The precision medicine issue

Vol. 121 Nov/Dec $9.99 USD No. 6 2018 $10.99 CAD

$2 million would save her life. The precision medicine issue Could you pay?

Should you? Medicine is becoming hyper-personalized, hyper-accurate ... and hyper-unequal. p.38

The AIs taking over Curing cancer How to plan your from doctors with customized digital afterlife vaccines p.22 p.46 p.76

ND18_cover.indd 1 10/3/18 4:16 PM Artificial intelligence is here. Own what happens next.

ND18 ETD19 Spread 16x10.875 D3.indd 2 9/28/18 10:59 AM March 25–26, 2019 St. “egis Hotel San Francisco, CA

Meet the people leading the next wave of intelligent technologies.

Andrew Kathy Daphne Sergey Harry Anagnost Gong Koller Levine Shum President and Cofounder and Founder and Assistant Executive Vice CEO, Autodesk CEO, WafaGames CEO, Insitro Professor, EECS, President, UC Berkeley Microsoft Artifi cial Intelligence and Research Group

EmTechDigital.com/2019

ND18 ETD19 Spread 16x10.875 D3.indd 3 9/28/18 10:59 AM 02 From the editor

It’s been nearly two decades since the help meet the ballooning healthcare first human genome was sequenced. needs of an aging population. That achievement opened up the prom The problem? As medicine gets ise of drugs and treatments perfectly more personalized, it risks getting more tailored to each person’s DNA. But per unequal. Our cover story is Regalado’s sonalized or “precision” medicine has gripping and troubling account ƒpage always felt a bit like flying cars, sexbots, 38 of the parents who raise millions of or labgrown meat­one of those things dollars to finance genetherapy cures for we’re perpetually being promised but their children’s ultrarare diseases. Are never quite getting. This issue of MIT they trailblazers for a technology that Technology Review makes the case that, will one day provide cheap, custom in fact, the age of precision medicine ized care to everyone, or harbingers of a has been slowly dawning on us all this future in which only the superwealthy time­and we’re unprepared. and crowdfunding whizzes are saved? What’s changing fastest now is the IVF combined with genetic screening sheer volume of medical data avail can weed fatal diseases out of a fam able, and the tools for analyzing it. As ily for good, but, Laura Hercher argues Antonio Regalado points out in his ƒpage 68 , it could also lead to two Gideon Lichfield is editor in chief of MIT Technology Review. opening essay ƒpage 8 , the number of genetically distinct human castes­one people getting their DNA tested is now rich and diseasefree, the other poor in the tens of millions and doubling and diseaseridden. The rich and well each year. educated won’t only be better able to By pairing DNA data with people’s afford boutique treatments; they’ll be medical records, algorithms can predict more likely to have the technology, and your risk of certain common diseases hence the data, that helps them avoid and suggest drugs and diets to ward falling ill in the first place. them off, as Ali Torkamani and Erik All this has more than merely med Topol explain ƒpage 20 . Cancer drugs ical consequences. Nathaniel Comfort are now being customized to individual warns ƒpage 16 that our growing ability patients, as Adam Piore reports ƒpage to find genetic correlations with things 46 . Epigenetic data can forecast how like intelligence is threatening to long you’ll live, writes Karen Weintraub revive the ugly dogma of eugenics. And ƒpage 80 , and new “senolytic” drugs what, asks Mary Madden ƒpage 34 , can might keep agerelated ailments at bay any of us do to keep tabs on how the for more of that time, reports Stephen S. oceans of data about us are being used, Hall ƒpage 84 . or misused? Not just DNA sequences but data We’ll face this question even in death. of all kinds is being scooped up and As Courtney Humphries reports ƒpage crunched in vastly greater quantities 72 , people now in their 30s will have than before. As Rachel Metz explains generated enough data by the time they ƒpage 56 , it’s becoming possible to track die to power quite convincing digital mental illness just by monitoring how avatars of themselves. So who will own you tap, type, and swipe on your phone. you when you’re gone? At least Simson Better treatments and healthier liv Garfinkel ƒpage 76 has some advice on ing aren’t the only benefits. Doctors like how to prepare. Rahul Parikh ƒpage 28 hope to be able This issue of the magazine, therefore, to spend more time getting to know spans the entire arc of human existence, their patients as algorithms take on the from before you’re born until after more routine tasks. In a UK trial, AI sys you die. Through it all runs a simple tems are already replacing physicians question. We know that in health care, for simple consultations, as Douglas human beings are unequal. But just how

Heaven reports ƒpage 22 . That could unequal are we willing to be? NIVEN NEPHI

ND18_front_editorial_letter.indd 2 10/3/18 2:43 PM The future of work is here. How will you take the lead?

Technology isn’t replacing people. It’s augmenting what people can accomplish. Is your organization prepared for what’s next?

Let’s get to work.

www.deloitte.com/us/futureofwork

Copyright © 2018 Deloitte Development LLC. All rights reserved.

Untitled-2 1 10/4/18 4:20 PM 04 The precision medicine issue

1 2 3 Health Illness Death

Introduction 16 38 72

8 Opening a door to eugenics A cure for one Never let me go Precision Social genomics threatens What are the ethics of developing You’re definitely temporary, but a medicine’s to revive genetic racism and treatments only a tiny number of digitally enhanced version of you long journey entrench discrimination. people will needor can pay for? doesn’t have to be. Skeptics say By Nathaniel Comfort By Antonio Regalado By Courtney Humphries drugs based on genetic 20 46 76 insights have Your genome, on demand One tumor at a time Six things to do with your data gone nowhere. A detailed genetic profile can Personalized cancer vaccines before you die But look closer predict your risk for all kinds of are an amazing breakthrough. Here’s how to make sure your and they’re diseases. But they might not be a sustain
loved ones can get into your everywhere. By Ali Torkamani and Eric Topol able business. By Adam Piore accounts after you’re gone. Or, By alternativelyhow to cover your Antonio 22 52 tracks. By Simson Garfinkel Regalado Dr. Bot will see you now evolution? What revolution? A chatbot might help you A skeptic pokes holes in the 80 avoid seeing your overworked façade of genomic drugs. Want to know when you’re The back page physician. By Douglas Heaven By Stephen S. Hall going to die?

88 Your life span is written into your Genes 28 56 DNA. We’re getting closer to pre
I wish AI won’t replace your doctor The mental-health cure inside dicting how much time you have they would Machines can crunch numbers, your phone left. By Karen Weintraub fi n d letting your doctor focus on you. How our smartphone obsession By By Rahul Parikh could help treat brain disorders. 84 Sarah Cooper By Rachel Metz Finally, the drug that keeps 30 you young Diversity for genomes 62 We can make mice youthful again. Our genetic data comes mostly Profi les in precision medicine Are people next? Anti
aging pio
from people of European ances
Advances in genetics have cre
neer Judith Campisi thinks so. try. That’s terrible for everybody. ated dilemmas that would have By Stephen S. Hall By David Rotman been unthinkable a few years ago. These people’s stories tell 34 the tale. Your doctor’s data experiment Getting medical help now often 68 comes with a condition: give us Unequal by design your data first. By Mary Madden What if only the poor inherited genetic disease? Sadly, that’s where

we’re headed. By Laura Hercher IMAGES GETTY BY IMAGE COVER

ND18_front_contents.indd 4 10/3/18 1:54 PM WORLD CLASS EXECUTIVE EDUCATION AT BROWN UNIVERSITY

Transforming Mid-Career Professionals An Executive Master’s degree from Brown University will prepare you to lead your organization, transform your eld, and build a powerful lifelong professional network.

You will join a vibrant learning community and apply your new knowledge and skills through a critical challenge project.

e results can be seen in our alumni who are proven leaders; impacting their organizations and the world.

IE Brown Executive MBA Executive Master in Cybersecurity Executive Master in Science and Technology Leadership brown.edu/professional Executive Master of Healthcare Leadership

Brown-34343 Print Ad_81875x105.indd 1 3/29/18 9:24 AM

BROWN.indd 1 5/31/18 5:38 PM 06 Masthead

Editorial Corporate Advertising sales Board of directors

Editor in chief Chief executive ocer and publisher Vice president, sales and brand Martin A. Schmidt Gideon Lichfield Elizabeth Bramson-Boudreau partnerships Whitney Espich Marii Sebahar Jerome I. Friedman Executive editor Director, human resources [email protected] Joichi Ito Megan McCarthy Hilary Siegel 415-416-9140 Israel Ruiz Editor at large Assistant to the CEO Senior director, brand partnerships David Schmittlein David Rotman Katie McLean Kristin Ingram Alan Spoon [email protected] Deputy editor Manager of information technology 415-509-1910 Michael Reilly Colby Wheeler Director, brand partnerships Customer service and News editor Oce manager Debbie Hanley subscription inquiries Niall Firth Linda Cardinal [email protected] Managing editor 214-282-2727 National Timothy Maher 800-877-5230 Licensing and Director, brand partnerships San Francisco bureau chief communities Ian Keller International Martin Giles [email protected] 903-636-1115 Vice president, licensing 203-858-3396 Senior editor, MIT News E-mail and communities Alice Dragoon Digital sales strategy manager customer_service@ Antoinette Matthews Ken Collina mittechnologyreview.info Senior editor, AI and robotics Client services manager [email protected] Will Knight Web Ted Hu 617-475-8004 www.technologyreview.com/ Senior editor, cyborgs Advertising services customerservice Rachel Metz Events [email protected] MIT Records (alums only) Senior editor, biomedicine 617-475-8004 617-253-8270 Antonio Regalado Senior vice president, Media kit Reprints events and strategic partnerships Senior editor, energy www.technologyreview.com/media [email protected] Amy Lammers James Temple 877-652-5295 Director of events programming Senior editor, infrastructure Licensing and permissions Laura Janes Wilson MIT Technology eview Elizabeth Woyke Insights [email protected] Reporters Senior events manager Karen Hao (AI), Charlotte Jee (news), Nicole Silva Vice president of international Mike Orcutt (blockchain), Erin Winick Content and program developer, events business development, head of (future of work) Kelsie Pallanck MIT Technology Review Insights Nicola Crepaldi Copy chief Events associate Linda Lowenthal Bo Richardson Senior editor MIT Technology eview Senior production director Mindy Blodgett One Main Street, 13th Floor James LaBelle Senior project manager Cambridge, MA 02142 Finance Anna Raborn Tel: 617-475-8000 Finance director Content manager Design Enejda Xheblati Jason Sparapani The mission of MIT Technology Review Chief creative ocer General ledger manager Director of consulting, Asia is to bring about better-informed Eric Mongeon Olivia Male Claire Beatty and more conscious decisions about technology through authoritative, Lead designer Accountant influential, and trustworthy journalism. Emily Luong Letitia Trecartin MIT Enterprise Forum,

Marketing and events designer Administrative assistant Inc. Technology Review, Inc., is an Kyle Thomas Hemingway Andrea Siegel independent nonprofit 501(c)(3) Art assistant Chairman and president corporation wholly owned by MIT; Emily Caulfield Consumer marketing Elizabeth Bramson-Boudreau the views expressed in our publications and at our events are not always shared Executive director and clerk Senior vice president, by the Institute. Product development Antoinette Matthews marketing and consumer revenue Director of product Doreen Adger Treasurer Enejda Xheblati Vanessa DeCollibus Director of analytics Project manager Tom Russell Director of chapter leadership and process Allison Chase Director of audience development Gaylee Duncan User interface/ Rosemary Kelly user experience designer Director of digital marketing Jon Akland and communications Engineers Josh Getman Shaun Calhoun, Molly Frey, Assistant consumer marketing manager Jason Lewicki Sanjeet Chowdhury Assistant product marketing manager Amanda Saeli

ND18_front_masthead.indd 6 10/2/18 12:27 PM Be a part of MIT Technology eview. Join our Global Panel.

You have opinions on today’s tech trends. So share them.

The MIT Technology eview Global Panel is a growing group of executives and innovators. As a panelist, you’ll be asked to take part in surveys, weighing in on a range of technology topics. And you’ll be among the first to see the results.

Apply at technologyreview.com/globalpanel

ND18 Insights Global Panel Ad 8x10.875 D4.indd 1 10/4/18 10:01 AM 08 Introduction

Skeptics say drugs based on genetic But look carefully and they’re by Antonio Regalado insights have underdelivered. everywhere. Look how far precision medicine has come

ometime this fall, the number doubling, roughly, every year since 2010. company GlaxoSmithKline to develop of people who have spit in a The figures are now growing by a million personalized drugs, starting with treat tube and sent their DNA to the each month, and the DNA repositories ments for Parkinson’s disease. The notion largest consumer DNA testing are so big that they’re enabling surprising is that targeted medicines could help the companies, like Ancestry and new applications. Consumers are receiv small subset of Parkinson’s patients with S23andMe, is likely to top 20 million. The list ing scientific predictions about whether a particular gene error, which 23andMe by now is certain to include some of your they’ll go bald or get cancer. Investigators can easily find in its database. classmates and neighbors. If you are just this year started using consumer DNA data Ever since the Human Genome tuning in, this figure will seem huge. And to capture criminals. Vast gene hunts are Project­the 13year, $3 billion effort you might wonder: how did we get here? under way into the causes of insomnia to decipher the human genetic code­ The answer is little by little. The number and intelligence. And 23andMe made a researchers and doctors have been pre of people getting DNA reports has been $300 million deal this summer with drug dicting the arrival of “precision medicine.”

ND18_intro.indd 8 10/4/18 11:30 AM The precision medicine issue 09



151



 132





 5

36

106

81

Genetic Rx The number of “personalized” drugs on the US market The growing tally of personalized or targeted has multiplied during the past decade medicines consists of those drugs whose label includes information about how genetic makeup can affect a person's response to a drug. Personalized Medicine Coalition

ND18_intro.indd 9 10/4/18 11:30 AM 10 Introduction

It’s a term with no agreedupon definition, these treatments come from different A time line although it suggests most strongly just the corners of biology, it’s what they have in of precision kinds of medicines that Glaxo and 23andMe common that’s important: each benefits medicine are pursuing: more targeted and more from detailed understanding of genetic Your genes and mine effective because they take into account information and tools to control it. differ. Can science create a person’s particular genetic makeup. To our thinking, these drugs display real drugs to match? President Bill Clinton, at the unveiling precision. The hep C pill, called Solvadi, of the genome’s first draft back in June consists of a chemical that is irresistible 2000, said the data would “revolutionize to the replicating virus, but when the drug the diagnosis, prevention, and treatment comes in contact with the virus’s genome, of most, if not all, human diseases.” replication quickly grinds to a halt. The Almost two decades after those big treatment for spinal muscular atrophy, promises, it is in vogue to question why meanwhile, is a genetic replacement part. precision medicine has not delivered more. With gene therapy, doctors can add fresh

1998 A report in the New York Times this sum DNA instructions to the child’s nerve TARGETED DRUGS mer, noting that deaths from cancer still cells. The dozen or so kids who’ve gotten The breast cancer treatment outnumber cures by a wide margin, asked: the therapy at a young age don’t develop Herceptin is approved for sale “Are We Being Misled About Precision the disease. in the US. It is the first cancer drug to target an underlying Medicine?” One reason for this seemingly All this traces back to even before the genetic defect responsible for slow progress is that not all precision med Human Genome Project. Think instead of producing tumors. icine involves drugs. As gene hunts gain the foundational act of the biotechnology in scope­the latest involve comparisons industry, 40 years ago. On September 6, of more than a million people’s DNA and 1978, Genentech announced “the success health records­an inconvenient fact about ful laboratory production of human insu many common diseases has emerged: they lin.” Before then, diabetics had injected don’t, by and large, have singular causes. insulin from pigs. It took around two tons Instead, many hundreds of genes play of pig parts to extract eight ounces ƒ227 small roles, and there is no obvious point grams of pure insulin. But Genentech 1999 at which to intervene with a pill. had found a way to splice the human ver PERSONALIZED MEDICINE So instead of drugs, we are seeing a new sion of the insulinproducing gene into The Wall Street Journal declares a “New Era of Personalized predictive science in which genetic risk E. coli bacteria, which then manufactured Medicine” based on genetic profiles may say which people should lower the hormone. Genentech still keeps the mapping of one-letter DNA differences between humans. Drug their blood pressure, which should steel 40yearold press release online. makers call it the start of a themselves for Alzheimer’s, and which To the pharmaceutical houses of the “grand experiment.” cancer patients aren’t going to benefit from 20th century, with their roots in commer chemotherapy and can skip the ordeal. To cial dye making and synthetic chemistry, be sure, these sorts of prognostics aren’t these new biotech drugs looked at first widely accepted, and it’s hard to get peo like a sideshow. They were hard to make ple to change their behavior. Yet for many and inconvenient to take ƒby injection, people, these predictions may begin to offer mostly . The pharma giants could easily a concrete route to precision health and believe their way of doing things would increased knowledge of their own biology. always dominate. Until well into the 1990s, 2000 Look beyond cancer, and some defin a single drug company, Merck, was more HUMAN GENOME PROJECT itive cures have arrived. As with those valuable than all biotech companies com The Human Genome Project and Celera Genomics, a company growing millions sending in their DNA, it’s bined. It probably seemed as if biotech started by entrepreneur J. Craig easy to miss the change before it’s every would never arrive­until it did. Of the Venter, both announce that a where. Here are just two medications of 10 bestselling drugs in the US during working draft of the genome sequence is complete. note: a drug that mops up hepatitis C in 2017, seven ƒincluding the top seller, the 90% of those who take it and an experi arthritis drug Humira are biotech drugs IN NT UE mental gene therapy that is curing a rare, based on antibodies. Antibodies embody O D C fatal, and previously untreatable childhood biological precision too. These tiny blood

disease, spinal muscular atrophy. Though proteins, normally part of our immune VENTER CORPORATION GENOMICS CELERA OF COURTESY TUMOR; RESEARCH CANCER FOR CENTER NCI OF COURTESY

ND18_intro.indd 10 10/4/18 1:58 PM The precision medicine issue 11

The number of patients who must take a best-selling drug for one of them to benefit

Seeking better Disease Benefit rate drugs Cancer The proportion of patients who actually benefit from a best- selling drug in each category Arthritis

Depression

Schizophrenia

Alzheimer’s

Schork, NJ, Nature; PubMed

Genetic information explodes Cost of sequencing a genome Number of people who have bought consumer DNA tests

$100M 25M Š‹Œ,‘’“,”•‘ ‘–,”””,”””* 75 20

50 15

25 10 Š˜,˜‘˜ ““”,”””

2001 2009 2017 2012 2015 2018*

NHGRI Company reports, Leah Larkin, ISOGG * DATA ESTIMATED

Drugs based on DNA Percentage of drugs in development that may be tailored Number of the 10 best-selling drugs in the US that are to a person’s genetic profile biological molecules

2000 2010 2017

all cancer drugs –‘š •“š drugs 0 1 7

Tufts; Personalized Medicine Coalition MIT Technology Review

ND18_intro.indd 11 10/4/18 11:30 AM 12 Introduction

response, fit­like a key in a lock­onto ƒwhich created Herceptin now imagines other molecules, like those dotting the sur what it calls “cancer vaccines,” tailored face of a cancer cell. And just like insulin, not just to broad subtypes of people but to they’re often constructed using DNA code the unique signature of a person’s tumor. retrieved from our bodies. The new approach involves collecting 2013 Insulin and antibodies are meant to information about the peculiarities of A STEP BACK work the same way on everyone. But no a person’s cancer through highspeed The US bars consumer DNA testing company 23andMe from calculating two people’s genomes are exactly the genome sequencing; using software to people’s chances of common same­about 1% of the DNA letters dif analyze and predict what a custom bio diseases and cancer, calling results inaccurate. It’s a major fer between any two of us. Those differ logical drug would look like ƒthey will be blow for consumer genetics. ences can explain why one person is ill and reverse images of antibodies, known as another isn’t, or why one person’s version antigens, that stimulate the immune sys of diabetes is different from another’s. tem ; and then quickly manufacturing it. Drugs that take into account these dif No two of these vaccines would be alike. ferences in genetic information are called Also, note this: if and when the US Food “targeted” drugs. and Drug Administration approves these The cancer drug Herceptin, an antibody vaccines, it won’t be greenlighting a par that reached the market in 1998, was among ticular compound. Instead, it will approve 2016 the first. It was effective, but mostly in peo a computerized process for turning DNA AI DOCTOR ple whose newly diagnosed breast cancer information into drugs. A team at Google uses an was growing because of specific genetic Medicine as programmatic and predict artificial-intelligence method called deep learning to diagnose damage­about 20% of cases. It depended able as a computer? The idea has begun symptoms of blindness by reading on the genome of the tumor itself. Herceptin to exert a potent appeal in Silicon Valley, retina scans. It performs as well came to market with the admonition that, where some of tech’s biggest names now as an ophthalmologist. to get it, you should first have a test to see see biology as “just a code” they can crack. if you would benefit. According to the US Marc Andreessen ƒbest known for invent National Cancer Institute, there are now ing the web browser is one of them. The more than 80 such targeted medicines for venture fund he cofounded, Andreessen cancer on the market. Horowitz or a16z, has set aside a total of Critics argue rightly enough that such $650 million since 2015 to put into biotech medications still do too little for too few investments. As the firm’s blog states with people at too great a cost ƒoften $10,000 awe, “You don’t just read the code of biology 2017 a month . In fact, on the whole, those who but you can also write, or design, with it.” GENE THERAPY The US approves three gene survive cancer still owe little to targeted Welcome to biotech, a16z. Yet they’re therapies in a four-month span. drugs. “The single biggest determinant of on to something. Even 40 years after Two of them prime immune cells who survives cancer is who has insurance,” Genentech’s insulin press release, genetic to kill blood cancers. The other is a one-time treatment for an Greg Simon, who leads the Biden Cancer engineering is a marvel worth rediscov inherited form of blindness. It Initiative, has said­not whether there’s a ering. The ability to see, understand, and costs $850,000. drug to match their mutation. Some think manipulate human genes and the proteins we are spending too much time searching they make is the great advance that is still under the lamplight shed by genetic tools. unfolding in all its immense complexity “Perhaps we had been seduced by the tech four decades later. Biology isn’t anywhere nology of gene sequencing­by the sheer as neat as a computer program, but little wizardry of being able to look at a cancer’s by little, we’re learning how to control it. genetic core,” a Pulitzer Prizewinning To enzymes and antibodies we’ve added cancer doctor, Siddhartha Mukherjee, gene therapy and gene editing. We haven’t 2018 wrote this summer. sequenced one genome­we’ve sequenced RISK PREDICTION He’s right that the impulse toward pre a million. An astute observer might realize Giant gene studies lead to “risk score” technology based on cision medicine, cost be damned, springs we’ve already come a long way. measuring millions of sites in from new technology. It’s what it can do. a person’s genome. Doctors say Antonio Regalado is MIT Technology it can predict heart disease and And so you can be sure even more per Review’s senior editor covering

other conditions. sonalization is on the horizon. Genentech biomedicine. BIOBANK UK OF COURTESY RETINA; OPHTHALMOLOGY BMC THERAPEUTICS; SPARK OF COURTESY WOJCICKI; ISETT/FLICKR STUART

ND18_intro.indd 12 10/4/18 1:58 PM The precision medicine issue 13

Big questions need big data Studies are using DNA data from more people than ever

1,500,000

1.3M 1,400,000

1,300,000

1,200,000

1,100,000

1,000,000

900,000

800,000

People 700,000

500K 600,000

500,000

300K 400,000

300,000

200,000

94 96 3,426 100,000

0

2002 2005 2010 2013 2015 2017 2018 Japanese A gene hunt Consumer The FDA Why are A massive Research- A search for scientists reveals crit- test company cracks down some peo- trove of gene ers identify the genes use a new ical muta- 23andMe on consumer ple fatter data from genes linked behind approach— tions that contributes test compa- than others? the UK Bio- to educa- insomnia the genome- increase the user data to nies oŽer- Clues from a bank permits tional suc- is the larg- wide risk of mac- a search for ing genetic genetic study simultaneous cess. They est genetic association ular degen- Parkinson’s health pre- are quickly analysis of warn against study ever. It study—to eration, a genes. dictions from oŽered to 2,000 human using the relies heavily hunt for the common DNA, calling consumers traits and results as on the con- causes of cause of the results in the form diseases. a “DNA IQ sumer DNA heart attack. blindness. unreliable. of “DNA diet” test.” database of tests. 23andMe.

