Sho Repo
In es iga ing he possible causal ole o co ee consump ion
wi h p os a e cance isk and p og ession using Mendelian
andomiza ion analysis
Amy E. Taylo
1,2
, Richa d M. Ma in
1,3,4
, Milan S. Geybels
5
, Jane L. S an o d
5,6
, I ene Shui
5
, Rosalind Eeles
7,8
,
Doug Eas on
9
, Zso ia Ko e-Ja ai
7
, Ali Amin Al Olama
9
, Sa a Benlloch
9
, Kenne h Mui
10
, G aham G Giles
11,12
,
F ed ik Wiklund
13
, Hen ik G onbe g
13
, Ch is ophe A Haiman
14
, Johanna Schleu ke
15,16
, Bø ge G. No des gaa d
17
,
Ru h C T a is
18
, Da id Neal
19
, No a Pashayan
12,20
, Kay-Tee Khaw
21
, William Blo
22
, S ephen Thibodeau
23
,
Ch is iane Maie
24,25
, Adam S Kibel
26,27
, Ceza y Cybulski
28
, Lisa Cannon-Alb igh
29
, He mann B enne
30,31,32
, Jong Pa k
33
,
Radka Kane a
34
, Jyo sna Ba a
35
, Manuel R Teixei a
36,37
, Ha de Pandha
38
, and he PRACTICAL Conso ium,
Jenny Dono an
3
and Ma cus R. Muna 
o
1,2
1
MRC In eg a i e Epidemiology Uni (IEU) a he Uni e si y o B is ol, B is ol, Uni ed Kingdom
2
School o Expe imen al Psychology and UK Cen e o Tobacco and Alcohol S udies, Uni e si y o B is ol, B is ol, Uni ed Kingdom
3
School o Social and Communi y Medicine, Uni e si y o B is ol, B is ol, Uni ed Kingdom
4
The NIHR B is ol Nu i ion Biomedical Resea ch Uni , Uni e si y Hospi als B is ol NHS Founda ion T us and he Uni e si y o B is ol
5
Di ision o Public Heal h Sciences, F ed Hu chinson Cance Resea ch Cen e , Sea le, WA
6
Depa men o Epidemiology, School o Public Heal h, Uni e si y o Washing on, Sea le, WA
Key wo ds: p os a e cance , co ee, Mendelian andomiza ion
Addi ional Suppo ing In o ma ion may be ound in he online e sion o his a icle.
Con lic s o in e es : Rosalind Eeles ecei ed an hono a ium om he geni o-u ina y Ame ican Socie y o Clinical Oncology (ASCO) mee ing 2016.
G an sponso : UK Cen e o Tobacco and Alcohol S udies, a UKCRC Public Heal h Resea ch: Cen e o Excellence ( o A.E.T. and M.R.M.);
B i ish Hea Founda ion, Cance Resea ch UK, Economic and Social Resea ch Council, Medical Resea ch Council, and he Na ional Ins i u e
o Heal h Resea ch, unde he auspices o he UK Clinical Resea ch Collabo a ion; G an sponso : Cance Resea ch UK ( o R.M.M. and
Ca oline Rel on (In eg a i e Cance Epidemiology P og amme)); G an numbe : C18281/A19169; G an sponso : Canadian Ins i u es o
Heal h Resea ch; (CRUK s udy and he PRACTICAL conso ium); G an sponso : Eu opean Commission’s Se en h F amewo k P og amme;
G an numbe : 223175 (HEALTH-F2-2009–223175); G an sponso : Cance Resea ch UK; G an numbe s: C5047/A7357, C1287/A10118,
C5047/A3354, C5047/A10692, C16913/A6135; G an sponso : Na ional Ins i u e o Heal h (NIH) Cance Pos -Cance GWAS ini ia i e ( he
GAME-ON ini ia i e); G an numbe : 1 U19 CA 148537–01; G an sponso : Eu opean Communi y’s Se en h F amewo k P og amme (iCOGS
in as uc u e); G an numbe : 223175 (HEALTH-F2-2009–223175) (iCOGS); G an sponso : Cance Resea ch UK; G an numbe s: C1287/
A10118, C1287/A 10710, C12292/A11174, C1281/A12014, C5047/A8384, C5047/A15007, C5047/A10692; G an sponso : Na ional Ins i u es
o Heal h; G an numbe : CA128978; G an sponso : Pos -Cance GWAS ini ia i e; G an numbe s: 1U19 CA148537, 1U19 CA148065 and
1U19 CA148112 (GAME-ON ini ia i e); G an sponso : Depa men o De ence; G an numbe : W81XWH-10–1-0341; G an sponso :
Canadian Ins i u es o Heal h Resea ch (CIHR) o he CIHR Team in Familial Risks o B eas Cance , Komen Founda ion o he Cu e, he
B eas Cance Resea ch Founda ion, and he O a ian Cance Resea ch Fund; G an sponso : VicHeal h and Cance Council Vic o ia (MCCS
coho ec ui men ); G an sponso : Aus alian NHMRC; G an numbe s: 209057, 251553 and 504711; G an sponso : Vic o ian Cance
Regis y (VCR) and Aus alian Ins i u e o Heal h and Wel a e (AIHW), including he Na ional Dea h Index and he Aus alian Cance
Da abase; G an sponso : U.K. Heal h Technology Assessmen (HTA) P og amme o he NIH Resea ch (P o ecT s udy); G an numbe s: HTA
96/20/99; ISRCTN20141297; G an sponso : Na ional Cance Resea ch Ins i u e (NCRI) o med by he Depa men o Heal h, he Medical Resea ch
Council and Cance Resea ch UK [NHS R&D Di ec o a e suppo ed P odigal s udy and he P oMPT (P os a e Mechanisms o P og ession and
T ea men )]; G an numbe : G0500966/75466; G an sponso : Cance Resea ch UK ( o R.A.E. and Z.K.J.); G an numbe : C5047/A7357;
G an sponso : NIHR Biomedical Resea ch Cen e a The Ins i u e o Cance Resea ch and Royal Ma sden NHS Founda ion T us ( o RAE and ZKJ);
G an sponso s: Na ional Ins i u e o Heal h Resea ch B is ol Nu i ion Biomedical Resea ch Uni based a Uni e si y Hospi als B is ol
NHS Founda ion T us and he Uni e si y o B is ol ( o R.M.M.); G an sponso : NIHR Senio In es iga o s ( o F.C.H., D.E.N. and J.L.D.);
G an sponso s: MRC and he Uni e si y o B is ol (In eg a i e Epidemiology Uni ); G an numbe s: G0600705, MC_UU_12013/6
DOI: 10.1002/ijc.30462
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and
ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed.
