired function (grip, gait, stair climb). Lloyd ran a triaxial-accelerometer fall-detection study with geriatrician Lou Lipitz and clinical-ops lead Kieran Dole that enrolled 60 high-risk subjects (each with ≥1 fall in prior 6 months) over 6 months.

• The study recorded 117 fall events; the pendant detected only ~17% of real falls with a ~17% false actuation rate, leading the team to abandon the fall-prevention program.
• The trial predated modern AI analytics; a separate MIT WiFi-sensing approach (Dina Katabi) also failed for fall assessment, underscoring the difficulty of the problem.

Myostatin Biology and Bimagrumab Development

Myostatin (Se-Jin Lee) is an endogenous muscle-growth inhibitor whose knockout produces "Arnold Schwarzenegger" mice (~1997 images). Bimagrumab (BYM338) targets ActRII because, in humans, myostatin is not the whole story—activins also signal through ActRII—and receptors signal via SMADs, suppressing MuRF-1/MAFbx/Atrogin-1 (David Glass's discovery).

• Follistatin is an endogenous pathway inhibitor but would cost ~$1M/month systemically; in mice, antibody-treated muscle hypertrophy exceeded pure myostatin knockouts (~30% gain).
• In humans, bimagrumab produced only ~4–8% muscle gain (max ~8%) versus 20–30%+ in rodents, and without proportional strength gains—a pattern also seen with IGF-1 agonists, androgen agonists, and SARMs.
• The team engaged WADA early; WADA has screened for bimagrumab for 10+ years given its obvious performance-enhancing potential.

Antibody Engineering

Antibodies were made via phage display with MorphoSys, screening thousands of candidates and narrowing them through developability; Fc-fusion technology (Brian Seed, Mass General) is illustrated by Enbrel/etanercept. High-affinity binders were required because myostatin/activins bind receptors at low-nanomolar to high-picomolar affinity—hard to beat with small molecules.

• A firefly-luciferase reporter assay screened candidates; fully human antibodies are preferred over humanized antibodies, and cross-reactivity with at least one of two toxicology species is required.
• Bimagrumab was immunogenic in mice, so researchers built a murinized construct (CDDH866) for mouse work.
• Legacy antibodies still in use include mouse OKT3 (anti-CD3) for transplant rejection and rabbit anti-thymocyte globulin; most antivenoms remain horse serum, with repeat-use serum sickness as a clinical limit.

Toxicology and Program Checkpoints

Toxicology priorities are monitorability and reversibility; irreversible cardiac/neurologic toxicity is program-ending, and idiosyncratic liver toxicity (Lamictal example, transaminases) is the most common reason drugs are pulled. Rodent studies of bimagrumab showed cardiac hypertrophy that tracked body-size gain (heart-to-body size remained normal) and resolved on washout.

• Bimagrumab was dosed weekly with peak/trough exposures exceeding projected human levels.
• Large pharmas have 2–4 major governance checkpoints; small companies raise capital at the equivalent stages. Peter Attia estimated ~$20–$30M (today) for IND-enabling + Phase 1 + runway to next round for a bimagrumab-style startup.
• Biologics attrition from first candidate to IND is roughly 5-to-1 to 6-to-1 (about 30% reach humans), lower than for small molecules because antibodies usually lack off-target toxicity (a CEP-inhibitor example raised BP off-target).
• IND (US) vs. CTA (Europe) is regulatory permission to dose humans; patents are typically filed once a candidate slate is identified.

Manufacturing, GMP, and Peptide Risks

GMP mandates high standards and extensive documentation on purity, activity, contamination, and sterility; manufacturing splits across parties account for a large share of branded-drug cost. Lloyd likens buying peptides from unverified sources to buying street opioids, warning about fentanyl/carfentanil contamination.

• BPC-157 is called the "poster child" of the wellness-industry grift: not encoded in the human genome, no known receptor, all data from one investigator, unreproduced.
• Lloyd argues the regulatory gray zone around wellness peptides creates direct consumer-safety risks that most clinicians underestimate.

Phase 1 and the "Lightning-Strike" Framework

Bob Schmutter's rule: test in healthy volunteers only when risk is no greater than the ~1 in 100,000 annual US lightning-strike probability; otherwise, test in patients who may benefit. Over ~30 years, Lloyd estimates a serious adverse event in healthy volunteers roughly once per decade.

