nu ien s
A icle
Die a y Ad anced Glyca ion End-P oduc s and Colo ec al
Cance Risk in he Eu opean P ospec i e In es iga ion in o
Cance and Nu i ion (EPIC) S udy
Elom K. Aglago 1, Ana-Lucia Mayén1, Vik o ia Knaze 2, Heinz F eisling 1, Ve onika Fedi ko 3,
Da id J. Hughes 4, Li Jiao 5, Anne Ki s ine E iksen 6, Anne Tjønneland 6, Ma ie-Ch is ine Bou on-Ruaul 7,8,
Joseph A. Ro hwell 7,8, Gianluca Se e i 7,8,9, Rudol Kaaks 10, Ve ena Ka zke 10, Ma hias B. Schulze 11,12 ,
Anna Bi uko 11 , Domenico Palli 13 , Sabina Sie i 14 , Ma ia San ucci de Magis is 15, Rosa io Tumino 16,
Ful io Ricce i 17,18, Bas Bueno-de-Mesqui a 19, Je oen W. G. De ksen 20 , Gu i Skeie 21 , Inge To hild G am 21 ,
To kjel Sandange 21, J. Ramón Qui ós22, Leila Luján-Ba oso 23 , Ma ia-Jose Sánchez 24,25,26,27, Pila Amiano 24,28,
Ma ía-Dolo es Chi laque 24,29, Au elio Ba ica e Gu ea 24,30,31, Ingege d Johansson 32 , Jonas Manje 33,34,
Au o a Pe ez-Co nago 35, Elisabe e Weide pass 36 , Ma c J. Gun e 1, Alicia K. Hea h 37 ,
Caspe G. Schalkwijk 38,† and Mazda Jenab 1,*,†


Ci a ion: Aglago, E.K.; Mayén, A.-L.;
Knaze, V.; F eisling, H.; Fedi ko, V.;
Hughes, D.J.; Jiao, L.; E iksen, A.K.;
Tjønneland, A.; Bou on-Ruaul ,
M.-C.; e al. Die a y Ad anced
Glyca ion End-P oduc s and
Colo ec al Cance Risk in he
Eu opean P ospec i e In es iga ion
in o Cance and Nu i ion (EPIC)
S udy. Nu ien s 2021,13, 3132.
h ps://doi.o g/10.3390/nu13093132
Academic Edi o : Rachel Mu phy
Recei ed: 10 Augus 2021
Accep ed: 2 Sep embe 2021
Published: 8 Sep embe 2021
Publishe ’s No e: MDPI s ays neu al
wi h ega d o ju isdic ional claims in
published maps and ins i u ional a il-
ia ions.
Copy igh : © 2021 by he au ho s.
Licensee MDPI, Basel, Swi ze land.
This a icle is an open access a icle
dis ibu ed unde he e ms and
condi ions o he C ea i e Commons
A ibu ion (CC BY) license (h ps://
c ea i ecommons.o g/licenses/by/
4.0/).
1
Nu i ion and Me abolism Sec ion, In e na ional Agency o Resea ch on Cance (IARC), 69372 Lyon, F ance;
[email p o ec ed] (E.K.A.); [email p o ec ed]c. (A.-L.M.); eislingh@ia c. (H.F.);
[email p o ec ed] (M.J.G.)
2Ea ly De ec ion, P e en ion, and In ec ions B anch, In e na ional Agency o Resea ch on Cance (IARC),
69372 Lyon, F ance; [email p o ec ed]
3
Depa men o Epidemiology, The Uni e si y o Texas MD Ande son Cance Cen e , Hous on, TX 77030, USA;
[email p o ec ed]
4Cance Biology and The apeu ics G oup (CBT), Conway Ins i u e, School o Biomolecula and Biomedical
Science (SBBS), Uni e si y College Dublin, D04 V1W8 Dublin, I eland; [email p o ec ed]
5Depa men o Medicine, Baylo College o Medicine, Hous on, TX 77030, USA; [email p o ec ed]
6Danish Cance Socie y Resea ch Cen e , 2100 Copenhagen, Denma k; ake@cance .dk (A.K.E.);
anne @cance .dk (A.T.)
7CESP, Facul éde Médecine—Uni e si éPa is-Saclay, UVSQ, INSERM, 94805 Villejui , F ance;
Ma ie-ch is ine.BOUTRON@gus a e oussy. (M.-C.B.-R.); Joseph.ROTHWELL@gus a e oussy. (J.A.R.);
Gianluca.SEVERI@gus a e oussy. (G.S.)
8Gus a e Roussy, 114, Rue Édoua d-Vaillan , CEDEX, 94805 Villejui , F ance
9Depa men o S a is ics, Compu e Science and Applica ions (DISIA), Uni e si y o Flo ence,
50121 Flo ence, I aly
10 Di ision o Cance Epidemiology, Ge man Cance Resea ch Cen e (DKFZ), 69120 Heidelbe g, Ge many;
[email p o ec ed] (R.K.); V[email p o ec ed] (V.K.)