His o y: Recei ed 27 Ap 2016; Accep ed 5 Sep 2016; Online 14 Oc 2016
Co espondence o: D . Amy Taylo , School o Expe imen al Psychology, 12a P io y Road, Uni e si y o B is ol, B is ol, UK, Tel.: 0117 928
8547, Fax: 0117 928 8588, E-mail: [email p o ec ed]
Cance Epidemiology
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC
In e na ional Jou nal o Cance
IJC
7
The Ins i u e o Cance Resea ch, London, SM2 5NG, Uni ed Kingdom
8
The Royal Ma sden NHS Founda ion T us , London, SW3 6JJ, Uni ed Kingdom
9
S angeways Labo a o y, Cen e o Cance Gene ic Epidemiology, Depa men o Public Heal h and P ima y Ca e, Uni e si y o Camb idge, Wo s
Causeway, Camb idge, Uni ed Kingdom
10
Ins i u e o Popula ion Heal h, Uni e si y o Manches e , Manches e , Uni ed Kingdom
11
Cance Epidemiology Cen e, Cance Council Vic o ia, 615 S Kilda Road, Melbou ne, VIC, Aus alia
12
Cen e o Epidemiology and Bios a is ics, Melbou ne School o Popula ion and Global Heal h, The Uni e si y o Melbou ne, Melbou ne, VIC, Aus alia
13
Depa men o Medical Epidemiology and Bios a is ics, Ka olinska Ins i u e, S ockholm, Sweden
14
Depa men o P e en i e Medicine, Keck School o Medicine, Uni e si y o Sou he n Cali o nia/No is Comp ehensi e Cance Cen e , Los Angeles, CA
15
Depa men o Medical Biochemis y and Gene ics, Uni e si y o Tu ku, Tu ku, Finland
16
Ins i u e o Biomedical Technology/BioMediTech, Uni e si y o Tampe e and FimLab Labo a o ies, Tampe e, Finland
17
Depa men o Clinical Biochemis y, He le Hospi al, Copenhagen Uni e si y Hospi al, He le Ring ej 75, He le , 2730, Denma k
18
Cance Epidemiology Uni , Nu ield Depa men o Clinical Medicine, Uni e si y o Ox o d, Ox o d, Uni ed Kingdom
19
Su gical Oncology (U o-Oncology: S4), Uni e si y o Camb idge, Addenb ooke’s Hospi al, Hills Road, Box 279, Camb idge, Uni ed Kingdom
20
Depa men o Applied Heal h Resea ch, Uni e si y College London, 1-19 To ing on Place, London, WC1E 7HB, Uni ed Kingdom
21
Camb idge Ins i u e o Public Heal h, Uni e si y o Camb idge, Fo ie Si e, Robinson Way, Camb idge, CB2 0SR, Uni ed Kingdom
22
In e na ional Epidemiology Ins i u e, 1455 Resea ch Bl d, Sui e 550, Rock ille, MD
23
Mayo Clinic, Roches e , MN
24
Depa men o U ology, Uni e si y Hospi al Ulm, Ulm, Ge many
25
Ins i u e o Human Gene ics, Uni e si y Hospi al Ulm, Ulm, Ge many
26
B igham and Women’s Hospi al/Dana-Fa be Cance Ins i u e, 45 F ancis S ee -ASB II-3, Bos on, MA
27
Washing on Uni e si y, School o Medicine, S . Louis, MO
28
In e na ional He edi a y Cance Cen e , Depa men o Gene ics and Pa hology, Pome anian Medical Uni e si y, Szczecin, Poland
29
Di ision o Gene ic Epidemiology, Depa men o Medicine, Uni e si y o U ah School o Medicine, Sal Lake Ci y, UT
30
Di ision o Clinical Epidemiology and Aging Resea ch, Ge man Cance Resea ch Cen e (DKFZ), Heidelbe g, Ge many
31
Di ision o P e en i e Oncology, Ge man Cance Resea ch Cen e (DKFZ), Heidelbe g, Ge many
32
Ge man Cance Conso ium (DKTK), Ge man Cance Resea ch Cen e (DKFZ), Heidelbe g, Ge many
33
Di ision o Cance P e en ion and Con ol, H. Lee Mo i Cance Cen e , 12902 Magnolia D , Tampa, FL
34
Molecula Medicine Cen e and Depa men o Medical Chemis y and Biochemis y, Medical Uni e si y So ia, 2 Zd a e S , So ia, 1431, Bulga ia
35
Aus alian P os a e Cance Resea ch Cen e-Qld, Ins i u e o Heal h and Biomedical Inno a ion and School o Biomedical Sciences, Queensland
Uni e si y o Technology, B isbane, QLD, Aus alia
36
Depa men o Gene ics, Po uguese Oncology Ins i u e, Po o, Po ugal
37
Biomedical Sciences Ins i u e (ICBAS), Po o Uni e si y, Po o, Po ugal
38
Facul y o Heal h & Medical Sciences, Uni e si y o Su ey, Guild o d, Su ey, GU2 7XH, Uni ed Kingdom
Co ee consump ion has been shown in some s udies o be associa ed wi h lowe isk o p os a e cance . Howe e , i is
unclea i his associa ion is causal o due o con ounding o e e se causali y. We conduc ed a Mendelian andomisa ion