ND18_intro.indd 13 10/4/18 11:30 AM Shape the future of MIT Technology eview Join our Beta Panel Did you know we’re redesigning the MIT Technology Review website?

Be a part of the process by joining our beta tester panel. As one of the fi rst people to test-drive the new website design, your feedback will be a vital part of making our new site a better user experience for all.

Apply at technologyreview.com/beta-test

ND18 Beta Testing Ad 8x10.875 D4.indd 1 9/28/18 4:00 PM The precision medicine issue 15

Let’s talk about health

It’s all about one thing: data. Your data can be used to predict disease, improve your lifestyle, and make your doctor smarter and less overworked. But data can be misused if it’s in the wrong hands. It can be flawed if it doesn’t come from a diverse range of people. And it can be twisted to bolster theories of racial superiority.

ND18_front_health.indd1 15 10/3/18 4:48 PM 16 Health OPENING A DOO TO EUGENICS

Want to predict aggression? Neuroticism? Risk aver- New ways of using By your genetic data NATHANIEL sion? Authoritarianism? Academic achievement? could bolster COMFORT This is the latest promise from the burgeoning field scientific racism of sociogenomics. and encourage There have been many “DNA revolutions” since discrimination. the discovery of the double helix, and now we’re in the midst of another. A marriage of the social and nat ural sciences, it aims to use the big data of genome sciencedata that’s increasingly abundant thanks to genetic testing companies like 23andMeto describe

ND18_health_economist.indd 16 10/3/18 11:20 AM Opening a door to eugenics 17

the genetic underpinnings of the sorts of complex as directtoconsumer DNA companies like 23andMe behaviors that interest sociologists, economists, polit that collect biographical and biometric as well as ical scientists, and psychologists. The field is led by a genetic data on all their clients. group of mostly young, often charismatic scientists Algorithms then chew up the data and spit out who are willing to write popular books and opeds, correlations between the trait of interest and tiny vari and to give interviews and highprofile lectures. This ations in the DNA, called SNPs †for single nucleotide work shows that the naturenurture debate never polymorphisms‡. Finally, sociogenomicists do the thing OPENING diesit is just cloned and raised afresh in a new world. most scientists do at the outset: they draw inferences Advocates of sociogenomics envision a prospect and make predictions, primarily about an individual’s that not everyone will find entirely benevolent: health future behavior. “report cards,” based on your genome and handed out Sociogenomics is not concerned with causation at birth, that predict your risk of various diseases and in the sense that most of us think of it, but with cor propensity for different behaviors. In the new social relation. The DNA data often comes in the form of sciences, sociologists will examine the genetic com genomewide association studies †GWASs‡, a means A ponent of educational attainment and wealth, while of comparing genomes and linking variations of SNPs. economists will envision genetic “risk scores” for Sociogenomics algorithms ask: are there patterns of spending, saving, and investment behavior. SNPs that correlate with a trait, be it high intelligence Without strong regulation, these scores could be or homosexuality or a love of gambling? used in school and job applications and in calculat Yesalmost always. The number of possible com ing health insurance premiums. Your genome is the binations of SNPs is so large that finding associations ultimate preexisting condition. with any given trait is practically inevitable. DOO Such a world could be exciting or scary †or both‡. The evolutionary biologist Graham Coop shows But sociogenomicists generally focus on the sunny side. that big data can lull us into a false sense of objec And anyway, they say with a shrug, there’s nothing tivity. The success of GWASs, he writes, “seems to we can do about it. “The genie is out of the bottle,” suggest that we’ll soon be able to settle debates about writes the educational psychologist Robert Plomin, whether behavioral differences among populations “and cannot be stuffed back in again.” are driven in part by genetics.” However, he adds, TO Is this what the science says, in fact? And if it is, “answering this question is a lot more complicated is it a valid basis for social policy? Answering these than it seems.” questions demands setting this new form of hered Coop offers what he calls a “toy” example of a mis itarian social science in contextconsidering not leading polygenic studya thought experiment. The merely the science itself but the social and histor hypothetical research question: Why do the English ical perspective. Doing so can help us understand drink more tea than the French? what’s at stake and what the real risks and benefits Coop’s imaginary researcher, Bob, uses data from EUGENICS are likely to be. existing databases like the UK Biobank. He counts up the average number of alleles †different forms of a WEID SCIENCE gene‡ associated with a preference for tea in English people and French people. “If the British, overall,” If this is “the science,” the science is weird. We’re used Coop writes, “are more likely to have alleles that to thinking of science as incrementally seeking causal increase tea consumption than French people, then explanations for natural phenomena by testing a series Bob might say that we have demonstrated that the of hypotheses. Just as important, good science tries difference between French and UK people’s prefer as hard as it can to disprove the working hypotheses. ence for tea is in part genetic.” Sociogenomics has no experiments, no null hypoth Being a conscientious scientist, of course, Bob eses to accept or reject, no deductions from the data would offer the usual assurances about the quality to general principles. Nor is it a historical science, of his data. He would piously insist that his results like geology or evolutionary biology, that draws on a do not show that all Brits who drink lots of tea do so longrunning record for evidence. because of their genesonly that the overall differ Sociogenomics is inductive rather than deductive. ence between the populations is partly genetic. Data is collected first, without a prior hypothesis, from Coop then walks us through the problems with longitudinal studies like the Framingham Heart Study, this thinking. It ignores the crucial fact that alleles twin studies, and other sources of informationsuch may behave differently in different genomes and in

ND18_health_economist.indd 17 10/3/18 11:20 AM 18 Health

different environments: “The issue is that GWAS stud Karl Pearson, Galton’s main protégé –who invented ies do not point to specific alleles for tea preferences, the correlation coefficient, a workhorse statistic of only to alleles that happen to be associated with tea GWASs and hence of sociogenomics—, was a socialist preference in the current set of environments experi who believed in separating sex from love. The latter enced by people in the UK Biobank.” In other words, should be spread around liberally, the former tightly we can’t be sure that a different group of people with regulated to control who bred with whom„that is, the same genetic variations would be equally avid tea for eugenic ends. drinkers. And even if they were, we still wouldn’t know The point is that eugenics was not, as some claim, it was those genes that made them love tea. merely an unfortunate bit of specious science. It was Bob, then, commits two fallacies. First, he confuses central to the development of biological statistics. This correlation and causation. The study does not show entanglement runs down the history of hereditarian that the putative tea drinking alleles affect tea drink social science, and today’s sociogenomicists, like it ing„merely that they are associated with it. They are or not, are heir to it. predictive but not explanatory. The second fallacy is Early in the 20th century, a vicious new strain of one I learned on the first day of class in college biosta eugenics emerged in America, based on the new science tistics: statistical significance does not equal biological of Mendelian genetics. In the context of Progressive era significance. The number of people buying ice cream reformist zeal, belief in a strong government, and faith at the beach is correlated with the number of people in science to solve social problems, eugenics became who drown or get eaten by sharks at the beach. Sales the basis of coercive social policy and even law. After figures from beachside ice cream stands could indeed prominent eugenicists canvassed, lobbied, and testified be highly predictive of shark attacks. But only a fool on their behalf, laws were passed in dozens of states would bat that waffle cone from your hand and claim banning “miscegenation” or other “dysgenic” mar that he had saved you from a Great White. riage, calling for sexual sterilization of the unfit, and “Complex traits are just that„complex,” Coop throttling the stream of immigrants from what certain concludes. “Most traits are incredibly polygenic, politicians today might refer to as “shithole countries.” likely involving tens of thousands of loci ˆi.e., SNPs At the end of the 1960s, the educational psychol or genes‹. These loci will act via a vast number of ogist Arthur Jensen published an enormous article in pathways, mediated by interactions with many envi the Harvard Educational Review arguing that Negro ronmental and cultural factors.” children –the term of the day— were innately less intel ligent than white children. His policy action item: A LONG TADITION separate and unequal school tracks, so that African American children would not become frustrated Sociogenomics is the latest chapter in a tradition of by being over challenged with abstract reasoning. hereditarian social science dating back more than What became known as “Jensenism” has resurfaced 150 years. Each iteration has used new advances in every few years, in books such as Charles Murray science and unique cultural moments to press for a and Richard Herrnstein’s The Bell Curve –1994— specific social agenda. It has rarely gone well. and the journalist Nicholas Wade’s A Troublesome The originator of the statistical approach that Inheritance –2014—. sociogenomicists use was Francis Galton, a cousin Given the social and political climate of 2018, of Charles Darwin. Galton developed the concept today would seem a particularly inauspicious time to and method of linear regression„fitting the best undertake a new and potentially vastly more powerful line through a curve„in a study of human height. expression of genetic determinism. True, the research Like all the traits he studied, height varies contin papers, white papers, interviews, books, and news uously, following a bell curve distribution. Galton articles I’ve read on the various branches of socioge soon turned his attention to personality traits, such nomics suggest that most researchers want to move as “genius,” “talent,” and “character.” As he did so, past the racism and social stratification promoted by he became increasingly hereditarian. It was Galton earlier hereditarian social scientists. They downplay who gave us the idea of nature versus nurture. In his their results, insist upon avoiding bald genetic deter mind, despite the “sterling value of nurture,” nature minism, and remain inclusive in their language. But, was “by far the more important.” as in the past, fringe groups have latched onto socio Galton and his acolytes went on to invent modern genomic research as evidence for their hostile claims biostatistics„all with human improvement in mind. of white superiority and nationalism.

ND18_health_economist.indd 18 10/3/18 1:25 PM Opening a door to eugenics 19

Indeed, the human biodiversity crowd Given the social and and other socalled “race realists” love sociogenomics. American Renaissance, a political climate of 2018, publication run by the avowed white supremacist Jared Taylor, features articles today would seem a about the possibilities of sociogenomics, as does the HBD Bibliography, an aggre particularly inauspicious gator of hereditarian materials. Steve Sailer, a wellknown and prolific writer in white time to undertake a new supremacist and human biodiversity cir cles, writes extensively about sociogenom and potentially vastly more ics on “race realist” sites such as Unz Review and VDARE. powerful expression of To be clear: I am not saying that socio genomicists are racists. I am saying that genetic determinism. their work has serious social implications outside the lab, and that too few in the field are taking those problems seriously. Genetics has an abysmal record for solv ing social problems. In 1905, the French psychologist Simon Binet invented a quan titative measure of intelligencethe IQ testto identify children who needed extra help in certain areas. Within 20 years, SOCIAL ISKS Binet was horrified to discover that people were being sterilized for scoring too low, out of a misguided fear Social genomics comes with its own large set of social that people of subnormal intelligence were sowing risksand number one on the list is failing to grapple feeblemindedness genes like so much seed corn. sufficiently with those risks. In the 2012 paper that What steps can we take to prevent sociogenomics has become the de facto manifesto of genoeconomics from suffering the same fate? How do we ensure that †the use of genetic data to predict economic behavior‡, polygenic scores for educational attainment are used Daniel Benjamin and his coauthors dedicated two full to offer extra help tailored to those who need itand sections to “pitfalls.” Every one of them is method ensure that they don’t become tools of stratification? ological and statisticalfalse positives, studies with Here’s one way: when the evolutionary biologist too few participants, and so forth. Most could be fixed Coop and his student Jeremy Berg published a GWAS with more data and better statistics. paper on the genetics of human height, they took Some in the field readily acknowledge the skel the extraordinary step of writing a 1,500word blog etons in the closet. “Eugenics is not safely in the post about what could and could not be legitimately past,” wrote Kathryn Paige Harden, a developmen inferred from their paper. tal behavior geneticist at the University of Texas, in Why isn’t this more common? The field needs more a New York Times oped earlier this year. Harden people like Coopand fewer cheerleaders. It needs lamented the rise of the socalled human biodiver scientists who reckon with the social implications of

Sociogenomics sity movement †referring to it as “the eugenics of the their work, especially its potential for harmscien is the latest altright”‡, with its ties to white supremacy and its tists who take seriously the social critique of science, chapter in specious claims to scientific legitimacy. Members of who understand their work in both its scientific and a history of hereditarian this movement, she wrote, “enthusiastically tweet historical contexts. It is such people who stand the social science and blog about discoveries in molecular genetics best chance of using this potent knowledge produc dating back more than 150 years. that they mistakenly believe support the ideas that tively. For scientists studying human social genomics, inequality is genetically determined; that policies doing so is a moral responsibility. like a more generous welfare state are thus impotent; Nathaniel Comfort is a professor of the history and that genetics confirms a racialized hierarchy of of medicine at Johns Hopkins. Jon Phillips human worth.” contributed research for this article.

ND18_health_economist.indd 19 10/3/18 11:20 AM 20 Health

n early 2018, it was estimated that over 12 million people had had their DNA analyzed by a directto consumer genetic test. A few months later, that number had grown to 17 million. Meanwhile, geneticists and data scientists have been improving our ability to convert genetic data into useful insights­forecasting which Ipeople are at triple the average risk for heart attack, or identifying women who are at high risk for breast cancer even if they don’t have a family history or a BRCA gene mutation. Parallel advances have dra matically changed the way we search for and make sense of volumes of data, while smartphones continue their unrelenting march toward becoming the de facto portal through which we access data and make informed decisions. Taken together, these things will trans form the way we acquire and use personal genetic information. Instead of getting tests reactively, on a doctor’s orders, people will use the data proactively to help them make decisions about their own health. With a few exceptions, the genetic tests used today detect only uncommon forms of disease. The tests identify rare variants in a single gene that causes the disease. But most diseases aren’t caused by vari ants in a single gene. Often a hundred or more changes in genetic letters collectively indicate the risk of common diseases like heart attack, diabetes, or prostate can cer. Tests for these types of changes have Your genome, recently become possible, and they pro duce what is known as your “polygenic” risk score. Polygenic risk scores are derived on demand from the combination of these variants, inherited from your mother and father, and can point to a risk not manifest in either parent’s family history. We’ve learned from studies of many polygenic risk scores for different diseases that they provide insights we can’t get from traditional, known risk factors such as smoking or high cholesterol ƒin the case of heart attack . Your polygenic How your detailed genetic profile can predict your risk of diseases and score doesn’t represent an unavoidable improve your health. fate­many people who live into their 80s and 90s may harbor the risk for a disease By Ali Torkamani and Eric Topol / Illustration by Nico Ortega without ever actually getting it. Still, these

ND18_health_Topol.indd 20 10/2/18 3:29 PM Opinion 21

scores could change how we view certain Yet despite growing evidence that rather than needlessly exposing them diseases and help us understand our risk polygenic risk scores are important, until selves to a medical treatment that has its of contracting them. recently there was no service allowing own risks. People tend to overestimate Genetic tests for rare forms of disease people to determine their own scores, the likelihood of catastrophic events, so if caused by a single gene typically give a even if they had invested in their own per polygenic scores are to achieve their full simple yes or no result. Polygenic risk sonal directtoconsumer genetic profiling. impact on health outcomes and health scores, in contrast, are on a spectrum We’re attempting to remedy that through care spending, we’ll need to find a way to of probability from very low risk to very the development of MyGeneRank, a free effectively communicate those tradeoffs. high risk. Since they’re derived from com mobile app that estimates users’ polygenic And finally there are the privacy con binations of genome letter changes that risk for heart attack and stroke from their cerns. We need to maintain our current are common in the general population, own genetic data. It also allows them to protections against genetic discrimina they’re relevant to everybody. The ques participate in a clinical trial to measure the tion so that people can benefit from their tion is whether we’ll find a way to make influence of polygenic risk information on own genetic information without having proper use of the information we get from people’s behavior, as reported by them, to worry that insurance companies will them. Can they inform us about changes to our lifestyle, or point to medications we should take or a screening test we should A polygenic risk score might tell you that get, that might improve our chances of staying healthy? you’re at high risk for breast cancer and Statin drugs are a good case study for spur you to get more intensive screening. this. They’re widely used, even though 95% of the people taking them who haven’t had heart disease or stroke get and their heath data, captured by mobile get access to that information and use it no benefit aside from a nice cholesterol sensors linked to their smartphones. to raise their rates or deny coverage. lab test. We can use a polygenic risk score There are still some issues and con You can’t change your genetic risk. But to reduce unnecessary statin use, which troversies we need to deal with. Equal you can use lifestyle and medical interven not only is expensive but also carries access is one major concern­especially tions to offset that risk. We can accelerate health risks such as diabetes. We know given that the majority of genetic studies breast cancer screening for women with that if you are in the top 20% of polygenic have been performed in populations of a high risk for the disease, and help peo risk for heart attack, you’re more than European ancestry. For now, it appears ple with borderline risk of heart disease twice as likely to benefit from statins as that the more powerful the predictions to make decisions about whether to take people in the bottom 20%; these people become, the less accurate they become statins or not. If we deliver and track the can also benefit greatly from improving with other populations. response to polygenic risk information, we their lifestyle ƒstop smoking, exercise In addition, genetic risk information can collect realworld evidence on how to more, eat more vegetables . So knowing is likely to make some people feel anx optimize the use of that data to give safe your polygenic risk might cause you to ious or fatalistic ƒor might give others a and effective health advice. take statins but also make some lifestyle false sense of security . Previous studies In the near future your smartphone changes. ƒAnd a recent largescale study suggest that genetic risk information has might feature technologies that moni in Finland showed that people with high a minimal influence on these psychologi tor your physiological, genetic, environ heartrisk scores responded with lifestyle cal states, but many of those studies were mental, and behavioral characteristics. improvements at a much higher rate than done when the variations in risk you could And this information could be linked to those with low risk scores. get via polygenic factors were marginal. virtual medical coaches and AI systems And it’s not just about heart disease. As our ability to separate people into that can synthesize all that information A polygenic risk score might tell you that increasingly different classes of genetic and deliver you insights about your own you’re at high risk for breast cancer and risk gets better, these issues may become health, on demand. spur you to get more intensive screening more prominent. and avoid certain lifestyle risks. It might Another challenge will be to convince Ali Torkamani is director of genomic tell you that you’re at high risk for colon people to forgo or delay medical inter informatics at the Scripps Research Translational Institute. Eric Topol cancer, and therefore you should avoid ventions if they have a low risk of a cer is a cardiologist and the author eating red meat. It might tell you that tain condition. This will require them to of books including the upcoming Deep Medicine: How Artificial you’re at high risk for type 2 diabetes, and agree that they’re better off accepting a Intelligence Can Make Healthcare therefore you should watch your weight. very low risk of a catastrophic outcome Human Again.

ND18_health_Topol.indd 21 10/2/18 3:29 PM 22

Dr. Bot will see you now

ND18_health_Dr_Bot.indd 22 10/3/18 11:19 AM 23

An AI chatbot might help you avoid having to make an appointment illustrated with your by by overworked Douglas Nicole physician. Heaven Ginelli

ND18_health_Dr_Bot.indd 23 10/3/18 11:19 AM 24 Health

y stomach is kill condition as simple as Googling When in doubt, the apps ing me!” your symptoms, but with many will always recommend seek “I’m sorry to more benefits. Unlike selfdiag ing a second, human opinion. hear that,” says nosis online, these apps lead you But by placing themselves a female voice. through a clinicalgrade triage between us and medical pro M“Are you happy to answer a few processthey’ll tell you if your fessionals, they shift the front questions?” symptoms need urgent atten line of health care. When the And so the consultation tion or if you can treat yourself Babylon Health app started begins. Where’s the pain? with bed rest and ibuprofen giving advice on ways to How bad is it? Does it come instead. The tech is built on a selftreat, half the company’s and go? There’s some delibera grab bag of AI techniques: lan patients stopped asking for an tion before you get an opinion. guage processing to allow users appointment, realizing they “This sounds like dyspepsia to to describe their symptoms in didn’t need one. me. Dyspepsia is doctorspeak a casual way, expert systems to Babylon is not the only app for indigestion.” mine huge medical databases, of its kindothers include Doctorspeak, maybe, but machine learning to string Ada, Your.MD, and Dr. AI. it’s not a doctor speaking. together correlations between But Babylon is the front The female voice belongs to symptom and condition. runner because it’s been inte Babylon, part of a wave of new Babylon Health, a London grated with the UK’s National AI apps designed to relieve based digitalfirst healthcare Health Service †NHS‡, showing your doctor of needless paper provider, has a mission state how such tech could change work and office visitsand ment it likes to share in a big, the way health services are reduce the time you have to bold font: to put an accessible run and paid for. Last year wait for medical advice. If and affordable health service in Babylon started a trial with you’re feeling unwell, instead the hands of every person on a hospital trust in London in of calling a doctor, you use your earth. The best way to do this, which calls to the NHS’s non phone to chat with an AI. says the company’s founder, Ali emergency 111 advice line are

The idea is to make seek Parsa, is to stop people from 40,000 people in London handled partly by Babylon’s AI. ing advice about a medical needing to see a doctor. have used the Babylon app. Callers are asked if they want to wait for a human to pick up or download the Babylon powered “NHS Online: 111” app instead. Around 40,000 people have already opted for the app. Between late January and early October 2017, 40% of those who used the app were directed to selftreatment options rather than a doctor around three times the propor tion of people who spoke to a human operator. But both the AI and the humans staffing

ND18_health_Dr_Bot.indd 24 10/3/18 11:19 AM Dr. Bot will see you now 25

the phone line told the same On average, people in the proportion of people to seek UK see a doctor six times a When the app started emergency care †21%‡. year, twice as often as a decade Now Babylon has also ago. From 2011 to 2015, the giving advice on colaunched the UK’s first dig average GP clinic’s patient ital doctor’s practice, called GP list grew by 10% and its num ways to self-treat, at Hand. People in London can ber of contacts with patients register with the service as they †by phone or in person‡ grew half of patients would with their local doctor. by 15.4%, according to a sur But instead of waiting for an vey by the King’s Fund. In a stopped asking for an appointment slot and taking survey by the British Medical time off work to see a physician Association in 2016, 84% of appointment, realizing in person, patients can either general practitioners said they chat with the app or talk to a found their workload either they didn’t need one. GP at Hand doctor on a video “unmanageable” or “exces link. And in many cases the sive,” with “a direct impact on call isn’t needed. The human the quality” of care they gave doctor becomes your last resort their patients. rather than your first. In turn, people often have to GP at Hand has proved pop wait days to get a nonurgent ular; some 50,000 people reg consultation. Many show up istered in the first few months, at hospital emergency depart among them Matt Hancock, the ments instead, adding even UK health minister. Babylon more strain to the system. “We now wants to expand across have the perception that it’s the UK. The service is also older people who turn up ªat available in Rwanda, where the emergency room«,” says Lee 20% of the adult population has Dentith, CEO and founder of already signed up, according to the Now Healthcare Group, a Mobasher Butt, a doctor and a healthtech company based in member of Babylon’s found Manchester, UK. “But it’s not. ing team. And it’s setting up It’s the 18 to 35yearolds who services in Canada, with plans are unwilling to wait a week for prohibitive costs in the US Bhatti, a general practitioner in to do the same in the US, the an appointment.” or the lack of medical profes East London. “A nurse might Middle East, and China. Population and life expec sionals in Rwanda, “all health be better at dressing a wound, tancy will continue to grow. By systems around the world are and a pharmacist might be bet Your doctor 2040, it is estimated, the UK stretched,” says Butt. “There’s ter for advice about a repeat will have more than 70 million not enough clinical resources. prescription. Anything that is overloaded people, one in four of whom There’s not enough money.” helps unload a very overloaded For 70 years, the NHS has pro will be over 65. Most other rich Which is where companies system, allowing doctors to do vided free medical care to any countries are also getting older. like Babylon come in. A chat what they are best at, is always one who needs it, paid for by At the same time, the next bot can act as a gatekeeper to welcome.” UK taxpayers. But it is showing few decades will see more peo overworked doctors. Freeing signs of strain. Two genera ple living with longterm ill up even more of the doctor’s Sometimes tions ago there were 50 million nesses such as diabetes and time, the AI can also han  Britons, and their average life heart disease. And better treat dle paperwork and prescrip AI is just better expectancy was not much over ment for diseases like cancer tions, and even monitor care Bhatti remembers how upset 60 years. There are now 66 mil means millions more people at home. lots of doctors were when lion, and most can expect to live will be living with or recov A chatbot can also direct patients first started bringing into their 80s. That stretches ering from them. people to the right provider. in printouts from their own the resources of a system that Of course, the UK is not “A GP is not always the best web searches. “How dare they has never been flush with cash. alone. Whether because of person to see,” says Naureen try and diagnose themselves!