11 Depa men o Molecula Epidemiology, Ge man Ins i u e o Human Nu i ion Po sdam-Rehb uecke,
14558 Nu he al, Ge many; [email p o ec ed] (M.B.S.); [email p o ec ed] (A.B.)
12 Ins i u e o Nu i ion Science, Uni e si y o Po sdam, 14558 Nu he al, Ge many
13
Cance Risk Fac o s and Li e-S yle Epidemiology Uni , Ins i u e o Cance Resea ch, P e en ion and Clinical
Ne wo k (ISPRO), 50139 Flo ence, I aly; d.palli@isp o. oscana.i
14
Epidemiology and P e en ion Uni , Fondazione IRCCS Is i u o Nazionale dei Tumo i di Milano, 20133 Milan,
I aly; [email p o ec ed]
15 Azienda Ospedalie a Uni e si a ia Fede ico II, 80131 Napoli, I aly; [email p o ec ed]
16 Hyblean Associa ion o Epidemiological Resea ch AIRE-ONLUS, 97100 Ragusa, I aly;
[email p o ec ed]
17 Depa men o Clinical and Biological Sciences, Uni e si y o Tu in, Regione Gonzole 10, 10043 O bassano,
I aly; [email p o ec ed]
18 Uni o Epidemiology, Regional Heal h Se ice ASL TO3, Via Sabaudia 164, 10095 G ugliasco, I aly
19 Depa men o De e minan s o Ch onic Diseases (DCD), Na ional Ins i u e o Public Heal h and he
En i onmen (RIVM), P.O. Box 1, 3720 BA Bil ho en, The Ne he lands; [email p o ec ed]
20 Julius Cen e o Heal h Sciences and P ima y Ca e, Uni e si y Medical Cen e U ech , U ech Uni e si y,
3584 CG U ech , The Ne he lands; H.W[email p o ec ed]
21
Facul y o Heal h Sciences, Depa men o Communi y Medicine, Uni e si y o T omsø, The A c ic Uni e si y
o No way, 9037 T omsø, No way; [email p o ec ed] (G.S.); inge [email p o ec ed] (I.T.G.);
[email p o ec ed] (T.S.)
22 Public Heal h Di ec o a e, 33402 As u ias, Spain; [email p o ec ed]g
Nu ien s 2021,13, 3132. h ps://doi.o g/10.3390/nu13093132 h ps://www.mdpi.com/jou nal/nu ien s
Nu ien s 2021,13, 3132 2 o 13
23 Uni o Nu i ion and Cance , Ca alan Ins i u e o Oncology—ICO; and Nu i ion and Cance G oup;
Epidemiology, Public Heal h, Cance P e en ion and Pallia i e Ca e P og am, Bell i ge Biomedical Resea ch
Ins i u e—IDIBELL, L’Hospi ale de Llob ega , A . G an ia 199-203, 08908 Ba celona, Spain;
[email p o ec ed]
24 Cen o de In es igación Biomédica en Red de Epidemiología y Salud Pública (CIBERESP), 28029 Mad id,
Spain; [email p o ec ed] (M.-J.S.); [email p o ec ed] (P.A.);
[email p o ec ed] (M.-D.C.); au [email p o ec ed] (A.B.G.)
25 Escuela Andaluza de Salud Pública (EASP), 18011 G anada, Spain
26 Ins i u o de In es igación Biosani a ia ibs.GRANADA, 18012 G anada, Spain
27 Depa men o P e en i e Medicine and Public Heal h, Uni e si y o G anada, 18071 G anada, Spain
28 Public Heal h Di ision o Gipuzkoa, BioDonos ia Resea ch Ins i u e, 20014 Donos ia-San Sebas ian, Spain
29 Depa men o Epidemiology, Mu cia Regional Heal h Council, IMIB-A ixaca, Mu cia Uni e si y,
30003 Mu cia, Spain
30 Na a a Public Heal h Ins i u e, 31008 Pamplona, Spain
31 Na a a Ins i u e o Heal h Resea ch (IdiSNA), 31008 Pamplona, Spain
32 Depa men o Radia ion Sciences, Oncology, Umeå Uni e si y, 907 36 Umeå, Sweden;
[email p o ec ed]
33 Depa men o Clinical Sciences, Malmö, Lund Uni e si y, 221 00 Lund, Sweden; [email p o ec ed]
34 Di ision o Su ge y, Malmö, Lund Uni e si y, 221 00 Lund, Sweden
35 Cance Epidemiology Uni , Nu ield Depa men o Popula ion Heal h, Uni e si y o Ox o d,
Ox o d OX3 7LF, UK; Au o a.Pe [email p o ec ed]
36 O ice o he Di ec o , In e na ional Agency o Resea ch on Cance (IARC), 69372 Lyon, F ance;
W[email p o ec ed]
37 Depa men o Epidemiology and Bios a is ics, School o Public Heal h, Impe ial College London,
London W2 1PG, UK; [email p o ec ed]
38
Depa men o In e nal Medicine, CARIM School o Ca dio ascula Diseases, Maas ich Uni e si y Medical
Cen e , 6229 HX Maas ich, The Ne he lands; c.schalkwijk@maas ich uni e si y.nl
*Co espondence: jenabm@ia c. ; Tel.: +33-472-73-80-82
† These au ho s con ibu ed equally o his wo k.
Abs ac :
Die a y ad anced glyca ion end-p oduc s (dAGEs) ha e been hypo hesized o be asso-
cia ed wi h a highe isk o colo ec al cance (CRC) by p omo ing in lamma ion, me abolic dys-
unc ion, and oxida i e s ess in he colonic epi helium. Howe e , e idence om p ospec i e
coho s udies is sca ce and inconclusi e. We e alua ed CRC isk associa ed wi h he in ake o
dAGEs in he Eu opean P ospec i e In es iga ion in o Cance and Nu i ion (EPIC) s udy. Di-
e a y in akes o h ee majo dAGEs: N
ε
-ca boxy-me hyllysine (CML), N
ε
-ca boxye hyllysine (CEL),
and
Nδ-(5-hyd o-5-me hyl-4-imidazolon-2-yl)-o ni hine (MG-H1)
we e es ima ed in 450,111 pa ic-
ipan s (median
ollow-up = 13 yea s,
wi h 6162 CRC cases) by ma ching o a de ailed published
Eu opean ood composi ion da abase. Haza d a ios (HRs) and 95% con idence in e als (CIs) o
he associa ions o dAGEs wi h CRC we e compu ed using mul i a iable-adjus ed Cox eg ession
models. In e se CRC isk associa ions we e obse ed o CML (HR compa ing ex eme quin iles:
HRQ5 s.Q1 = 0.92, 95% CI = 0.85–1.00)
and MG-H1 (HR
Q5 s
.
Q1
= 0.92,
95% CI = 0.85–1.00),
bu no
o CEL
(HRQ5 s.Q1 = 0.97,
95% CI = 0.89–1.05). The associa ions did no di e by sex o ana omical
loca ion o he umo . Con a y o he ini ial hypo hesis, ou indings sugges an in e se associa ion
be ween dAGEs and CRC isk. Mo e esea ch is equi ed o e i y hese indings and be e di e en-
ia e he ole o dAGEs om ha o endogenously p oduced AGEs and hei p ecu so compounds
in CRC de elopmen .