analysis o in es iga e he causal e ec s o co ee consump ion on p os a e cance isk and p og ession. We used wo gene ic
a ian s obus ly associa ed wi h ca eine in ake ( s4410790 and s2472297) as p oxies o co ee consump ion in a sample
o 46,687 men o Eu opean ances y om 25 s udies in he PRACTICAL conso ium. Associa ions be ween gene ic a ian s and
p os a e cance case s a us, s age and g ade we e assessed by logis ic eg ession and wi h all-cause and p os a e cance -
speci ic mo ali y using Cox p opo ional haza ds eg ession. The e was no clea e idence ha a gene ic isk sco e combining
s4410790 and s2472297 was associa ed wi h p os a e cance isk (OR pe addi ional co ee inc easing allele: 1.01, 95% CI:
0.98,1.03) o ha ing high-g ade compa ed o low-g ade disease (OR: 1.01, 95% CI: 0.97,1.04). The e was some e idence ha
he gene ic isk sco e was associa ed wi h highe odds o ha ing nonlocalised compa ed o localised s age disease (OR: 1.03,
95% CI: 1.01, 1.06). Amongs men wi h p os a e cance , he e was no clea associa ion be ween he gene ic isk sco e and
all-cause mo ali y (HR: 1.00, 95% CI: 0.97,1.04) o p os a e cance -speci ic mo ali y (HR: 1.03, 95% CI: 0.98,1.08). These
esul s, which should ha e less bias om con ounding han obse a ional es ima es, a e no consis en wi h a subs an ial
e ec o co ee consump ion on educing p os a e cance incidence o p og ession.
Wha ’s new?
Does co ee consump ion educe p os a e cance isk? I ’s biologically plausible ha i could, bu s udies showing a link ha e
elied on obse a ional e idence, which could be a ec ed by con ounding ac o s. These au ho s se ou o isola e co ee’s
con ibu ion. They ocused on wo gene ic a ian s ha co espond wi h ca eine in ake, and used hem as p oxies o co ee
d inking. Alleles a e no a ec ed by beha io o demog aphic ac o s, no can beha io changes a e diagnosis change
whe he a pe son ca ies an allele. The au ho s ound no co ela ion be ween ei he o he alleles and p os a e cance isk,
sugges ing d inking co ee may no p o ec agains p os a e cance .
Cance Epidemiology
Taylo e al. 323
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC
Co ee consump ion has been epo ed o be in e sely associ-
a ed wi h p os a e cance isk,
1,2
and p og ession o ad anced
disease and mo ali y.
2–6
In a ecen me a-analysis o 12 case
con ol and 9 coho s udies, he odds o p os a e cance
amongs indi iduals in he highes ca ego y o co ee con-
sump ion we e 0.91 imes ha in he lowes ca ego y.
2
E i-
dence is, howe e , mixed; no all s udies ha e ound s ong
e idence o a link be ween co ee and p os a e cance .
7,8
A
p o ec i e e ec is biologically plausible, gi en co ee’s abun-
dance o compounds wi h an i-oxidan and an i-inflamma o y
e ec s
8
and epo ed e ec s on insulin le els.
9
Howe e , in e -
ing causali y om obse a ional da a is di ficul due o o en
in ac able p oblems o con ounding and e e se causali y. Fo
example, co ee consump ion is associa ed wi h socioeconomic
s a us, alcohol consump ion and smoking.
10
Mendelian andomiza ion, which uses gene ic a ian s
ha a e associa ed wi h exposu es o in e es as p oxies o
measu ed exposu es, may help o s eng hen causal in e ence
abou po en ially modifiable exposu es.
11
Due o he way ha
alleles a e andomly assigned du ing game e o ma ion and
concep ion, alleles ha a e associa ed wi h co ee consump-
ion should no be associa ed wi h li es yle and demog aphic
ac o s which dis o he obse a ional ela ionship be ween
co ee and p os a e cance .
11
Fu he mo e, as i is no possi-
ble o change he ge mline geno ype ha an indi idual is
bo n wi h, e e se causali y is no an issue in such analyses.
Gene ic a ian s which demons a e obus associa ions
wi h ca eine in ake ha e been iden ified in ecen genome-
wide associa ion s udies (GWAS) o co ee consump ion.
12–14
Two key gene ic loci a e close o he cy och ome P450 1A1/2
(CYP1A1/CYP1A2) and a yl hyd oca bon ecep o (AHR)
genes, which a e known o play a unc ional ole in ca eine
me abolism.