ND18_health_Dr_Bot.indd 25 10/3/18 11:19 AM 26 Health

video consultations with your study posted online in June and “ How do we make this existing doctor, and it also acts coauthored with researchers as an AI pharmacist. Users can at Imperial College London, a job that people want buy their drugs from the Now Stanford University, and the Healthcare Group’s drug Northeastern Medical Group, to do? I don’t think ... delivery service. It’s a kind of Babylon put its AI through a Amazon for medicines. version of the final exam of consulting from their “This is a service that the Royal College of General patients really want, that they Practitioners †RCGP‡, which kitchen is why people didn’t previously have, and that British GPs must pass in is now being provided to them order to practice unsuper get into medicine. through the NHS 365 days a vised. Babylon’s AI scored 81%, year, 24 hours a day, for free,” 9% higher than the average They come to meet Butt says of Babylon. “And the grade achieved by UK medi brilliant thing is it doesn’t cost cal students. patients.” the NHS a single penny more The RCGP was quick to to deliver that.” distance itself from Babylon’s Not only will the AI in hype, however. “The potential these apps get smarter; it will of technology to support doc get to know its users better. tors to deliver the best possible “We’re building in the ability patient care is fantastic, but for patients to manage their at the end of the day, comput health not only when they’re ers are computers, and GPs sick, but also when they’re not are highly trained medical sick,” says Butt. The apps will professionals: the two can’t become constant companions be compared and the former for millions of us, advising us may support but will never and coaxing us through every replace the latter,” said RCGP day health choices. vice chair Martin Marshall in a statement. “No app or algo Death by rithm will be able to do what a GP does.” Don’t think you can negate my time most diseases are diag chatbot? Others level far more six years at medical school with nosed, a £10 problem has Not everyone is happy about all serious charges, suggesting your one hour on the internet.” become a £1,000 one,” says this. For a start, there are safety that Babylon has focused on But she likes to see it from the Parsa. “We wait until we break concerns. Parsa compares what making its service acces  patients’ perspective: “Well, down before going to a doctor.” Babylon does with your medi sible and affordable at the don’t think you can negate my Catching a disease early slashes cal data to what Facebook does expense of patients’ safety. six years of living with this ill the cost of treating it. with your social activities One Twitter user with the ness with your onehour lecture These apps first hit the amassing information, build handle DrMurphy11 †he’s an at medical school.” market as private health ser ing links, drawing on what it NHS consultant who told me When a patient does meet vices. Now they are starting to knows about you to prompt he needs to remain anony  a doctor face to face, the AI integrate with national health some action. Suggesting you mous because of the corporate can still help by suggesting care providers and insurers. make a new friend won’t kill culture there‡ has coined the diagnoses and possible treat For example, Ada users can you if it’s a bad recommenda hashtag #DeathByChatbot. In ments. This is useful even when share their chatbot sessions tion, but the stakes are a lot videos showing interactions a doctor is highly skilled, says with their NHS doctor, and higher for a medical app. with the app, DrMurphy11 sug Butt, and it’s “really critical” in the company is now working According to Babylon, its gests that Babylon’s AI misses poorer countries with a short with a handful of GP practices chatbot can identify medical obvious diagnoses and fails age of competent doctors. to enable the chatbot to refer conditions as well as human to ask the right questions. AI can also help spot seri them to the doctor. Another doctors do, and give treat  “I have no concerns about ous conditions early. “By the app, Now Patient, provides ment advice that’s safer. In a health tech or AI in general,”

ND18_health_Dr_Bot.indd 26 10/3/18 11:19 AM Dr. Bot will see you now 27

he says. “No doctor wants to service, they might have gone regularly, a digital€first service make mistakes, and any system for something that was 100 per€ may not be the best place to that helps minimize the risk of cent safe, but that could mean be,” he says. harm from human error will be you send everyone to hospital, And Butt insists that they welcomed.” But he’s worried which is not what a real doctor exclude nobody. “The service is that companies are misleading or nurse would do.” available to everyone,” he says; doctors and the public with Another fear is that digital€ it just may not suit some peo€ marketing claims that vastly first services will create a two€ ple, such as those with severe oversell their current tech. tiered health€care system. For learning difficulties or visual Babylon has also met with example, GP at Hand advises impairments, who would strug€ criticism in Rwanda, where it people with serious medical gle with the app. runs the Babyl service, for not issues to think twice about taking local epidemiology into signing up to a practice that People still account. In an interview with offers mostly remote access the BBC, Rwanda’s minister of to doctors. That might seem come in handy health claimed that the Babyl prudent, but it has led to For Bhatti, having a local doc€ app included no questions about accusations that GP at Hand tor who knows you is a cru€ malaria, for example although is effectively cherry€picking cial part of the health system. Babylon disputes this . younger patients with less “Knowing your doctor saves Still, while Babylon may not complexŒand less expen€ lives,” she says. “Doctors will be as good as a real doctor and siveŒhealth€care needs. Since pick up things because there’s such apps are always careful British GP practices get per€ continuity.” She thinks this is to recommend you see a real patient funding from the NHS, just as much an issue for doc€ doctor when in doubt , playing cherry€picking would mean the tors as for patients. “How do it too safe would defeat the pur€ rest of the health€care system we make this a job people want pose. “We wanted to re€create is left to do more with less. to do?” she says. “I don’t think the same pragmatic approach For some GPs, this isn’t people working flexibly, con€ that a clinician takes,” says Butt. acceptable. “We take every€ sulting from their kitchen, is

“If we just had a group of non€ Digital-first services may body,” says Bhatti. But Oliver why people come to medicine. clinical people building the scare off sicker patients. Michelson, a spokesperson for They come to meet patients.” the NHS, accepts that GP at Not even Butt envisions Hand has to issue some form chatbots replacing human doc€ of caveatŒit can’t realistically tors entirely. “Care is not just welcome everyone. “They are about diagnosing or prescrib€ not denying people access but ing medicine,” he says. “It’s saying that if you’re going to about knowing your patient is need to come into your GP going to be able to cope with the chemotherapy you’re pro€ posing for them, knowing that their family will be able to offer them the support that they’re going to need for the next few months. Currently there is no software that’s going to be able to replace that.”

Douglas Heaven is a freelance writer based in London. His most recent story for MIT Technology Review was “Can you spot the cryptocrime in this picture?” in our May/June issue.

ND18_health_Dr_Bot.indd 27 10/3/18 2:56 PM 28 Health

A machine can collate environmental data, genetic AI can’t replace data, and patient history way better than I can.

doctors. By Rahul Parikh But it can make them better.

everal years ago Vinod doubt that for certain doctors, whose work Consider what AI could do for asthma, Khosla, the Silicon Valley is highly focused on diagnosis radiologists the most common chronic medical disease investor, wrote a provoc or pathologists, for example , that break in childhood. Six million American kids ative article titled “Do through may prove an existential threat. suffer from it. In 2013, they collectively We Need Doctors or A decade ago, for example, researchers missed 14 million days of school. The cost Algorithms?” Khosla showed that AI was as good as radiologists of medications, visits to the doctor and argued that doctors were at detecting breast cancer. emergency room, and hospitalizations no match for artificial intelligence. Doctors But for physicians like me in primary nears $60 billion a year. Sbanter with patients, gather a few symp care, managing 1,500 to 2,000 patients, I diagnose asthma via a rule of thumb toms, hunt around the body for clues, and AI presents an opportunity. I went to that’s been handed down over time: if send the patient off with a prescription. medical school to connect with people you’ve had three or more wheezing epi This sometimes accidentally, maybe leads and make a difference. Today I often feel sodes and the medicines for asthma help, to the correct treatment, but doctors are like an overpaid bookkeeper instead, tak you have the disease. Once it’s diagnosed, acting on only a fraction of the available ing in information and spitting it back to I ask the parents to remember as best information. An algorithm, he wrote, could patients, prescribing drugs and adjusting they can how often they administer med do better. doses, ordering tests. But AI in the exam icines to their child. I ask: What seems to I’m a pediatric and adolescent physician room opens up the chance to recapture trigger episodes? Is the child exposed to in the San Francisco Bay Area, where entre the art of medicine. It could let me get anyone who smokes at home? I can also preneurs like Khosla have been knocking to know my patients better, learn how a review their records to count how many on the doors of doctors for years with their disease uniquely affects them, and give visits to the emergency room they’ve had, pilot technologies and software and hard me time to coach them toward a better or the number of times they’ve refilled ware. I can say with some authority that outcome. their prescriptions. Khosla’s is the voice of a savvy outsider who knows what he knows which isn’t health care. Yes, AI could help us diagnose and treat It’s not that we don’t have the data; disease. It can collate and serve up broad swaths of data in a clear and concise way, it’s just that it’s messy. We spend a great cutting down on the imprecise judgments deal of our time trying to make sense of it. that doctors make because of the pressures and complexity of our practices. There’s no

ND18_health_End_of_Doctors.indd 28 10/1/18 8:26 AM First person 29

But even with the most accurate recall and cellular level. The genes, proteins, asthma related emergency room visits to by parents and patients, and the most enzymes, and other drivers of asthma are a DallasŸFort Worth hospital. They pulled accurate electronic records, it’s still just highly diverse, even if their environmental data from patient records, along with air retrospective knowledge. There’s no pro triggers overlap. A number of experts now pollution data from EPA sensors, Google active, predictive strategy. think of asthma in the same way they think searches, and tweets that used terms like It’s not that we don’t have the data; of cancer an umbrella term for a disease “wheezing,” or “asthma.” The Google and it’s just that it’s messy. Reams of data that varies according to the tumor’s loca Twitter data were tied to the user’s loca clog the physician’s in box. It comes in tion and cellular characteristics. Ian Adock tion data. many forms and from disparate direc of the National Heart & Lung Institute at If I had this kind of data I could say, tions: objective information such as lab Imperial College, London, studies the link “Alexa, tell me which asthma patients I results and vital signs, subjective con between asthma and the environment. He need to worry about today.” I could give a cerns that come in the form of phone and his team have been collecting biolog heads up to the affected families. And if I messages or e mails from patients. It’s all ical samples from asthma patients’ blood, also had some genetic data like Adock’s, I fragmented, and we spend a great deal urine, and lung tissue and organizing the could diagnose asthma before the patient of our time as physicians trying to make genetic and molecular markers he finds suffered three bouts of wheezing, by order sense of it. Technology companies and into subtypes of asthma. The hypothe ing blood tests and comparing the results fledging startups want to open the data sis is that with that kind of knowledge, against those molecular markers. spigot even further by letting their direct patients can be given the drug that works This kind of time saving intelligence to consumer devices phone, watch, best for them. frees me to spend more time with my blood pressure cuff, blood sugar meter AI might also help to manage asthma patients. One study showed that asth

The author recently came across this drawing by a seven- year-old depicting her idea of a visit to the doctor—she’s on the exam table, and the doctor’s facing the other way.

matic children only took or received their inhaled medications about half of the time. AI might allow me more time to person ally interact with those kids, and get bet ter results. Lots of questions lie ahead. Are patients willing to share more of their personal data with us? If the AI shows your care is better one way, but you or your doctor feel differ ently, will an insurance company accept it? What if the algorithm misses something or is applied incorrectly? Who is liable, the

URPHY MD  MD URPHY doctor or the machine’s maker? M .

G Not long ago, in the Journal of the American Medical Association, I saw a col HOMAS HOMAS T send continuous streams of numbers flares. For many patients, asthma gets orful picture drawn by a child in crayon. It directly to us. We struggle to keep up worse as air pollution levels rise, as hap portrayed her pediatrician, eyes glued to with it, and the rates of burnout among pened this past summer when brush fires the computer, while she sat on the exam , COPYRIGHT COPYRIGHT , doctors continue to rise. swept through Northern California. AI table, looking wide eyed. I hope that AI How can AI fix this? Let’s start with could let us take environmental infor will soon allow me to turn my attention AG, AGE  diagnosis. While the clinical manifesta mation and respond proactively. In 2015, back to that little girl.

tions of asthma are easy to spot, the dis researchers published a study show Rahul Parikh is a pediatrician in the

DRAWING BY BY DRAWING ease is much more complex at a molecular ing they could predict the number of San Francisco Bay area.

ND18_health_End_of_Doctors.indd 29 10/1/18 4:25 PM 30 Health

Carlos D. Bustamante’s hunt for genetic variations between Making genomic populations could correct health medicine relevant disparities and drive drug discovery.

By David Rotman for everyone Photographs by Christie Hemm Klok

In the 15 years since the Human Perhaps his optimism is due to his First of all, it has nothing to do with Genome Project first exposed our DNA personality‰few sentences go by with­ political correctness. It has everything blueprint, vast amounts of genetic data out a “fantastic” or “extraordinarily to do with human biology and the fact have been collected from millions of exciting.” But it is also his recognition as that human populations and the great people in many different parts of the a population geneticist of the incredible diaspora of human migrations have left world. Carlos D. Bustamante’s job is opportunity that understanding differ­ their mark on the human genome. The to search that genetic data for clues to ences in human genomes presents for genetic underpinnings of health and everything from ancient history and improving health and fighting disease. disease have shared components across human migration patterns to the rea­ David Rotman, MIT Technology human populations and things that are sons people with different ancestries are Review’s editor at large, discussed with unique to different populations. so varied in their response to common Bustamante why it’s so important to diseases. include more people in genetic studies How does that play out? Bustamante’s career has roughly and understand the genetics of different Diabetes is a great example. If we look spanned the period since the Human populations. at the genetics of diabetes, they are dif­ Genome Project was completed. A pro­ ferent in different parts of the world. In fessor of genetics and biomedical data How good are we at making sure that the early 2010s, the Broad ˜Institute of science at Stanford and 2010 winner of a the genomic data we’re collecting is MIT and Harvard™ did a study with the MacArthur genius award, he has helped inclusive? National Institute of Genomic Medicine to tease out the complex genetic varia­ I’m optimistic, but it’s not there yet. in Mexico to study the genetics of dia­ tion across different populations. These In our 2011 paper, the statistic we betes. Sure enough, they found a genetic variants mean that the causes of dis­ had was that more than 96% of par­ variant that has a 25% frequency in eases can vary greatly between groups. ticipants in genome­wide association Mexico that you don’t see in European, Part of the motivation for Bustamante, studies were of European descent. In East Asian, or African populations. It is who was born in Venezuela and moved the follow­up in 2016, the number went largely seen only in the Americas, and to the US when he was seven, is to use from 96% to around 80%. So that’s it underscores a large part of ethnic dis­ those insights to lessen the medical dis­ getting better. Unfortunately, or per­ parity in diabetes. parities that still plague us. haps fortunately, a lot of that is due We’ve done research on seemingly But while it’s an area ripe with to the entry of China into genetics. A innocuous traits like blond hair. There potential for improving medicine, it’s lot of that was due to large­scale stud­ is no more striking phenotype. Some also fraught with controversies over ies in Chinese and East Asian popula­ people have blond hair and some peo­ how to interpret genetic differences tions. Hispanics, for example, make up ple don’t. And the cause of blond hair in between human populations. In an era less than 1% of genome­wide associa­ Melanesia is completely different from still obsessed with race and ethnic­ tion studies. So we need to do better. the cause in Europe‰and that’s blond ity‰and marred by the frequent misuse Ultimately, we want precision medicine hair. So why do you think diabetes, heart of science in defining the characteris­ to benefit everybody. disease, all these other complex traits tics of different groups‰Bustamante will have identical causes in all humans? remains undaunted in searching for the Aside from a fairness issue, why is It doesn’t make sense. nuanced genetic differences that these diversity in genomic data important? It turns out the highest prevalence groups display. What do we miss without it? of asthma ˜in the US™ is in individuals

ND18_health_Q_A_Bustamante.indd 30 10/3/18 5:46 PM ND18_health_Q_A_Bustamante.indd 31 10/3/18 2:47 PM 32 Health

of Puerto Rican ancestry, followed by individuals of African­American ances­ try, followed by European ancestry. The people with the lowest rate of asthma are those of Mexican ancestry. You have two of the Hispanic populations at the opposite ends of the spectrum.

Why is detailing these genetic differ ences helpful for medicine? If the genetic etiology of disease is dif­ ferent, it gives us an opportunity to dis­ cover new drug targets. It gives us new biology that then can be used even for those that don’t necessarily suffer from the disease in that way. It’s important for drug discovery. If you think of it like looking for oil, we’ve only been looking for oil in the North Sea. There are plenty of other places to search, and that bene­ fits everyone. Secondly, we’re finding that poly­ genic risk scores ˜disease­risk pre­ “We can’t use genetics for the purpose of trying to dictions based on genetic tests™ for European ancestry don’t translate eas­ define the stories we tell about ourselves.” ily into other populations. If we don’t have broad representation in medical and population genetics, then we run In a global context there is no model of We can’t use genetics for the purpose of the risk of widening health disparities, three, or five, or even 10 human races. trying to define the stories we tell about which will be a terrible outcome for pre­ There is a broad continuum of genetic ourselves. Social determinants of health cision medicine and precision health. variation that is structured, and there are often far more important than genetic are pockets of isolated populations. determinants of health, but that doesn’t So aren’t you disappointed by the lack Three, five, or 10 human races is just mean genetic determinants aren’t of progress in including more popula not an accurate model; it is far more of important. So you’ve got to embrace the tions in genomic data? a continuum model. complexity and figure out how we trans­ I’m actually super­excited. We’ve done Humans are a beautifully diverse late this to a broad general public. a great job of mining for drug targets in species both phenotypically and genet­ I’m actually an optimist. I think the Europe. Iceland led the way, Britain led ically. This is very classic population world is becoming a less racist place. If the way, and now Finland. So we’re tap­ genetics. If I walk from Cape Horn all you talk to the next generation of people, ping all those resources‰awesome. But the way to the top of Finland, every vil­ millennials on down, those abhorrent what about Latin America? What about lage looks like the village next to it, but ideologies are thrown away. That means Africa? What about South Asia? All of at the extremes people are different. it gives us a space to now think about those places have tons to contribute to what role does genetics play in health our understanding of health and disease. But as a population geneticist? and diseases and human evolution in It is both a moral obligation and a I don’t find race a meaningful way to ways that we can soberly understand and missed scientific opportunity if we don’t characterize people. bring to bear on important problems. go to work in those populations. We can’t allow genetics to get You walk a tricky line, though, don’t hijacked by identity politics. If you begin Many genetic researchers have long you? You’re pointing out the impor to allow politics and other interests to argued that race has no basis in sci tance of variance between different come in, you just muddy the waters. You ence. But the debate doesn’t seem to populations, but you don’t want to need to let the data lead. You need to let go away. reinforce old categories of race. outcomes lead. And the rest will follow.

ND18_health_Q_A_Bustamante.indd 32 10/3/18 5:46 PM Q+A 33

DATA BIAS IN DNA STUDIES 2009 2016 373 studies 2,511 studies Precision medicine is getting more pre- 1.7 million samples 35 million samples cise for some but leaving many others behind. And those left behind are often people with Latin American, African, Native American, and other ancestries 96% 81% that are underrepresented in genomic European ancestry European ancestry databases.