Keywo ds:
ad anced glyca ion end-p oduc s; die a y glyca ion compounds; colo ec al cance ; di-
e a y exposu e
1. In oduc ion
Colo ec al cance (CRC) is he hi d mos common cance globally, and he second
leading cause o cance - ela ed dea hs [
1
]. The incidence o CRC ollows a geog aphical
dis ibu ion pa e n, wi h he highes igu es obse ed in Wes e n coun ies [
2
], mos
likely a ibu able o he “mode n” li es yle and die ich in ene gy-dense p ocessed oods
Nu ien s 2021,13, 3132 3 o 13
wi h poo nu i ional alue [
3
–
5
]. The Wes e n die is a subs an ial sou ce o ad anced
glyca ion end-p oduc s (AGEs), an expansi e g oup o molecules p oduced by i e e sible
non-enzyma ic combina ion o educing suga s and p o eins, lipids, o nucleic acids [
6
].
The ypical Wes e n die can also p omo e endogenous o ma ion o AGEs by supplying
educing suga s and AGE p ecu so s such as eac i e dica bonyls, i.e., me hylglyoxal,
glyoxal, glycolaldehyde, and glyce aldehyde [
7
]. Die a y AGEs (dAGEs) a e known o
hei p o-in lamma ion and p o-oxida ion p ope ies in he colon and ha e been epo ed
in di e se colonic pa hologies, such as in lamma o y bowel diseases [
8
]. A ound 70–90%
o AGEs inges ed a e unabso bed [
9
,
10
] and emain in he gas oin es inal ac whe e hey
can in e ac di ec ly wi h colon epi helial cells. The human colon is, he e o e, po en ially
exposed o AGEs om he die , bu also om he sys emic milieu by way o ci cula ing
AGEs [11,12].
AGEs ha e been hypo hesized o be associa ed wi h CRC de elopmen [
13
], mos ly
due o hei abili y o p omo e umo cell g ow h
in i o
[
13
]. A body o mechanis ic
e idence has linked AGEs o CRC h ough s imula ion o he p o-in lamma o y esponse
ia he ac i a ion o he ecep o o AGEs (RAGE) [
14
], an inc ease in colonic ba ie
pe meabili y—allowing close in e ac ion o AGEs wi h colonic epi helium—and conse-
quen ial leakage o bac e ial oxins in o he sys emic ci cula ion [
15
]. No wi hs anding
hese nume ous plausible mechanisms, no p e ious epidemiological s udies ha e in es i-
ga ed he ela ionship be ween dAGEs and CRC, p obably due o he lack o de ailed ood
composi ion da abases o hese compounds. The de elopmen o ood composi ion ables
o es ima ing die a y AGEs is ecen , and ew ables exis , mainly o Japanese oods [
16
]
and mo e ecen ly o Eu opean oods [
17
]. Due o he la ge numbe o di e en AGEs, he
ables de eloped ocused on he majo compounds, speci ically N
ε
-(ca boxyme hyl)lysine
(CML). The Eu opean ood composi ion able p o ides da a on CML and, addi ionally,
on wo o he majo dAGEs: N
ε
-(ca boxye hyl)lysine (CEL) and N
δ
-(5-hyd o-5-me hyl-4-
imidazolon-2-yl)-o ni hine (MG-H1).
Conside ing he po en ial di ec in e ac ion o dAGEs wi h he colonic epi helium
and hei nume ous CRC-p omo ing e ec s, we hypo hesize a posi i e CRC isk associ-
a ion wi h highe dAGEs consump ion. We e alua ed ou hypo hesis using in o ma ion
on die a y in ake o CML, CEL, and MG-H1 in he p ospec i e Eu opean P ospec i e
In es iga ion in o Cance and Nu i ion (EPIC) coho .
2. Ma e ials and Me hods
2.1. S udy Pa icipan s
We used da a om he EPIC s udy, a la ge p ospec i e coho wi h o e hal a million
pa icipan s (n= 521,324) om 10 Eu opean coun ies (Denma k, F ance, Ge many, G eece,
I aly, he Ne he lands, No way, Spain, Sweden, and he Uni ed Kingdom) [
18
]. In b ie ,
pa icipan s aged be ween 35 and 75 yea s we e ec ui ed om 1992 o 2000 in 23 pa ici-
pa ing cen e s. An h opome ic measu es, socio-demog aphic in o ma ion, and li es yle
and die a y in ake da a we e collec ed a ec ui men om all pa icipan s. S anding heigh ,
weigh , and wais and hip ci cum e ences we e measu ed, wi h sel - epo ing excep ions
in F ance, No way, and Ox o d. Body mass index (BMI, in kg/m2) was calcula ed.
2.2. E hical Conside a ions
E hical app o al o he EPIC s udy was ob ained om he E hical Commi ee o he
In e na ional Agency o Resea ch on Cance (IARC) and local e hical commi ees. All
pa icipan s p o ided w i en consen o pa icipa e in he s udy.
2.3. Die a y Assessmen and dAGEs Es ima ion
Usual die was collec ed a baseline using a combina ion o coun y- o cen e -speci ic
ques ionnai es ha ha e been alida ed o e lec local con ex s [
19
,
20
]. Die a y da a we e
collec ed du ing in e iews in G eece, Spain, and Naples and Ragusa (I aly), whe eas
in o he EPIC cen e s sel -adminis a ed ques ionnai es we e used. Quan i a i e die a y
Nu ien s 2021,13, 3132 4 o 13
ques ionnai es we e used in Ge many, G eece, he Ne he lands, and No he n I aly; semi-
quan i a i e ood equency ques ionnai es we e employed in Denma k, No way, Naples,
Umea, and he UK; and in Malmo a combina ion o a non-quan i a i e ood- equency
ques ionnai e and a ood eco d was u ilized.