12,14
CYP1A2 is he p ima y enzyme esponsible
o me abolizing ca eine, whils AHR con ols ansc ip ion
o CYP1A2.
15
Combining a ian s in hese egions in o a
mul iple allelic gene ic isk sco e inc eases he p opo ion o
a iance in ca eine consump ion explained and hence
inc eases powe .
10
I is impo an o no e ha hese a ian s
a e likely o a ec consump ion h ough hei e ec s on ca -
eine me abolism (i.e., slow me abolism o ca eine esul s in
educed consump ion), so hese ins umen s may ha e
opposing e ec s on blood ca eine le els; he allele in AHR
which inc eases co ee consump ion was associa ed wi h low-
e blood ca eine in a GWAS o blood me aboli es.
16
Al hough hese a ian s appea ela ed o ca eine in ake in
gene al a he han co ee consump ion specifically, hey
demons a e obus associa ions wi h co ee consump ion.
12
Gi en ha many o he p oposed mechanisms o he p o ec-
i e e ec o co ee a e ela ed o nonca eine compounds,
2
hese gene ic ma ke s a e likely o be in o ma i e ins umen s
o hese analyses.
Using a ian s in hese wo loci as ins umen s o co ee
consump ion, we pe o med a Mendelian andomiza ion
analysis in 46,687 p os a e cance cases and con ols om
he PRACTICAL conso ium o in es iga e whe he co ee
consump ion is causally associa ed wi h p os a e cance isk,
s age, g ade and mo ali y. I co ee consump ion causes a
educ ion in p os a e cance isk o p og ession ia com-
pounds o he han ca eine, we migh expec o see an in e se
ela ionship be ween numbe o co ee consump ion inc eas-
ing alleles and hese ou comes.
Ma e ials and Me hods
S udies
We used da a on p os a e cance cases and con ols om 25
s udies in he PRACTICAL Conso ium (PRos a e cance
AssoCia ion g oup To In es iga e Cance Associa ed aL e a-
ions in he genome, p ac ical.ccge.medschl.cam.ac.uk). Men
included in he analysis we e o Eu opean geno ypic ances y.
Full de ails o he indi idual pa icipa ing s udies ha e been
published p e iously
17,18
and a e a ailable a : h p://www.
na u e.com/ng/jou nal/ 45/n4/ex e /ng.2560-S1.pd . All s ud-
ies me he app op ia e e hical c i e ia o each coun y in
acco dance wi h he Decla a ion o Helsinki.
Geno yping
The wo ca eine- ela ed single nucleo ide polymo phisms
(SNPs) ( s4410790 in AHR and s2472297 nea CYP1A1/
CYP1A2) we e impu ed using a HapMap 2 CEU e e ence
panel om a Cus om Infinium geno yping a ay (iCOGS).
This a ay was designed o he Collabo a i e Oncological
Gene-en i onmen S udy (COGS) and consis ed o 211,155
SNPs (de ails a : h p://ec.eu opa.eu/ esea ch/heal h/medical-
esea ch/cance / p7p ojec s/cogs_en.h ml).
Full de ails o he geno yping and impu a ion ha e been
published p e iously.
17,18
A e quali y con ol, excluding SNPs
wi h low call a es (<95%) o SNPs ha de ia ed om Ha dy
Weinbe g Equilib ium in con ols (P<1310
27
), 201,598
SNPs emained. These SNPs we e used o impu e 2.6 million
SNPs; poo ly impu ed SNPs (R
2
<0.3) we e excluded.
19
Gene ic isk sco es o co ee consump ion
Gene ic isk sco es we e c ea ed by summing he numbe o
co ee consump ion inc easing alleles ( he mino allele o
s2472297 and majo allele o s4410790) o he wo SNPs,
assuming an addi i e gene ic model. We used allele dosages
om impu a ion (which ange on a con inuous scale om 0
o 2 o each gene ic locus) o indica e he numbe o co ee
inc easing alleles. This accoun s o unce ain y in he impu-
a ion o each geno ype.
Cance s age and g ade
Cance s we e ca ego ised in o low o high g ade, acco ding
o Gleason sco e (low g ade 6, high g ade 7). Cance s
we e ca ego ised in o clinically localised and nonlocalised,
using TNM s aging (T1/T2/N0/NX/M0/MX o localised, T3/
T4/N1/M1 o nonlocalised) o SEER s aging, whe e TNM
s aging was no a ailable (“local” o localised, “ egional” o
“dis an ” o nonlocalised).
Cance Epidemiology
324 Co ee and p os a e cance
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC
All cause and p os a e cance speci ic mo ali y
Analyses we e limi ed o s udies o which mo ali y ollow-
up amongs cases was a leas 90% comple e and had a leas
fi e p os a e cance dea hs ( o he p os a e cance -specific
mo ali y analysis). Indi iduals wi h i al s a us eco ded as
“unknown” we e excluded om hese analyses. Indi iduals
wi h an unknown cause o dea h we e assumed no o ha e
died o p os a e cance .