• TGN1412 (>20 years ago) caused acute cytokine release with deaths, prompting the sentinel-patient model—no more dosing six subjects simultaneously.
• Lloyd and Peter Attia both describe being serial paid research subjects as broke medical students—Attia donating plasma via ~4-hour lymphoplasmapheresis at NIH/NCI for ~$200.

BEMA / Bimekizumab First-in-Human Study

Phase 1 of bimekizumab/BEMA was run by clinicians Dan Rrook and Ronan Rubinoff in healthy older adult volunteers without diagnosable disease or confounding medications. Novartis (Switzerland) ran the study in the US; muscle mass was assessed by DEXA and MRI with strength testing, plus soluble muscle proteins CK, aldolase, and LDH.

• Three on-target adverse effects were observed: muscle spasms/cramps, acne (more common in younger subjects), and first-dose-related reproducible GI diarrhea symptoms.
• Highest dose tested was approximately 50–100 mg/kg; the study dosed about three times after the single-dose portion, following the principle of dose-escalating as high as possible without exceeding animal-tested exposures (including contaminants).
• Target exposures were informed by cell culture experiments showing muscle cell hypertrophy.

First-in-Human Study Logistics & Geography

The speaker's three biggest challenges of any clinical study are recruitment, recruitment, and recruitment—alongside qualified investigators and a supportive regulatory environment. Common countries for FIH studies today: Germany, US, Australia, New Zealand, with China rising due to a favorable regulatory environment allowing investigator-initiated studies with less supporting data than a typical IND.

• Australia uses a clinical trial notification (not approval) process: ethics committees assess safety, regulators assess drug quality.
• If the FDA oversees a study it must be in the US; to run a US study after a foreign one, sponsors submit an IND with all foreign data.
• Japan formally requires an ethnic-sensitivity Phase 1 in Japanese subjects (defined narrowly, allowing studies in Hawaii/California); China generally requires one too and is easiest to conduct in-country.

Bimagrumab Clinical Development at Novartis

Bimagrumab originated in Chris Lo's lab, moved to David Glass's department, and was taken into the clinic by Lloyd. Novartis ran ~16 Phase 2 studies across indications after acquiring the asset.

• Bimagrumab reliably increased muscle mass by 4–8% in humans versus 20–30%+ in rodents, but without proportional strength gains—a pattern also seen with IGF-1 agonists, androgen agonists, and SARMs.
• A Novartis sarcopenia meta-analysis found a 4–8% muscle increase yielded only a 9-meter improvement in six-minute walk distance; other assessments included timed up-and-go and the Short Physical Performance Battery.
• A protein-intake study tested half RDA, RDA, and 1.5x RDA (1.2 g/kg = 2x RDA); higher intake within that range produced more muscle, while at half RDA, bimagrumab prevented expected muscle loss—suggesting substrate-limited rather than drug-limited effects.
• Final Novartis study in type 2 diabetics: 10 mg/kg monthly × 12 doses over 48 weeks (maximal effect-size dose), increasing muscle mass, substantially decreasing fat mass, and lowering HbA1c by ~7–8% absolute.
• Novartis out-licensed/spun out bimagrumab after judging effect size insufficient commercially; a planned nutritional co-development ended when Novartis sold its nutrition unit to Nestle.

Versanis Bio and the BELIEVE Trial

Versanis Bio (founded ~February 2021) licensed bimagrumab from Novartis and raised $70M from ~53 of ~100 investors targeted, with Atlas Venture and Medicxi as key backers, focusing on sarcopenic obesity. After Novo Nordisk's late-2021 semaglutide obesity readout, Versanis repositioned bimagrumab as a GLP-1/incretin add-on (Jay Bradner: "skating to where the puck is going to be"); mouse studies showed additive efficacy with semaglutide, tirzepatide, and liraglutide.

• BELIEVE enrolled ~500 people (specifically 507); originally 6 months of bimagrumab alone, redesigned to a combination and extended to 72 weeks treatment, 48-week primary endpoint, plus 6-month follow-up = 104 weeks total.
• The trial cost ~$70M; Series B closed in two tranches before acquisition.
• Nine arms: four combination arms (low/high bimagrumab × low/high semaglutide), placebo, and individual drug arms, with dose responses across all arms.
• Primary endpoint was body weight (regulatory standard); DEXA collected in every patient but too expensive as primary. Lloyd advocates waist circumference over BMI/body weight via a white paper.
• High-dose combination at 72 weeks: 22–23% body-weight loss and 45.7% loss of starting body fat—described as first medical therapy producing fat loss equivalent or superior to bariatric surgery.
• Functional tests: timed up-and-go, SPPB, 30-second chair stand, grip strength (selected for BELIEVE as most clinically linked).
• An unexpected adverse outcome was ~20% LDL elevation, attributed to on-target activin receptor activity in the liver possibly affecting LDL clearance/synthesis.
• Glucose did not rise and decreased independent of semaglutide; results published in Nature Medicine, with six-month off-drug data due at EASD in September.
• Eli Lilly (not Novo Nordisk) acquired Versanis and now owns bimagrumab; Lilly paused one program but continues a combination study with tirzepatide on ClinicalTrials.gov.