To es ima e in akes o indi idual AGEs, we used he da abase o p o ein-bound
AGEs de eloped o 190 ood i ems selec ed om he Du ch coho o EPIC and he Du ch
Na ional Food Consump ion su ey [
17
]. These oods we e ma ched o he EPIC ood lis
by name and desc ip o s, especially conside ing p epa a ion and p ocessing o expand he
EPIC Nu ien Da abase (ENDB) wi h ex a ood componen s, a p ocedu e used o o he
nu ien s/an i-nu ien s and desc ibed in de ail elsewhe e [
21
,
22
]. Fo complex oods wi h
mul iple ing edien s, he oods we e decomposed in o speci ic ing edien s o ood i ems o
gene a e EPIC dAGEs composi ion da a o each ood i em. The ea e , o each pa icipan ,
daily in akes o CML, CEL, and MG-H1 we e es ima ed.
∑
dAGEs was calcula ed as he
sum o indi idual dAGEs (CML+CEL+MG-H1) and used o pic u e o e all AGEs in ake
pa e ns. The main ood con ibu o s o die a y CML and MG-H1 we e ( om highes o
lowes con ibu ion): ce eals and ce eal p oduc s, mea s and mea p oduc s, cakes, and
biscui s (Supplemen a y Figu e S1) [
23
]. Fo CEL, he main con ibu o s we e mea and
mea p oduc s, ce eal and ce eal p oduc s, cakes, and biscui s. Dai y p oduc s, ish and ish
p oduc s, and non-alcoholic d inks we e also ele an die a y sou ces o he h ee dAGEs.
We excluded pa icipan s om G eece (n= 26,048) due o da a use es ic ions, hose
diagnosed wi h cance a baseline (n= 25,184), hose wi h missing ollow-up in o ma ion
(n= 4148)
o die a y ques ionnai e da a (n= 6259), hose in he highes o lowes 1% o
ene gy in ake e sus ene gy equi emen s (n= 9573), and a pa icipan who wi hd ew
om EPIC. Ou inal da ase included 450,111 pa icipan s, among whom 318,686 we e
women (71%).
2.4. Iden i ica ion o CRC Cases
Cance cases we e asce ained om cance egis ies in Denma k, I aly, he Ne he -
lands, No way, Spain, Sweden, and he Uni ed Kingdom, o by using a combina ion o
sou ces, including heal h insu ance eco ds, oncology and pa hology eco ds, o , in he
speci ic cases o F ance and Ge many, h ough ac i e ollow-up o he pa icipan s and
hei ela i es. CRC cases we e i s inciden , and his ologically con i med by a pa hol-
ogis . We used he In e na ional Classi ica ion o Diseases o Oncology (ICD-O, codes
C18–C20) o de ine he cases. Colon cance s we e de ined as umo s ha occu ed in he
cecum, appendix, ascending colon, hepa ic lexu e, ans e se colon, splenic lexu e, o he
descending o sigmoid colon (C18.0–C18.7), and o e lapping and/o unspeci ied o igin
umo s (C18.8 and C18.9). Rec al cance s we e de ined as umo s ha occu ed a he
ec o-sigmoid junc ion (C19) o ec um (C20).
2.5. S a is ical Analyses
Daily in akes o CML, CEL, and MG-H1 we e na u al log- ans o med and hei s an-
da dized esiduals we e compu ed by eg essing he ln- ans o med alues on pa icipan
ene gy in ake and cen e . Ln- ans o med dAGEs we e di ided in o quin iles, wi h he
i s quin ile used as he e e ence in all ou analyses. Cox p opo ional haza ds eg ession
models s a i ied by age a ec ui men (one-yea ca ego ies), sex, and cen e we e used
o compu e haza d a ios (HRs) and 95% con idence in e als (CIs) o he associa ion
be ween indi idual dAGEs and CRC isk. Time a en y was age a ec ui men , while exi
ime was se as he age a which any o he ollowing i s occu ed: CRC diagnosis, dea h,
emig a ion, o las da e a which ollow-up was conside ed comple e. To es he ends o
he associa ions, we an Cox models using median alues o each ca ego y as a con inuous
a iable. Analyses we e also conduc ed using con inuous a iables o dAGEs (pe ln(SD)
inc emen ). No de ia ion om he p opo ional haza ds assump ion was obse ed a e
assessing Schoen eld esiduals. Th ee main models we e un. Model 1 was s a i ied
by age (1-yea ca ego ies), sex, and cen e . Model 2 was addi ionally adjus ed o BMI
Nu ien s 2021,13, 3132 5 o 13
(con inuous), heigh (con inuous), and li es yle ac o s, including educa ion (none; p ima y;
echnical and p o essional; seconda y, highe ), physical ac i i y (inac i e; mode a ely in-
ac i e; mode a ely ac i e; ac i e), smoking s a us and in ensi y (ne e ; cu en smoke s,
ciga e es/day: 1–15, 16–25, >26; o me smoke s who qui : <=10, 11–20,
>20 yea s;
oc-
casional), and o al ene gy in ake (kcal/day, con inuous). Model 3 was u he adjus ed
o he Medi e anean die sco e o conside he die as a whole, and because his sco e
has also been speci ically associa ed wi h CRC isk [
24
]. We conside ed missing da a as
a sepa a e ca ego y o physical ac i i y (1.9%), educa ion (2.3%), and smoking (3.3%).
Res ic ed cubic splines we e used o model possible nonlinea ends [
25
,
26
]. Linea i y
o he associa ions was es ed using he likelihood a io es , compa ing he model wi h
only he linea e m wi h he model including bo h he linea and he cubic spline e ms.
The e was no indica ion o nonlinea associa ions in any o ou analyses. Analyses (using
Model 3) by ana omical subsi es o he colo ec um we e also un o ec al and colon
cance , and, speci ically, o p oximal and dis al colon cance s. Po en ial di e ences in
he associa ions by umo si es, i.e., ec al s. colon o p oximal colon s. dis al colon,
we e es ed using compe ing isk analyses [
27
,
28
]. S a i ied analyses by coun y, BMI, sex,
and yea s o ollow-up we e conduc ed and mul iplica i e in e ac ions we e included in
he model o e alua e po en ial he e ogenei y. To model he possible impac o e e se
causa ion, we an sensi i i y analyses by excluding he i s 2 yea s o ollow-up.