Co ee and ea consump ion
Da a on co ee and ea consump ion we e a ailable o ou o
he s udies (ESTHER, FHCRC, MCCS and UKGPCS). Ou o
hese s udies, in o ma ion on whe he co ee o ea was ca ein-
a ed o deca eina ed was only a ailable in UKGPCS. In o ma-
ion abou equency o co ee and ea consump ion was
collec ed in ca ego ies, bu o he pu poses o analysis was
ecoded o numbe o consumed cups pe day using he mid-
poin o each ca ego y. Fu he de ails o he coding o hese
a iables and how co ee and ea da a we e collec ed in each
s udy a e a ailable in Suppo ing In o ma ion (Table S1).
S a is ical analysis
Analyses we e conduc ed in S a a ( e sion 14). Associa ions
be ween he gene ic isk sco e and consump ion o co ee, ea
and co ee and ea combined we e assessed using linea
eg ession, adjus ing o he op eigh p incipal componen s
ha eflec he gene ic s uc u e o he popula ion ( o con ol
o con ounding by popula ion s a ifica ion). Robus s an-
da d e o s we e calcula ed o accoun o he igh skewed
na u e o he co ee and ea a iables. Analyses we e con-
duc ed wi hin each o he ou s udies wi h co ee and ea
consump ion da a a ailable and combined in a andom
e ec s me a-analysis using he me an command in S a a.
Associa ions be ween he co ee- ela ed SNPs and p os a e
cance isk (case/con ol s a us) we e assessed using logis ic
eg ession. Fo hese analyses, we only included s udies con-
ibu ing bo h cases and con ols (N523, P oMPT and
WUGS excluded). Wi hin p os a e cance cases, we used
logis ic eg ession o in es iga e associa ions o hese SNPs
wi h high g ade compa ed o low g ade and nonlocalised
compa ed o localised cance . Fo he nonlocalised s
Figu e 1. Associa ions o gene ic isk sco e wi h ea and co ee consump ion in ESTHER, FHCRC, MCCS and UKGPCS. [Colo igu e can be
iewed a wileyonlinelib a y.com]
Cance Epidemiology
Taylo e al. 325
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC
localised analysis, we excluded s udies wi h no nonlocalised
cance s (N52). In men diagnosed wi h p os a e cance , we
used Cox p opo ional haza ds eg ession o in es iga e whe h-
e he ca eine- ela ed SNPs we e associa ed wi h all-cause
mo ali y and p os a e cance -specific mo ali y. Fo hese
analyses, we used age a diagnosis as he s a da e and age o
dea h o age o las ollow up ( o indi iduals who we e s ill
ali e a he end o he s udy) as he censo ing da e. All analyses
o he associa ions be ween he co ee- ela ed gene ic a ian s
and p os a e cance we e adjus ed o gene ic p incipal compo-
nen s and s udy and obus s anda d e o s we e used o
accoun o clus e ing by s udy. To in es iga e be ween-s udy
he e ogenei y we calcula ed es ima es sepa a ely o each s udy
and combined hese in a fixed e ec s me a-analysis using he
me an command in S a a. Be ween-s udy he e ogenei y was
low (I
2
34%), so we epo he combined es ima es.
Resul s
A o al o 46,687 men o Eu opean ances y om 25 s udies
in he PRACTICAL conso ium con ibu ed o he analyses
(see Suppo ing In o ma ion Table S2). Mean age a p os a e
cance diagnosis was 65 yea s (SD 8) wi h mean age ac oss
he s udies anging om 59 o 72 yea s. Reflec ing he a ie y
o clinical popula ions ac oss he included s udies, he
p opo ion o men wi h nonlocalised cance anged om 0%
o 65% and wi h high g ade cance om 28% o 84%.
Associa ion o co ee SNPs wi h co ee and ea
consump ion
Da a on co ee and/o ea consump ion we e a ailable o
4,722 indi iduals (2,591 con ols and 2,131 cases). Associa-
ions be ween he gene ic isk sco e and ea and co ee con-
sump ion we e in he expec ed di ec ions and o simila
magni ude o hose obse ed in co ee consump ion
GWAS
12,13
(Fig. 1). In he combined es ima e, each addi ional
co ee consuming allele was associa ed wi h a 0.10 cup (95%
CI: 0.02, 0.19) inc ease in combined co ee and ea consump-
ion. Associa ions wi h co ee (0.06, 95% CI: 20.03, 0.15) and
ea (0.06, 95% CI: 0.003, 0.11) indi idually we e consis en bu
weake . The e was e idence o he e ogenei y in hese es ima es
be ween s udies (I
2
>33%).
Associa ion o co ee SNPs wi h p os a e cance isk, s age
and g ade
The e was no clea e idence ha he co ee- ela ed SNPs
we e associa ed wi h p os a e cance case s a us o ha ing
high g ade compa ed o low g ade disease (Table 1). The
odds a ios (OR) o p os a e cance and high g ade disease
Table 1. Associa ions o co ee ela ed SNPs wi h p os a e cance isk, s age and g ade
NOR
1
95% CI p alues I-squa ed (%)
s4410790
Con ols 23,034 – – –
All p os a e cance s 22,721 1.00 0.99 1.02 0.64 0
Localised 14,908 – – –
Nonlocalised 4,850 1.03 0.99 1.08 0.12 0
Low g ade 9,622 – – –
High g ade 9,293 1.00 0.95 1.06 0.92 21
s2472297
Con ols 23,034 – – –
All p os a e cance s 22,721 1.01 0.97 1.05 0.67 19
Localised 14,908 – – –
Nonlocalised 4,850 1.03 0.99 1.08 0.13 0
Low g ade 9,622 – – –
High g ade 9,293 1.01 0.96 1.07 0.63 12
Gene ic isk sco e
Con ols 23,034 – – –
All p os a e cance s 22,721 1.01 0.98 1.03 0.58 2
Localised 14,908 – – –
Nonlocalised 4,850 1.03 1.01 1.06 0.02 0
Low g ade 9,622 – – –
High g ade 9,293 1.01 0.97 1.04 0.68 12
Analyses a e adjus ed o p incipal componen s and s udy and obus s anda d e o s used o accoun o wi hin s udy clus e ing. Fo he case con-
ol analyses, he ollowing s udies did no con ibu e da a: P oMPT, WUGS. Fo analyses o p os a e cance s age, he ollowing s udies did no con-
ibu e da a: CPCS1, CPCS2, EPIC- No olk, QLD. Fo analyses o p os a e cance g ade, he ollowing s udies did no con ibu e da a: MEC, UTAH.
