RESEARCH ARTICLE
Beyond Wish ul Thinking: In eg a ing
Consume P e e ences in he Assessmen
o Die a y Recommenda ions
Xa ie I z
1☯
*, Pascal Le oy
2☯
, Vincen Réquilla
3☯
, Louis-Geo ges Sole
2☯
1Na u al Resou ces Ins i u e Finland (Luke), Economics & Socie y Resea ch Uni , Helsinki, Finland,
2Ins i u Na ional de la Reche che Ag onomique, INRA-ALISS, UR 1303, I y-su -Seine, F ance,
3Toulouse School o Economics, Uni e si é Toulouse Capi ole (INRA), Toulouse, F ance
☯These au ho s con ibu ed equally o his wo k.
*[email p o ec ed]
Abs ac
Con enience, as e, and p ices a e he main de e minan s o ood choices. Complying wi h
die a y ecommenda ions he e o e imposes a “ as e cos ”on consume s, po en ially hinde -
ing adop ion o hose ecommenda ions. The s udy p esen s and applies a new me hodol-
ogy, based on economic heo y, o quan i y his as e cos and assess he heal h and
wel a e e ec s o di e en die a y ecommenda ions. Then, by compa ison o hose e ec s,
we iden i y socially desi able ecommenda ions ha a e mos compa ible wi h consume
p e e ences (i.e., ha bes balance heal h bene i s agains ” as e cos ”) and should be p io i-
ized o p omo ion. The me hodology p oceeds in h ee-s eps: i s , an economic-beha -
io al model simula es how whole die s would change i consume s complied wi h die a y
ecommenda ions; second, an epidemiological model es ima es he numbe o dea hs
a oided (DA) due o he die a y change; hi d, an e iciency analysis weighs he heal h ben-
e i s agains he as e and policy cos s o each ecommenda ion. The empi ical model is cal-
ib a ed using F ench da a. We ind ha ecommenda ions o educe consump ion o ed
mea and so -d inks, o aise consump ion o milk p oduc s and ish/sea ood impose ela-
i ely mode a e as e cos s. By compa ison, ecommenda ions ela ed o F&V consump ion
and, o a lesse ex en , bu e /c eam/cheese, snacks, and all mea s impose la ge as e
cos s on consume s. The F&V ecommenda ion is he cos lies o consume s o comply
wi h, bu i also educes die - ela ed mo ali y he mos , so ha a la ge budge could be allo-
ca ed o p omo ing F&V consump ion while keeping his policy cos -bene icial. We conclude
ha p omo ion o mos die a y ecommenda ions imp o es social wel a e. Ou amewo k
complemen s he p og amming models a ailable in nu i ion and public heal h: hose mod-
els a e bes used o iden i y die a y a ge s, ollowing which ou amewo k iden i ies cos -
bene icial ways o mo ing owa ds hose a ge s.
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 1/17
a11111
OPEN ACCESS
Ci a ion: I z X, Le oy P, Réquilla V, Sole L-G
(2016) Beyond Wish ul Thinking: In eg a ing
Consume P e e ences in he Assessmen o Die a y
Recommenda ions. PLoS ONE 11(6): e0158453.
doi:10.1371/jou nal.pone.0158453
Edi o : P a ibha V. Ne u ka , College o T opical
Ag icul u e and Human Resou ces, Uni e si y o
Hawaii, UNITED STATES
Recei ed: No embe 18, 2015
Accep ed: June 16, 2016
Published: June 30, 2016
Copy igh : © 2016 I z e al. This is an open access
a icle dis ibu ed unde he e ms o he C ea i e
Commons A ibu ion License, which pe mi s
un es ic ed use, dis ibu ion, and ep oduc ion in any
medium, p o ided he o iginal au ho and sou ce a e
c edi ed.
Da a A ailabili y S a emen : The ood in ake da a
INCA2 a e eely a ailable om he open da a
pla o m o he F ench go e nmen : (h ps://www.da a.
gou . / /da ase s/donnees-de-consomma ions-e -
habi udes-alimen ai es-de-le ude-inca-2-3/). The ood
consump ion da a a e owned by a hi d-pa y and can
be pu chased om: Kan a Wo ldpanel F ance
(pos al add ess: 2 ue F ancis Péd on, BP 3, 78241
Chambou cy Cedex, F ance; phone: +33 (0)1 30 74
80 80; websi e: www.kan a wo ldpanel.com).
Funding: This esea ch was pa ially unded by
p ojec s ANR-ALID-OCAD, FOODPOL-DIDIT, and
In oduc ion
Die modeling has been widely used in he las decades o assess nu i ional ecommenda ions
and die a y guidelines. Mainly based on Linea P og amming (LP), hose models aim a cha -
ac e izing op imal die s op imizing an objec i e unc ion (e.g., minimizing die cos ) subjec o
a se o cons ain s (e.g., nu i ional equi emen s). The main a ionale o he use o such mod-
els is ha hey help sol e complex p oblems ha a ise because indi iduals need nu ien s bu
ea oods, and, as nu ien s a e no e enly dis ibu ed in oods, he e exis s a la ge a ie y o
possible die a y pa e ns compa ible wi h a gi en se o nu i ional equi emen s [1].
LP models ha e been used o di e en pu poses in nu i ion and public heal h [1]: assess-
men o he di icul y o complying wi h and compa ibili y o a ious nu ien - and ood-based
ecommenda ions [2–5]; cha ac e iza ion o leas -cos die s mee ing a lis o nu i ional
equi emen s [6]; ansla ion o nu ien ecommenda ions in o ood plans [7].
An impo an d awback o LP models, howe e , is ha hey may p oduce un ealis ic die s
as hey ail o cap u e consume s’p e e ences [8]. Hence, Henson [6] o he UK and Con o i
and D’Amicis [9] o I aly ound ha i was possible o compose heal hy die s which cos only
20 o 30% o obse ed cos and we e composed o a e y small numbe o ood i ems. These
esul s imply ha ood choices a e no d i en solely, o e en mainly, by he sa is ac ion o
nu i ional needs [10,11] bu ha many o he conside a ions come in o play. Thus, models
based only on nu i ional aspec s may p oduce die s ha a e incompa ible wi h consume p e -
e ences, which ob iously co e many o he cha ac e is ics o ood p oduc s and die s.