All he analyses we e ca ied ou using S a a 14.0 (S a aCo p., College S a ion, TX,
USA). We conside ed wo-sided p- alues below 0.05 as s a is ically signi ican .
3. Resul s
Table 1summa izes selec ed baseline cha ac e is ics o he s udy pa icipan s by
quin iles o
∑
dAGEs. Pa icipan s in he highes quin ile consumed mo e p ocessed mea s,
cakes, biscui s, ce eals and ce eal p oduc s, and legumes and less ui . They also ended o
consume less suga , con ec ione y, and alcohol.
In minimally adjus ed models (Model 1), CML (HR compa ing highes o lowes quin-
ile, HR
Q5 s
.
Q1
= 0.83, 95% CI = 0.77–0.90, p o end <0.001) and MG-H1
(HRQ5 s.Q1 = 0.84,
95% CI = 0.77–0.90, p o end <0.001) showed in e se associa ions wi h CRC, whe eas
no signi ican associa ion was obse ed o CEL (HR
Q5 s
.
Q1
= 0.92, 95% CI = 0.85–1.00,
p o end = 0.064) (Table 2). The signi icance o he associa ions was a enua ed a e ull
adjus men (Model 3), wi h HR (95% CI) o 0.97 (0.94–0.99) and 0.97 (0.95–1.00) o CML
and MG-H1, espec i ely.
Analyses by umo subsi es showed no signi ican he e ogenei y be ween colon and
ec al cance s, al hough he associa ion wi h ec al cance was s a is ically signi ican o
CML (HR pe ln(SD) inc emen : ec al cance HR
ln(SD)
= 0.93, 95% CI = 0.88–0.97) and
MG-H1
( ec al cance HR
ln(SD)
= 0.94, 95% CI = 0.90–0.99) (Table 3). No signi ican di e -
ence in he associa ion be ween he indi idual dAGEs and CRC isk was obse ed by sex
(Supplemen a y Table S1). In s a i ied analyses, dAGEs–CRC isk did no di e by coun y,
and ended o be es ic ed o pa icipan s wi h BMI <30 kg/m
2
(Supplemen a y Table S2).
When he ollow-up ime o cases was conside ed, dAGEs–CRC showed a g adien in he
associa ion be ween highe CRC isk obse ed and lowe ollow-up. Excluding pa icipan s
wi h ollow-up less han 2 yea s did no ma e ially change he esul s (no shown).

Nu ien s 2021,13, 3132 6 o 13
Table 1. Selec ed cha ac e is ics o he s udy pa icipan s (by quin iles o die a y AGEs), EPIC coho s udy, 1992–2014.
Quin iles o ∑AGEs In ake
Quin ile 1 Quin ile 2 Quin ile 3 Quin ile 4 Quin ile 5
Rec ui men and ollow-up
Age a ec ui men , yea s 50.9 ±9.9 50.5 ±9.7 50.2 ±9.7 50.1 ±9.7 51.5 ±9.7
Follow-up, yea s 14.1 ±4.2 14.1 ±4 14.1 ±3.9 14.1 ±3.8 14.2 ±4.1
An h opome y
BMI, kg/m225.2 ±4.2 25.3 ±4.2 25.3 ±4.2 25.3 ±4.2 25.1 ±4.2
Socio-demog aphic and li es yle *
Educa ion s a us, %
None 3.3 3.5 3.5 3.7 3.6
P ima y school 25.5 24.9 24.9 24.8 25.0
Technical o p o essional 23.0 24.0 23.9 23.3 22.7
Seconda y school 20.6 20.6 21.3 22.3 21.0
Highe educa ion 24.7 24.4 24.3 24.2 25.1
Smoking s a us, %
Ne e 38.2 41.4 43.0 44.1 46.0
Cu en , 1–<16 ciga e es/day 12.8 12.3 11.9 11.5 9.58
Cu en , 16–<=20 ciga e es/day 8.1 6.75 6.1 5.5 4.19
Cu en , >20 ciga e es/day 2.4 1.57 1.3 1.2 0.88
Fo me , qui <=10 yea s 9.7 9.89 9.7 9.7 9.15
Fo me , qui 11–<20 yea s 8.1 8.42 8.4 8.4 8.56
Fo me , qui >20 yea s 7.8 7.83 8.1 8.0 9.22
Cu en , pipe-ciga -occasional 9.5 8.46 8.5 8.6 9.31
Physical ac i i y s a us, %
Inac i e 21.2 19.8 19.2 18.6 19.0
Mode a ely inac i e 33.3 33.7 33.3 32.9 33.4
Mode a ely ac i e 25.2 26.7 27.0 27.5 27.0
Ac i e 18.5 18.0 18.4 18.5 19.0
Daily die a y in ake
Ene gy in ake, kcal 2052 ±775 2084 ±639 2091 ±585 2092 ±548 2063 ±512
Red mea , g 44.4 ±40.3 44.6 ±37 43.3 ±35.2 42.0 ±34.1 38.9 ±33.1
P ocessed mea , g 30.6 ±30.7 33.1 ±30 34.1 ±30.2 34.1 ±30.1 34.7 ±32.6
Fib e, g 20.1 ±8.1 22.0 ±7.4 23.1 ±7.2 23.9 ±7.3 25.1 ±7.9
Dai y p oduc s, g 345 ±271 332 ±237 326 ±229 324 ±224 341 ±226
Fish and shell ish, g 36.7 ±37.2 38.6 ±37.1 38.4 ±36.7 37.6 ±35.9 38.2 ±34.7
Cakes and biscui s, g 29.0 ±34.9 39.3 ±40.5 45.0 ±43.4 49.2 ±45.8 47.9 ±46.5
Ce eal and ce eal p oduc s, g 171 ±101 207 ±105 224 ±106 237 ±108 260 ±121
F ui s, nu s, and seeds, g 242 ±220 235 ±183 231 ±169 228 ±160 225 ±156
Vege ables, g 205 ±146 198 ±127 193 ±122 190 ±121 200 ±128
Legumes, g 10.2 ±18.7 13.2 ±21.7 14.8 ±23.7 16.2 ±25.7 16.8 ±27.2
Po a oes and o he ube s, g 100 ±86.3 96 ±75.1 92.0 ±69.8 89 ±68.3 93 ±69.2
Egg and egg p oduc s, g 17.6 ±18.7 18.2 ±17 18.2 ±16.5 18.1 ±16.6 17.8 ±17.1
Fa , g 79.9 ±35.5 81.3 ±29.8 81.1 ±27.6 80.6 ±26.5 78.4 ±25.9
Suga and con ec ione y, g 50.3 ±76.3 44.3 ±45.4 41.8 ±38.7 39.8 ±36.2 37.6 ±33.4
Alcohol, g 18.9 ±25 13.1 ±16.5 10.9 ±14.1 9.3 ±12.5 8.3 ±11.3
Medi e anean die sco e, %
Low 32.8 26.2 23.9 22.8 21.2
Medium 44.6 46.5 47.1 47.9 49.3
High 22.6 27.3 28.9 29.3 29.6
Abb e ia ion: AGE, ad anced glyca ion end-p oduc s; EPIC, Eu opean P ospec i e In es iga ion in o Cance and Nu i ion;
mean ±s anda d
de ia ion is p esen ed, unless o he wise s a ed; quin iles we e calcula ed using o al ene gy, cen e -s anda dized
esiduals, and log- ans o med ∑AGEs alues. * Pe cen ages do no add up o 100% because o missing alues.