1
Associa ions a e pe co ee consump ion inc easing allele.
Cance Epidemiology
326 Co ee and p os a e cance
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC

pe addi ional co ee inc easing allele in he gene ic isk sco e
we e 1.01 (95% CI: 0.98 o 1.03) and 1.01 (95% CI: 0.97 o
1.04) espec i ely. Howe e , he e was sugges i e e idence
ha he gene ic isk sco e o co ee consump ion was associ-
a ed wi h highe odds o nonlocalised disease (OR pe co ee
inc easing allele 1.03, 95% CI: 1.01 o 1.06).
Associa ion o co ee SNPs wi h all-cause and p os a e
cance -speci ic mo ali y
The 15,555 men who con ibu ed o he all-cause mo ali y
analysis we e ollowed up o an a e age o 6.8 yea s, du ing
which 4,081 died. The 14,010 men who con ibu ed o he
p os a e-cance specific analysis we e ollowed up o an
a e age o 7.1 yea s du ing which 1,754 died o p os a e can-
ce . The e was no clea e idence ha he indi idual co ee
ela ed SNPs o he gene ic isk sco e o co ee consump ion
we e associa ed wi h all-cause mo ali y (haza d a io pe co -
ee inc easing allele o he gene ic isk sco e: 1.00 (95% CI:
0.97 o 1.04)) o wi h p os a e cance mo ali y: HR 1.03
(95% CI: 0.98 o 1.08) (Table 2). The e was no e idence o
sugges ha he p opo ional haza ds assump ion o Cox
eg ession was no me in his analysis.
Discussion
We pe o med a Mendelian andomiza ion analysis in a la ge
p os a e cance case con ol s udy o in es iga e whe he co -
ee consump ion causally influences p os a e cance incidence
and p og ession. We ound no clea e idence o sugges ha
co ee consump ion is causally associa ed wi h isk o p os a e
cance , disease g ade o mo ali y amongs men diagnosed
wi h p os a e cance .
Ou findings sugges ha obse a ional associa ions indi-
ca ing ha co ee consump ion educe p os a e cance isk
and p og ession
1,3,4,20
may no be causal and could be
explained by esidual con ounding o by o he li es yle o
demog aphic ac o s. Gi en ha he associa ions be ween he
gene ic isk sco e and blood ca eine le els may be null o in
he opposing di ec ion o co ee consump ion,
16
we canno
use hese esul s o d aw s ong conclusions abou any po en-
ial ole o ca eine in he de elopmen o p os a e cance .
Ou finding o a weak posi i e associa ion be ween he
co ee gene ic isk sco e and inc eased isk o nonlocalised
disease is in he opposi e di ec ion o obse a ional e idence
sugges ing ha co ee may educe isk o disease p og es-
sion.
3
In e es ingly, his aises he possibili y ha highe co -
ee, ea o ca eine consump ion o , con e sely, ha lowe
blood ca eine le els (due o as e ca eine me abolism) could
be associa ed wi h p og ession o mo e se e e disease. How-
e e , gi en ha he case defini ion o p os a e cance
(including s age o cance cases) and quali y o su i al
ollow-up da a di e ed be ween s udies, we canno ule ou
he possibili y ha his esul could be due o selec ion bias.
This finding equi es eplica ion in u he s udies be o e any
conclusions can be made wi h espec o causali y.
The e a e se e al limi a ions o hese analyses. Fi s , as
a o emen ioned, he e is he e ogenei y be ween s udies in
e ms o case defini ion, ea men ecei ed, classifica ion o
s age, g ade and mo ali y ollow up. Second, as discussed
p e iously and shown by he associa ions in he ou PRAC-
TICAL s udies wi h ca eine consump ion da a, hese gene ic
ins umen s a e no specific o co ee and associa e wi h con-
sump ion o o he ca eina ed be e ages (e.g., ea), and e en
wi h deca eina ed co ee.
10,13
Al hough we did no find
s ong e idence o an associa ion wi h co ee specifically in
ou subsample, co ee consump ion is widesp ead in mos
Eu opean and No h Ame ican popula ions, so i is likely
ha co ee is consumed a high enough le els in he ull sam-
ple o he gene ic ins umen o be su ficien ly s ongly asso-
cia ed wi h co ee.
