Eric Topol on the Creative Destruction of Medicine
What this covers
Eric Topol of Scripps Research Institute, author of The Creative Destruction of Medicine, joins Russ Roberts to examine how genomic science and digital technology can remake American medicine. The conversation moves between the failures of current practice and the technical possibility of individualized (precision) medicine Medicine). Topol argues that medicine has been catastrophically simplified by treating all patients within a diagnostic category as statistical averages, resulting in unnecessary screenings, over-treatment, and medication that helps very few people while harming many. Roberts raises structural questions about incentives, expertise, and the regulatory capture that locks in incumbent advantage. The two map out how cheap whole genome sequencing, known drug-gene interactions, wearable sensors, and social networks could shift the system from prevention through mass screening to predictive prevention through individual biology.
The discussion ranges across specific harms: PSA and mammography screening that injure far more people than they help, statins prescribed to fix lab numbers rather than to prevent heart attacks, cancer drugs chosen by tumor location rather than genetic driver mutation, and carbamazepine prescribed without genotyping despite a known fatal reaction that genotyping would predict. Topol shows how eminence-based medicine—expert consensus rather than robust evidence—dominates practice, and proposes that diseases like diabetes are not two types but over twenty when examined at the molecular level. Roberts argues the pharmaceutical industry and medical education are insulated from creative destruction by high barriers to entry, allowing outdated curricula and the failing blockbuster-drug model to persist. Topol counters that patient activism and the convergence of cheap sequencing, predictive analytics, and connectivity create an inflection point that no incumbent can suppress indefinitely.
Topol argues that medicine has been catastrophically 'dumbed down' by treating every patient as the statistical average, and that the convergence of cheap genome sequencing, biosensors, and digital connectivity now enables a shift to individualized, prevention-oriented care that will replace the failing mass-medicine and blockbuster-drug model.
- Population-level screening and uniform dosing produce net harm and waste because individuals differ genetically
- Cheap sequencing and known drug-gene interactions already exist but are not implemented in daily care
- Digital 'superconvergence' (sequencing, sensors, predictive analytics, social activism) makes individualized prevention feasible
Genomic sequencing and testing are now affordable and clinically actionable but remain culturally unimplemented in practice.
- The cost of sequencing an individual genome has dropped more than a millionfold, to about $4000 today, and is expected to fall below $2000 and toward $1000 per whole genome by year end, though understanding an individual properly requires sequencing a trio (the person plus siblings or parents), bringing analyzed costs to roughly $12-15,000.
“It's reduced more than a millionth-fold. The cost of sequencing for an individual would be about $4000 today.”
- Carbamazepine (Tegretol) carries a roughly 1-in-1000 chance of a potentially fatal Stevens-Johnson syndrome reaction that can be accurately predicted by genotyping, yet the US does not screen for it; in Taiwan a prescription cannot be obtained without a prior genotype, demonstrating that actionable genomic information exists but is not implemented in US practice.
“in Taiwan, for example, you can't get a prescription for the drug unless you have a genotype first”
- While Topol would not recommend genome sequencing for healthy people, for patients with serious undiagnosed illness sequencing should be done far earlier than traditional 'diagnostic odysseys'; a 16-year-old girl with a severe neurologic condition who had visited ten medical centers over 16 years at a cost of millions was diagnosed via sequencing that pinpointed the two causative gene mutations.
“this was a 16-year old girl with a pretty severe neurologic abnormality, condition, we were able to find exactly the two gene mutations that were causative”
- A $99 23andMe saliva scan of common genome variants, while limited, reveals roughly 25-30 major drug interactions with strong supporting data—such as personalized Warfarin dosing where one person needs only 2mg versus the average 5-7mg—making it a bargain even though it captures only hundreds of thousands of variants rather than the full 6-billion-letter genome.
“it gets you something like 25-30 major drug interactions about you. So, just that alone is a bargain in my view.”
- Once you know which variant to look for, the actual genotype test can be performed for less than a dollar by hospitals running their own certified CLIA labs, so cost is not the real barrier to drug-genomic screening; the barrier is the absence of a cultural mindset embracing genomics.
“The actual genotype, once you know where to look, in fact it could be done for less than a dollar, the actual test. It's so inexpensive.”
Current medicine treats everyone identically despite individual biological differences requiring personalized approaches.
- Because the tools to define each individual did not exist until now, medicine defaulted to treating everybody the same—same screening, same drug, same dose for the same condition—even though many drugs don't work in many people, resulting in waste and imprecise use of treatments.