Nu ien s 2021,13, 3132 7 o 13
Table 2.
Haza d a ios and 95% con idence in e als (CI) o colo ec al cance isk associa ed wi h indi idual die a y AGEs
(quin iles and con inuous), EPIC coho s udy, 1992–2014.
Die a y AGE NCases Median In ake Model 1 Model 2 Model 3
CML, mg/day
Quin ile 1 1391 1.90 1.00 (Re .) 1.00 (Re .) 1.00 (Re .)
Quin ile 2 1259 2.41 0.93 (0.86–1.00) 0.94 (0.87–1.02) 0.98 (0.90–1.06)
Quin ile 3 1210 2.75 0.91 (0.85–0.99) 0.94 (0.87–1.01) 0.98 (0.91–1.07)
Quin ile 4 1120 3.16 0.85 (0.79–0.92) 0.88 (0.81–0.95) 0.93 (0.86–1.01)
Quin ile 5 1182 4.02 0.83 (0.77–0.90) 0.87 (0.80–0.94) 0.92 (0.85–1.00)
p o end <0.001 <0.001 0.023
pe ln(SD) inc ease 0.94 (0.91–0.96) 0.95 (0.92–0.97) 0.97 (0.94–0.99)
CEL, mg/day
Quin ile 1 1214 1.37 1.00 (Re .) 1.00 (Re .) 1.00 (Re .)
Quin ile 2 1271 1.71 1.00 (0.92–1.08) 1.00 (0.92–1.08) 1.02 (0.95–1.11)
Quin ile 3 1219 1.93 0.95 (0.88–1.03) 0.95 (0.88–1.03) 0.99 (0.91–1.07)
Quin ile 4 1268 2.21 0.99 (0.91–1.07) 1.00 (0.92–1.08) 1.04 (0.96–1.13)
Quin ile 5 1190 2.85 0.92 (0.85–1.00) 0.92 (0.85–1.00) 0.97 (0.89–1.05)
p o end 0.064 0.072 0.630
pe ln(SD) inc ease 0.97 (0.95–1.00) 0.97 (0.95–1.00) 0.99 (0.96–1.01)
MG-H1, mg/day
Quin ile 1 1388 13.0 1.00 (Re .) 1.00 (Re .) 1.00 (Re .)
Quin ile 2 1250 16.7 0.94 (0.87–1.01) 0.94 (0.87–1.02) 0.98 (0.90–1.06)
Quin ile 3 1183 19.3 0.91 (0.84–0.99) 0.93 (0.86–1.00) 0.97 (0.90–1.06)
Quin ile 4 1120 22.5 0.88 (0.81–0.95) 0.89 (0.82–0.97) 0.94 (0.87–1.03)
Quin ile 5 1221 29.9 0.84 (0.77–0.90) 0.86 (0.80–0.94) 0.92 (0.85–1.00)
p o end <0.001 <0.001 0.033
pe ln(SD) inc ease 0.94 (0.92–0.97) 0.95 (0.93–0.98) 0.97 (0.95–1.00)
Abb e ia ions: AGE, ad anced glyca ion end-p oduc ; CML, N
ε
-ca boxy-me hyllysine; CEL, N
ε
-ca boxy-e hyllysine; EPIC, Eu opean
P ospec i e In es iga ion in o Cance and Nu i ion; MG-H1, N
δ
-(5-hyd o-5-me hyl-4-imidazolon-2-yl)-o ni hine; Model 1 is s a i ied by
age (1-yea ca ego ies), sex, and cen e ; Model 2 is addi ionally adjus ed o BMI, heigh , educa ion, physical ac i i y, smoking, and ene gy
in ake; Model 3 is Model 2 addi ionally adjus ed o he Medi e anean die sco e.
Table 3.
Haza d a ios and 95% con idence in e als (CI) o he isk o colo ec al cance in ana omical subsi es associa ed
wi h die a y AGEs (quin iles and con inuous), EPIC coho s udy, 1992–2014.
Median
In ake
Colon Cance Rec al Cance
All P oximal Colon Dis al Colon
N
Cases HR (95% CI ) N
Cases HR (95% CI ) N
Cases HR (95% CI ) N
Cases HR (95% CI )
CML, mg/day
Quin ile 1 1.90 873 1.00 (Re .) 399 1.00 (Re .) 397 1.00 (Re .) 518 1.00 (Re .)