21,22
Whils we canno a ibu e any e ec s
o hese a ian s o co ee specifically, lack o a nega i e asso-
cia ion o hese SNPs wi h p os a e cance ou comes s ill p o-
ides e idence agains co ee being p o ec i e o p os a e
Table 2. Associa ions o co ee ela ed SNPs wi h all-cause and p os a e cance -speci ic mo ali y in p os a e cance cases
NNdea hs
Yea s a
isk (1000s) HR
1
95% CI p alues I-squa ed (%)
s4410790
All-cause 15,555 4,081 106 1.01 0.98 1.03 0.70 0
P os a e cance -speci ic 14,010 1,754 100 1.02 0.98 1.07 0.35 7
s2472297
All-cause 15,555 4,081 106 1.00 0.92 1.08 0.95 0
P os a e cance -speci ic 14,010 1,754 100 1.04 0.96 1.13 0.33 29
Gene ic isk sco e
All-cause 15,555 4,081 106 1.00 0.97 1.04 0.91 0
P os a e cance -speci ic 14,010 1,754 100 1.03 0.98 1.08 0.22 34
Analyses a e adjus ed o p incipal componen s and s udy and obus s anda d e o s used o accoun o wi hin s udy clus e ing. Fo analyses o
all-cause mo ali y, he ollowing s udies con ibu ed da a: CAPS, CPCS1, EPIC, ESTHER, FHCRC, IPO-Po o, MAYO, MEC, PPF-UNIS, Poland, SEARCH,
TAMPERE, UKGPCS, UTAH, WUGS. Fo analyses o p os a e cance mo ali y, he ollowing s udies con ibu ed da a: CAPS, CPCS1, EPIC, ESTHER,
FHCRC, MAYO, MEC, PPF-UNIS, SEARCH, TAMPERE, UKGPCS, UTAH.
1
Associa ions a e pe co ee consump ion inc easing allele.
Cance Epidemiology
Taylo e al. 327
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC
cance . Thi dly, we we e also unable o es he associa ion o
hese ins umen s wi h po en ial con ounde s o he co ee-
p os a e cance ela ionship wi hin hese samples so canno
ule ou he possibili y o pleio opy ( ha he gene ic a ian s
ac on p os a e cance h ough pa hways un ela ed o co ee/
ca eine consump ion). SNPs in hese gene egions (AHR and
CYP1A1/2) ha e been iden ified in GWAS o blood p essu e,
bladde cance and Pa kinson’s disease,
23,24
al hough hese
may be explained by downs eam e ec s o ca eine o co ee
consump ion o me abolism. We know ha CYP1A2 me abo-
lises o he xenobio ic subs a es o he han ca eine and
al hough nei he o he SNPs used in his analysis we e ound
o associa e wi h blood me aboli es (o he han ca eine) a
genome wide significance le el,
16
we canno ule ou he pos-
sibili y ha associa ions wi h p os a e cance occu ia
me abolism o hese o he compounds. In addi ion, ciga e e
smoking inc eases ca eine me abolism ia induc ion o
CYP1A2,
25
so i is possible ha e ec s could di e in smok-
e s and nonsmoke s. In he subsample wi h in o ma ion on
smoking da a, we ound no clea e idence ha he associa-
ion o he gene ic isk sco e wi h p os a e cance di e ed
be ween e e and ne e smoke s (Suppo ing In o ma ion
Fig. 1). Howe e , i is unlikely ha we had su ficien powe
o de ec an in e ac ion. Finally, s a is ical powe o de ec
associa ions in Mendelian andomiza ion s udies is subs an-
ially lowe han con en ional obse a ional analyses.
Al hough poin es ima es a e e y close o he null o mos
findings, we canno ule ou he possibili y ha co ee may
ha e small e ec s on p os a e cance . Fo example, he me a-
analysis o co ee and p os a e cance conduc ed by Lu and
colleagues in 2014 epo s an OR o 0.96 o p os a e cance
isk o he highes (a leas 4 cups pe day) compa ed o
he lowes ca ego ies o consump ion (gene ally <1 cup pe
day).
2
This would equa e o an OR close o 0.999 o p os a e
cance isk pe addi ional 0.06 cups o co ee consumed. Ou
analysis was only powe ed o de ec ORs in he egion o
0.98 pe addi ional 0.06 cups o co ee consumed.
In conclusion, ou findings do no suppo a causal ole o
co ee consump ion in p os a e cance incidence o g ade and
sugges ha obse a ional findings ha co ee consump ion is
associa ed wi h a educed isk o p os a e cance may be due
o con ounding by o he li es yle ac o s. Fu he in es iga-
ion o ou finding ha he gene ic isk sco e was posi i ely
associa ed wi h isk o nonlocalised disease is equi ed in
samples which also ha e da a on co ee consump ion, and
which ha e g ea e powe o in es iga e a subsequen impac
on p os a e cance specific mo ali y.
Acknowledgemen s
No unding body has influenced da a collec ion, analysis o i s in e p e a-
ions. The iews exp essed a e hose o he au ho s and no necessa ily hose
o he NHS, he NIHR o he Depa men o Heal h. This publica ion is he
wo k o he au ho s, who se e as he gua an o s o he con en s o his
pape .
Re e ences
1. Cao S, Liu L, Yin X, e al. Co ee consump ion and
isk o p os a e cance : a me a-analysis o p ospec-
i e coho s udies. Ca cinogenesis 2014;35:256–61.
2. Lu Y, Zhai L, Zeng J, e al. Co ee consump ion
and p os a e cance isk: an upda ed me a-analy-
sis. Cance Causes Con ol 2014;25:591–604.
3. Wilson KM, Bal e K, Molle E, e al. Co ee and
isk o p os a e cance incidence and mo ali y in
he Cance o he P os a e in Sweden S udy. Can-
ce Causes Con ol 2013;24:1575–81.