These p oblems ha e been ecognized in he esea ch li e a u e and pa ially add essed
h ough he addi ion o pala abili y and social accep abili y cons ain s. A eal p og ess has
been o conside , as a s a ing poin o he analysis, he sel -selec ed die s obse ed in ep esen-
a i e samples o he popula ion, and use objec i e unc ions such as he minimiza ion o
depa u e om hose cu en ly obse ed die a y pa e ns [12–16]. This app oach elies on he
idea ha obse ed ood choices e eal consume p e e ences and he unde lying ade-o s, o
ins ance be ween pala abili y and cos s, which in luence consume s’decisions [17]. P e e ences
changing only slowly, la ge depa u e om obse ed die s seem un ealis ic [14].
Howe e , e en wi hin his imp o ed amewo k, he objec i e unc ion o he p og amming
p oblem emains a bi a y, which implies ha he subs i u ion possibili ies among oods a e
exogenously de ined by he modele and nei he heo e ically no empi ically g ounded. Thus,
we a gue ha he p og amming models used o da e in nu i ion and public heal h o assess
die a y guidelines do no in eg a e consume p e e ences in a sa is ac o y manne , and ha
such a limi a ion has impo an consequences. Fi s , he e is li le eason o belie e ha he sub-
s i u ions among oods simula ed by hose models ( o ins ance, he change in whole die
induced by a ise in ui s and ege ables (F&V) consump ion) e lec he beha io al adjus -
men s ha would be made by” eal”consume s. Second, LP models do no allow es ima ion o
he ull cos o die a y changes induced by he adop ion o ecommenda ions. This cos is no
only inancial, i.e. due o he change in expendi u e o adop he op imized die , as i also
includes he loss o well-being (o u ili y) c ea ed, a leas in he sho e m, by he decision o
comply wi h new die a y guidelines by consuming less p e e ed oods (”I decide o ea mo e
b occoli o comply wi h he ‘5-a-day‘ ule, al hough I do no like b occoli much”). This sho -
e m loss o hedonic ewa ds om a die a y adjus men is hence o h e e ed o as a as e cos .
Thi d, as hey do no pe mi calcula ion o as e cos s, LP models canno suppo he no ma i e
analysis o die a y ecommenda ions, o ins ance by applying he cos -bene i o cos -e ec-
i eness echniques widely used o assess he social desi abili y o ea men s, d ugs and o he
componen s o heal h ca e sys ems. This p e en s a anking o al e na i e ecommenda ions,
as well as an assessmen o he social desi abili y o in es ing mo e o less in hei p omo ion.
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 2/17
E aNe SUSDIET (Founda ion Daniel and Nina
Ca asso) in F ance, p ojec s Vas uu oo umi and
E aNe SUSDIET (MAKERA und) in Finland, as well
as EU Ho izon 2020 P og amme unde G an
Ag eemen n° 633692 (SUSFANS). The unde s had
no ole in s udy design, da a collec ion and analysis,
decision o publish, o p epa a ion o he manusc ip .
Compe ing In e es s: The au ho s ha e decla ed
ha no compe ing in e es s exis .
O e coming hose limi a ions appea s o be a ele an challenge o nu i ion and public
heal h. Many s udies show ha s anda d die a y ecommenda ions, o ins ance ela ed o
F&V, so d inks, o snacks, a e poo ly adop ed in many coun ies, especially among disad an-
aged and less educa ed people [18–21]. This low le el o compliance wi h nu i ional ecom-
menda ions migh be explained by ood p ices and budge a y cons ain s [22], o nu i ion
knowledge [23], bu we a gue ha i is also, and pe haps mainly, due o he cos ha compli-
ance imposes on consume s in e ms o as e and con enience.
Hence, he goal o his a icle is o p opose a new modeling app oach o quan i y his as e
cos and iden i y die a y ecommenda ions compa ible, as much as possible, wi h consume s’
p e e ences. In e ec , ou con ibu ion add esses, a leas pa ially, he challenge ecen ly o -
mula ed by Webb and By d-B edbeene [11] o o e come consume ine ia o die a y guidance
by “gi [ing] consume s con ol wi h nu i ion messages ha a e ealis ic, posi i e, easy o
unde s and, and ac ionable wi hou an expec a ion ha consume s will su ende oods hey
lo e”. Ou die modeling me hod is based on he economic heo y o he consume and pa am-
e e ized using mic o-le el da a on eal ood pu chases in F ance. As i is pa amoun o p opose
simple messages o consume s, we simula e and compa e he impac s o di e en ood-based
ecommenda ions. Finally, we pe o m a cos -bene i analysis o es ablish how di e en ecom-
menda ions should be p io i ized.
Me hods and Da a
The o mula ion and p omo ion o die a y ecommenda ions emains he mos popula policy
ins umen o induce consume s o make heal hie ood choices [24]. Ye he e ec o such ec-
ommenda ions and hei e ec on social wel a e is di icul o iden i y ex-an e. Ou esea ch
seeks o de elop and apply a ool o ill his knowledge gap [25]. To do so, we ha e de eloped a
h ee-s ep me hodology:
1. An economic-beha io al model p edic s how whole die s would change i consume s com-
plied wi h a gi en ecommenda ion; o example o consume mo e F&V.
2. An epidemiological model es ima es he numbe o dea hs a oided (DA) due o he die a y
change.
3. A cos -bene i assessmen o he ecommenda ion is ca ied ou by balancing he as e cos
o compliance wi h he ecommenda ion and he policy cos o inducing consume s o
change hei beha iou s agains he mone ized alue o he heal h bene i om imp o ed
die s.
The economic model o die choice
The s anda d economic heo y o consume beha iou assumes ha an indi idual chooses he
amoun s o goods she is going o consume in o de o maximise a unc ion–named”u ili y”—
subjec o a budge cons ain . The u ili y unc ion desc ibes he p e e ences o a consume
which, in he case o ood choices, ela e o he as e o he goods, hei con enience, and many
o he a ibu es. The budge cons ain akes in o accoun he p ices o goods and a ailable
income. This op imisa ion p og am is e e ed o as he”nu i ionally uncons ained p oblem”.
I s solu ion, in he case o ood choices, de ines which goods a e ea en and in which quan i ies.