Quin ile 2 2.41 774 0.95
(0.86–1.05) 357 0.95
(0.82–1.10) 317 0.87
(0.75–1.02) 485 1.02
(0.90–1.17)
Quin ile 3 2.75 786 1.01
(0.91–1.11) 366 1.01
(0.87–1.17) 319 0.92
(0.79–1.08) 424 0.95
(0.83–1.09)
Quin ile 4 3.16 759 0.99
(0.89–1.09) 345 0.95
(0.82–1.11) 351 1.04
(0.89–1.21) 361 0.81
(0.70–0.94)
Quin ile 5 4.02 805 0.98
(0.89–1.09) 389 0.99
(0.85–1.15) 342 0.97
(0.83–1.13) 377 0.81
(0.70–0.93)
p o end 0.989 0.944 0.564 <0.001
pe ln(SD) inc ease 0.99
(0.96–1.02)
1.00
(0.95–1.05)
0.99
(0.94–1.04)
0.93
(0.88–0.97)
CEL, mg/day
Quin ile 1 1.37 763 1.00 (Re .) 336 1.00 (Re .) 357 1.00 (Re .) 451 1.00 (Re .)
Quin ile 2 1.71 830 1.06
(0.95–1.17) 380 1.10
(0.94–1.27) 365 1.00
(0.86–1.16) 441 0.99
(0.86–1.13)
Quin ile 3 1.93 789 1.01
(0.91–1.12) 348 1.02
(0.87–1.19) 353 0.98
(0.84–1.14) 430 0.94
(0.82–1.08)
Nu ien s 2021,13, 3132 8 o 13
Table 3. Con .
Median
In ake
Colon Cance Rec al Cance
All P oximal Colon Dis al Colon
N
Cases HR (95% CI ) N
Cases HR (95% CI ) N
Cases HR (95% CI ) N
Cases HR (95% CI )
Quin ile 4 2.21 828 1.08
(0.97–1.19) 402 1.18
(1.01–1.37) 323 0.92
(0.79–1.07) 440 0.99
(0.86–1.14)
Quin ile 5 2.85 787 1.01
(0.91–1.12) 390 1.12
(0.96–1.31) 328 0.93
(0.79–1.08) 403 0.89
(0.77–1.03)
p o end 0.697 0.083 0.194 0.166
pe ln(SD) inc ease 1.00
(0.97–1.04)
1.03
(0.99–1.09)
0.97
(0.92–1.02)
0.96
(0.92–1.00)
MG-H1, mg/day
Quin ile 1 13.0 861 1.00 (Re .) 384 1.00 (Re .) 386 1.00 (Re .) 527 1.00 (Re .)
Quin ile 2 16.7 806 1.00
(0.91–1.10) 367 1.02
(0.88–1.19) 352 0.98
(0.84–1.13) 444 0.95
(0.83–1.09)
Quin ile 3 19.3 770 1.01
(0.91–1.11) 351 1.02
(0.87–1.18) 342 1.01
(0.86–1.17) 413 0.93
(0.81–1.07)
Quin ile 4 22.5 735 0.98
(0.89–1.09) 348 1.03
(0.89–1.20) 306 0.92
(0.78–1.08) 385 0.89
(0.77–1.02)
Quin ile 5 29.9 825 0.99
(0.90–1.10) 406 1.06
(0.91–1.23) 340 0.94
(0.81–1.10) 396 0.81
(0.71–0.94)
p o end 0.793 0.457 0.326 0.003
pe ln(SD) inc ease 0.99
(0.96–1.02)
1.03
(0.98–1.07)
0.96
(0.92–1.01)
0.94
(0.90–0.99)
Abb e ia ions: AGE, ad anced glyca ion end-p oduc ; CI, con idence in e al; CML, N
ε
-ca boxy-me hyllysine; CEL, N
ε
-ca boxy-
e hyllysine; EPIC, Eu opean P ospec i e In es iga ion in o Cance and Nu i ion; HR, haza d a io; MG-H1, N
δ
-(5-hyd o-5-me hyl-
4-imidazolon-2-yl)-o ni hine; models we e adjus ed o body mass index, heigh , educa ion, physical ac i i y, smoking, ene gy in ake and
Medi e anean die sco e and s a i ied by age (1-yea ca ego ies), sex, and cen e ; p o he e ogenei y be ween colon and ec al cance was
0.391, 0.849, and 0.825 o CML, CEL, and MG-H1, espec i ely; p o he e ogenei y be ween p oximal and dis al colon cance was 0.878,
0.793, and 0.804 o CML, CEL, and MG-H1, espec i ely.
4. Discussion
In his la ge p ospec i e s udy, we ound ha die a y in akes o CML and MG-H1, bu
no CEL, we e in e sely associa ed wi h he isk o CRC. Ou analyses did no iden i y any
he e ogenei y in hese indings by ana omical subsi e o he umo wi hin he colo ec um,
by sex, o by ollow-up ime.
