The Economist May 7th 2022 TechnologyQuarterlyThequantifiedself 7
Killerapps, saving lives
I
t is a trendlinethat gives you pause: a spikenearingthetopof
the blue band that marks your normal bloodsugarrange,fol
lowed by a sharp dip. The culprit is obvious:thesourdoughtoast
for breakfast half an hour before. A generouspieceofbanana
bread the following morning leaves a completelydifferenttrace.
The bloodsugar line runs nicely unperturbedrightdownthemid
dle of the normal zone all the way through tolunchtime.
These are the sort of revelations thatacontinuousglucose
monitor, a coinsized device attached to theskin,beamstoyour
smartphone’s screen. The device lasts for twoweeksandhasa tiny
needle that gets just under the skin. Everyfewminutesitmea
sures the concentration of sugar in the fluidbetweenthecells
there—a good proxy for what is going on inthebloodstream.
Almost as soon as the first such continuousglucosemonitors
started replacing fingerprick blood testsfordiabeticsin2014,
they began to show up on the arms of nondiabeticgeeksinSili
con Valley. They were looking for ways to “hack”theirmetabolism
into delivering, for instance, more energyorbrainclarity.Their
adhoc experiments were soon replicatedbyformalresearchers
doing broader studies of metabolism. Those,inturn,haveledto
the possibility of personalised nutrition.
Such metabolic studies have changedthescientificthinking
on what a healthy diet looks like. It hasturnedoutthatmany
seemingly healthy people often have largepostmealspikesin
blood sugar, which have been linked to the developmentofpredi
abetes. Without any intervention prediabetesturnsintodiabetes
in 3770% of cases within four years. The sugardipsthatoftenfol
low the spikes were recently found to be problematic,too,because
they make people feel hungry. People whoare“bigdippers”con
sume about 300 more calories a day than thosewhoarenot.
In 2015 researchers in Israel showed thatanaibasedalgorithm
they had developed could predict someone’sbloodsugarreaction
to various foods. The algorithm’s inputsincludedbloodtests,
sleep, exercise, height and weight, which allaffectdailymetabolic
variations. They also included the compositionofthegutmicro
biome, the trillions of bacteria residing in thegutwhosecollective
job is to process what we eat. Microbiome analysisisdonebyshot
gun genomic sequencing of everything foundina stoolsample.
In the past five years startups in Ameri
ca, Europe and Asia have launched ai
based personalisednutrition apps that
build on these discoveries. One of them,
Zoe, sends customers a set of specially for
mulated muffins. By knowing exactly what
is in the food being eaten, and measuring
the changes in blood sugar and fat that
come about in response, the company can
create a predictive model of its customers’
metabolism. Its algorithm then whips up a
bespoke catalogue of foods and meals,
with predicted bloodsugar reactions to
each. Tushar Vashisht, cofounder of
HealthifyMe, an Indian startup that pro
vides digital coaching for weight loss, says
the trove of data from customers who can
afford various connected devices and
blood tests as inputs for their bespoke
plansisusefulfarbeyondthosecustomers.Itcanhelptobuild
appsforpeoplewhodonothavesuchdevices,whichwouldrely
onaiderivedproxiesofweight,bloodsugarandsoon.
Forsuchsystemstokeepmetabolismsinbalance,theyhaveto
beadheredto.Knowingwhatishappeninginsideyourbodyisno
helpifyoudonothingto changethepatternofbehaviourin
volved.Justbeingtoldthatitisinyourinterestsisnot,typically,
enough.Soappssoldaswaysofachievinghealthgainsontheba
sisofmeasurementsmadebywearablestypicallyincorporatea
varietyofbehaviouralnudgestokeeptheuserfocused.
Theaiinnovationinsuchpersonaliseddietsmakesthemeasi
ertomaintaininthelongtermbecauseit givespeopleoptionson
howtomakethefoodsthatthealgorithmsaysareparticularlybad
forthema littlelessbad.Thealgorithms
maysuggestsmalltweaks,suchassprin
klingsomenutsonthaticecreamorgoing
fora longwalkaftereatingit.Januaryai,
another personalisednutrition startup,
saysithasderivedthenutritioncontents
of16m grocerystoreitems, recipes and
menusoflocalrestaurants,whichmakesit
easierforuserstoplanandtrackmeals.
Itisstillearlydays,butresultsreported
byusersofsuchprecisionnutritionpro
grammeslookencouraging.Userssaythey
arelosingweight,havehigherenergylev
elsandaresleepingbetter.Somediabetic
usersnolongerneedmedication.Studies
ofseveralappsareunderwaytoconfirm
andquantifythesebenefits.
However,thoughtfulsuggestionsthat
makecomplianceeasieraretheexception.
Better measurement does little good unlesstheappsthat
interpret the data are well designed
Dealing with the data
App-trition
Health-care apps released by year*
Source:IQVIAInstitute *Available on Apple Store and Google Play
120
90
60
0
0
2010 2018161412
No longer available