In his con ex , he adop ion o a nu i ional ecommenda ion, such as ea ing a minimum
quan i y o F&V pe day, is concep ualized as he in eg a ion o an addi ional cons ain in he
p e ious p og am. The addi ional cons ain leads he consume o modi y he choices in
o de o comply wi h his new cons ain and hus choose a modi ied se o goods (o he same
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 3/17
goods bu in di e en quan i ies). We call his new op imisa ion p og am he”nu i ionally
cons ained p oblem”. Compa ison o he solu ions o hose wo p og ams p o ides wo key
esul s:
•Fi s , he impac o he adop ion o a nu i ional ecommenda ion on he en i e die , and
hence a ull cha ac e iza ion o he subs i u ions among oods ha he ecommenda ion has
induced.
•Second, an es ima e o he loss o u ili y, o as e cos , ha he consume incu ed in he sho
e m by adop ing he nu i ional ecommenda ion. Adop ion always educes u ili y o he
consume because, i i was no he case, he consume would ha e complied wi h he ecom-
menda ion in he uncons ained si ua ion.
An empi ical di icul y a ises because he u ili y unc ion is no obse ed. Howe e , assum-
ing a ional beha io as is s anda d in mos analyses o consume choices, obse ed consump-
ion, gi en ma ke p ices and income, is jus he solu ion o he”nu i ionally uncons ained
p og am”. This p ope y is used o in e p e e ences om ac ual consump ion da a. Once he
u ili y has hus been e ealed and summa ized in he o m o p ice and income elas ici ies, eco-
nomic heo y is used o de e mine how a nu i ional cons ain a ec s choices and u ili y. To
ge he in ui ion o how he model wo ks, le ’s imagine a change in p ices. We know ha he
a ional consume would adjus he consump ion as a esul o hose p ice a ia ions, and ha
his esponse is empi ically quan i iable. When we simula e changes in consump ion induced
by compliance wi h a nu i ional ecommenda ion, p ices a e held cons an , bu we all back
on he economic amewo k by in oducing “shadow p ices”, de ined as he se o p ices ha
would ha e o p e ail o he nu i ionally-uncons ained indi idual o choose he exac same
bundle o goods as he nu i ionally-cons ained indi idual. In o he wo ds, i he shadow
p ices we e he cu en ma ke p ices, he consume would spon aneously con o m o he
nu i ional ecommenda ion. Fo ins ance, he shadow p ice o ed mea o he ecommenda-
ion o dec ease ed mea consump ion by 5% is he inc ease in he p ice o ed mea ha
would be necessa y o consume s o adop he ecommenda ion, all else being held cons an .
The di e ence be ween ma ke p ice and shadow p ice would in ha case be nega i e and akin
o a p ice ax ha would discou age consump ion o ed mea . Mo eo e , he shadow p ice di -
e s om he ma ke p ice o any ood i em which con ains ed mea (e.g. eady meals). How-
e e , because consume s subs i u e oods wi h one ano he , he op imal solu ion ha
minimizes he u ili y cos o complying wi h he ecommenda ion also in ol es changes in all
he oods ha may subs i u e o complemen ed mea consump ion. The economic heo y
hen gene a es impo an insigh s:
•Fo any gi en ood, he wedge be ween shadow p ice and ma ke p ice is p opo ional o he
pe -uni nu i ional con en s (e.g. con en s in F&V) o he goods appea ing in he nu i ional
cons ain (e.g. an inc ease in F&V consump ion), and depends on demand elas ici ies.
Shadow p ices a e hus empi ically compu able.
•F om he se o shadow p ices, we can deduce he change in consump ion o each good. In
pa icula , compliance wi h a nu i ional cons ain induces b oad changes in he die , e en
o hose goods ha do no appea di ec ly in he cons ain , because o ela ionships o sub-
s i u abili y and/o complemen a i y wi h he goods en e ing he cons ain . Tha is o say
ha , o example, mea consump ion migh be a ec ed by a ecommenda ion o inc ease
F&V consump ion. This ela ionship o subs i u abili y o complemen a i y be ween goods is
exp essed by he c oss p ice elas ici ies o demand in he ini ial uncons ained si ua ion. Fo
example, he c oss p ice elas ici y o mea demand wi h espec o he p ice o ui s measu es
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 4/17
he esponsi eness o mea demand o a change in he p ice o ui s. Hence, i we a e able o
es ima e he p ice elas ici ies desc ibing he beha io o he uncons ained indi idual, i is
possible o in e he die a y adjus men s esul ing om compliance wi h he nu i ional ec-
ommenda ion. We no e ha such elas ici ies, based on econome ic analyses o ood demand,
a e o en used o assess he impac s o a nu i ional ax on ood choices [25]. Compa ed o
hese app oaches de eloped o assess he e ec o p ice a ia ions on consump ion and nu i-
en in akes, in his pape we conside he dual p oblem which consis s o de e mining he
p ice sys em and he compensa ion alue such ha a nu i ional ecommenda ion can be
adop ed wi hou loss o u ili y.
•The as e cos o complying wi h a nu i ional cons ain is measu ed by he”compensa ing
a ia ion”(CV) o he die a y adjus men , de ined as he amoun o addi ional money a
household would need o each i s ini ial u ili y a e complying wi h he ecommenda ion.
The CV is calcula ed as he di e ence be ween obse ed expendi u e ( he ood budge in he
uncons ained si ua ion) and he co esponding expendi u e (i.e., ood budge ) ha would be
necessa y o hold u ili y o i s ini ial le el when he nu i ion cons ain is imposed. This as e
cos should be in e p e ed as a measu e o he sho - e m loss o u ili y o he consume
which is a way o e alua e how cos ly/di icul i is o de ia e om he uncons ained si ua-
ion, igno ing he long- e m heal h bene i s o compliance, which a e conside ed sepa a ely
in he analysis (i.e., hei measu emen in e ms o dea hs a oided is p esen ed in he nex
sec ion and hei alua ion is discussed in he “Cos -bene i ”sec ion).
The heo e ical and ma hema ical backg ounds o he model a e p esen ed in g ea e de ail
elsewhe e [25].