Ou indings we e con a y o ou ini ial hypo hesis ha dAGE exposu e could p o-
mo e CRC de elopmen . This hypo hesis was based on conside able expe imen al e idence
sugges ing cance -p omo ing cha ac e is ics o hese compounds. Th ee main mecha-
nisms ha e been pos ula ed: i s , AGEs may bind o he RAGE ecep o in colonocy es
and, subsequen ly, p omo e and sus ain in lamma ion and oxida i e s ess [
29
,
30
]; second,
hey may modi y he composi ion o he mic obiome owa ds mic obial gene a ha a e
dele e ious o gu heal h [
31
]; and inally, hey may inc ease gu pe meabili y, he eby
allowing bac e ial ansloca ion and inc eased exposu e o colonocy es o oxic bac e ial
compounds [
32
]. In
in i o
en e ocy e models, cells ea ed wi h AGEs ha e shown highe
exp ession o RAGE and an inc ease in in lamma o y ac o s, such as IL-8, IL-1
β
, and
nuclea ac o -kappa B (NF-
κ
B) [
13
,
33
], sugges ing ha dAGEs may p oduce simila e ec s
in he gu . Ne e heless, a main condi ion o his o occu
in i o
is ha CML, CEL, and
MG-H1 need o each he colon un ans o med, in a p o ein-bound o m which could be
ecognized by RAGE and be able o in e ac wi h he cell su ace o colonocy es. This is be-
cause se e al s udies ha e epo ed ha ee AGEs o hose a ached o single amino acids
a e no as ecognizable by RAGE as p o ein-bound AGEs [
34
,
35
]. A ecen s udy using a
dynamic
in i o
model showed ha p o ein-bound dAGEs can su i e in es inal diges ion
and emain in he gas oin es inal ac [
36
]. Zenke e al. [
37
], using a model wi h casein,
ha e shown in a ecen s udy ha unglyca ed p o eins could also in e ac wi h RAGE. This
p o ides addi ional e idence suppo ing he complexi y o he AGEs, pa icula ly hei
in e ac ions in he gas oin es inal milieu. I is e iden ha mo e knowledge is needed on
Nu ien s 2021,13, 3132 9 o 13
he ole o he mic obiome and in es inal condi ions in he conse a ion o deg ada ion o
p o ein-bound dAGEs, and how his may a ec RAGE-speci ic in lamma ion.
Recen g owing e idence sugges ed ha he human gu mic obiome can me abolize
dAGEs, possibly as much as 40% o inges ed CML [
38
]. CML has been shown o be
me abolized by he mic obiome in o se e al sub-p oduc s, including biogenic amines and
a y acids, no ably N-ca boxyme hylcada e ine, N-ca boxyme hylaminopen anoic acid,
N-ca boxyme hyl-
∆
1-pipe ideinium ion, and 2-amino-6-( o mylme hylamino)hexanoic
acid [
39
,
40
]. Less is known abou he speci ic ac ions o hese ca abolic p oduc s wi hin he
colo ec um, o he possible downs eam molecules ha can be p oduced om hem. I is
possible ha hese compounds may no be ecognized by RAGE, and, hence, no induce an
in lamma o y esponse wi hin he gu . Ne e heless, dAGEs ha e been associa ed wi h
educed di e si y and ichness o he gu mic obiome, which is hough o be conduci e o a
CRC-p omo ing en i onmen [
15
]. Thus, he possible mic obial me abolism o dAGEs may
no en i ely nega e hei dele e ious p ope ies. I is also no ewo hy ha he h ee dAGEs
ha we assessed a e hough o bind o a single domain o RAGE (V domain), whe eas
o he AGE compounds, such as pen osidine, which a e much less abundan in he die ,
could bind o addi ional domains (V and C1 domains) [
41
], po en ially igge ing a s onge
in lamma o y esponse. This sugges s ha u u e s udies should also conside s udying
he po en ial dele e ious p ope ies o o he less abundan dAGEs and hei possible CRC
isk associa ions.
We a e unsu e why we obse ed in e se CRC isk associa ions wi h CML and MG-
H1, whe eas CEL demons a ed no associa ion. One possible explana ion may ela e
o he di e en die a y sou ces o hese compounds. In EPIC, CML and MG-H1 sha e
e y simila ood sou ces (e.g., mos ly ce eals and ce eal p oduc s), while CEL is de i ed
o a g ea e ex en om mea s [
23
,
42
]. Ce eals and ce eal p oduc s a e majo sou ces
o die a y ibe , which has been p e iously associa ed wi h lowe CRC isk and could
pa ially explain he in e se associa ion obse ed wi h CML and MG-H1 [
43
]. Ano he
po en ial explana ion o he di e en ial CRC isk associa ions o hese compounds may
be he chemical pa hways h ough which hey a e p oduced. Al hough he h ee AGEs
a e all likely de i ed om educing suga s, hei main p ecu so s a e eac i e dica bonyl
compounds, pa icula ly glyoxal (GO) and me hylglyoxal (MGO). CML o igina es om
GO, whe eas CEL and MG-H1 a e mainly p oduced om MGO [
44
]. In addi ion, CEL
and MG-H1 ( he wo MGO-de i ed AGEs) di e in hei amino acid con en , as CEL is
p oduced om lysine, while MG-H1 de i es om a ginine [
45
]. In e es ingly, eac i e
dica bonyl compounds a e hough o possess a glyca ing po en ial ha may be housands
o imes highe compa ed o ha o suga s such as glucose o uc ose [
46
] and, hence, he
CRC isk associa ion o hese die a y compounds and hei p ecu so s may also wa an
u he s udy.
The impo ance o s udying dAGEs in CRC de elopmen lies in se e al po en ial
public heal h measu es ha may be aken o con ol hei exposu es a wide popula ion
le els. Fo example, dAGEs may be di ec ly a ge ed h ough adop ion o speci ic cooking
me hods (e.g., s eaming) o educe he dAGE con en o speci ic p ocessed oods, o possibly
wi h an i-glyca ion die a y compounds (e.g., polyphenols) [
47
] o coun e any possible
ad e se e ec s [48,49].
Al hough, o da e, no la ge p ospec i e s udies ha e explo ed he associa ion be ween
dAGEs and CRC de elopmen , he ole o ci cula ing AGEs in CRC de elopmen has been
explo ed in h ee sepa a e case-con ol s udies nes ed wi hin p ospec i e coho s, wi h
disco dan indings. Two o he h ee s udies epo ed in e se associa ions wi h CRC [
50
,
51
],
whe eas one epo ed a posi i e associa ion in male smoke s [
52
]. I is no ewo hy ha ,
among hese p e ious s udies, wo [
50
,
52
] es ima ed AGEs using ELISA, which is now
ecognized as p o iding biased AGE es ima ion [
53
]. Ne e heless, aken oge he wi h
ou s udy, hese obse a ions sugges ha he ole o AGEs in CRC de elopmen is likely o
be qui e complex, and u he s udies o o he AGE compounds beyond he h ee s udied
he e a e also wa an ed.