By far, most of the data in genome-wide association studies, which have been crit- ical in spotting genetic variants tied to common diseases, comes from people with European ancestry. In 2011, Carlos D. Bustamante and his colleagues called 4% 19% out the disparities and the resulting threat Non-European ancestry Non-European ancestry that genomic medicine “will largely benefit a privileged few.” In subsequent years, the collection of genomic data has exploded, but the disparities remain. In 2016, Alice Popejoy, who was a PhD student at the 3% University of Washington and is now a postdoc in Bustamante’s lab, updated the results in the journal Nature, finding little 0.15% 0.57% progress for most population groups. 0.06% 0.06% One result of this lack of data is that genetic tests may be less relevant and accurate for people from underrepre- sented groups. Increasingly popular con- sumer genetic tests can be misleading 14%     or just plain wrong, and medical genetic NATURE

, tests for some common diseases are often inconclusive. Likewise, Popejoy

URCHARD says, false positives and false negatives B in genetic diagnoses are more common AND , A in people with non-European ancestry, VEG A because the results are interpreted using

, DE L databases that are incomplete or biased

ANT E toward European ancestry. M TA

Asian  ; B US ;  African 3% NATURE , Mixed ancestry ON RT Hispanic + Latin American

Pacific Islander 1% Arab + Middle Eastern 0.28%

EJOY AND FULL E AND EJOY 0.54%

P 0.08% PO Native peoples 0.05%

ND18_health_Q_A_Bustamante.indd 33 10/3/18 5:46 PM 34 Health

I nodded politely and brought the When you’re sick, you’re vulnerable—and that’s when your doctor tablet back to my chair. From the insti pressures you into participating in a data-gathering experiment. tutional perspective, this was a totally

By Mary Madden reasonable request for verification. But it was also a clear instance of surveillance, and the power dynamics between me and the administrative authority were not at all equal. I was in pain and in no mood to argue. Need medical By agreeing to use the tablet, I’d already consented to a form of data collection I wasn’t entirely comfortable with. I had help? Sorry, not never heard of the branded tablet the office was using, and the logo assuring me that it was “antibacterial” didn’t ease my concerns until you sign away about letting scores of other patients han dle a device into which I’d put my private data. The awkward software interface did little to suggest that my data would be dealt your privacy with carefully; worse than the clunky visual design, there was no indication of whether or not the tablet was internetconnected, and there was no explanation of how my data would be stored or protected once it entered their system. So what did I do? I dutifully entered my info anyway­immediate physical Last summer I found myself running late for a doctor’s needs have a way of leapfrogging over data privacy concerns, even for people appointment I’d waited months to get. Even though the like me who feel strongly about maintain back injury I had sustained three months earlier was ing control over how their information is finally starting to improve, I was eager to get an expert collected and used. opinion from an orthopedic surgeon. When I arrived, breathless and apologetic, the doctor’s office was filled Not the first time this happened with patients­many with much more serious injuries Mary Madden is As I scrambled to consult my phone for than mine­who had also waited months to see the a technology records of my grandparents’ cause of renowned specialist. As I was about to take my seat, I researcher and death and the appropriate medical term writer. She was called back to the front desk: Could I also please leads a project to describe the blood condition that runs answer some questions about my personal health his with the Data and in my family, I realized that this was prob Society Research tory using the office’s new tabletbased system? Institute to ably the fourth time over the past year As a social science researcher who has studied dig understand that I’d been asked to enter some version the social and of this data digitally in other systems­in ital privacy and security issues for much of my career, cultural impacts I was less than thrilled to be a guinea pig for their new of data-driven addition to various paper versions of the technologies on same information. Instead of making the datamanagement system. But … I had waited so long health equity for this appointment, and I had already kept the doctor and well-being. patient experience more efficient and less stressful, it made me feel as though waiting, and maybe this would save me time at future doctor’s offices were crowdsourcing their appointments with other doctors? At that moment, as work to stressedout patients with little if in response to my frustrated realization that there explanation of why. was no clear way to opt out and still receive the care I When I’d finished digitally detailing needed, my back muscles tightened up. my health history, the final screen seemed to mock me with one last request: Could

ND18_illness_essay_Madden.indd 34 10/1/18 8:26 AM Essay 35

transparent system than the one I’d used at the orthopedic surgeon’s office , only a little more than half of American adults definitively said they’d be comfortable sharing their data. Health data is one of the few catego ries of information that enjoy a robust ƒif outdated set of privacy protections by law in the US, but the definition of what even counts as health data is rapidly evolving. More and more companies are looking to use diagnostic insights from socialmedia data and other nonregulated categories that currently exist in the lucrative mar ketplace of predictive analytics. The cur rent Wild West environment allows health data brokers to create risk scores that are sold to insurance companies that in turn use these metrics to charge higher rates to the most vulnerable among us. Not only is this bad for patient privacy, but it further exacerbates inequalities in our society.

Care shouldn’t require data consent Americans’ concerns about the sanctity of their health data have been cited as one reason that Google and Apple have recently partnered with the likes of the Opt in and you risk losing control over how your American Heart Association and doctors health data is used. Opt out and you risk losing from Massachusetts General Hospital. access to the medical care you need. Such household names can help allay patients’ fears about entrusting their data to Big Tech. But we’re now at the point where the stakes are growing much higher when we make decisions to share our data with I please acknowledge that I’d received a health data systems have been vulnerable a platform or participate in a study. When copy of the office’s privacy practices? ƒI to numerous ransomware attacks, genetic we opt in, we risk losing control over how hadn’t. But what were the consequences testing companies have opened up their our health data is used and who can profit of opting out at this point? And what about customers’ data to use by pharmaceutical from it. When we opt out, we risk losing people who were much less comfortable companies, and the market for health data access to the care we need. with technology than I was? How were is massive and growing. In the era of datadriven medicine, they dealing with questions or concerns I’ve spent more than a decade studying systems for handling data need to avoid about this process? Americans’ attitudes to different kinds anything that feels like manipulation­ of digital information, and I have seen whether it’s subtle or overt. At a mini The banality of Big Brother repeatedly that health data is one of the mum, the process of obtaining consent In the internet age, it’s become repetitive most sensitive categories. In a study I con should be separated from the process of and banal to simply agree to terms of ser tributed to at the Pew Research Center, obtaining care. vice that we don’t fully understand. And respondents were asked whether they If you don’t want to hand over your while it would be nice to think that my would participate in a webbased system information right away, or if you have doctors and their thirdparty software ven that their doctor’s office used to manage concerns about the security of your doc dors will forever treat my health data with patient records. Even in this scenario tor’s datagathering efforts­you should

BENEDIKT LUFT BENEDIKT the utmost care, the reality is that digital ƒwhich notably involved a much more be able to see the doctor anyway.

ND18_illness_essay_Madden.indd 35 10/1/18 8:26 AM Get MIT Technology eview delivered to your inbox.

THE DOWNLOAD Your daily dose of what’s up in emerging technology

CLOCKING IN A daily look at the workplace of the future

CHAIN LETTER Blockchains, cryptocurrencies, and why they matter

THE ALGORITHM News and views on the latest in artificial intelligence

Stay in the know: technologyreview.com/inbox

ND18 Newsletter Ad 8x10.875 D1.indd 1 9/28/18 4:00 PM The precision medicine issue 37

Then there’s illness

Now it’s about money. Whatever ails you, there’s a growing chance we can cure it—for a price. Who can pay? And should they? If some people can excise genetic disease from their bloodlines, will we end up creating two human races, one sick and one healthy? Or are both the promise and the perils of genomic medicine being wildly overhyped?

ND18_front_illness.indd2 37 10/3/18 4:49 PM 38 A

By Antonio Regalado c u r e for Illustrations by One day, gene therapy may help with the rarest of Sébastien Thibault diseases. Some parents aren’t waiting.

ND18_illness_cure.indd 38 one 10/4/18 11:46 AM A 39 c u r e for

one ND18_illness_cure.indd 39 10/4/18 11:46 AM 40 Illness

like Canavan, are caused by errors in a single gene. Gene therapy suggests the ultimate bug fix“just give people work ing DNA instructions. The problem for the Canavan kids is that there have been too few patients for anyone to bring the research out of the lab and put dollars ennie and Gary Landsman launched behind it. The same is true for countless an appeal to save their sons on other diseases you’ve never heard of, some of which are known to affect fewer than Thanksgiving of 2017. By the end of 50 people on the globe. J the weekend the family, who live in “The simple math is that there are a very Marine Park, Brooklyn, had raised $200,000. limited number of patients. That is what created this crazy, crazy paradigm,” says In a moving threeminute video posted online, they sit on an Eric David, an executive with BridgeBio, overstuffed leather couch. Jennie glances away from the cam a biotech in Palo Alto, California, that spe era, betraying little emotion, as Gary talks. “We need your help, cializes in treatments for rare diseases. “Families are saying, ‘Oh my God, no one we really do,” Gary says, his voice breaking. The Landsmans’ is going to pay for this. I have to fund it two sons“Benny, then 18 months, and Josh, four months“ myself.’” both have a fatal genetic brain disorder called Canavan disease. Gene therapy already has a reputation as medicine’s gnarliest economic case. The Benny, limp on his mother’s lap, is already affected by nerve problem is who will pay. Even those few loss. Josh isn’t yet. But he will be if nothing is done. treatments approved for sale carry price tags as high as $1 million. Underlying the Canavan is an “ultrarare” disease. So rare, in fact, that there is no reliable under standing of how many children are born with it. Relatively few researchers study Canavan, and no drugs are approved to treat it. There isn’t even a single clinical trial open for some lastditch remedy, the kind people battling cancer can turn to. Doctors told Jennie there was not much “The simple math is that there are a very limited to be done. She should go home and make number of patients. That is what created this crazy, her boys comfortable until they died. crazy paradigm.” The Landsmans refused to accept that advice. Instead, Jennie hit Google and from August to December 2017, the US unheardof prices is the cost of years of started emailing scientists. Here’s what Food and Drug Administration ŒFDAŽ research, human tests, and paperwork to she learned: there may be a way to fix the approved three such therapies, two for win the FDA’s signoff, all in tiny markets genetic error in the boys’ brains. But the blood cancer and one for an inherited cause with small pools of patients. Costly, too, is family would have to pay for it themselves. of blindness. Companies are investigating the still unwieldy process of manufactur And it would be expensive. treatments for hemophilia and muscular ing trillions of viruses, into each of which “We need one and a half million dol dystrophy. “Once just a theory,” said the a gene is placed so it can be conveyed into lars, and our goal is to get it in the next six FDA’s chief, Scott Gottlieb, in July, gene people’s cells. The result? A growing gap months,” Jennie says in the video. therapy “may have the potential to treat between the list of diseases that could be The Landsmans had discovered gene and cure some of our most intractable and treated with gene therapy and those that therapy, technology that uses viruses to add vexing diseases.” actually are. healthy genes to cells with defective ones. The technology’s medical logic is espe I learned of six cases where parents After several decades caught in scientific cially irresistible for parents of children financed clinical trials for gene therapy backwaters, gene therapy has entered a with the rarest diseases on earth. These are in which their own children were treated. golden age. During a span of four months, the 7,000 or so conditions that typically, These include Karen Aiach, who started a

ND18_illness_cure.indd 40 10/4/18 11:46 AM A cure for one 41

biotechnology company, Lysogene, based Janson, a physician at the University of National Institutes of Health ŒNIHŽ, said outside Paris; it funded the trial in which Illinois College of Medicine. “Until then, in a speech this year. “Shouldn’t we think her daughter was treated for Sanfilippo we are stuck on our own trying to help a about ways to do that in a fashion that syndrome. A connected Hollywood cou couple of kids.” scales to hundreds or maybe thousands ple, the Grays, raised $7 million to pay for Some scientists who know of the of diseases? So what would that take?” a trial that infused genecarrying viruses Landsmans’ plan fear it represents a new Nobody really knows. And the into two of their daughters and several form of boutique medicine“a way to give Landsmans can’t wait for Washington, other children with a rare form DC, or drug companies to fig of Batten disease. More than 20 ure it out. At today’s rate of other parentfinanced trials are new drug approvals for rare in the planning phase. diseases“about 15 a year“it Other families are avoid could take 1,000 years for com ing the rigors of formal studies panies to get around to all of and trying to secure untested them. With two sons slipping gene therapies as emergency away at home, Jennie and Gary treatment. In Florida, a sin are measuring time in months gle boy was treated with a instead. Josh has a big smile but Canavan gene therapy in 2017 never learned to crawl. He’ll after his parents paid for the soon become like Benny, who experiment. They did it under moves his arms only weakly and an exemption in federal rules communicates by glancing at called “expanded access,” pictures Velcroed to a felt pad. which can allow unapproved “He’s never said ‘mommy,’” drugs to be offered to specific Jennie told me. But he can still patients “whose life is imme ask for her“one of the pictures diately threatened.” pinned up there is hers. That experiment fell into a Jennie says she hopes that gray zone, not quite research all Canavan kids will someday and not quite medicine. It is the benefit and that the research same pathway the Landsman ers helping her “will become family is trying to follow, with famous.” But she did not raise the help of Paola Leone, a gene all that money to fund an exper therapist in New Jersey, and iment or to become a philan Christopher Janson, a neurol thropist. “This is not a clinical ogist in Chicago. Leone and trial,” Jennie told me. “This is Janson asked the FDA last June private. This is for Benny and to permit emergency use of Josh.” their own Canavan gene therapy in up those with fat checkbooks or a knack for to five children they have designated in viral fundraising campaigns special access ______Perfect _ _ __ timing advance. The first two names on the wait to cuttingedge breakthroughs. A different anavan disease is rare, ing list: Benny and Josh Landsman. perspective is that it’s just a preview of the but it’s significantly more According to the FDA, the strategy personalized genetic medicine that will common among people of is not as unusual as it sounds. Of the increasingly be available more generally. Ashkenazi Jewish descent, approximately 700 genetherapy trials it In the future, health officials believe, like the Landsmans. Like oversees, 77 fall into the desperatecase it could become commonplace for sci CTaySachs, it’s enough of a threat that category, according to an agency spokes entists to detect a genetic mutation and prospective parents in this population person. It is not known in how many of whip up a custom DNA antidote for one are tested to see if they are silent carriers these cases the families are covering the person. “Those 7,000 diseases are ones of the gene error. About one in 40 are. A costs, but that is entirely legal, too. “We where we know the molecular defect for series of medical miscommunications, would love to do it in a broader, systematic most of them. We know exactly what the Jennie says, led her to mistakenly believe fashion that would lead to a drug treat initial glitch was that has led to this out she had tested negative. Since it takes ment, but we don’t have the money,” says come,” Francis Collins, the director of the two mutated gene copies, one from each

ND18_illness_cure.indd 41 10/4/18 11:46 AM 42 Illness

parent, to cause the illness, they saw no potential, and their treatment, though it reversed. “Then I was prepared to say ‘Yes’ reason to test Gary. had some effect, was no cure. to the family that came along,” she says. Jennie and her pediatrician were slow Leone had been working toward a new to pick up on Benny’s symptoms. Her Canavan gene therapy. But her last federal ______“There __ is a lot we can do” sister said the toddler seemed “mushy,” grant had run out in January 2018. In her eone met the Lands but the Landsmans’ doctor said not to lab, I saw a scientist curse at an oldmodel mans in New York, worry. By then she was pregnant with Mac that was slow to load an image. The near the 9/11 memo Josh. The disastrous diagnoses rial, in September unfolded over a few days last of 2017. Gary con summer. In late July, a blood Lfessed that if he had a choice test finally showed Benny had between fighting and fleeing, Canavan. Two weeks later, on he wanted to flee. Many par Gary’s birthday, they learned ents institutionalize children their newborn had it, too. with Canavan. Gary wished he As Jennie remembers it, she could take Jennie far away and spent weeks in depression, star never come back. “Excruciating ing at sunbeams coming under pain,” recalls Leone. “Eyes that the awning and trying to “live in had cried so much they were the moment.” But when I visited hard to see.” Leone, the gene therapist, at Jennie wanted to know what the Rowan University School of they could do. Leone told her: Osteopathic Medicine in New “There is a lot we can do, but Jersey, she showed me emails the first thing is how much Jennie had sent her between the it will cost. I can tell you it’s two boys’ diagnoses. “Can you approximately $1.5 million.” help?” she had asked. “We can do it,” Jennie said The idea of gene ther without blinking. apy traces back to 1970, but Leone tallied the costs. They only recently have scientists would need to hire a company mastered its components. In to chemically synthesize healthy 2017, doctors at Nationwide copies of the gene that’s broken Children’s Hospital, in Ohio, in Canavan, set aside payments described in The Lancet how for neurosurgeons, and hire they had prevented a group of consultants to prepare a request infants from developing spinal to the FDA. The biggest single muscular atrophy, a nerve dis expenditure would be manufac order that, like Canavan, is fatal. turing. Making the viral parti The key elements: a virus that infects the shoestring budget is nothing new. “When I cles“they’re grown in thin sheets bathed right cells Œnerves, in this caseŽ, immense started this work,” she says, “people looked in components of cow blood“remains a doses, and timing. Give a onemonthold at me and said, ‘You must be out of your delicate craft, and there are long waiting infant the missing gene, and the nerve mind to work on a rare disease“you are lists at production centers. Leone believed damage doesn’t begin. It now appears to never going to find any money and no one it would cost at least $1 million just to make scientists“and parents“that similar strat is going to be interested.’” enough virus to treat Benny, Josh, and per egies must be capable of saving kids with Leone keeps pictures and memorials to haps a few others. other inherited nervoussystem diseases. Canavan kids she has known in her office. The Landsmans didn’t have the money. Leone was a logical person for Jennie to Over the years, she had told many of their The family is squarely middle class. “But approach. Between 2001 and 2005, Leone parents there was no chance at a cure. there’s money everywhere, isn’t there?” and Janson had, with government funding, But the Landsmans’ timing was perfect. Jennie reasoned. She was right. Their video, treated 13 children with Canavan in one By the fall of 2017, Leone had given the posted to Facebook and later GoFundMe, of the first attempts to change the genetic new gene therapy to enough mice to see a crowdfunding site, went viral. By now, code inside a person’s brain. At that time, what she calls dramatic effects. The dis they’ve been on TV and in People maga scientists were unsure of the concept’s ease seemed to have greatly slowed, even zine. Eight thousand donors have already

ND18_illness_cure.indd 42 10/4/18 11:46 AM A cure for one 43

given more than $1.5 million. “This was all he had a pathological SLC6A1 mutation. Part of Gray’s job is to reset parents’ local people in a small Jewish community Freed had been working as an equity ana expectations. Gene therapy is not as simple in Brooklyn,” says Ilyce Randell, a Canavan lyst in Denver, Colorado, but quit the day as packaging a gene into a virus. Many dis patient advocate who has been in con of the diagnosis. “I stood up from my desk eases can be poor candidates“for instance, tact with the Landsmans and has funded and never looked back,” she says. those in which a gene is overactive rather Leone’s work in the past. “It was a perfect Until recently, many children with clus than broken. Often scientists have ground storm“everyone rallied.” ters of unusual symptoms would remain work to do, such as engineering a mouse But if the Landsman children end up undiagnosed. Starting in 2010, genome to mimic the condition. Bypassing these benefiting, she says, it will be because of sequencing became inexpensive enough steps can be perilous. If a child’s body has research under way long before they were to employ as a routine diagnostic tool. been missing a vital molecule since birth, born. “To make it seem like they bought Now, more often than not, even mysteri for example, adding it may provoke a vio a cure for a million bucks“that is mis ous inherited disorders can be linked to lent immune response. “We have gotten leading,” she says. “What is true is they a genetic misspelling. “Now you can walk this wrong in the lab and we have killed came at the right time. Ten years ago you out of a hospital with a genetic cause,” mice,” says Gray. “Gene therapy is not a couldn’t say, ‘I’ll raise money and get my Freed told me. “I think pretty soon kids pill you can stop taking.” kid treated.’ Three years ago you couldn’t will walk out the door with a solution.” Gray’s bestknown client is Lori Sames, do it. The science was not ready.” Without treatment, Freed’s son will whose daughter suffers from giant axonal come to experience a violent form of sei neuropathy. The disease affects only about ______Unfair _ _ __ system zure called a “drop attack.” The victim 80 people in the world. Sames managed n August, many of the families and remains conscious but frozen and can to raise $6 million, which she funneled to key researchers in the raredisease topple to the ground, unable to break the Gray and into animal tests. In 2016, her world arrived in the security line at the impact. “It’s coming, but we are going to daughter became the fifth child treated in NIH in Bethesda, Maryland. During a get the cure before it gets to that point,” a study of Gray’s gene therapy at the NIH. twoday meeting, cosponsored by the said Freed, who came to the meeting in Gray told me that if a gene looks like a IFDA, scientists gave talks whose topics a power suit and positioned herself near good candidate, and a family has money to support laboratory work, he will agree to take on their cause, no matter how rare At today’s rate of new drug approvals for rare the disease. “This is the most unfair sys diseases—about 15 a year—it could take 1,000 years tem imaginable,” he admitted. “If you don’t for companies to get around to all of them. have money, it won’t happen.” To some bioethicists, when parents fund treatments it has the potential to cre ate sharp ethical dilemmas. “There is a fair ness issue if only the people who have the money get to be first in line,” says Mildred Cho, a bioethicist at Stanford University, who has consulted on similar cases. “And teetered between remarkable results of the stage. “We are going to find the cure there is a scientific integrity issue, because tailored therapies and what the organiz for him. Our secondary mission is to help those with the money may not be the most ers called “unanswered questions” about those who come after us.” appropriate or the best candidates. These how these could ever reach patients at That evening I spotted Freed perched decisions should be objective.” affordable prices. on a stool at a Bethesda eatery, speaking I asked Sames if she had created a con The event attracted parents hoping to to a researcher named Steven Gray. A flict of interest by paying for research. The find genetherapy specialists who would soft spoken southerner and gene therapy question makes “the hairs on my arms treat their children. One, Amber Freed, specialist at the University of Texas stand up,” she said. “Anyone who suggests wore a name tag reading “SLC6A1,” the sci Southwestern Medical Center, Gray has it’s corrupt that parents privately fund entific designation of a littlestudied gene. become the goto scientist for parents like development of a treatment for a child, Freed told a story that was by now becom Freed. During the conference, he showed in an attempt to save the child“well, I ing familiar. After months crisscrossing a slide listing 23 rare diseases for which think it’s irrational and rather insane. If the country trying to diagnose her son’s he is trying to develop genetic treatments. the parents don’t drive it, it’s never going unexplained symptoms, she finally had his Gray says he finds the kids’ stories tragic to happen. Wake up.” Sames adds that the genome sequenced. In May, she learned and a powerful motivator. fundraising she did never guaranteed her