New Direction: Cancer Prevention via Ribotoxic Stress

BELIEVE deliberately enrolled metabolic syndrome (pre-diabetic) patients to enable diabetic endpoints; Lloyd argues the LDL rise is monitorable/treatable and would not have blocked Phase 3 if the program had remained standalone. Predicted paradigm: induction with combination/injectable therapies to move patients from obese to non-obese, then maintenance with an oral agent (e.g., oral GLP-1 agonist).

• mTOR inhibition extends lifespan from yeast to dogs, likely primates; expected modestly geroprotective in humans. Selectivity is the barrier: rapalogs are mTORC1-selective but downregulate mTORC2 with sustained exposure; Restore Bio combined catalytic + allosteric mTOR inhibitors to manage this.
• In young rodents, fasting downregulates mTOR as expected; in old rodents, fasting did not—questioning whether intermittent fasting produces the same biology in older people.
• The guest's new company focuses on cancer prevention; there are no medical billing codes for preventive cancer visits, creating institutional hurdles.
• Cardiovascular disease is the #1 cause of US adult morbidity/mortality; the gap is primary prevention using existing drugs rather than new CV drugs.
• Sorafenib (multikinase inhibitor for renal cell and hepatocellular carcinoma) causes skin cancers in ~10% of older patients, mostly basal cell and squamous cell carcinomas.
• Sorafenib inhibits a sensing kinase triggering ribotoxic stress—a cell-death pathway targeted by potent toxins (diphtheria toxin, ricin, sarin).
• Hypothesis: gently activating ribotoxic stress could prevent ~50% of cancers, starting with skin cancer (nearly as common as all other cancers combined).

Planned Cancer Prevention Study

Phase 2 would enroll 100–120 adults aged 50+ with ≥5 prior skin cancers (~50% chance of another within a year). Broad Institute collaboration tested the tool compound against 1,000 cancer cell lines; melanomas were by far the most sensitive.

• Skin has the highest mutational burden of any normal organ (UV exposure), though mutational burden alone didn't fully explain melanoma sensitivity.
• An Australian study showed intensive sunscreen use prevents melanoma, supporting testability of pharmacological melanoma prevention.
• The Swedish Obesity Study (bariatric surgery cohort followed for decades) shows treating obesity prevents multiple cancers—motivating the obesity-prevention paradigm.
• Many other companies are now developing pathway inhibitors inspired by this data, expanding beyond Eli Lilly.

Key Numbers and Surprises

• 7,000 unmet indications compiled by the New Indications Discovery Unit; ~3-year mortality ~90% in frail elderly.
• Fall-detection study: 60 subjects, 117 falls, 17% sensitivity, 17% false-positive rate on the pendant.
• Bimagrumab muscle gain: ~30% in mice vs. 4–8% in humans.
• Tufts CSDD: ~$2–4B and ~10 years per approved drug today, vs. older ~$1B/10-year estimate.
• $20–$30M to get a bimagrumab-like program to Phase 2a readout (Peter Attia estimate).
• ~30 years, ~1 serious healthy-volunteer SAE per decade (Lloyd's personal observation).
• 1 in 100,000 lightning-strike risk as the Schmutter threshold for healthy-volunteer studies.
• BEMA Phase 1 highest dose: 50–100 mg/kg; 3 doses after single-dose portion.
• Novartis final T2D study: 10 mg/kg monthly × 12 doses over 48 weeks; ~7–8% absolute HbA1c reduction.
• Versanis: $70M raised from ~53 of ~100 investors; BELIEVE enrolled ~500 people at ~$70M cost.
• BELIEVE high-dose 72-week combination: 22–23% body-weight loss and 45.7% fat-mass loss; ~20% LDL elevation.
• ~10% skin-cancer rate in older sorafenib patients; ~50% projected cancer-prevention effect; ~100–120 subjects in planned Phase 2.