“because we didn't have the tools until now to define each individual, the default mode has been: Treat everybody the same”
- Treating everyone as average is wrong because no two people are the same—not even identical twins—so with sequencing, sensors, and other tools now available it is time to move toward individualized rather than population-based medicine.
“because we didn't have the tools until now to define each individual, the default mode has been: Treat everybody the same”
- Prescribing the same diet and exercise to everyone is misguided because individuals differ in nutritional needs and suitable exercise types; some people tolerate carrying a little extra weight while others suffer many issues from it, and disease-specific prevention (like sun protection for melanoma risk) should be geared to each individual's risk, starting in childhood when most chronic illnesses take root.
“I don't think everybody has to do the same type of exercise and eat the same food.”
- The notion that there are only two types of diabetes is inane; viewed at the molecular level—which genes and pathways are involved—there are more than 20 types, and this pathway-based understanding plus supercomputers, predictive analytics, and machine learning will let us break down complex diseases despite the difficulty of teasing apart causes in complex systems.
“This inane idea that there's two types of diabetes, when there's more than 20 types. When you look at it from a molecular basis--which genes, which pathways.”
Patient activism and consumer access to validated self-care tools can overcome entrenched medical system resistance to change.
- Early, unvetted medical technologies can produce dangerous false alarms—an alpha-version handheld ultrasound showed Topol a large mitral valve leak that was actually a software bug, and an early Navigenics report stated a nonsensical 108% heart-attack risk—so consumers should wait until devices and genomic tools are fully tested and the bugs are out before relying on them.
“I had this really big leak in my mitral valve... they had a software problem.”
- The future of self-care is consumer access to validated diagnostic data: a person could confirm an infection with a dipstick or smartphone-attached imaging (e.g. a tenfold-resolution image of a child's eardrum sent to an algorithm), receive a yes/no answer, and—once validated—be authorized to obtain the appropriate antibiotic directly, democratizing care and putting patients in charge.
“You could then have an algorithm that goes right to your phone that tells you what to take, and that, once it's validated, should give you the leeway to be in charge.”
- Sensitivity to toxins and radiation varies enormously between individuals—some people can undergo frequent CT, PET and nuclear scans without ever getting cancer while others are exquisitely sensitive, and the same applies to mercury from seafood—so individual toxicity thresholds will eventually be defined, though it is not currently a high research priority.
“Some people, you could send them for scans, CT scans, PET scans, nuclear scans--they could go every week and they'll never get a cancer. Whereas other people are exquisitely sensitive.”
- We have more identification chips for our pets than for our people.
“We've got more chips for our pets than we have for our people.”
- Physicians should be less promiscuous with antibiotics because overuse drives antibiotic resistance and, as recently learned, drastically disrupts our microbiomes, which carries its own disease risks beyond the intrinsic DNA effects.
“it is incumbent on physicians to be less promiscuous with antibiotics because that's what led to all sorts of problems of resistance. We've now learned in recent years how it's had drastic influence on our microbiomes”
- The Vitamin D deficiency scare is terribly overhyped; recent studies have undercut its importance, the optimal level for an individual is not really known, and most people are not Vitamin D deficient—it is one item on a long list of things that need to be individualized rather than treated as a population-wide fad.
“Oh, it's been terribly overhyped... most people are not Vitamin D deficient.”
- The most exhilarating prospect in medicine is moving from treating heart attacks to predicting them days or weeks in advance and gearing up so they never happen, and more broadly being able to recognize each human being, digitize what makes them tick, and render specific individualized treatments and preventions—transitioning out of today's economically and risk-imbalanced mess.
“the dream of being able to prevent one, the sense that it was coming days or weeks before and gearing up so it doesn't happen, that to me as a singular project is the most exciting”
- Because entrenched medical systems are hard to change, the path forward is patient activism: 'the power of the people is greater than the people in power'—just as pre-social-network HIV activism forced definitive treatments through Congress, social-networking-era activism is exponentially more powerful and could rally patients to make deeply seated medical systems more malleable and progressive.
“the power of the people is greater than the people in power”
- Given drug costs reaching hundreds of thousands or even a million dollars per treatment per year (e.g. cystic fibrosis drug Kalydeco at $294,000/year for life for patients with a specific mutation), drugs should be sold on a 'guarantee to succeed' model—if the drug doesn't work, the patient doesn't pay—mirroring consumer returns of products that don't perform.
“I want to see a guarantee to succeed model... It doesn't work, then I'm not paying for it.”
Evidence-based medicine is often eminence-based opinion lacking adequate underlying evidence for universal prescriptions.