ND18_illness_cure.indd 43 10/4/18 11:46 AM 44 Illness

daughter, Hannah, a spot in the NIH trial. ______Calling _ _ __ Bill Gates 20 open studies exploring gene treatments Hannah had to pass a lung function test f the 7,000 rare diseases, for that disease, which could be the tech like others to get in. “We were no differ around 90% currently have nology’s first blockbuster. The markets for ent than any other candidate,” she says. “I no treatment whatsoever. ultrarare diseases haven’t drawn as much wept the day she passed the test.” Since Gene therapy could poten commercial interest. “Imagine a company then the trial has been moving forward at tially help with many, and in with 75 employees that exists to treat 75 a “glacial” government pace, according to Othe future, new technologies like gene edit people. You can see the problem,” says Sames, and other parents are mad at her. ing could, in theory, make it possible to fix Eric David, the executive with BridgeBio. “They are hurt“their child is failing before nearly any genetic mutation. Christopher In April 2018, however, something hap their eyes“and they are angry, angry their Austin, chief of an NIH branch responsible pened to make biotechnology executives kid is not injected,” she says. “But there is for new therapies, says eventually there take a fresh look. The Swiss drug company nothing I can do.” may be as many different treatments as Novartis announced that it would buy the Some families are managing to move there are unique DNA flaws. To Austin, that genetherapy company AveXis for $8.7 even faster to a treatment than Sames means madetoorder, hyperpersonalized billion. AveXis had just one drug in the did. The Hollywood couple, film pro medicine isn’t some ethical mistake to clinic“it owned rights to the treatment ducer Gordon Gray and his wife, Kristen, avoid; it is the next step forward. “All of us for spinal muscular atrophy that had been were able to get two daughters treated need to think deeply that this is possible tested at Nationwide Children’s Hospital. at Nationwide Children’s Hospital about now,” he says. It’s something “that people The acquisition price was immense for a one year after the girls were found to have have thought about for decades“and now treatment used, at that time, on only about Batten Cln6, an inherited nervoussystem it seems to be coming true.” 15 kids, and for a disease that affects one disease believed to affect just a few hundred Exactly who will pay to discover, in 10,000 births. kids. Kristen Gray says the couple paid for develop, and deploy this Noah’s ark of “My jaw hit the floor. I don’t even know the trial in its entirety. They also formed medicines is not clear. Lori Sames told me what $8 billion is,” says Jerry Mendell, the a company to take commercial rights to she sometimes fantasizes about approach doctor who led the trial. Mendell didn’t the treatment. ing Bill Gates, whose foundation is trying hold shares in AveXis, but one of his cen Few parents, though, are able to raise millions or start a company. On GoFundMe, hundreds of appeals mention gene therapy, but most raise only a few thousand dollars. One woman from Texas appealed for funds because she has muscular dystrophy; she has gathered just $35. The medical pos sibilities are out there, “but I don’t think there is the regulatory infrastructure or “All of us need to think deeply that this is possible now. the funding infrastructure to really make It’s something that people have thought about for it happen,” says Steven Gray, the gene decades—and now it seems to be coming true.” therapist from Texas. Another obstacle is that most of the to eradicate malaria and polio. Leone envi ter’s former employees, Brian Kaspar, did. key components of gene therapy are pat sions a different solution: a global institute Kaspar, who joined the company, is now ented“including the viruses, the produc of cures, with access to manufacturing, $400 million richer. “In my mind, the tion tricks, and engineered genes. That hospital beds, and agreements in place AveXis deal“there is a before and there means parents, and the scientists who to streamline the “biblical” work of deal is an after,” says David. “After that, peo help them, are often working in a cloud ing with insurers and regulators. “So any ple who would not have looked at gene of legal uncertainty. Leone says to treat new disease, any new genetic mutation, therapy for a disease quite that rare said, the Landsman boys she will have to buy we’d have everything set up,” she says. ‘Wow“if I can get a trial going, maybe I $250,000 worth of trial insurance. “I “We would bring patients from all over can be worth a billion dollars too.’” could have been stopped with a phone call, the world for treatment.” One reason AveXis was worth so much but I wasn’t. People have been very kind,” Biotech companies have raced into gene is that the treatment seemed to be an out she says. “But I will tell you, there are so therapy, but so far, much of their effort has right cure. That could let Novartis charge many pieces in the patent puzzle … it’s been aimed at more common genetic con $2 million per patient, and perhaps more. like a contemporary symphony, one that ditions like hemophilia. The US govern To Walter Kowtoniuk, a principal at the is atonal. It makes you want to scream.” ment’s clinicaltrial website lists more than investment company Third Rock Ventures,

ND18_illness_cure.indd 44 10/4/18 11:46 AM A cure for one 45

in Boston, such medical successes mean ______A _ _ticking __ clock agency responded with a notice, called a diseases previously thought to affect too avid says the formal clinical “clinical hold,” delaying the experiment. few people to attract companies are sud trial of his company’s Canavan At the time of writing, in October 2018, denly drawing intense interest. He says he treatment, different in design Jennie was counting on December at the has been “shocked” by the “massive com from Leone’s, won’t begin for latest. Janson, the doctor running the trial, petition” to gain control of gene therapy another year, maybe two. For thought sometime in 2019 was more likely. programs. Dhis company it’s important to plan carefully He and Leone have plans to submit a new That’s created a situation in request following a meeting which desperate bids to treat with FDA officials. He has also children can rapidly turn profit started testing the treatment on able. In October, the Gray fam monkeys, a costly safety step ily“which had helped form a he predicts regulators may virtual company around the insist on. Batten Cln6 treatment“sold Nerves are fraying. At the rights to another biotech Benny’s age, Canavan patients nology company, Amicus, for often have a steep decline in $100 million. Some investors brain function. Even gene ther are starting to think Canavan apy might not reverse it. “My looks pretty interesting too. It’s blood pressure is really going widely known among Canavan up,” Janson says. “We proba parents that a couple in Florida bly lost at least three to four spent more than $1 million months.” to get their child treated in a When I visited her, Jennie oneperson study organized by had the idea of going to the the University of Massachusetts FDA meeting and bringing her and the University of Florida. I kids. She wanted to know what reached the boy’s father, who I thought. If the regulators saw asked to remain anonymous but them, how could they say no? did say his son seems to have Janson doesn’t think it’s a good benefited. idea. “I think we have to go The Florida experiment within the system,” he says. helped launch Canavan out of “We aren’t a drug company. We the tooraretocare category. don’t have unlimited resources Early in 2018, David’s com to lobby the FDA.” pany, BridgeBio, entered an I asked Janson if he thought agreement to license the treat it was fair that the Landsmans’ ment from the University of kids could end up getting Massachusetts and created a subsidiary, and not rush, since that would jeopardize treated while some other family without Aspa Therapeutics, which he now leads. But its investments and its aim of getting a a surprise GoFundMe success would not Kowtoniuk says other investors have been treatment approved. be. “Unfortunately, there are a lot of things angling to take over the project because the Jennie Landsman’s children, though, in society that are not fair,” he said. “There risk seems much lower now that one boy can’t wait that long. A selffinanced experi are parents who want to see me in my has been treated. “There is a battle, literally ment is probably the only way her two kids neurology clinic and can’t because they a battle, to license the technology,” he says. can get gene therapy in time. When I vis don’t have insurance. We have a problem “I think there is such a tidalwave shift in ited the Landsman home, I stood behind in society.” what is going on right now.” Josh, who was belted into a high chair, as Precision medicine, it seems, is just The growing biotech interest could his mother showed him pictures of lunch another example: “There is no easy answer mean the end of the parentled scrambles. foods: chicken, macaroni and cheese, corn. to your question, because the system is not David told me he doesn’t think the epoch Jennie followed his gaze. set up to deal with this.” of parentfinanced gene therapy will last She had hoped the boys would be Antonio Regalado is MIT Technology very long. “It’s transitional,” he says. “I treated by now. The team submitted its Review’s senior editor covering think it’s going to be for a limited time.” proposal to the FDA in June 2018, but the biomedicine.

ND18_illness_cure.indd 45 10/4/18 11:46 AM 46 Illness

ND18_illness_genentech.indd 46 10/3/18 3:27 PM 47

One Personalized tumor cancer at a vaccines time are a scientifi c breakthrough, but can they be a sustainable business?

by illustrations

Adam Piore Selman Design

he first time someone pitched sitting next to him and mutter, “Over my Genentech’s senior leader dead body. A vaccine will never work.” ship on a personalized cancer That was in 2012. Cancer immunother vaccine, it did not go well. “I apy, which uses a person’s own immune thought there was going to system to attack tumors, is now one of Tbe a riot,” Ira Mellman, then Genentech’s medicine’s most promising fields, and one head of research oncology, recalls. of the greatest breakthroughs in oncol From across the table, he watched the ogy in decades. But it took a long time scientific review committee grimly shaking to get there. Until the recent advent of their heads as his team member and long a new class of blockbuster immunology time collaborator Lélia Delamarre made drugs, the field was notorious for ques her case. Then he overheard the head of tionable science, hype, and spectacular clinical development turn to the person disappointments.

ND18_illness_genentech.indd 47 10/3/18 3:27 PM 48 Illness

And what Mellman and his team were No wonder the Genentech leadership proposing that day went further than tur was so skeptical. bocharging immune cells to make them After that calamitous first pitch meet better able to attack cancers. They were ing, Mellman and Delamarre retreated talking about a vaccine precisely tailored to their laboratories. They returned a few to stimulate the immune system to react to months later with more exciting data: they specific tumors. If it worked, the approach had identified specific targets on cancer could, in some cases, be even more potent cells, targets that would readily be attacked than other types of immunotherapy. But by immune cells. They also had fresh, con it faced a series of daunting hurdles. If vincing research from a growing number Genentech, a San Francisco‹based biotech of other academic groups on the feasibil company owned by the Swiss pharma giant ity of their approach. And, critically, they Roche, were to attempt to develop a vac had a preliminary plan for how Genentech cine that could attack individual tumors, it itself might take the first tentative steps wouldn’t just have to accept new scientific toward making tailor made treatments an advances; it would also have to embrace an economically viable product. entirely new and untested business model. This time the reception was different. That’s because the vaccine Mellman and The committee signed off on an explora Delamarre envisioned could not be man tion that would culminate in 2016 with ufactured the traditional way, in large a $310 million deal with BioNTech, a batches that could be packaged in bulk, German company that has a technique warehoused, and dispensed off the shelf for producing personalized vaccines to at your local pharmacy. target tumors. Last December, the part When Mellman and Delamarre said ners launched a massive round of human “personalized,” they really meant it. The testing, targeting at least 10 cancers and composition of each vaccine would be enrolling upwards of 560 patients at sites based on the characteristics of each around the globe. patient’s tumor DNA. The company would At Genentech headquarters, Mellman have to, in essence, make a separate treat and Delamarre’s small team has grown by ment for every single patient. now into an army of hundreds, consisting Nor would this be the kind of drug you not just of lonely lab workers but supply could order up with a prescription in hand chain specialists, regulatory experts, diag and get in a few days, like Genentech’s nosticians, and a whole host of consultants, highly successful cancer drugs Herceptin all focused on the laborious task of figuring and Avastin. To create this drug, the com out how the production of their promis pany would have to orchestrate a multi ing new product–should it continue to The company would step process for each patient, performed demonstrate the powerful effects seen have to, in essence, at multiple sites. Each patient would need so far–might be scaled up in a way that a biopsy, the tumor tissue would have won’t bankrupt the company. make a separate to undergo full genome sequencing, the “It’s never been done, so we are learn treatment for every results would require complex compu ing as we go,” says Sean Kelley, the project tational analysis, and the individual vac team leader overseeing the effort. single patient. cines would then need to be designed and Nor are Genentech and BioNTech the queued up for manufacture. Theoretically, only companies now pushing into this if the vaccines were to be produced on a new territory. In late 2017, Moderna, a bio large scale, this would have to happen tech based in Cambridge, Massachusetts, hundreds of times a week. And it would announced that, in partnership with phar have to happen fast. maceutical giant Merck, it intended to If any single step in the process went start human trials with a vaccine tar awry, if a shipping mistake occurred or a geting solid tumors. Another company, batch was contaminated, it could prove Neon Therapeutics, founded by research deadly–because cancer doesn’t wait. ers at Dana Farber Cancer Institute and

ND18_illness_genentech.indd 48 10/3/18 3:27 PM One tumor at a time 49

that involve removing immune cells from the body, genetically engineering them, and then reinfusing them in the hopes of triggering a robust response. Other cancer immunologists have focused on developing drugs to turn off molecular switches on the immune system’s T cells that can interfere with their ability to attack. But until recently, the scientific tools simply didn’t exist to take the sophisticated personalized approach Genentech is now pursuing–an approach that requires sci entists to fully characterize an individual cancer tumor, identify the most attackable mutations, and then design a personalized vaccine that would provoke the immune system to target them. The problem was identifying the right target molecules on the tumor cell, or–as researchers thought of them–the antigens that would catch the attention of the immune cells. “It was so much work to identify anti gens in the past,” says Robert D. Schreiber, director of immunotherapy at Washington University. “You could do all this work, and then you end up with one antigen from one individual that is not necessarily ever seen again in any other individual.” That all changed with the advent of cheap genetic sequencing. In 2008, five years after the Human Genome Project published the sequence of the first human genome, scien Washington University, treated its first tists published the first genome sequence patient in phase 1 trials in May with a Fighting against yourself of a cancerous cell. Soon after, scientists similar vaccine derived using a different cientists have been intrigued began to compare the DNA in tumor cells method. It raised $100 million in an IPO for decades by the possibil and healthy cells to characterize the myr this summer, driven largely by optimism ity that cancer’s greatest iad ways that they differed. These studies over its approach. strength–its ability to mutate confirmed that all cancer cells contain hun The technology for the first truly and evolve–might also be one dreds–if not thousands–of mutations, personalized cancer vaccine is not yet Sof its greatest vulnerabilities. most of which are unique to each tumor. proven. And these therapies are all likely Mutations in cellular DNA are, after In 2012, a team of German researchers, to be expensive, Mellman acknowledged all, what cause cancer in the first place, led by scientists at BioNTech, sequenced a recently, sitting in a spacious conference by prompting the cells carrying them to widely used mouse tumor cell line designed room outside his office at Genentech’s grow and proliferate uncontrollably. As far to mimic human melanoma cells. They headquarters in South San Francisco. But back as the 1940s, some researchers were identified 962 mutations and used RNA he insists that if it’s all done right, the extra arguing that it might be possible to put the sequencing to identify 563 that were costs and thinner margins will be more immune system’s cellular bloodhounds onto expressed in genes. The group then created than offset by the sheer number of people the scent of a specific tumor by somehow vaccines made of protein fragments that who would use the treatment. priming them with a vaccine that helped it contained 50 of the mutations and injected “You can imagine a scenario where every recognize the tumor’s mutations. A num them into mice to see if this would prime single cancer patient would benefit from ber of researchers have experimented and the immune system to respond. About one this vaccine,” he says. “That’s unheard of.” continue to experiment with techniques third–16 of the mutations–were detected

ND18_illness_genentech.indd 49 10/3/18 3:27 PM 50 Illness

by the immune system, and five of those generated an immune response designed specifically to attack any cell found to har bor such mutations. It was concrete evidence suggesting that genome sequencing could be used to design an effective cancer vaccine capable of putting the immune system on the trail of multiple mutations at the same time– and that such a vaccine might indeed provoke the immune system to attack a tumor. The race was on to answer the next logical questions: Why is it that the human immune system can be stimulated to attack some mutations and not others? And how can we figure out which mutations are most likely to be vulnerable? At the urging of Mellman, Delamarre took Genentech’s own lab mice and sequenced their tumor cells, identifying 1,200 individual mutations not present in normal tissue. Then she measured how T cells naturally responded to them. Of those 1,200 mutations, she found, the mice’s immune system had begun to mount attacks against only two. To answer why only those two mutations appeared to attract an immune response, Delamarre took a closer look at the inter action between the cancer DNA and a key component of the mouse immune system known as major histocompatibility com plex, which in humans is called the human had ignored faced down and were hidden in warehouses, and grassy courtyards. On a leukocyte antigen system ¦HLA§. The HLA grooves in the cellular surface or obscured sunny morning this past August, cheerful complex comprises 200 different proteins on the edges of the HLA. The immune sys groups of men and women in shirtsleeves that protrude from cellular surfaces like tem attacked those two mutations because and T shirts strolled casually through a microscopic thumbtacks on a poster board. they were the easiest to detect. By inject courtyard outside the company cafeteria. When passing immune cells detect the ing mice with a vaccine designed to target A band was setting up, getting ready to presence of a protein fragment that doesn’t those two mutations, she could enhance regale the lunchtime crowd with some belong–a piece of an unwanted virus or their bodies’ ability to fight the tumors. blues, while nearby some kitchen work bacterium, or a mutation–they sound the Together, these findings were what ers prepared outdoor grills to cook food alarm and cause the body to attack it. helped her and Mellman convince for employees. Delamarre had determined that roughly Genentech’s review committee that a Much of this is paid for by cancer seven of the 1,200 tumor mutations she’d cancer vaccine was worth pursuing. drugs. Genentech won approval for its identified were displayed on the cellular first cancer treatments in 1997, and since surface by HLA. When she examined the then the company has fielded no fewer structure of these seven protein fragments, Facing the music than 15 of them. something got her attention: in the two enentech’s headquarters, But a cancer vaccine is unknown that the immune system had recognized, in an industrial park just territory. The initial human trials that the mutations were prominent on the cel off California’s Highway Genentech and BioNTech launched last lular surface, facing up toward passing 101, is a sprawling campus year are shaping up as a test not just immune cells. Those the immune system Gof glass buildings, hulking of the vaccine’s efficacy but of the two

ND18_illness_genentech.indd 50 10/3/18 3:27 PM One tumor at a time 51

partners’ ability to scale up the new tech turnaround time has been at clinics and nology. By design, the geographic scope labs where patient biopsies themselves and the number of conditions targeted in are collected and analyzed–a problem, the trial are broad–so far Genentech and since individual vaccines can’t be manu BioNTech have opened sites in the US, the factured until the samples are received. UK, Belgium, Canada, and Germany, and The issue, Fine believes, is that patients they are likely to expand to other nations with metastatic cancer may have problems around the globe. getting to the doctor in a timely manner Producing the vaccines even for the because they are too sick. And many col small number of patients in early trials lection sites don’t yet have a procedure for “was an extremely challenging process,” flagging their samples as urgent, which says BioNTech CEO Ugur Sahin, a vet means they can get lost in the stack with eran cancer researcher who cofounded other biopsies. the company in 2008. “Everything was Getting the vaccines back to the patients driven by pipetting and by people on the themselves has also proved problematic. bench producing the vaccines,” he says. At least one vaccine has been held up at “So we had a very small capacity.” customs in New York City. BioNTech has been able to automate For now, the problems are manage some functions and reduce the time it takes able and informative because the num to manufacture each vaccine from three ber of patients is relatively small. But all months to about six weeks. It is shooting these problems will have to be solved if to get that down to four weeks by the end the vaccines are ever to go mainstream. of the year. “You’re not going to be able to wait six The company can now produce hun months for a vaccine if you have a patient dreds of vaccines in a year–it aims to reach with fast progressing pancreatic cancer,” 1,500 over the next year. But if Genentech says Kelley. and BioNTech are ever to bring the prod Genentech officials declined to spec uct to market, they will need to be able to ulate about the eventual price of the vac produce between 10,000 and 20,000 a cine, insisting it was too early to know. year, Sahin says. “It’s going to be more expensive,” says In San Francisco, teams from Genentech Kelley. “This will cost us much more to and BioNTech track progress in a desig make per person.” nated space, consisting of a suite of rooms. The cost of sequencing might come On the walls, there are huge charts spelling down, building out a manufacturing net out the patient status, the manufacturing work would increase efficiencies, and new If any single step in and supply chain, the duration and sched assays might be developed, or new technol the process went awry, ule for each activity. “The key thing is that ogies that allow the cheaper manufacture on paper it can look like a very coordinated of the vaccines themselves. “We’ve done if a shipping mistake process, but if any of those steps break estimates, and we feel that right now it occurred or a batch down, then you can be in a situation where is viable, but we would like it to become, you have to start over,” Genentech’s Sean obviously, more and more viable,” he says. was contaminated, Kelley notes. For now, though, one of the most prom it could prove deadly. A number of unanticipated chal ising advances in cancer research remains lenges have arisen. Early on, the team an experimental treatment. It might be a was surprised to discover that workers at medical breakthrough, but it is facing a BioNTech were contractually prohibited familiar logistical challenge: how to get from working on weekends–so there was the product cheaply and quickly where it no one to receive patient tissue samples needs to go. arriving then. Adam Piore is the author of The Gregg Fine, a senior medical direc Body Builders: Inside the Science of the Engineered Human, about how tor who is overseeing the trials, says he bioengineering is changing modern has been surprised by how variable the medicine.

ND18_illness_genentech.indd 51 10/3/18 3:27 PM 52 Illness

The benefits of genomic drugs are exaggerated, hurting patients and The skeptic: What the practice of medicine, says one precision medicine high-profile oncologist.

By Stephen S. Hall revolution? Portraits by John Clark

Vinay Prasad is relatively young 35 and studies and the data they generate. In all the time, we shouldn’t be funding still climbing the academic ladder he’s the following conversation with veteran this. That’s wrong, because science an associate professor of medicine at medical writer Stephen S. Hall, he takes needs more funding“needs a lot more Oregon Health & Sciences University in aim at “precision oncology,” the gaps in funding than what we’re currently Portland , but he has already established direct­to­consumer genetic testing, and investing. an outsize reputation as a “professional what it really costs to bring a new drug scold” for his sharp critiques of contem­ to market. Does it hurt the patient? porary biomedical research, including I would say inflated rhetoric about the personalized medicine. In commentar­ Proponents have been promising a value of medical practices, technolo­ ies in high­profile medical and scientific revolution in personalized medicine gies, or science harms patients because journals, and in a Twitter account with for decades. What’s the reality? it distorts their understanding of what a some 25,000 followers, Prasad has ques­ I would say, and I think many people therapy or intervention might do. And tioned the evidence or lack thereof will agree, that the promises that were by distorting the understanding, it robs to support the use of precision oncol­ made around the time of the Human them of autonomy. I’ll give you just one ogy, the practice of selecting drugs for Genome Project have largely not mate­ example. patients on the basis of specific muta­ rialized, and that the impact of per­ Sometimes cancer patients are on tions in their tumors. He has also criti­ sonalized medicine has probably been medications that add real side effects to cized the inflated cost of cancer drugs exaggerated. their life, but they believe that there’s and the financial conflicts of interests going to be some survival benefit by bedeviling contemporary research. What’s the danger of exaggerating the taking this medicine. Every person is Prasad brings several unique per­ promises? making kind of a daily decision: Do I spectives to the role of medical I think we have a schizophrenia in stick with this medicine or not? Are the scold. Born in Euclid, Ohio, outside science and medicine. On the one side effects worth it to me or not? And Cleveland, to an immigrant couple hand, people who are good scien­ if that decision is made in a very impar­ from India, he developed an interest tists understand that science is diffi­ tial way, with a good understanding of in philosophy in college before attend­ cult. You should not be, nor will you what the drug does, that’s the right way. ing medical school at the University of be, having breakthroughs all the time. But if that decision is made under the Chicago. As a practicing oncologist, the Breakthroughs are rare. Science is hard. cloud of hype, when it’s surrounded and prolific Prasad has generated a boat­ It takes years of slogging to understand marinated in hype and misinformation, load of peer­reviewed papers, gather­ very fundamental pathways. then I think what we’re really doing is ing evidence to suggest, among other On the other hand, we often are that we’re preventing the person from things, that genomic­based evidence tempted to“and I see experts continue making the decision compatible with hasn’t made much of an impact on can­ to“make grandiose promises, and have their wishes. We’re kind of taking away cer patients. As a sometimes prickly a lofty, unrealistic vision for what might that choice. And I do fear that that hap­ online persona, he has been faulted for be achieved in the next few years. pens quite often. unleashing expletive­laden putdowns That harms the public understand­ but has also attracted a robust audi­ ing of science, because the public You recently published a study indi ence for what he calls “tweetorials,” comes to believe that unless you guys cating that most cancer patients don’t which dissect the design of high­profile and gals are producing breakthroughs benefit from personalized genomic

ND18_illness_Q_A_Prasad.indd 52 10/2/18 4:36 PM ND18_illness_Q_A_Prasad.indd 53 10/2/18 4:36 PM 54 Illness

medicine, even though it’s been in practice since at least 2006. Why do you think that’s the case? Some people have said that study is pessimistic. It’s neither pessimistic nor optimistic; it is simply the most realis­ tic estimate of how many people have benefited from genome­driven ther­ apies. There clearly are some situa­ tions in cancer where drugging a single cancer­causing gene is important, and that should not be taken away. Those clearly do exist. The problem is that they simply don’t exist for the majority of patients who will be diagnosed with metastatic cancer. The purpose of our paper was to docu­ ment what that number is, and what has been the change over time. I’ve heard the rhetoric that we’re reaching expo­ nential growth, or that ›precision oncol­ ogy˜ is taking off, or there’s an inflection point. We simply don’t see that evidence if you look objectively at the data.