Heal h impac assessmen
The heal h e ec s o consump ion changes induced by he adop ion o a nu i ional ecom-
menda ion a e assessed wi h he DIETRON model. The DIETRON model p o ides es ima es
o he numbe o dea hs a oided due o die - ela ed ch onic diseases. As explained by Sca bo -
ough e al. ([26], p. 711):” he DIETRON model uses age- and sex-speci ic es ima es o ela i e
isk d awn om me a-analyses o ials, coho s udies and case–con ol s udies, o es ima e
he impac on ch onic disease mo ali y o coun e ac ual popula ion die a y scena ios”. The
inpu s o he DIETRON model a e changes in in akes o he ollowing oods and nu ien :
ui s, ege ables, ibe s, o al a , monounsa u a ed a y acids (MUFA), polyunsa u a ed a y
acids (PUFA), sa u a ed a y acids (SFA), choles e ol, sal , and ene gy. The exac pa hways o
speci ic diseases (e.g., s oke) and in e media e isk ac o s a e desc ibed in he pape ha i s
p esen ed he model [27]. The s udies co e ed by he me a-analysis o isk ac o s a e clea ly
lis ed in Table 1 o he same e e ence, while he ela i e isk a ios used in DIETRON a e pub-
lished in Table A2 o he Appendix o e e ence [26].
The ou pu o he economic model is a ec o o changes in consump ion o goods when he
consume con o ms o a nu i ional ecommenda ion. To link i o DIETRON, we use a com-
posi ion able ha p o ides o each good i s con en in he di e en oods and nu ien s en e -
ing he DIETRON model.
Cos -bene i analysis
The o e all e ec o each ecommenda ion on social wel a e is es ablished by compa ing he
mone a y alue o i s cos s and bene i s. In his amewo k, he exp essions”wel a e imp o ing”
and “cos -bene icial” e e o si ua ions whe e he heal h bene i s ou weigh he cos s as de ined
in his sec ion. The long- e m heal h bene i om compliance wi h a ecommenda ion is
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 5/17

mone ized by mul iplying he numbe o DA, calcula ed om he heal h impac assessmen
ou lined in he p e ious sec ion, by he alue o a s a is ical li e, which is in e p e ed as he
e o , in e ms o he esou ces used, ha socie y is willing o make in o de o educe he isk
o dea h [28]. On he cos side, he sho - e m as e cos imposed on consume s, which was dis-
cussed abo e in he p esen a ion o he economic model, ep esen s a i s componen . How-
e e , we also need o include he cos s o implemen a ion o policies and in e en ions
necessa y o b ing abou compliance wi h each ecommenda ion. This is impo an because,
while he model calcula es he e ec s o die a y adjus men s unde an “as i ”assump ion, i.e.
Table 1. Changes in ood consump ion induced by he imposi ion o die a y cons ain s ( igh pe cen age in each column) and baseline con ibu-
ion o each ood g oup o he cons ained ood (le pe cen age in each column) o he “Lowe -a e age”consume ype.*
F&V +5% Red mea
-5%
All mea s
-5%
Sal y/Swee
a p od.
-5%
So d inks
-5%
Milk p od.
+5%
Bu e , c eam
& cheese -5%
Fish &
sea ood
+5%
%%%%%%%%%%%%% %%%
Red mea 0.0 -9.1 89.7 -5.5 22.7 -8.2 0.0 2.1 0.0 0.3 0.0 -1.5 0.0 0.6 0.0 -0.9
O he mea s 0.0 6.2 0.0 0.7 38.8 -6.4 0.0 3.4 0.0 0.4 0.0 0.1 0.0 7.0 0.0 -0.1
Cooked mea s 0.0 -3.3 0.0 0.8 32.2 -1.3 0.0 -2.4 0.0 -0.1 0.0 -1.1 0.0 0.8 0.0 -0.2
Mea agg ega e 0.0-0.389.7-0.793.7-5.20.01.20.00.20.0-0.60.03.60.0-0.3
Milk p oduc s 0.0 -4.3 0.0 0.7 0.0 3.3 0.0 1.2 0.0 0.5 100.0 5.0 0.0 -0.9 0.0 0.0
Cheeses, bu e s, esh c eams 0.0 -2.9 0.0 0.1 0.0 4.2 0.0 -4.8 0.0 -0.1 0.0 0.2 100.0 -5.0 0.0 0.1
Dai y pd s 0.0-4.00.00.60.03.40.00.00.00.3100.04.1100.0-1.70.00.0
Fish 0.0 9.7 0.0 1.7 0.0 7.5 0.0 2.4 0.0 0.1 0.0 0.2 0.0 -0.8 96.1 5.3
Eggs 0.0 -7.6 0.0 -0.8 0.0 -3.3 0.0 1.3 0.0 0.5 0.0 -0.7 0.0 -9.8 0.0 -1.0
Animal p oduc s 0.0-2
.389.70.393.71.10.00.50.00.30.02.40.0-0.596.10.2
G ains 0.2 -6.2 0.0 -1.0 0.0 -0.3 0.0 -3.2 0.0 -0.2 0.0 -2.3 0.0 -3.2 0.0 -1.4
Po a oes 0.0 -27.6 0.0 -0.8 0.0 -4.5 0.0 4.5 0.0 -1.5 0.0 -1.2 0.0 8.5 0.0 -1.0
S a chy ood 0.2-16.10.0-0.90.0-2.20.00.40.0-0.80.0-1.80.02.20.0-1.2
F ui s—F esh 40.7 -1.1 0.0 1.5 0.0 2.7 0.0 -1.8 0.0 0.3 0.0 -0.4 0.0 -3.3 0.0 1.0
F ui s—P ocessed 2.8 27.0 0.0 0.2 0.0 -3.2 0.0 3.6 0.0 1.0 0.0 -5.2 0.0 -8.3 0.0 -0.5
F&V juices 6.3 4.0 0.0 0.8 0.0 -0.3 0.0 7.9 0.0 0.8 0.0 -1.8 0.0 -0.6 0.0 0.4
Vege ables—F esh 32.6 9.5 0.0 -0.5 0.0 -0.3 0.0 0.0 0.0 0.1 0.0 0.4 0.0 6.7 0.0 0.0
Vege ables—P ocessed 9.9 18.4 0.0 0.0 0.0 -2.7 0.0 1.5 0.0 0.5 0.0 -1.6 0.0 4.4 0.0 -0.4
F ui s—D y 0.5 -6.0 0.0 1.4 0.0 11.7 0.0 2.7 0.0 -1.7 0.0 6.6 0.0 -2.5 0.0 1.5
F&V agg ega e 92.75.90.00.50.00.80.0-0.50.00.30.0-0.30.01.10.00.4
Ready meals 4.2 -11.7 10.1 -1.1 6.3 -3.6 0.0 -1.6 0.0 -0.6 0.0 -1.3 0.0 1.4 3.8 -2.9