4. Shafique K, McLoone P, Qu eshi K, e al. Co ee
consump ion and p os a e cance isk: u he e i-
dence o in e se ela ionship. Nu J 2012;11:42.
5. Wilson KM, Kaspe zyk JL, Ride JR, e al. Co ee
consump ion and p os a e cance isk and p o-
g ession in he Heal h P o essionals Follow-up
S udy. J Na l Cance Ins 2011;103:876–84.
6. Discaccia i A, O sini N, Wolk A. Co ee con-
sump ion and isk o nonagg essi e, agg essi e
and a al p os a e cance –a dose- esponse me a-
analysis. Ann Oncol 2014;25:584–91.
7. Jacobsen BK, Bjelke E, K ale G, e al. Co ee
d inking, mo ali y, and cance incidence: esul s
om a No wegian p ospec i e s udy. J Na l Can-
ce Ins 1986;76:823–31.
8. Bosi e C, S amp e MJ, Suba AF, e al. Co ee con-
sump ion and he isk o o e all and a al p os a e
cance in he NIH-AARP Die and Heal h S udy.
Cance Causes Con ol 2013;24:1527–34.
9. Tunnicli e JM, Shea e J. Co ee, glucose homeo-
s asis, and insulin esis ance: physiological mech-
anisms and media o s. Appl Physiol Nu Me ab
2008;33:1290–300.
10. McMahon G, Taylo AE, Da ey Smi h G, e al.
Pheno ype Refinemen S eng hens he Associa-
ion o AHR and CYP1A1 Geno ype wi h Ca -
eine Consump ion. PloS One 2014;9:e103448
11. Da ey Smi h G, Eb ahim S. ’Mendelian andomi-
za ion’: can gene ic epidemiology con ibu e o
unde s anding en i onmen al de e minan s o
disease? In J Epidemiol 2003;32:1–22.
12. Co nelis MC, Monda KL, Yu K, e al. Genome-
wide me a-analysis iden ifies egions on 7p21
(AHR) and 15q24 (CYP1A2) as de e minan s o
habi ual ca eine consump ion. PLoS Gene 2011;
7:e1002033.
13. Co ee and Ca eine Gene ics Conso ium,
Co nelis MC, By ne EM, e al. Genome-wide
me a-analysis iden ifies six no el loci associa ed
wi h habi ual co ee consump ion. Mol Psychia y
2015;20:647–56
14. Amin N, By ne E, Johnson J, e al. Genome-wide
associa ion analysis o co ee d inking sugges s
associa ion wi h CYP1A1/CYP1A2 and NRCAM.
Mol Psychia y 2012;17:1116–29.
15. Hukkanen J, Jacob P, III, Peng M, e al. E ec o
nico ine on cy och ome P450 1A2 ac i i y. B J
Clin Pha macol 2011;72:836–8.
16. Shin SY, Fauman EB, Pe e sen AK, e al. An a las
o gene ic influences on human blood me abo-
li es. Na Gene 2014;46:543–50.
17. Ko e-Ja ai Z, Eas on DF, S an o d JL, e al. Mul i-
ple no el p os a e cance p edisposi ion loci con-
fi med by an in e na ional s udy: he
PRACTICAL Conso ium. Cance Epidemiol Bio-
ma ke s P e 2008;17:2052–61.
18. Eeles RA, Olama AA, Benlloch S, e al. Iden ifica-
ion o 23 new p os a e cance suscep ibili y loci
using he iCOGS cus om geno yping a ay. Na
Gene 2013;45:385–91. 91e1-2.
19. Da ies NM, Gaun TR, Lewis SJ, e al. The e ec s
o heigh and BMI on p os a e cance incidence
and mo ali y: a Mendelian andomiza ion s udy
in 20,848 cases and 20,214 con ols om he
PRACTICAL conso ium. Cance Cause Con ol
2015;26:1603–16.
20. Yu X, Bao Z, Zou J, e al. Co ee consump ion
and isk o cance s: a me a-analysis o coho
s udies. BMC Cance 2011;11:96
21. F eedman ND, Pa k Y, Abne CC, e al.Associa-
ion o co ee d inking wi h o al and cause-specific
mo ali y. N Engl J Med 2012;366:1891–904.
22. Hea he ley SV, Mullings EL, Tidbu y MA, e al.
The die a y ca eine and heal h s udy: adminis a-
ion o a la ge pos al su ey in B is ol. Appe i e
2006;47:257–79.
23. Le y D, Eh e GB, Rice K, e al. Genome-wide
associa ion s udy o blood p essu e and hype en-
sion. Na Gene 2009;41:677–87.
24. Bu de T, Hall PN, Has ing E, Hindo LA,
Junkins HA, Klemm AK, MacA hu J, Manolio
TA, Mo ales J, Pa kinson H, Wel e D. The
NHGRI-EBI Ca alog o published genome-wide
associa ion s udies. A ailable a : www.ebi.ac.uk/
gwas. Accessed 01/04/2016, e sion 1.0.
25. Benowi z NL, Peng M, Jacob P, III. E ec s o cig-
a e e smoking and ca bon monoxide on chlo -
zoxazone and ca eine me abolism. Clin
Pha macol The 2003;74:468–74.
Cance Epidemiology
328 Co ee and p os a e cance
In . J. Cance : 140, 322–328 (2017) V
C2016 The Au ho s In e na ional Jou nal o Cance published by John Wiley & Sons L d on behal o
UICC