Does that mean you’re reluctant to use them in your own practice? does what you think it does. Precision American Medical Association˜ about Of course I use genome therapies. I love medicine should be held to the same a month or two ago. It points out that ›them˜. Where they work, they work standard. there are some limitations to that direct­ well. In fact, I would increase the fund­ One of the differences is that preci­ to­consumer BRCA testing. The test is ing to research them. But at the same sion medicine is very, very seductive. actually only for three mutations that are time, I think we should be realistic Some of its bio­plausibility is just such very common in the Ashkenazi Jewish about their prospects. We’re also doing a compelling story that I think we do population, but not perhaps the most that same kind of analysis right now for see this temptation by proponents that common BRCA mutations among all immunotherapy drugs and cytotoxic it shouldn’t be assessed in the same people with deleterious mutations. And drugs and different kinds of drugs. Can way. It’s so plausible, it should just be thus there are some unintended conse­ we more accurately compare what has adopted“that kind of attitude. That quences. A woman with a family history been the impact of these different types kind of attitude might paradoxically lead who may be worried will send off that of therapies? us to adopt potentially more things that test, get a negative result, and feel reas­ ultimately turn out not to do what you sured. But that person may have a dele­ In a recent article, you suggested think they should do. terious BRCA mutation. It may actually that if adopted prematurely, the use be counterproductive. of precision medicine might actu Do you think directtoconsumer mar ally increase the risk of inappropriate keting by companies like 23andMe has If genomic testing and these other medical care. How so? made it seem as though personalized aspects of personalized medicine are Every day there are new potential medicine has arrived already? not currently predictive of outcomes treatments or therapies or strategies Yes, I think the constant rhetoric that for individual patients, are the drug to treat any disease, and they all have this is wonderful has shifted the public companies and medical institutions some degree of bio­plausibility. When perception. In terms of the direct­to­ taking advantage of consumers by it comes to a new cancer drug, bio­ consumer advertising, we actually have pushing these methods? plausibility is just not enough. You a paper on the BRCA breast cancer gene It’s a big category, and there are some should also test it and prove that it test that appeared in ›the Journal of the things that are very well validated. But

ND18_illness_Q_A_Prasad.indd 54 10/2/18 4:36 PM Q+A 55

I think there are some things that are I think that the cleanest estimate that would if you were not receiving that not. And the consumer doesn’t always I’ve seenand I’m a little bit person money. That’s the concern. I think we know which ones are which, and that’s ally biasedis the estimate that Sham should try to curb the financial conflicts the challenge. Even some of the people Mailankody and I put out in JAMA of forprofit companies in the health in the field apparently seem to forget Internal Medicine, where we estimate care space. which ones are which, and that’s what I that it costs something like $800 million There are some legitimate questions try to remind them of. in R&D to bring a cancer drug to mar here about the role of financial conflicts ket. The industry estimate is $2.6 bil in this space. Does it distort the impar When you remind them, it sounds like lion. There’s a big difference there. But tiality around adjudicating medical prac you get pretty strong pushback. at the end of the day, this is one of those tices? I fear it does. I appreciate pushback when it’s about few things in life where you don’t have the technical merits of any of these to settle for estimates. Since the indus Given the implications of the kind of arguments. Where I think pushback is try repeatedly uses the cost of R&D as critiques that you have been publish counterproductive is when pushback a justification for the high priceand ing pretty prolifically, why aren’t more becomes personal or when pushback is unsustainable priceof drugs, I think people saying the same thing? about the intention. it’s probably fair game for governments I ask myself that all the time. These There are a number of people who to ask them to show the data. Let’s just questions feel very obvious to me. have voiced concern that one or more put all the data on the table and let’s see There are a lot of people who do care. precision therapies don’t have the data. what it really costs. A lot of them are general internal And sometimes I feel as if the argument medicine folks. I think we see it a little devolves into the people who want that One of the other things you’ve sug less in the specialties. And I think we therapy saying, “Well, we want what’s gested is that the expert panels that see it much more in the younger crop best for patients. And you people who advise the FDA have financial con of physicians than the older crop, in are saying that we don’t have data, you flicts of interest. Is that compromis the sense that people who have done this, practiced for many years in this environment and who have found their niche in the environment, they’re com Precision medicine is very, fortable where they are, and they don’t really feel the urge to comment about very seductive. The temptation these more problematic areas. But peo ple who are younger, and approach this is that it shouldn’t be assessed field with fresh eyes, feel as if these in the same way as other things are problematic. You don’t always sound like a scold. treatments. I’m very optimistic about science, that we will improve outcomes. I just think that we would benefit from a lot more empir apparently don’t want what’s best for ing the quality of medicines that icism and impartiality in the process. patients.” I think we have to recognize consumers are getting? That’s what I feel is missingempiri we all want what’s best for patients. This I just want to clarify my view here, cism, impartiality, and more modest rhet is an argument about the evidence. And which is that I wholeheartedly sup oric. I think those three things would go I get personally frustrated when I see port collaboration between academic like 90% of the way. people try to pervert the argument in investigators and forprofit companies. that way. The additional complexity and chal Is it true, as reported by The Cancer lenge is when you have payments made Letter, that you’ve closed your Twitter You’ve also criticized the high cost to physicians personally. I think those account? of drugs, and you recently argued paymentsand they’ve been shown No, it’s not true at all! I’m on Twitter, that industry estimates of the cost todo affect our perception of prod •VPplenarysesh. I believe that there of bringing a new drug to market ucts. If you’re receiving a lot of money are a number of inaccuracies in the are wildly exaggerated. What does it from a manufacturer, you may not view Cancer Letter stories about me. I’ll save really cost? their product as impartially as you that for another day.

ND18_illness_Q_A_Prasad.indd 55 10/3/18 11:11 AM 56 Illness

A psychiatrist i n e v e r y pocket

BY Our RACHEL METZ obsession PHOTOGRAPHY BY with JESSICA CHOU smartphones There are about 45 million people in the may US alone with a mental illness, and those actually illnesses and their courses of treatment can vary tremendously. But there is something be most of those people have in common: a a smartphone. A startup founded in Palo Alto, boon California, by a trio of doctors, including for the former director of the US National Institute of Mental Health, is trying to treating prove that our obsession with the tech disorders nology in our pockets can help treat some of today’s most intractable medical prob like lems: depression, schizophrenia, bipolar depression disorder, post traumatic stress disorder, and substance abuse. and Mindstrong Health is using a smart schizophrenia. phone app to collect measures of people’s

ND18_illness_Mindstrong_updated.indd 56 10/3/18 5:15 PM Mindstrong founder and CEO Paul Dagum

ND18_illness_Mindstrong_updated.indd 57 10/3/18 5:15 PM 58 Illness

lead to far more accurate ways to track these problems over time. If Mindstrong’s method works, it could be the first that manages to turn the technology in your pocket into the key to helping patients with a wide range of chronic brain dis orders and may even lead to ways to diagnose them before they start. Digital fi ngerprints

Before starting Mindstrong, Paul Dagum, its founder and CEO, paid for two Bay AreaŸbased studies to figure out whether there might be a systemic measure of cognitive ability or disability hidden in how we use our phones. One hundred and fifty research subjects came into a clinic and underwent a standardized neu rocognitive assessment that tested things like episodic memory how you remember events and executive function mental skills that include the ability to control impulses, manage time, and focus on a

Cofounder Tom Insel, a psychiatrist and former task the kinds of high order brain func director of the National Institute of Mental Health tions that are weakened in people with mental illnesses. The assessment included neuropsy cognition and emotional health as indi chological tests that have been used for cated by how they use their phones. Once decades, like a so called timed trail tracing a patient installs Mindstrong’s app, it test, where you have to connect scat monitors things like the way the person tered letters and numbers in the proper types, taps, and scrolls while using other order a way to measure how well peo apps. This data is encrypted and analyzed ple can shift between tasks. People who remotely using machine learning, and the have a brain disorder that weakens their results are shared with the patient and the attention may have a harder time with this. patient’s medical provider. app patients, too, can use it to message Subjects went home with an app that The seemingly mundane minutiae of their care provider . measured the ways they touched their how you interact with your phone offers For years now, countless companies phone’s display swipes, taps, and keyboard surprisingly important clues to your men have offered everything from app based typing , which Dagum hoped would be an tal health, according to Mindstrong’s therapy to games that help with mood unobtrusive way to log these same kinds research revealing, for example, a and anxiety to efforts to track smartphone of behavior on a smartphone. For the next relapse of depression. With details activities or voice and speech for signs of year, it ran in the background, gathering gleaned from the app, Mindstrong says, depression. But Mindstrong is different, data and sending it to a remote server. a patient’s doctor or other care manager because it’s considering how users’ phys Then the subjects came back for another gets an alert when something may be ical interactions with the phones not round of neurocognitive tests. amiss and can then check in with the what they do, but how they do it can As it turns out, the behaviors the patient by sending a message through the point to signs of mental illness. That may researchers measured can tell you a lot.

ND18_illness_Mindstrong_updated.indd 58 10/3/18 5:15 PM Psychiatrist in every pocket 59

B D 5 1 2 3

6 4 A C

The trail-tracing test requires subjects to connect scattered letters and numbers in the right sequence. E

“There were signals in there that were Mindstrong can first determine your base measuring, correlating predicting, in line by looking at how you use your handset fact, not just correlating with the neuro and combining those characteristics with F cognitive function measures that the neu general measures. Even when you’re just ropsychologist had taken,” Dagum says. using the smartphone’s keyboard, Dagum For instance, memory problems, which says, you’re switching your attention from are common hallmarks of brain disor one task to another all the time for exam Tom Insel, a psychiatrist who had spent 13 ders, can be spotted by looking at things ple, when you’re inserting punctuation years as director of the National Institute including how rapidly you type and what into a sentence. of Mental Health before he joined Verily errors you make such as how frequently He became convinced the connec in 2015. you delete characters, as well as by how tions presented a new way to investigate Verily was trying to figure out how fast you scroll down a list of contacts. human cognition and behavior over time, to use phones to learn about depression in a way that simply isn’t possible with or other mental health conditions. But typical treatment like regularly visiting Insel says that at first, what Dagum pre a therapist or getting a new medication, sented more a concept than a show of taking it for a month, and then checking actual data didn’t seem like a big deal. back in with a doctor. Braindisorder treat “The bells didn’t go off about what he had The assessment ment has stalled in part because doctors done,” he says. included classic simply don’t know that someone’s having Over several meetings, however, Insel trouble until it’s well advanced; Dagum realized that Dagum could do something neuropsychological believes Mindstrong can figure it out he believed nobody in the field of mental tests that have been much sooner and keep an eye on it 24 health had yet been able to accomplish. He used for decades, hours a day. had figured out smartphone signals that In 2016, Dagum visited Verily, correlated strongly with a person’s cogni like a so-called timed Alphabet’s life sciences company, where tive performance the kind of thing usu trail-tracing test. he pitched his work to a group including ally possible only through those lengthy

ND18_illness_Mindstrong_updated.indd 59 10/3/18 6:17 PM 60 Illness

lab tests. What’s more, he was collecting these signals for days, weeks, and months on end, making it possible, in essence, to 21.7% look at a person’s brain function contin male uously and objectively. “It’s like having a 18.3% continuous glucose monitor in the world overall 14.5% of diabetes,” Insel says. Why should anyone believe that what female Mindstrong is doing can actually work? Dagum says that thousands of people are using the app, and the company now has five years of clinical study data to confirm its science and technology. It is continuing to perform numerous studies, and this past March it began working with patients and doctors in clinics. In its current form, the Mindstrong app Prevalence of mental illness that patients see is fairly sparse. There’s a among US adults graph that updates daily with five different signals collected from your smartphone swipes and taps. Four of these signals are measures of cognition that are tightly tied 13.4 to mood disorders such as the ability to make goal based decisions , and the other measures emotions. There’s also an option 10.9 to chat with a clinician. For now, Insel says, the company is working mainly with seriously ill people 2005 2016 who are at risk of relapse for problems like depression, schizophrenia, and sub US suicide rates (per 100,000) stance abuse. “This is meant for the most severely disabled people, who are really needing some innovation,” he says. “There are people who are high utilizers of health care and they’re not getting the benefits, so we’ve got to figure out some way to anxiety major bipolar get them something that works better.” disorders depression disorder Actually predicting that a patient is headed

18.1% 6.9% 2.6%

Prevalence of mental illness among US adults, by diagnosis

Sources: National Institute of Mental Health; National Alliance on Mental Illness

ND18_illness_Mindstrong_updated.indd 60 10/3/18 5:15 PM Psychiatrist in every pocket 61

toward a downward spiral is a harder task, but Dagum believes that having more people using the app over time will help cement patterns in the data. There are thorny issues to consider, of course. Privacy, for one: while Mindstrong says it protects users’ data, collecting such data at all could be a scary pros  pect for many of the people it aims to help. Companies may be interested in, say, including it as part of an employee wellness plan, but most of us wouldn’t want our employers anywhere near our mental health data, no matter how well protected it may be. Spotting problems The Palo Alto startup wants before assessing your mental health they start to fit into your regular life. depressive disorder might share just one A study in the works at the University of numerous symptoms: they could both of Michigan is looking at whether feel depressed, but one might feel like Mindstrong may be beneficial for peo sleeping all the time, while the other is ple who do not have a mental illness but hardly sleeping at all. We don’t know do have a high risk for depression and how many different illnesses are in the suicide. Led by Srijan Sen, a professor of category of depression, Insel says. But psychiatry and neuroscience, the study Participants log their mood each day over time Mindstrong may be able to use tracks the moods of firstyear doctors and wear a Fitbit activity tracker to log patient data to find out. The company is across the country a group that is known sleep, activity, and heartrate data. About exploring how learning more about these to experience intense stress, frequent 1,500 of the 2,000 participants also let a distinctions might make it possible to sleep deprivation, and very high rates of Mindstrong keyboard app run on their tailor drug prescriptions for more effec depression. smartphones to collect data about the tive treatment. ways they type and figure out how their Insel says it’s not yet known if there cognition changes throughout the year. are specific digital markers of, say, audi Sen hypothesizes that people’s mem tory hallucinations that someone with ory patterns and thinking speed change schizophrenia might experience, and the We don’t know in subtle ways before they realize they’re company is still working on how to pre how many dižerent depressed. But he says he doesn’t know dict future problems like posttraumatic illnesses are in how long that lag will be, or what cog stress disorder. But he is confident that nitive patterns will be predictive of the phone will be the key to figuring it the category of depression. out discreetly. “We want to be able to do depression. Insel Insel also believes Mindstrong may this in a way that just fits into somebody’s hopes Mindstrong lead to more precise diagnoses than regular life,” he says. today’s often broadly defined mental can use patient data Rachel Metz is MIT Technology health disorders. Right now, for instance, Review’s senior editor covering the to find out. two people with a diagnosis of major cyborg beat.

ND18_illness_Mindstrong_updated.indd 61 10/3/18 6:17 PM 62 Illness

When I was younger I watched my mother’s decline, and eventual death, from Huntington’s disease. It was a ter rifying experience. So naturally when my wife and I learned that I had inher ited the mutation that causes the dis ease, we were devastated. We both understood the odds for any children we might have. As the child of a parent with Huntington’s, you either have the mutation and will surely develop the disease or you don’t, in Profi les in which case your family is free of it for ever. Since Huntington’s is a dominant disease, meaning you need only one copy of the mutated gene to fall ill, 50% of the precision children of people with Huntington’s face the same fate as their sick parent. After receiving my test results, my wife and I decided we’d never have bio medicine logical children. Years later we heard about a new pro cedure called preimplantation genetic Advances in DNA testing diagnosis, whereby embryos generated for in vitro fertilization can be screened and gene editing have given for Huntington’s. By implanting only people choices that would healthy embryos, we could reduce our child’s risk of inheriting the disease have been impossible a few from 50% to essentially zero. So we decades ago. Here, in their changed our minds. Our city had an IVF clinic that had own words, are the stories of recently established a pre implantation four people confronted with diagnosis program, but it hadn’t yet had a successful pregnancy as a result these dilemmas. of genetic screening. We would be its guinea pigs. We got lucky. In 2006 my wife gave birth to healthy twins who did not inherit the mutation. Now that more than 12 years have passed, I can say that using this technology to have healthy kids was one of the most powerful ways I’ve had to fight back against my diagnosis. We still don’t have any disease modifying treatments for Huntington’s, although I’m hopeful there will be in time for me. But either way, pre implantation genetic diagnosis gives parents affected by genetic diseases a cure for their kids, and for all future

generations. ALLEN IAN

ND18_illness_5_profiles.indd 62 10/4/18 10:16 AM Je Carroll ASSOCIATE PROFESSOR, Western Washington University Bellingham, Washington

ND18_illness_5_profiles.indd 63 10/3/18 7:21 PM ND18_illness_5_profiles.indd 64 64 only treat thesymptoms of cystic fibrosis but work paidoff. We now have drugsthatdon’t CF doesn’t have me.” I was wrong. lived my life with thephilosophy: “I have CF. function, Ihadto apply for disability. Ihad nologist noticedasignificantdrop inmy lung ule took atoll on my health. myWhen pulmo two different multimilliondollar companies. time I was 36, I’d beenastore manager for I endedupinretail management, and by the medical insurance trumpedacollege degree. to stop studying only becausemy needfor employed through college, andultimately had parents feared. Ibegan working at16, stayed $750,000 for drugresearch anddevelopment. Foundation andto date have raised more than They devoted their lives to theCystic Fibrosis never heard of cystic fibrosis, were inpanic. expectancy was 14 years. My parents, who had ulates how cellsprocess salt. The average life defective gene thatchanges aprotein thatreg with cystic fibrosis, adiseasecausedby a I was bornin1977. At age two I was diagnosed Lora Moser The good news was thatmy parents’ hard But thelonghours andinconsistent sched The diseasedidn’t get to measearly asmy CYSTIC FIBROSIS PATIENT, for many of us, those giftsare outof reach. and physical battle with an unknown outcome. 26% of my lungfunction. Eachday isamental antibiotics for three weeks in August. Ilost This landedmeinthehospital onintravenous been off thedrugsIneedfor eightmonths. ing andearly mortality causedby CF. these advances, sincethey canendthesuffer embryos without themutation. Isupport edit theCF mutationout of embryos, or select cies. It’s considered “double dipping.” hibited from beingonprivate medicalpoli insurance, sinceMedicare recipients are pro But becauseI’m onMedicare, Ican’t usehis insurance policy provided by hisemployer. a month. My husbandhasaprivate medical price of thedrugsIneedexceeds $38,000 ment isbeyond my family’s means. The retail as CFTR, causedby agenetic mutation. together to address amissingprotein, known These drugs, tezacaftor andivacaftor, work attack thediseaseitself atthemolecular level. Modern medicine gives usmany gifts. But But thoseadvances won’t helpme. I’ve I know we may becloseto beingableto The badnews isthatthecostof thistreat Illness Austin, Texas

PHIL KLINE 10/4/18 3:51PM Profiles 65

that way we could select embryos based his first steps, Elizabeth said something Othman Laraki on whether they carried the mutated striking: “I am so grateful for this child. gene. I wanted to ensure that our children This specific child. If we had gone through CEO, COLOR GENOMICS would be free of that cancer risk. embryo selection, it would have been a San Francisco, California Elizabeth was just coming off birth different child, who would not have been My grandmother died from breast cancer. control, so our doctor told us it was Kamal. This child who we love and adore My mother, who survived two breast can- unlikely that she would get pregnant for would not exist.” cers, got tested and discovered that she a while, and suggested we return for To me, that changed everything. I is a BRCA2 mutation carrier, explaining the testing in a few months. Thinking believe that this is a deeply personal our family’s history. I later got tested and nothing of it, we went back to life as choice—without a cosmically right or discovered that I, too, was a carrier of the usual. Almost immediately, Elizabeth got wrong answer. However, the thought that same mutation: 1466DelT. One typo with pregnant. our choice would have caused our beauti- consequences. At first I was distressed that we ful child to not exist convinced me that— I learned of this almost 15 years ago. were expecting a child without knowing at least for my BRCA2—we were willing to I’m now the CEO of a major genetics whether he or she carried my mutation. let fate call the shots. company. But I came to terms with it and put the Today, we have three wonderful boys: When my wife, Elizabeth, and I started question out of my mind—after all, there Kamal (five), Rami (four), and Zak (one). to discuss having children, I raised the

INTERMEYER wasn’t much I could do about it. All of them may or may not have the W issue of my mutation. We decided that Fast-forward a couple of years, and BRCA2 mutation, and we would have it no

GUTTER CREDIT HERE CREDIT GUTTERWINNI we’d do pre-implantation genetic testing: as we watched our first child, Kamal, take other way.