Oil, ma ga ine, condimen s 0.0 12.0 0.0 0.1 0.0 -1.2 0.0 11.1 0.0 0.4 0.0 1.0 0.0 9.3 0.0 0.1
Sal - a p oduc s 0.0 -20.7 0.1 1.2 0.1 10.3 8.1 -8.5 0.0 -1.4 0.0 -2.7 0.0 -28.5 0.0 -0.3
Suga - a p oduc s 2.9 2.1 0.0 0.1 0.0 0.3 91.9 -4.7 0.0 0.2 0.0 -0.2 0.0 -2.7 0.0 -0.2
So d inks 0.0 -18.4 0.0 0.7 0.0 5.3 0.0 2.8 100.0 -5.0 0.0 -3.5 0.0 -3.2 0.0 -0.1
Wa e 0.0 -20.0 0.0 1.8 0.0 10.0 0.0 -0.7 0.0 -0.2 0.0 -0.2 0.0 3.0 0.0 0.1
Alcoholic be e ages 0.0 12.9 0.0 0.3 0.0 -0.4 0.0 3.7 0.0 -0.2 0.0 0.2 0.0 1.6 0.0 -0.5
*Food g oups a e defined as in [29]: Red mea (bee and eal); o he mea s (poul y, po k, lamb, e c.); cooked mea s (ham, pâ é, sausages, bacon, e c.);
milk p oduc s (milk, yoghu , dai y desse s, e c.); cheese, bu e and c eam; fish and sea ood; eggs; g ain p oduc s (b ead, pas a, ice, whea flou , and
ce eals); po a oes; esh ui s; p ocessed ui s (canned & compo e); ui s & ege ables juices; esh ege ables including legumes; p ocessed ege ables
including legumes (canned, ozen); d ied ui s & nu s; eady meals (pizza, saue k au , cassoule , e c.); oils & ege able a ; sal - a p oduc s (finge ood,
chips, c acke s, appe ize s); suga - a p oduc s (candy, chocola e, cookies, pas y, ice c eam, jam, e c.); so d inks (sodas, lemonade, sy ups, e c.); mine al
and sp ing wa e s; alcoholic be e ages.
doi:10.1371/jou nal.pone.0158453. 001
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 6/17
assuming compliance wi h he nu i ional ecommenda ion, in p ac ice beha iou al change
equi es public in es men in social ma ke ing campaigns and o he ypes o in e en ions.
A p oblem a ises, howe e , because quan i ying he cos o a policy inducing consume s o
change hei consump ion o a gi en ood o nu ien by a ce ain amoun is e y di icul . We
ci cum en he issue by de ining a cos -bene i h eshold cha ac e izing he maximum amoun
ha could be in es ed o p omo e a gi en ecommenda ion while ensu ing ha he bene i s
ou weigh he cos s. This is achie ed by balancing he heal h bene i s gene a ed by a nu i ional
ecommenda ion (deno ed Bh) agains he cos o indi iduals (deno ed CC and measu ed by
he sho - e m as e cos s) and he cos o public sec o in e en ions (C
P
). The cos -bene i
h eshold is hen simply calcula ed as C
P
=Bh–CC.
Empi ical assessmen
The empi ical implemen a ion o he economic model equi es di e en se s o da a. To de e -
mine he ini ial consump ion o oods and he economic pa ame e s (elas ici ies), we use he
sou ce o da a and esul s o he mos ecen a ailable econome ic analysis o ood demand in
F ance conduc ed by Allais e al. [29]. Tha s udy is based on da a om a ep esen a i e panel
o F ench households (KANTAR Wo ldpanel) o e a i e-yea pe iod om 1996 o 2001. The
pa icipa ing households eco d weekly all hei pu chases o oods, using ba code scanning
echnology whene e possible, bu oods wi hou ba codes a e also eco ded. The in o ma ion
p o ided includes he cha ac e is ics o he pu chased p oduc (e.g., b and, size), he quan i y
pu chased as well as ela ed expendi u e. KANTAR also p o ides he main socio-economic
cha ac e is ics o he panel households, including household size, egion o esidence and
income class. Each annual ound o he da a se con ains in o ma ion on app oxima ely 5,000
households, wi h an annual o a ion o oughly one hi d o he pa icipan s.
Fo he sake o consis ency, we had o use he same 22 p oduc agg ega es as s udy [29]
om which he elas ici ies a e d awn. Those elas ici ies in [29] a e epo ed o ou ep esen a-
i e ypes o households di e en ia ed by income qua iles and labeled in wha ollows as
“Modes ”,“Lowe a e age”,“Uppe a e age”, and “Well-o ”. The nu ien con en s o he 22
ood agg ega es a e calcula ed by combining he ood composi ion da abase o he F ench die-
a y in ake su ey INCA2 and a e age adul in akes o he componen oods o each agg ega e
d awn om INCA2 (which s ands o “É ude Indi iduelle Na ionale des Consomma ions Ali-
men ai es 2006–7”), which a e eely a ailable om he open da a pla o m o he F ench go -
e nmen (h ps://www.da a.gou . / /da ase s/donnees-de-consomma ions-e -habi udes-
alimen ai es-de-le ude-inca-2-3/). The model is applied o es ima e he a ia ions in house-
holds’pu chases induced by he adop ion o eigh ood-based ecommenda ions aken one a a
ime. We p o ide some de ails abou he choice o ecommenda ions a he beginning o he
esul s sec ion. The e ec s o die a y adjus men s a e calcula ed unde an “as i ”assump ion,
i.e. assuming ha he households comply wi h a 5% change in he cons ain le el ( o example
in he case o F&V, we assume ha a consume has o inc ease his consump ion o F&V by
5%).