- Much of what is called 'evidence-based medicine' is really 'eminence-based medicine'—a group of experts sitting in a room forming opinions that all patients with a given diagnosis should receive a certain treatment—and this approach does not work because the underlying evidence is scant.
“Our evidence-based medicine is--it's a nice buzz word, but a lot of it is eminence-based medicine, where a bunch of people sit around in a room, these experts, and they just make an opinion about what everyone should do.”
- For people without heart disease who have high cholesterol, taking a statin lowers the LDL blood test reliably but only prevents a heart attack or death in 1 or 2 people out of 100, meaning 98 of 100 are medicated solely to fix the lab number; the often-cited '33% reduction in heart attacks' merely reflects going from 3 to 2 per hundred, which is weak rather than overwhelming evidence.
“in terms of reducing heart attacks or preventing deaths, we only benefit 1 or 2 people out of 100”
- Universal recommendations that all men get a PSA or all women get a mammogram induce net harm at the population level: roughly 250,000 US men per year have a false-positive PSA leading to repeated painful biopsies and emotional hardship without cancer, and mammography data show an overall lack of benefit; these harms outweigh the few people detected early, so blanket screening should be replaced by individualized risk-based strategies.
“there are 250,000 men each year in the United States alone that have a false positive PSA. And they undergo serial biopsy”
Medical technology should target specific mutations or patient characteristics rather than blockbuster drugs for everyone.
- Cancer drugs are typically prescribed based on the organ where the disease appears (prostate, lung) rather than the tumor's causative mutation, which works poorly and is wasteful; instead, sequencing the tumor relatively inexpensively to find the driver mutation allows genome-guided treatment.
“you have a disease of a particular organ, like prostate or lung, and you get a drug based on that. Well, that of course doesn't work very well”
- The pharmaceutical industry is imperiled by a lack of innovation and failure to embrace individualized medicine; the blockbuster model of mass medicine (drugs meant for everyone, 'put in the water supply') is over for good, and the industry will transition toward using genomics and sensors to develop drugs effective for most patients rather than the 2% the old drugs help.
“the whole blockbuster model of mass medicine, those days are over for good.”
- Drug development woes came from trying to make one drug for everyone; developing a drug for a specific mutation gives a clear biologic basis, can be done precisely and quickly, fast-tracked through the FDA with trivial patient numbers needed to show overwhelming efficacy and a highly favorable benefit-risk ratio—and a drug targeting a rare mutation may end up helping far more people than just those with that mutation, making the 'long tail' an ideal pharmaceutical model.
“when you are developing a drug for a specific mutation, that's a whole different look. Now you've got the biologic basis and it can be done precisely and quickly.”
Unnecessary screening and procedures cause harm by triggering cascades of intervention in asymptomatic or low-risk people.
- The stethoscope (invented 1816) lets clinicians listen only indirectly to the heart and abdomen, whereas modern imaging lets you see everything directly; mastering heart sounds—which Topol spent countless hours teaching—is now obsolete 'dinosaur stuff', though reimbursement structures and the medical profession's resistance to change have slowed the shift.
“we listen so indirectly to lub-dub for the heart... Whereas now you can see everything. That is such a dramatic difference.”
- An asymptomatic person who gets an unnecessary calcium-score CT scan showing coronary calcium can be pulled onto an 'express train' of angiograms and multiple stents despite having no symptoms, illustrating how unnecessary procedures and the inability to differentiate individuals drive cascades of intervention.
“that just led to this express train ride that went through getting an angiogram and then having all these stents put in to multiple arteries. And this fellow had no symptoms.”
Digital convergence of sequencing, sensors, and connectivity now makes individualized medicine technically feasible at scale.
- An unprecedented 'superconvergence' of digital infrastructure (broadband, connectivity, social networks), multi-layer biological sequencing (DNA, RNA, proteins, metabolites, epigenome of individuals and their pathogens), and real-time physiologic sensors creates a unique inflection point that will transform medicine, even though this infrastructure has barely been harnessed in medicine so far.
“this term I use about the 'superconvergence' the likes of which we've never had in our history, this sets off this unique time, this true great inflection of medicine”
- Wearable sensors are insufficient for true prevention; instead, nanosensors in the blood that detect genomic signals—free cancer DNA before a tumor forms, circulating cancer cells, immune activation signatures, or a cell shedding from an artery lining that precedes a heart attack—and relay them to a smartphone could enable prevention of heart attacks, cancer, and asthma attacks before any symptoms appear.
“we actually are using nanosensors that can pick up a genomic signal. So whether that be from a free cancer DNA, before a cancer has really taken form”