ND18_illness_5_profiles.indd 65 10/3/18 7:21 PM ND18_illness_5_profiles.indd 66 10/3/18 7:21 PM Profi les 67

Paula Amato, MD ASSOCIATE PROFESSOR Oregon Health & Science University, Portland, Oregon

Last year, I took part in an experiment wasting. The baby passed away at age that a lot of people think was ethically one. The mother subsequently com„ questionable. I was part of a team at pleted two physically burdensome Oregon Health & Science University rounds of IVF at a cost of tens of thou„ that used CRISPR gene editing to cor„ sands of dollars. She made a total of four rect a disease„causing gene mutation embryos, only one of which was chro„ in human embryos. In other words, we mosomally normal and unaffected with were “editing” humans. SMA. We transferred that embryo, but Why would we do this? With our unfortunately it did not take. work we were able to correct a mutation Such cases are not at all unusual. The in a gene called MYBPC3. This mutation type of gene editing we’re researching causes a deadly heart condition known would complement pre„implantation as hypertrophic cardiomyopathy. Our diagnosis by reducing the number of work was potentially a first step toward cycles of IVF required. It would relieve eliminating the disease from that family patients of the associated physical bur„ and all its descendants. den and costs. And it would rescue the Some people argue that we shouldn’t affected embryos. pursue our research, and that instead Another patient of mine, who car„ women should simply undergo a ried the BRCA gene mutation, which pre„implantation genetic diagnosis, increases the risk of breast and ovar„ which could identify any embryos with ian cancer, came to see me for IVF and the mutation before they’re implanted. pre„implantation genetic testing to This sentiment is most likely uttered by avoid passing on the gene to her chil„ people who have never treated an IVF dren. She was conflicted“if her parents patient. had made a similar choice, she wouldn’t One recent patient of mine and her be here today. And she was right. They husband easily conceived their first would have selected a different embryo. baby, who unfortunately was born with But what if instead, she Žthe very same a disease called spinal muscular atro„ person‘ could have been born, just with„ phy ŽSMA‘, a rare genetic neuromus„ out the BRCA mutation? cular disorder characterized by loss of That’s what our research promises. motor neurons and progressive muscle That’s why we’re doing it. IAN ALLEN IAN

ND18_illness_5_profiles.indd 67 10/3/18 7:21 PM 68 Next

Their kids, Matthew and Olivia were Designer babies aren’t futuristic. They’re already here. told, might not be as lucky. They would have a 5050 chance of inheriting the By Laura Hercher gene variant that causes dystonia and, if they did, a 30% chance of developing the disease. The risk of a severely affected child was fairly small, but not insignificant. My friends learned there was an alter native. They could undergo in vitro fertil Are we designing ization and have their embryos genetically tested while still in a laboratory dish. Using a technology called preimplantation genetic inequality into our testing, they could pick the embryos that had not inherited the DYT1 mutation. It would be expensiveƒcosts for IVF genes? in the US average over $20,000 for each try, and testing can add $10,000 or more. And it would require an unpleasant two week process of ovarian stimulation and egg harvesting. “It wasn’t the way I saw myself making a baby,” Olivia told me. But they wanted what the procedure could offer them: a guarantee that dystonia was elimi nated for the next generation, and beyond. Matthew and Olivia don’t think of them selves as having a “designer baby.” That term has negative associations, suggesting At first, Matthew assumed the weakness in his knee was something trivial, discretionary, or unethi cal. They weren’t choosing eye color or try the sort of orthopedic nuisance that happens when you ing to boost their kid’s SAT score. They were turn 30. It was weeks before he consulted a doctor, and looking out for the health and well being months before it occurred to him that there could be a of their future child, as parents should. connection between his worsening limp and a cousin’s Public opinion on the use of assisted shoulder problem when they were kids. DNA testing con reproductive technology consistently draws firmed it: Matthew, like his cousin, had a genetic form of Laura Hercher a distinction between preventing disease dystonia, a condition where muscles contract uncontrol is director and picking traits. The Johns Hopkins lably. Their grandfather most likely had dystonia as well. of research Genetics and Public Policy Center, which at the Sarah I’d met Matthew only a few months earlier, when he’d Lawrence College contacted over 6,000 people through sur married my friend’s daughter, Olivia, in one of those hip Program in Human veys and focus groups from 2002 to 2004, Genetics. old New York hotels with an elegant downtown vibe. summed up its findings this way: “In gen Since I was the only genetic counselor of their acquain eral, Americans approve of using reproduc tance, they brought their questions to me. With their tive genetic tests to prevent fatal childhood permission, I am sharing their story. I have changed disease, but do not approve of using the their names to preserve their privacy. same tests to identify or select for traits Matthew was lucky. His was a mild version of DYT1 like intelligence or strength.” The dystonia gene is in a gray zoneƒsome people born dystonia, and injections of Botox in his knee helped. with it live perfectly healthy livesƒyet But the genetic mutation can cause severe symptoms: presumably few parents would criticize contractures in joints or deformities in the spine. Many Matthew and Olivia’s choice to weed it out. patients are put on psychoactive medications, and some All embryo testing does fit the require surgery for deep brain stimulation. “designer” label in one important way, however: it is not available to everybody.

ND18_health_essay_designer_babies.indd 68 9/27/18 2:39 PM Essay 69

where coverage is mandated. Even policies that cover fertility treatment are inconsis tent in what they reimburse. Coverage for preimplantation genetic testing is down right Kafkaesque. Under many policies, testing the embryos is covered, but the IVF procedure itself is not, because the couples are not infertile. “The analogy I like to use,” says James Grifo, director of the Division of Reproductive Endocrinology and Infertility at NYU Langone Health, “is if you were having coronary bypass surgery and they didn’t pay for cracking the chest.” At least part of the reason the IVF indus try is growing is not that more people can afford it but that those who can are paying for new kinds of services. Egg banking, for example, is now aggressively marketed to younger women as an insurance policy against agerelated infertility. In 2011, egg banking did not even exist as a category in the CDC’s annual report on IVF; by 2016, storing eggs or embryos was the purpose of 25% of all IVF cycles. Elite companies like Facebook offer egg freezing as a perk, but for most people it remains a luxury. Cost isn’t the only barrier. Reproductive We risk creating a society where some groups, technology is less acceptable in racial, because of culture or geography or poverty, ethnic, and religious groups where being bear a greater burden of genetic disease. seen as infertile carries a stigma. Language barriers can reduce awareness and refer rals. Geography also plays a role, since IVF clinics cluster in areas of greatest demand. Presumably, many people would make Matthew and Olivia opted in to what with IVF clinics, Olivia’s own brother and the same decision as Matthew and Olivia if is a quiet but significant trend. Although his wife got news of a gene that increased given the option, but many don’t have that the number of couples using this tech risk for cancer in their kids. “If you could choice. Our discomfort around designer nology remains small, it is growing rap get rid of it, why wouldn’t you?” he asked. babies has always had to do with the fact idly. According to the Society for Assisted Cost was not a concern for these that it makes the playing field less levelƒ Reproductive Technology, the number of couples, but it is an obstacle for many taking existing inequities and turning US IVF attempts with singlegene testing Americans. The Centers for Disease them into something inborn. If the use of rose from 1,941 in 2014 to 3,271 in 2016, Control and Prevention £CDC¤ estimates preimplantation testing grows and we an increase of almost 70%. that 1.7% of babies born in the US today don’t address these disparities, we risk This is only the beginning. As the price are conceived using IVF. It’s much higher creating a society where some groups, of genetic testing of all kinds drops, more in countries that publicly fund assisted because of culture or geography or poverty, adults are learning about their genetic reproductive technology: 4% in Belgium, bear a greater burden of genetic disease. makeup as part of routine medical care and 5.9% in Denmark. A 2009 study found What could change society more pro discovering specific genetic risks before that 76% of the medical need for assisted foundly than to take genetic diseaseƒ pregnancy. But these people are still most reproduction in the US is unmet. something that has always epitomized our likely to be affluent and educated, like Insurance doesn’t normally cover IVF shared humanityƒand turn it into some

BENEDIKT LUFT BENEDIKT Olivia and Matthew. While they consulted in the US, except for a handful of states thing that only happens to some people?

ND18_health_essay_designer_babies.indd 69 9/27/18 2:39 PM B:8.4375” T:8.1875” S:7.1875” B:10.75” T:10.5” S:10”

QUAL.indd 1 5/31/18 5:35 PM

Ad #: M18QUAL003_04 Bleed: 8.4375” x 10.75” AD: None Headline: Foundational inventions... Trim: 8.1875” x 10.5” CW: None Visual: Copy Live: 7.1875” x 10” CD: None Space/Color: Magazine CMYK Gutter: None AP: -- Publication: MIT EmTech PP: None Project #: PR10665_NYC-CWW Document Name: CNY_M18QUAL003_04_01.indd AQR Retouch #: None Document Path: CPNY:ME Production:Qualcomm Snapdragon:Qualcomm Snapdragon Production:Magazine:M- Studio QA Client: Qualcomm Technologies 18QUAL003:CNY_M18QUAL003_04_01.indd Division: None Font Family: None Proofreader Product: QUALCOMM Ink Name: Cyan, Magenta, Black Job #: 10881834-1252-F0 Link Name: 11_08_2017 QCLogo_INVENTING TAG BLUE_OL.eps (Up to Date; 46.35%) Art Director Print/Export Time: 1-26-2018 6:48 PM CopyWriter Print Scale: 100% User Name: Joe.Congo Acct Exec lug Name: CRAFT MagNwp S Proof #: 1 Release QA PM: Tennille Martin InDesign Version: CC 2018 Print Prod The precision medicine issue 71

What about death?

In the age of big data, death isn’t an end, but more like a ... reformatting. Why not do some consulting work after you die? Or stick around to nag the grandkids? But beware—they might do things with your digital self that you don’t like. Take an epigenetic test and put your date of demise on your calendar. Meanwhile, some new drugs might keep you young until then.

ND18_front_death.indd3 71 10/3/18 4:46 PM ND18_death_augmented_realities.indd 72 10/3/18 1:43 PM 73

Never let me go

YOU’  E DEFINITELY TEMPO AY. BUT A DIGITALLY ENHANCED VE  SION OF YOU DOESN’T HAVE TO BE.

ossein Rahnama knows a CEO of a major financial company who wants to live on after he’s dead, and Rahnama thinks he can help him do it. Rahnama is creating a digital avatar for the HCEO that they both hope could serve as a virtual “consultant” when the actual CEO is gone. Some future company exec utive deciding whether to accept an acquisition bid might pull out her cell phone, open a chat window, and pose the question to the late CEO. The digital avatar, created by an artificialintelligence platform that analyzes personal data and correspondence, might detect that the CEO had a bad relationship with the acquiring company’s execs. “I’m not a fan of that company’s leadership,” the avatar might say, and the screen would go red to indicate disapproval.

BY COU TNEY HUMPH  IES / PHOTOG  APH BY TONY LUONG

ND18_death_augmented_realities.indd 73 10/3/18 1:43 PM 74 Death

Creepy? Maybe, but Rahnama believes capture the personality of a specific per into a neural network built with Google’s we’ll come to embrace the digital afterlife. son. There’s no software that can interact, opensource machinelearning framework, An entrepreneur and researcher based communicate, and make decisions the way TensorFlow. The bot was, by Kuyda’s own at Ryerson University in Toronto, and a you do. Rahnama says the CEO’s avatar admission, not very precise or polished, visiting faculty member at MIT’s Media will be a “decision support tool,” but it but when it answered questions, it often Lab, he’s building an application called won’t be capable of running the company. sounded uncannily like her friend. Augmented Eternity; it lets you create “There is one thing that is missing in Kuyda says the main complication with a digital persona that can interact with AI today, and that is context,” he says. trying to create digital versions of the dead people on your behalf after you’re dead. Most chatbots simply offer responses is that people are complicated. “We’re While most older people haven’t based on the content of a conversation, extremely different when we talk to dif amassed enough digital detritus to build but our communication changes depend ferent people,” she says. “We’re basically a working artificial intelligence, Rahnama ing on who we’re talking to, where we are, like twenty thousand personalities at once.” posits that in the next few decades, as we and what time of day it is. The need to For example, Mazurenko had said things to continue to create our digital footprints, include this kind of context was the basis her that he might have left out of a conver millennials will have generated enough for Rahnama’s company, Flybits ƒfor which sation with his parents. She could consult data to make it feasible. Even as we speak, he was named one of this publication’s with his family and other friends to figure the digital remains of the dead accumu 35 Innovators Under 35 in 2012 . Flybits out which information was too sensitive late. Something like 1.7 million Facebook provides a platform that lets companies to share. Could any company realistically users pass away each year. Some online tailor their communications to customers do the same? accounts of the dead are deleted, while on the basis of contextual cues. A bank, for Rahnama obviously thinks so. He says others linger in perpetual silence. “We example, might offer different messages Augmented Eternity will take a step toward are generating gigabytes of data on a daily through its mobile app depending on your accommodating various personalities by basis,” Rahnama says. “We now have a lot purchase history, your calendar schedule, tailoring the conversation according to of data, we have a lot of processing power, or whether you’re walking or taking a train. context and letting users control what we have a lot of storage capability.” With The contextual part was something data is accessible to whom. So someday enough data about how you communicate Rahnama found useful when he started his daughter might consult with his dig and interact with others, machinelearning Augmented Eternity. If you’re going to con ital family persona, while a former stu algorithms can approximate your unique struct a digital self, it’s not enough to know dent could ask questions of his academic personality­or at least some part of it. that somebody said something. You have persona. He sees it as one way of leaving And what would the digital “you” look to know the context in which it was said­ a legacy­a way to keep contributing to like? Well, what do you want it to look was the person joking? Annoyed? Reacting society instead of fading to black. like? It might be a textbased chatbot like to today’s news? These same kinds of the CEO’s or an audio voice like Siri or a clues end up being crucial when piecing digitally edited video or a 3D animated together a digital personality, which is why character in a virtualreality environment. the Augmented Eternity platform takes It might be embedded in a humanoid robot. data from multiple sources­Facebook, Twitter, messaging apps, and others­and T’S NOIT JUS T analyzes it for context, emotional content, FOR T HE DEAD and semantics. But a digital avatar might also come in A similar concept grabbed headlines handy even when you’re still around. AI a few years ago when Russian software could help transform your professional WENTTY T HOUSAND developer Eugenia Kuyda created a chatbot expertise from a scattered written record PERSONALIT IES A T ONCE representation of her best friend, Roman to a representation of your knowledge that We’re not there quite yet. It’s hard enough Mazurenko, who died in late 2015. Kuyda people can interact with. A lawyer who to create software agents that can carry on made the bot by plugging Mazurenko’s charges hundreds of dollars an hour could a naturalsounding conversation, let alone personal messages with friends and family let people consult a digital avatar instead,

ND18_death_augmented_realities.indd 74 10/3/18 1:43 PM Augmented Eternity 75

for a much lower price. Celebrities, politi companies can use ƒor exploit our data. If cians, and other public figures could out digital remains are like “the informational source some of their public interaction to corpse of the deceased,” they write, they digital versions of themselves. AI would “may not be used solely as a means to an allow us to consult experts with whom end, such as profit, but regarded instead as we’d never be able to meet in real life. The an entity holding an inherent value.” ability to represent and share expertise, Rahnama says, “can actually contribute to new business models on the internet.” Rather than speaking with a generic Siri Creepy? or Alexa, you could ask an eminent scien OLD A BLACKH MIRROR tist, a politician, or a coworker. And why Maybe, UP T O NA T URE attend a business meeting when you could Just about every discussion of the digital send your avatar? afterlife, Öhman points out, mentions “Be Another startup, Eternime, based in but Right Back,” an episode of the British show Mountain View, California, offers to incor Black Mirror, in which a bereaved young porate your personal information into ahnama widow interacts with a digital avatar of her “an intelligent avatar that looks like you” late husband. Over the course of the epi and that will “live forever and allow other sode, she progresses from sending a few people in the future to access your mem believes hesitant texts to a chatbot to purchasing ories.” Its founder, Marius Ursache, has a lifelike robot in her husband’s image. been promoting the idea for years, and we’ll all What’s often overlooked in discussions more than 40,000 people have signed up about the show is the role of the company to Eternime’s waiting list, but the self that created the avatar. In real life, Öhman funded company has still launched only come to says, we should be skeptical of such com limited beta versions. Ursache thinks the panies. The power of the digital dead to problem is less technical than behavioral: embrace manipulate the living is enormous; who “People don’t invest much time in activi better to sell us a product than someone ties that will pay off in decades,” he says. we’ve loved and lost? Thus our digital Whether or not it takes off as a busi the digital representations might be more talkative, ness, Rahnama hopes Augmented Eternity pushy, and flattering than we are­and will start conversations about privacy and afterlife. if that’s what their makers think is best, data ownership. “The reason I like this who’s going to stop them? research project is that it addresses a lot In the Black Mirror episode, the avatar of key ethical questions around data sci periodically elicits more of the dead hus ence and AI,” he says. “Like, who is going band’s data and upsells his widow on more to own my information after I pass away?” expensive representations of him, until it In a paper published in Nature Human becomes so lifelike that she can’t “kill” it. Behavior earlier this year, ethicists Carl The rhetoric around immortal digital selves Öhman and Luciano Floridi from the focuses on our desire to be remembered. Oxford Internet Institute argue that we But wouldn’t most of us want our loved need an ethical framework for the bur ones to be able to let us go? geoning digital afterlife industry. Should we treat digital remains by the same code that Courtney Humphries is a freelance writer who covers science and museums use for human remains? Doing the environment for a variety of so would severely limit the ways in which publications.

ND18_death_augmented_realities.indd 75 10/3/18 1:43 PM 76 Death

Today there’s a debate as to whether How to make sure your loved ones can get into all your accounts. tech companies should put back doors in Or, alternatively—how to cover your tracks. their crypto technology so law enforcement

By Simson Garfinkel can get access to data on devices they seize during an investigation. Short of that, it’s easy to backdoor your encryption yourself: just write down your hard drive’s master password, put the paper in an envelope, and seal it. Do the same with your Bitcoin Six things to do wallet. Make sure it’s well hidden but in a location that’s known to your loved ones.

with your data #2 Sign up for Inactive Account Manager If you have a Gmail account, use Inactive Account Manager to specify an email before you die address that will be automatically notified three months after your Google account goes inactive. Google defines “activity” broadly: if you check Gmail, log in to a Google website, or perform a search with a Chrome browser that’s logged into your account, Google will assume you’re not dead. But when your digital heartbeat stops, this approach ensures that someone you trust can access your Gmail account, Google Photos, and other data.

What would happen to your digital estate if you died, #3 Download your medical records Your doctor is supposed to keep copies suddenly, before finishing this paragraph? Would your of your test results and other records, but survivors be able to find what you left behind? it’s a good idea to keep your own. Ask for There is nothing hypothetical about this for many peo copies and scan them. You might also be ple: the problem emerges, wholly formed, when tragedy able to get your records directly if your strikes. What’s worse, more than half of Americans don’t healthcare provider participates in the have a will, let alone one that’s up to date, according to Simson Garfinkel US government’s Blue Button Connector, a 2016 Gallup Poll. As a result, most survivors lack a is a science which lets you download PDF files for road map to the deceased’s assets Œphysical and digitalŽ writer living yourself and a special format for other in Arlington, or even, in some cases, the legal authority to proceed. Virginia, and healthcare providers Œshould you wish Fortunately, there are many things you can do now, coauthor of The to give it to themŽ. Computer Book: without a lawyer, to make things easier for your survivors. From the Abacus My elderly father keeps a copy of his to Artificial records on a USB stick that he carries with Intelligence, him at all times. It comes in handy when he #1 Build a back door 250 Milestones Fifteen years ago, if you died and your next of kin got in the History sees a specialist who might not have access of Computer to his primary care provider’s computer. your laptop, that person was pretty much guaranteed Science, access to your data. Then, in 2003, Apple introduced published Yes, there’s a risk the stick could fall into this November the wrong hands, but he’s decided that the full disk encryption, designed to protect your data from by Sterling risk of medical professionals not having a thief, but also keeping it out of the reach of your survi Milestones. access to his records is greater. vors. Cryptocurrencies pose a similar problem: if no one has access to your digital wallet, then any value there is #4 Use a password manager lost“there’s no Bitcoin central control to complain to. It used to be straightforward to identify the deceased’s accounts by waiting for bank

ND18_death_essay_data_after_you_die.indd 76 10/1/18 4:16 PM Essay 77

authorized individuals to request that the user’s account be either “memori€ alized” or removed. Be aware: memo€ rialized accounts can be managed by a legacy contact who has to be specified in advance , but that person can’t log into the Facebook account, remove or change past posts, or read private messages. In one famous case, parents of a 15€year€ old German girl who died after being hit by a subway train were unsuccessful in trying to force Facebook to open the girl’s account so that they, the parents, could determine if she had experienced cyber€bullying or depression, or if her death really was a tragic accident. Twitter’s policy is similar: after you die, a family member can contact the company and ask that your account be deleted, according to a help page on its website. Twitter will also, if requested, remove spe€ cific imagery or messages sent just before or after an individual’s death. But Twitter will not give family members access to a deceased user’s private messages. So if you’re storing something on Facebook that you’d like people to have access to after you’re gone, you should If you’re an avid Facebook or Twitter user, download that data regularly and store it take some time to read their data-after-death where your loved ones will have accessŒ policies. You might not like what you fi nd. for example, in Google Drive.

#6 Be careful what you wish for I gave much of this advice at a cyberse€ curity training seminar a few months ago, statements and tax bills to arrive by snail One way that couples can simply access and almost everybody in the room thought mail. These days, two thirds of Americans each other’s accounts is by sharing their I was crazy. The people thereŒmostly do their banking online according to a passwords. This is getting harder as web€ menŒsaid they’d never share their pass€ 2017 survey by the American Bankers sites implement two€factor authentication, words with their spouses. Association , and many people no longer but it’s still possible by registering multi€ And maybe they’ve got a point. Family receive paper statements. This signifi€ ple second factors like a FIDO Universal members should be careful about taking cantly increases the chance that your bank 2nd Factor device and giving one to each extraordinary measures to crack open these accounts or retirement accounts might partner. encrypted digital crypts, warns Ibrahim be declared “abandoned” in the event Baggili, associate professor of computer that you die. #5 Ponder the complexities of social media science at the University of New Haven So use a password manager like If you are an avid user of Facebook or and an expert in digital forensics. “This 1Password or LastPass. Now make sure Twitter, take some time to read their data€ person I knew died, and his wife managed that your spouse, or lawyer, or children, or after€death policies. You might not like to finally break into his e€mails and iPad parents, or somebody has some way to get what you find. and found all sorts of things about him that to your accounts so they can, for example, When Facebook is notified that one she did not want to know,” says Baggili. save any cherished photos or easily delete of its users has become medically inca€ “She really loved him, and it changed her

BENEDIKT LUFT BENEDIKT your accounts after you’re gone . pacitated or died, the company allows whole perspective on him.”

ND18_death_essay_data_after_you_die.indd 77 10/4/18 12:16 PM ND18 Subscriptions Spread 16x10.875 D2.indd 2 9/28/18 4:01 PM Stay Ahead. Stay Connected. Subscribe Today.

Get access to the latest in innovation, emerging technology, and the conversations shaping the world around you.

Print Unlimited Daily Magazine Online Access Newsletter 1 year (six bi- Complete Get the day’s monthly issues) access to top headlines of MIT Technology technologyreview. delivered Review print com with the right to your edition exploring stories, inbox with the very latest interviews and MIT Technology in technology and videos you won’t Review’s daily innovation. find anywhere email newsletter, else during your The Download. subscription term.