To simula e he heal h e ec s, changes in ood pu chases a household le el, as calcula ed by
he economic model, a e ansla ed in o changes in indi idual in akes, dis inguishing be ween
males and emales and using he INCA2 die a y in ake da abase. This is accomplished unde
he assump ion ha (i) all household membe s expe ience he same ela i e changes in in akes,
and (ii) he ela i e changes in consump ion o ood-a -home and ood-away- om-home a e
equal. Va ia ions in nu ien in akes a e hen calcula ed om a ia ions in ood in akes by
using he nu i ional coe icien s o he 22 agg ega es. Finally, he changes in nu ien s a e ed
in o DIETRON model so as o es ima e he heal h e ec s o he die a y adjus men s.
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 7/17
The pa ame e s o he DIETRON model a e de i ed om wo ld-wide me a-analyses o die-
a y isk ac o s and a e no coun y speci ic, so ha adap ing he DIETRON model o F ance
only equi es calib a ion o he ini ial mo ali y le els, by ele an causes. This is achie ed by
using he INSERM da a on mo ali y in F ance a ibu able o majo die - ela ed diseases
(INSERM s ands o “Ins i u Na ional de la San é e de la eche che Médicale”). We limi he
s udy o indi iduals be ween he age o 25 and 74 and he e o e ocus on he e ec s o die a y
changes on p ema u e dea hs. The ch onic diseases conside ed in DIETRON accoun o
sligh ly mo e han one hi d o o al F ench mo ali y.
Finally, he heal h bene i is quan i ied by applying a mone a y alue o he educed mo al-
i y igu es calcula ed by DIETRON. Es ima es o he alue o a s a is ical li e a y subs an ially
ac oss coun ies and policy domains. In he ollowing analysis, we use a conse a i e alue o a
s a is ical li e, based on he cos h eshold o a Quali y Adjus ed Li e Yea ha is applied in he
UK o in es iga e he cos -e ec i eness o medical ca e (e.g., d ugs, p ocedu es). As explained
in I z e al. [25] his p o ides a alue anging om €240k o €360k.
In addi ion o he mean alue o he di e en ela i e isks, he DIETRON model also p o-
ides he dis ibu ions o each ela i e isk pa ame e . Twen y-six o hose ollow log-no mal
dis ibu ions and six ollow no mal dis ibu ions, while all a e independen . We pe o m
Mon e Ca lo simula ions by d awing a se o pa ame e s one million imes. To be mo e p ecise,
o a gi en ype o consume , we d aw a se o 32 pa ame e s, use his se o e alua e he num-
be o DA o each nu i ional ecommenda ion and epea he ope a ion one million imes. In
he esul s sec ion, we p o ide he median as well as he 2.5 and 97.5 pe cen iles o DA o each
ecommenda ion.
The R codes o simula e bo h he economic model and he epidemiological model a e a ail-
able upon eques om he au ho s.
Resul s
Compa ison o ou da a desc ibing he a e age die o F ench consume s wi h con en ional
die a y no ms con i ms p e ious indings ha he F ench die is ich in a , pa icula ly o he
sa u a ed kind ha o igina es p ima ily om animal p oduc s [14,15]. This jus i ies he analy-
sis o h ee ood-based ecommenda ions aimed a educing consump ion o : all mea s; ed
mea , because o i s pa icula ly high con en in sa u a ed a y acid (SFA); and SFA- ich dai y
p oduc s, which include bu e , c eam, and cheese. On a e age, in ake o calcium in he F ench
popula ion complies wi h die a y e e ence in ake, bu is oo low o some consume g oups
(mainly women and he low-income). This sho coming could be add essed by aising con-
sump ion o he o he dai y p oduc s (hence o h e e ed o as “milk p oduc s”). Snacking has
also been iden i ied in F ance as a po en ially g owing issue, pa icula ly among young con-
sume s [30], and we he e o e conside ecommenda ions o educe consump ion o bo h sal y
and swee snacks (hence o h e e ed o as “sal y/swee - a p oduc s”). Suga -swee ened be e -
ages a e inc easingly ecognized as an impo an de e minan o body weigh [31] and a ecom-
menda ion o educe hei consump ion is app op ia e in F ance gi en conce ns o e obesi y
and o e weigh in he popula ion. Consump ion o F&V alls well below he ecommended i e
po ions a day, and his is pa icula ly ue o low-income consume s. A F&V ecommenda-
ion is he e o e included. Finally, he e is mode a e bu consis en e idence ega ding he
heal h bene i s o inc eased ish consump ion [32], and we he e o e analyse he impac o a
ela ed ecommenda ion.
Table 1 p esen s he simula ed die a y adjus men s o he “lowe a e age”household ype.
Al hough esul s o he o he ypes a e no epo ed he e, we ha e ound ha hey a e b oadly
simila o hose p esen ed below. This simila i y in he way he ou household ypes espond
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 8/17
o ecommenda ions is explained by small di e ences in he p ice and income elas ici ies o
hose household ypes. Those elas ici ies, which we used o calib a e he model, a e epo ed in
Tables 6–10 o he supplemen a y ma e ial o e e ence [29]. Fo each cons ain , p esen ed in
a sepa a e column, Table 1 depic s i s he baseline con ibu ion o each ood g oup o he con-
s ained ood, hen he consump ion changes ha would esul om compliance wi h he
cons ain .
Tu ning o he changes in die and ocusing i s on he”all mea s” ecommenda ion, we
no e ha he i e pe cen dec ease ha is imposed exogenously co esponds o an absolu e
decline o oughly eigh g ams pe day. This quan i a i ely small educ ion in mea consump-
ion igge s ela i ely impo an die a y adjus men s. S a chy p oduc s a e complemen s o
mea s and hus hei consump ion dec eases wi h he educ ion in mea consump ion (-2.2%).
On he con a y, dai y p oduc s, which p o ide p o eins, a e mea subs i u es and hus hei
consump ion inc eases wi h he exogenously-imposed educ ion in mea consump ion
(+3.4%). Looking a he esul s a a lowe le el o p oduc agg ega ion e eals ha complex
adjus men s also ake place wi hin b oad ood ca ego ies. Thus, consume s choose o compen-
sa e he dec ease in mea consump ion by aising consump ion o ish (+7.5%) bu also, and
less expec edly, by educing ha o eggs (-3.3%).