Subscribe today at technologyreview.com/now

ND18 Subscriptions Spread 16x10.875 D2.indd 3 9/28/18 4:01 PM Drop dead Close Tomorrow, : PM The kitchen Snooze

ND18_death_prediction.indd 80 10/3/18 12:29 PM 81

the ultimate unanswerable question we all face:It’s When will I die? If we knew, would we live differently? So far, science has been no more accurate at predicting life span than a $10 fortune teller. But that’s starting to change. The measures being developed will never get good enough to forecast an exact date or time of death, but insurance com­ panies are already finding them useful, as are hospitals and palliative care teams. Your life span is written “I would love to know when I’m going to in your DNA, and we’re die,” says Brian Chen, a researcher who is learning to read the code. chief science officer for Life Epigenetics, a company that services the insurance industry. “That would influence how I approach life.” The work still needs to be made more Want practical, and companies have to figure out the best uses for the data. Ethicists, mean­ while, worry about how people will cope to know with knowing the final secret of life. But like it or not, the death predictor is coming. when The clock teve Horvath, a UCLA biostatistician Swho grew up in Frankfurt, Germany, you’re describes himself as “very straight,” while his identical twin brother is gay. So he had a personal interest when, a few years ago, going a colleague asked him for help analyzing biological data from the saliva of twins with opposite sexual orientations. The to die? colleague was trying to detect chemical changes that would indicate whether cer­ By Karen Weintraub tain genes were turned on or off. The hypothesis was that these so­called epigenetic changes, which alter the activ­ ity of DNA but not the DNA sequence itself, might help explain why two peo­ ple with identical genes differ in this way. But Horvath found “zero signal” in the epigenetics of the twins’ saliva. Instead, what caught his attention was a power­ ful link between epigenetic changes and

ND18_death_prediction.indd 81 10/3/18 12:29 PM 82 Death

aging. “I was blown away by how strong gains or loses methyl chemical groups in muscle, too, to see whether exercise

the signal was,” he says. “I dropped most •CH3–. Horvath’s insight was to measure makes a bigger difference there. other projects in my lab and said: ‘This is these increases and decreases in methyla­ Horvath’s own clock is not inspiring. the future.’” tion, find the 300 to 500 changes that mat­ He was surprised in analyzing his urine Horvath became particularly intrigued ter most, and use those to make his clocks. to find that he was epigenetically tracking by how certain chemical changes to cyto­ His findings suggest that the speed of the five years older than his chronological sine“one of the four DNA bases, or “let­ clock is strongly influenced by underlying age. A few years later, he tested his blood ters” of the genetic code“make genes genes. He estimates that about 40% of the and was relieved to find the results more more or less active. Given someone’s ticking rate is determined by genetic inher­ in line with his years, but still, he says, “I actual age, looking for these changes in itance, and the rest by lifestyle and luck. would say I’m not blessed in terms of epi­ that person’s DNA can tell him whether Morgan Levine, who completed post­ genetic aging.” the person’s body is aging unusually fast doctoral research in Horvath’s lab and At age 50, he says his work is motivated or slowly. His team tested this epigenetic now runs her own lab at Yale, is starting to by self­interest““I’m as desperate as any­ clock on 13,000 blood samples collected compare an individual’s epigenetic profile one else to find ways of slowing aging.” decades ago, from people whose sub ­ with the profile of cells from the lining of But he also keeps in mind the social and sequent date of death was known. The results revealed that the clock can be used to predict mortality. Because most common diseases“can­ “After five years of research, there is nobody who disputes that epigenetics predicts life span.” cer, heart disease, Alzheimer’s“are dis­ eases of aging, the ticking of Horvath’s clock predicts how long someone will live a healthy umbilical cord. The more people financial costs of an aging population. “We and how much of that life will be free of deviate from that standard, the worse off need to find ways to keep people healthier these diseases •though it doesn’t foretell they are likely to be. She thinks she will longer,” he says. which ones people will get–. “After five eventually be able to compare various He hopes that refinements to his clock years of research, there is nobody who epigenetic age measures to predict even will soon make it precise enough to reflect disputes that epigenetics predicts life in childhood who is going to be at great­ changes in lifestyle and behavior. Investors span,” he says. est risk of which diseases“when it’s still and biotech companies are spending hun­ Aging eight or more years faster than early enough to change that future. “Your dreds of millions of dollars right now on your calendar age equates to twice the typ­ genes aren’t your fate, but even less so with drugs that might slow aging and defer dis­ ical risk of dying, while aging seven years things like epigenetics,” she says. “There ease. But how will we know what’s effec­ slower is associated with half the risk of definitely should be things we can do to tive? Those working on drug discovery death, Horvath says. His lab has developed delay aging if we can just figure out what can’t wait 50 years to find out. Horvath a new version that is such a precise life they are.” hopes his clock will provide the answers. span predictor they named it after the Grim A few likely contenders are totally Reaper: DNAm GrimAge. The epigenetic unsurprising. Eating a healthy diet includ­ The business of death prediction clock is more accurate the younger a person ing lots of vegetables and fish is associated ompanies like Reinsurance Group is. It’s especially inaccurate for the very old. with slower epigenetic aging. Feel older Cof America are already looking into “At this point, we don’t have any evi­ when you’re sleep deprived? It’s probably using the epigenetic clock to tweak and dence that it’s clinically useful, because not a coincidence. Horvath has shown personalize risk assessments for life insur­ there are big error bars,” Horvath says. that people with insomnia are more likely ance. Right now, rates are based largely on Besides, there’s no pill to reverse the to show accelerated epigenetic aging. demographics“people’s gender and age“ effects. But though it will never be per­ “Everything you associate with a healthy and a few health metrics, such as whether fectly accurate, Horvath and his clock are lifestyle does relate to the new biomark­ they smoke. The clock adds another useful getting closer than anyone else ever has ers in the expected way, which is a boring data point. to answering the question that hangs over result, but it’s scientifically very exciting,” Such personalization raises questions us all“and determining whether there is he says. about fairness. If your epigenetic clock is anything we can do to change the answer. More unexpectedly, he finds that reg­ ticking faster through no fault of your own, ular exercise won’t add much more than a should you be charged a higher rate for Slow the ticking few months to your life. But those measure­ life insurance? The Genetic Information s we age, the cytosine at hundreds of ments are only on the DNA in blood, and Nondiscrimination Act of 2008“known

Athousand of spots in our DNA either Horvath says he’d like to look at changes as GINA“protects against discrimination CASAREZ DAMON

ND18_death_prediction.indd 82 10/3/18 1:44 PM The death predictor 83

Horvath, 50, on the basis of genes. But it doesn’t address says his work epigenetics. is motivated by self-interest. There’s also the issue of privacy. Your “I’m as desperate likely life span or true biological age is as anyone else to find ways of information that many consider intensely slowing aging.” personal. For now, regulations and privacy policies don’t even consider the possibil­ ity of such information. But as the science quickly progresses, questions about how to use and protect this data will become ever more pressing. Can Horvath’s clock and other tech­ nologies being developed to predict death ever be accurate enough to be truly use­ ful? “I haven’t seen any of these purported predictive algorithms be precise in terms of timing of death“to the contrary,” says Diane Meier, a professor of geriatrics and “After five years of research, there is nobody who disputes that epigenetics predicts life span.” palliative medicine at the Icahn School of Medicine at Mount Sinai in New York City. “People live for a really long time with a very high burden of disease and frailty,” she says. Gal Salomon, CEO of Clew Medical, an Israeli company that uses artificial intelli­ gence to identify medical risks in hospi­ tals, says he initially resisted the idea of developing a death predictor, thinking it unethical. Then he realized that doctors could use the technology “to understand where we need to stop.” An algorithm Clew developed can help doctors and fam­ ily members make the decision to switch from aggressive to palliative care, he says, overruling the typical instinct to provide heroic live­saving measures. The system, which for the moment is used only in hos­ pitals, can also alert a family that the end is near, he says. Atul Butte, a professor at the University of California, San Francisco, who studies quality of care, says the jury is still out about whether this kind of machine learn­ ing from patterns of care actually provides better treatment. But there’s no doubt, he adds, that medical care is headed in that direction. “Five to 10 years from now, the health system that doesn’t use this data to improve their medical delivery is going to be deemed archaic,” he says.

Karen Weintraub is a freelance writer based in Cambridge, Massachusetts.

ND18_death_prediction.indd 83 10/3/18 1:44 PM 84 Death

Anti-aging pioneer Judith Campisi explains how a recent Finally, the breakthrough could ward off drug that keeps age-related disease.

By Stephen S. Hall you young Portraits by Christie Hemm Klok

Judith Campisi has been a leading figure Why should we suddenly get excited drugs, which eliminate senescent in the biology of aging since the early about antiaging drugs again? cells. That’s a pretty broad claim. 1990s, when her research on the basic There are now tools available to biomed­ If we think of aging as a driver for mul­ mechanisms of cancer revealed an unex­ ical scientists that simply didn’t exist tiple age­related pathologies, the idea pected finding‰that cells enter a phase when I was a graduate student or even would be that a new generation of phy­ known as senescence that prevents a postdoc. So we’re finally able to do sicians‰we call them geriatricians them from becoming cancerous. More experiments that were either considered today‰will take a much more holistic than 25 years later, the insight has led impossible in some cases or were just approach, and the interventions will to a new kind of drug that may slow or dreams 20 or 25 years ago. The other also be more holistic. That’s the idea‰ modestly reverse human aging. thing that has changed is that the field it would revolutionize the way we’re Campisi’s research is on the role of of senescence‰and the recognition that thinking about medicine nowadays. cellular senescence in cancer and other senescent cells can be such drivers of And just to remind you, 80% of patients age­related diseases. Senescent cells aging‰has finally gained acceptance. in the hospital receiving acute medi­ undergo a transition into a twilight state Whether those drugs will work in peo­ cal attention are over the age of 65. So where they are still active but no longer ple is still an open question. But the first the idea is that senolytics would be one dividing; research by Campisi and oth­ human trials are under way right now. weapon that geriatricians will have in ers showed that this was a strategy to their arsenal of weapons to treat aging derail incipient cancers, which are char­ How specifically does senescence holistically as opposed to one disease acterized by runaway cell division and contribute to aging? at a time. growth. But she and others also discov­ The correct way to think about senes­ ered that these senescent cells accumu­ cence is that it’s an evolutionary bal­ There is a debate about whether late as we grow older, secreting an array ancing act. It was selected for the good there’s a biological limit to the human of molecules that promote the tissue purpose of preventing cancer‰if ˜cells™ life span, about 115 years, or whether degradation associated with aging. don’t divide, ˜they™ can’t form a tumor. maximum life span could be extended In the past five years, this insight has It also optimizes tissue repair. But the as long as 130, possibly 150 years. led to the pursuit of a new class of drugs downside is if these cells persist, which What do you think? known as senolytics, which eliminate happens during aging, they can now At present, we simply don’t know senescent cells and, in animal experi­ become deleterious. Evolution doesn’t enough to know whether it will even be ments, restore more youthful charac­ care what happens to you after you’ve possible to extend maximum human life teristics. Campisi, a professor at the had your babies, so after around age 50, span. Average life span? No problem‰ Buck Institute for Research on Aging in there are no mechanisms that can effec­ it’s already been done. But maximum Novato, California, cofounded a com­ tively eliminate these cells in old age. life span? We just don’t know. pany called Unity Biotechnology in 2011, They tend to accumulate. So the idea If you look at C. elegans, a little which launched a human trial of its first became popular to think about eliminat­ worm, the world record for extending senolytic drug last July. ing them, and seeing if we can restore the life span of that animal is 10­fold. She recently discussed her work with tissues to a more youthful state. For humans that would be unbelievable, Stephen S. Hall, a journalist who has right? A thousand years. But if you go been following anti­aging work for more You’ve suggested that health care up the evolutionary scale just a little bit, than two decades. could be transformed by senolytic to the fruit fly Drosophila, it’s maybe

ND18_death_Q_A_Campisi.indd 84 10/2/18 4:48 PM ND18_death_Q_A_Campisi.indd 85 10/2/18 3:06 PM 86 Death

twofold. And then if you go to a mouse, most of the really high­profile papers extend its life span maybe 20%, some­ times 30%. So think about the difference between a mouse and a human. We’re something like 97% genetically identical, meaning we have the same genes. And yet there’s a 30­fold difference in our life span. So it seems to me that in order for evolution to evolve a 30­fold differ­ ence in life span with so few really clear genetic differences, evolution maybe had to tweak hundreds, if not thousands, of genes. It’s unlikely at the present time that we will find a single drug that’s going to be able to do what evolution did.

Some Silicon Valley enthusiasts have been saying that lifespan extension up to 500 or 1,000 years is feasible. Well, it’s religion. It’s not science. I mean, that’s all I can say. It’s based on belief, not based on any data. People are certainly welcome to believe whatever they want to believe. But it doesn’t make good shape! There are lots of people are hoping for, because that will be the it true! at her age who are confined to wheel­ kind of intervention that will be broadly chairs. That’s aging, and that’s terrifying. applicable and will be very desirable. You’ve frequently emphasized that I am optimistic that we’re on the cusp of The conflict is with those who think that aging is a complex process, and that understanding enough about that pro­ we’re going to live to be 200 or 300 or modifying it is not going to be quick or cess to be able to intervene. And that more years old. That’s not realistic at easy. Yet we all yearn for a solution. people like us, who are not at that point, this point. Again, don’t confuse aging and death. will benefit. I am optimistic that we will experience But we’re still going to die. I’ll Let’s say we are successful at slowing medical interventions that will extend‰ remind you of the mouse models, where down or reversing aging, or extend the buzzword now is “health span.” I we eliminate senescent cells. There’s a ing health span. Are there any social think what terrifies people‰certainly significant increase in median life span, or cultural impacts that you have con what terrifies me‰is watching, for but there’s no increase in maximum life cerns about? example, my mom, who is well into her span. In a way, the mice died health­ No. In my lifetime, the population of 90s. She’s losing cognitive function, she ier. I think that’s the goal, and I think the earth has not quite doubled, but it’s doesn’t walk as well‰and she’s in pretty that that’s what the venture capitalists getting there. That’s unsustainable. The truth of the matter is, not having people die is not going to add much to the pop­ ulation of the earth the way the current Aging is terrifying. I’m optimistic rate at which we’re producing new peo­ ple is ruining the earth. So I think that that we’re on the cusp of this is ridiculous. So I really don’t see a downside to understanding enough about it this. There are problems, but I don’t think extending health span is going to to be able to intervene. exacerbate those problems.

ND18_death_Q_A_Campisi.indd 86 10/2/18 4:48 PM 87

June 11–12, 2019 MIT Media Lab Cambridge, MA

AI and robotics are changing the future of work. Are you ready?

EmTechNext.com/2019

Statement of Ownership, Management and Circulation. (Required by U.S.C. 3685.) (1). Publication Title: MIT Technology Review, (2). Publication No. 535-940, (3). Filing Date: 9/19/2018, (4). Issue Frequency: Bi-monthly. (5). No. of Issues Published Annually: 6. (6) Annual Subscription Price: $29.95 (7). Complete Mailing Address of Known O”ce of Publication: One Main St, 13th Floor, Cambridge, MA 02142 (8). Complete Mailing Address of Headquarters or General Business O”ce of Publisher: One Main St, 13th Floor, Cambridge, MA 02142 (9). Full Names and Complete Mailing Addresses of Publisher, Editor, and Managing Editor. Publisher: Elizabeth Bramson-Boudreau, MIT Technology Review, One Main St, 13th Floor, Cambridge, MA 02142. Editor: Gideon Lichfield, MIT Technology Review, One Main St, 13th Floor, Cambridge, MA 02142. Managing Editor: Timothy Maher, MIT Technology Review, One Main St, 13th Floor, Cambridge, MA 02142. (10). Owner: Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02138. (11). Known Bondholders, Mortgagees, and Other Security Holders Owning or Holding 1 Percent or More of Total Amount of Bonds, Mortgages, or other Securities: None. (12). Tax Status: The purpose, function, and nonprofit status of this organization and the exempt status for federal income tax purposes: Has Not Changed During Preceding 12 months. (13). Publication Title: MIT Technology Review. (14) Issue Date for Circulation Data Below: July/August 2018. (15). Extent and Nature of Circulation: (a) Total No. of Copies (Net Press Run): Average No. Copies Each Issue During Preceding 12 Months: 207,334; No. Copies of Single Issue Published Nearest to Filing Date: 217,046. (b) Legitimate Paid and/or Requested Distribution (1) Mailed Outside-County Paid Subscriptions Stated on PS Form 3541: Average No. Copies Each Issue During Preceding 12 Months: 154,454, No. Copies of Single Issue Published Nearest to Filing Date: 154,722. (2) In-County Paid/Requested Mail Subscriptions Stated on PS Form 3541: Average No. Copies Each Issue During Preceding 12 Months: 0, No. Copies of Single Issue Published Nearest to Filing Date: 0. (3) Sales Through Dealers and Carriers, Street Vendors, Counter Sales, and Other Paid Distribution Outside the USPS: Average No. Copies Each Issue During Preceding 12 Months: 11,572, No. Copies of Single Issue Published Nearest to Filing Date: 9,936. (4) Paid Distribution by Other Mail Classes Through the USPS: Average No. Copies Each Issue During Preceding 12 Months: 0, No. Copies of Single Issue Published Nearest to Filing Date: 0. (c) Total Paid and/or Requested Distribution: Average No. Copies Each Issue During Preceding 12 Months: 167,026, No. Copies of Single Issue Published Nearest to Filing Date: 164,658. (d) Nonrequested Distribution: (1) Outside-County Nonrequested Copies included on PS Form 3541: Average No. Copies Each Issue During Preceding 12 Months: 705, No. Copies of Single Issue Published Nearest to Filing Date: 780. (2) In-County Nonrequested Copies included on PS Form 3541: Average No. Copies Each Issue During Preceding 12 Months: 0, No. Copies of Single Issue Published Nearest to Filing Date: 0. (3) Nonrequested Copies Distributed Through the USPS by Other Classes of Mail: 0 Average No. Copies Each Issue During Preceding 12 Months: 0, No. Copies of Single Issue Published Nearest to Filing Date: 0 (4) Nonrequested Copies Distributed Outside the Mail: Average No. Copies Each Issue During Preceding 12 Months: 4,581, No. Copies of Single Issue Published Nearest to Filing Date: 6,203. (e) Total Free or Nominal Rate Distribution: Average No. Copies Each Issue During Preceding 12 Months: 5,286, No. Copies of Single Issue Published Nearest to Filing Date: 6,983. (f) Total Distribution: Average No. Copies Each Issue During Preceding 12 Months: 172,313, No. Copies of Single Issue Published Nearest to Filing Date: 171,641. (g) Copies not Distributed: Average No. Copies Each Issue During Preceding 12 Months: 48,671, No. Copies of Single Issue Published Nearest to Filing Date: 59,657. (h) Total: Average No. Copies Each Issue During Preceding 12 Months: 220,984, No. Copies of Single Issue Published Nearest to Filing Date: 231,298. (i) Percent Paid: Average No. Copies Each Issue During Preceding 12 Months: 96.9%, No. Copies of Single Issue Published Nearest to Filing Date: 95.6%. This Statement of Ownership will be printed in the November/December issue of this publication. I certify that all the information furnished on this form is true and complete. Electronic Copy Circulation: Requested and Paid Electronic Copies: Average No. Copies Each Issue During Preceding 12 Months: 13,651 No. Copies of Single Issue Published Nearest to Filing Date: 14,252 Total requested and paid copies: Average No. Copies Each Issue During Preceding 12 Months: 167,026 No. Copies of Single Issue Published Nearest to Filing Date: 164,658. Total Requested Copy Distribution (Line 15F) + Requested Paid Electronic Copies: Average No. Copies Each Issue During Preceding 12 Months: 172,312 No. Copies of Single Issue Published Nearest to Filing Date: 171,641. Percent Paid and /or Requested (16B divided by 16c x 100): Average No. Copies Each Issue During Preceding 12 Months: 96.9% No. Copies of Single Issue Published Nearest to Filing Date: 95.9%. Doreen Adger, SVP of Consumer Revenues and Marketing.

ND18_death_Q_A_Campisi.indd 87 10/3/18 9:11 AM 88 The back page

Genes I wish By Sarah Cooper they would fi nd

Sarah Cooper is a writer, comedian, and creator of the satirical blog TheCooperReview.com.

MIT Technology Review (ISSN 1099-274X), November/December 2018 issue, Reg. US Patent Office, is published bimonthly by MIT Technology Review, 1 Main St. Suite 13, Cambridge, MA 02142-1517. Entire contents ©2018. The editors seek diverse views, and authors’ opinions do not represent the official policies of their institutions or those of MIT. Periodicals postage paid at Boston, MA, and additional mailing offices. Postmaster: Send address changes to MIT Technology Review, Subscriber Services, PO Box 5001, Big Sandy, TX 75755, or via the internet at www.technologyreview.com/customerservice. Basic subscription rates: $59.94 per year within the United States; in all other countries, US$69. Publication Mail Agreement Number 40621028. Send undeliverable Canadian copies to PO Box 1051, Fort Erie, ON L2A 6C7. Printed in USA. Audited by the Alliance for Audited Media.

ND18_death_backpage.indd 88 10/1/18 4:17 PM Blockchain. Hype? Hope? The future is in between. The future is here.

May 2, 2019 MIT Media Lab Cambridge, MA

technologyreview.com/blockchain2019

Blockchain Ad 8x10.875 D2.indd 1 9/27/18 1:40 PM THE LEXUS LS 500. LIVE IN THE NEW. RARE We didn’t merely try something different. We crafted an entirely original experience. With accents like Kiriko glass, designed by hand then delicately etched with thousands of cuts to catch the eye and transform in the light. A 416-horsepower1 twin-turbo engine paired with a 10-speed FORM Direct-Shift automatic transmission takes you from 0 to 60 in just 4.6 seconds.1,2 All this is complemented by cutting-edge technology with one of the largest Head-Up Displays in the industry.3 The Lexus LS 500 isn’t simply unique. It redefines what a flagship luxury sedan can be. lexus.com/LS | #LexusLS

Options shown. 1. Ratings achieved using the required premium unleaded gasoline with an octane rating of 91 or higher. If premium fuel is not used, performance will decrease. 2. 2019 LS vs. 2018/2019 competitors. Information from manufacturers’ websites as of 5/21/2018. 3. Performance figures are for comparison only and were obtained with prototype vehicles by professional drivers using special safety equipment and procedures. Do not attempt. ©2018 Lexus

TECH Ad.indt 1 10/2/18 4:03 PM

JOB DESCRIPTION IDEA TEAM Client: Lexus Account Executive: Chris Loughlin Headline: Project Manager: Alix Knight Job Number: 420LEXLS-P93865 Project Manager: Shayna Fung Altair Number: Producer 1: David Voliva Bill to Job Number: 420LEXCO-P93685 Art Studio: Veronica Osborn Version: A Ad Unit: Publication(s): Vanity Fair, MIT Technology Review, Surface, Lifestyles Magazine Shot List: LEX-LSH-MY18-0079, LEX-LSG- MY18-0041.04 Bleed size: 8.375 x 11.125 Trim size: 8 x 10.875 Safety size: 7 x 9.875 Gutter: Vendor: Orora Visual # of Colors: 4/c Insertion Date:

Client - Team One Job # - 150617_001 Ver. - AD01A Cyan Magenta Yellow Black