To a oid con usion in in e p e ing he esul s, we no e ha he dec ease in consump ion o
he “Mea agg ega e”ca ego y is di e en om 5% ( he a ge o a dec ease in o al consump-
ion o mea ). This is because a ood ca ego y ( eady meals) also con ains some mea . Then, he
change in “All mea s”consump ion akes in o accoun he changes in consump ion o he
“Mea agg ega e”as well as he changes in he consump ion o he o he ood ca ego ies which
con ain some mea . This ema k applies o o he ecommenda ions (e.g., on F&V).
The die a y adjus men s induced by compliance wi h he cons ain s a e he e ogeneous.
Compa ed o simula ed adjus men s o he”all mea s”cons ain , imposi ion o he cons ain
on ed mea p oduces smalle consump ion changes. This is unde s andable as his cons ain
is less demanding in he sense ha i es ic s a smalle ac ion o he die ( he dec ease in ed
mea consump ion is abou 3 g/day) and subs i u ion wi h o he mea s occu s leading o a
small dec ease in agg ega e mea consump ion (-0.7%). O e all, he “ ed mea ”cons ain
a ec s consump ion o he di e en ood g oups in he same di ec ion as, bu wi h a lowe
magni ude han, he “all mea ”cons ain .
The ecommenda ion o inc ease consump ion o F&V by 5% induces he la ges adjus -
men s in he die . Thus, consump ion o s a chy p oduc s (-16.1%), dai y p oduc s (-4.0%),
eady meals (-11.7%), oil and ma ga ine (+12.0%), and sal - a p oduc s (-20.7%) a e s ongly
a ec ed. Wi hin he F&V ca ego y, and in e ms o absolu e quan i y, he bigges inc ease is o
esh ege ables whe eas esh ui consump ion dec eases by a e y small amoun . Howe e ,
he la ges pe cen age inc eases a e o p ocessed p oduc s, meaning ha he adop ion o he
ecommenda ion aises he ela i e impo ance o p ocessed p oduc s wi hin he F&V ca e-
go y. Rega ding he o he ecommenda ions, i is wo h no ing ha an inc ease in consump-
ion o milk p oduc s leads o a dec ease in consump ion o mea (pa icula ly ed mea ),
s a chy p oduc s, and sal - a p oduc s. A dec ease in consump ion o bu e -c eam and cheese
induces an inc ease in consump ion o mea , pa icula ly “o he mea s”, oil and ma ga ine, and
s a chy p oduc s; i also esul s in a (la ge) dec ease in he consump ion o sal - a p oduc s.
An inc ease in ish consump ion leads o a dec ease in consump ion o eady meals, s a chy
oods, and ed mea . A decline in consump ion o sal /swee - a p oduc s leads o an inc ease
in oil and ma ga ine consump ion and a dec ease in consump ion o c eam-bu e and cheese.
Finally, a dec ease in consump ion o so d inks has a ela i ely small impac on consump ion
o he di e en ood p oduc s.
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 9/17
Re e ences
1. Bu iss J, B iend A, Da mon N, Fe guson E, Maillo M, Lluch A. Die modelling: How i can in o m he
de elopmen o die a y ecommenda ions and public heal h policy. Nu Bull 2014; 39(1):115–125.
2. Da mon N, Fe guson EL, B iend A. Impac o a cos cons ain on nu i ionally adequa e ood choices
o F ench women: an analysis by linea p og amming. Jou nal o nu i ion educa ion and beha io
2006; 38(2):82–90. PMID: 16595286
3. Maillo M, Vieux F, Amio MJ, Da mon N. Indi idual die modeling ansla es nu ien ecommenda ions
in o ealis ic and indi idual-speci ic ood choices. Am J Clin Nu 2010 Feb; 91(2):421–430. doi: 10.
3945/ajcn.2009.28426 PMID: 19939986
4. Macdia mid JI, Kyle J, Ho gan GW, Loe J, Fy e C, Johns one A, e al. Sus ainable die s o he u u e:
Can we con ibu e o educing g eenhouse gas emissions by ea ing a heal hy die ? Am J Clin Nu 2012
Sep; 96(3):632–639. doi: 10.3945/ajcn.112.038729 PMID: 22854399
5. Ma in A. Nu i ional ecommenda ions o he F ench popula ion. Sciences des Alimen s 2001; 21:317-
317-458.
6. Henson S. Linea p og amming analysis o cons ain s upon human die s. Jou nal o Ag icul u al Eco-
nomics 1991; 42(3):380–393.
7. Masse G, Monsi ais P, Maillo M, Da mon N, D ewnowski A. Die op imiza ion me hods can help ans-
la e die a y guidelines in o a cance p e en ion ood plan. J Nu 2009 Aug; 139(8):1541–1548. doi: 10.
3945/jn.109.104398 PMID: 19535422
8. S igle GJ. The cos o subsis ence. Jou nal o a m economics 1945; 27(2):303–314.
9. Con o i P, D'Amicis A. Wha is he cos o a heal hy die in e ms o achie ing RDAs? Public Heal h
Nu 2000; 3(03):367–373.
10. Glanz K, Basil M, Maibach E, Goldbe g J, Snyde D. Why Ame icans ea wha hey do: as e, nu i ion,
cos , con enience, and weigh con ol conce ns as in luences on ood consump ion. J Am Die Assoc
1998; 98(10):1118–1126. PMID: 9787717
11. Webb D, By d-B edbenne C. O e coming consume ine ia o die a y guidance. Ad Nu 2015 Jul 15;
6(4):391–396. doi: 10.3945/an.115.008441 PMID: 26178023
12. Da mon N, Fe guson EL, B iend A. A cos cons ain alone has ad e se e ec s on ood selec ion and
nu ien densi y: an analysis o human die s by linea p og amming. J Nu 2002 Dec; 132(12):3764–
3771. PMID: 12468621
13. Da mon N, Fe guson E, B iend A. Do economic cons ain s encou age he selec ion o ene gy dense
die s? Appe i e 2003; 41(3):315–322. PMID: 14637330
14. S ini asan CS, I z X, Shanka B. An assessmen o he po en ial consump ion impac s o WHO die a y
no ms in OECD coun ies. Food Policy 2006 2; 31(1):53–77.
15. Shanka B, S ini asan CS, I z X. Wo ld heal h o ganiza ion die a y no ms: A quan i a i e e alua ion o
po en ial consump ion impac s in he Uni ed S a es, Uni ed Kingdom, and F ance. Applied Economic
Pe spec i es and Policy 2008; 30(1):151–175.
16. A noul M, Jones PJ, T an e RB, Ti in R, T aill WB, Tzanopoulos J. Modelling he likely impac o
heal hy ea ing guidelines on ag icul u al p oduc ion and land use in England and Wales. Land Use Pol-
icy 2010; 27(4):1046–1055.
17. WHO Die a y No ms: A Quan i a i e E alua ion o Po en ial Consump ion Impac s in he Uni ed S a es.
2006 Annual Mee ing, Augus 12–18, 2006, Queensland, Aus alia: Ci esee ; 2006.
18. K ebs-Smi h SM, Guen he PM, Suba AF, Ki kpa ick SI, Dodd KW. Ame icans do no mee ede al
die a y ecommenda ions. J Nu 2010 Oc ; 140(10):1832–1838. doi: 10.3945/jn.110.124826 PMID:
20702750
19. Guen he PM, Juan W, Lino M, Hiza HA, Fungwe TV, Lucas R. Die quali y o low-income and highe -
income Ame icans in 2003–2004 as measu ed by he Heal hy Ea ing Index-2005. The FASEB Jou nal
2009; 23(1_Mee ingAbs ac s):540.5.
20. de Ab eu D, Guessous I, Vauche J, P eisig M, Waebe G, Vollenweide P, e al. Low compliance wi h
die a y ecommenda ions o ood in ake among adul s. Clinical Nu i ion 2013; 32(5):783–788. doi: 10.
1016/j.clnu.2012.11.022 PMID: 23260749
21. Maillo M, Da mon N, Vieux F, D ewnowski A. Low ene gy densi y and high nu i ional quali y a e each
associa ed wi h highe die cos s in F ench adul s. Am J Clin Nu 2007 Sep; 86(3):690–696. PMID:
17823434
22. D ewnowski A, Da mon N. The economics o obesi y: die a y ene gy densi y and ene gy cos . Am J
Clin Nu 2005 Jul; 82(1 Suppl):265S–273S. PMID: 16002835
23. Wa dle J, Pa men e K, Walle J. Nu i ion knowledge and ood in ake. Appe i e 2000; 34(3):269–275.
PMID: 10888290
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 16 / 17

24. Capacci S, Mazzocchi M, Shanka B, Macias JB, Ve beke W, Pé ez-Cue o FJ, e al. Policies o p omo e
heal hy ea ing in Eu ope: a s uc u ed e iew o policies and hei e ec i eness. Nu Re 2012; 70
(3):188–200. doi: 10.1111/j.1753-4887.2011.00442.x PMID: 22364161
25. I z X, Le oy P, Réquilla V, Sole L. Economic assessmen o nu i ional ecommenda ions. J Heal h
Econ 2015 1: ; 39:188–210.
26. Sca bo ough P, Allende S, Cla ke D, Wick amasinghe K, Rayne M. Modelling he heal h impac o
en i onmen ally sus ainable die a y scena ios in he UK. Eu J Clin Nu 2012; 66(6):710–715. doi: 10.
1038/ejcn.2012.34 PMID: 22491494
27. Sca bo ough P, Nnoaham KE, Cla ke D, Capewell S, Rayne M. Modelling he impac o a heal hy die
on ca dio ascula disease and cance mo ali y. J Epidemiol Communi y Heal h 2012 May; 66(5):420–
426. doi: 10.1136/jech.2010.114520 PMID: 21172796
28. T eich N. La aleu de la ie humaine en économie. Fu u ibles 2015(404: ):61–71.
29. Allais O, Be ail P, Nichèle V. The e ec s o a a ax on F ench households’pu chases: a nu i ional
app oach. Am J Ag ic Econ 2010; 92(1):228–245.
30. Dubuisson C, Lio e S, Tou ie M, Du ou A, Calamassi-T an G, Vola ie J, e al. T ends in ood and
nu i ional in akes o F ench adul s om 1999 o 2007: esul s om he INCA su eys. B J Nu 2010;
103(07):1035–1048.
31. Te Mo enga L, Malla d S, Mann J. Die a y suga s and body weigh : sys ema ic e iew and me a-analy-
ses o andomised con olled ials and coho s udies. BMJ 2012 Jan 15; 346:e7492. doi: 10.1136/bmj.
e7492 PMID: 23321486
32. Nesheim MC, Nes le M. Ad ice o ish consump ion: challenging dilemmas. Am J Clin Nu 2014 May;
99(5):973–974. doi: 10.3945/ajcn.114.086488 PMID: 24670941
33. Popkin BM, Hawkes C. Swee ening o he global die , pa icula ly be e ages: pa e ns, ends, and pol-
icy esponses. The Lance Diabe es & Endoc inology 2015.
34. Jou J, Techakehakij W. In e na ional applica ion o suga -swee ened be e age (SSB) axa ion in obe-
si y educ ion: ac o s ha may in luence policy e ec i eness in coun y-speci ic con ex s. Heal h Policy
2012; 107(1):83–90. doi: 10.1016/j.heal hpol.2012.05.011 PMID: 22727243
35. Capacci S, Mazzocchi M. Fi e-a-day, a p ice o pay: An e alua ion o he UK p og am impac accoun -
ing o ma ke o ces. J Heal h Econ 2011 1; 30(1):87–98. doi: 10.1016/j.jhealeco.2010.10.006 PMID:
21129797
36. DGCA. Scien i ic epo o he 2015 Die a y Guidelines Ad iso y Commi ee (DGCA). 2015.
37. Lus ig RH, Schmid LA, B indis CD. Public heal h: The oxic u h abou suga . Na u e 2012; 482
(7383):27–29. doi: 10.1038/482027a PMID: 22297952
38. Malho a A. Sa u a ed a is no he majo issue. BMJ 2013 Oc 22; 347: 6340. doi: 10.1136/bmj. 6340
PMID: 24149521
Consume P e e ences and Die a y Recommenda ions
PLOS ONE | DOI:10.1371/jou nal.pone.0158453 June 30, 2016 17 / 17