1125
ORIGINAL RESEARCH
Blood lipids and p os a e cance : a Mendelian
andomiza ion analysis
Ca oline J. Bull1,2,3,4, Ca olina Bonilla1,2,3, Je M. P. Holly3,4, Clai e M. Pe ks3,4, Neil Da ies1,2,3, Philip
Haycock1,2,3, O iana Hoi Yun Yu5, J. B en Richa ds5,6, Rosalind Eeles7,8, Doug Eas on9, Zso ia Ko e-
Ja ai7, Ali Amin Al Olama9, Sa a Benlloch9, Kenne h Mui 3,10,11, G aham G. Giles12,13, Robe J.
MacInnis12,13, F ed ik Wiklund14, Hen ik G onbe g14, Ch is ophe A. Haiman15, Johanna
Schleu ke 16,17, Bø ge G. No des gaa d18, Ru h C. T a is19, Da id Neal20,21, No a Pashayan9,20,21,22,
Kay-Tee Khaw23, Jane L. S an o d24,25, William J. Blo 26, S ephen Thibodeau27, Ch is iane Maie 28,29,
Adam S. Kibel30,31, Ceza y Cybulski32, Lisa Cannon-Alb igh 33, He mann B enne 34,35,36, Jong Pa k37,
Radka Kane a38, Jyo sna Ba a39, Manuel R. Teixei a40,41, Agnieszka Micheal42, Ha de Pandha42,
Geo ge Da ey Smi h1,2, Sa ah J. Lewis1,2,3,a & Richa d M. Ma in1,3,43,a The PRACTICAL conso iumb
1School o Social and Communi y Medicine, Uni e si y o B is ol, B is ol, Uni ed Kingdom
2MRC/Uni e si y o B is ol In eg a i e Epidemiology Uni , Uni e si y o B is ol, B is ol, Uni ed Kingdom
3In eg a i e Cance Epidemiology P og amme, Uni e si y o B is ol, B is ol, Uni ed Kingdom
4IGFs and Me abolic Endoc inology G oup, School o Clinical Sciences No h B is ol, Uni e si y o B is ol, B is ol BS10 5NB, Uni ed Kingdom
5Depa men o Medicine, Di ision o Endoc inology, Epidemiology Bios a is ics and Occupa ional Heal h McGill Uni e si y, Mon eal, Quebec, Canada
6Depa men o Twin Resea ch, King’s College London, London SE1 7EH, Uni ed Kingdom
7The Ins i u e o Cance Resea ch, London SM2 5NG, Uni ed Kingdom
8The Royal Ma sden NHS Founda ion T us , London SW3 6JJ, Uni ed Kingdom
9Cen 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, S angeways Labo a o y, Wo s
Causeway, Camb idge, Uni ed Kingdom
10Wa wick Medical School, Uni e si y o Wa wick, Co en y CV4 7AL, Uni ed Kingdom
11Ins i u e o Popula ion Heal h, Uni e si y o Manches e , Manches e M13 9PL, Uni ed Kingdom
12Cance Epidemiology Cen e, The Cance Council Vic o ia, 615 S Kilda Road, Melbou ne, Vic o ia, Aus alia
13Cen 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, Vic o ia, Aus alia
14Depa men o Medical Epidemiology and Bios a is ics, Ka olinska Ins i u e, S ockholm, Sweden
15Depa 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, Cali o nia
16Depa men o Medical Biochemis y and Gene ics, Uni e si y o Tu ku and Tyks Mic obiology and Gene ics, Depa men o Medical
Gene ics, Tu ku Uni e si y Hospi al, Tu ku, Finland
17Ins i u e o Biomedical Technology/BioMediTech, Uni e si y o Tampe e and FimLab Labo a o ies, Tampe e, Finland
18Depa men o Clinical Biochemis y, He le and Gen o e Hospi al, Copenhagen Uni e si y Hospi al, He le Ring ej 75, He le DK-2730, Denma k
19Cance Epidemiology, Nu ield Depa men o Popula ion Heal h, Uni e si y o Ox o d, Ox o d, Uni ed Kingdom
20Su gical Oncology (U o-Oncology: S4), Uni e si y o Camb idge, Addenb ooke’s Hospi al, Box 279, Hills Road, Camb idge, Uni ed Kingdom
21Cance Resea ch UK Camb idge Resea ch Ins i u e, Li Ka Shing Cen e, Camb idge, Uni ed Kingdom
22Depa 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
23Camb 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
24Di ision o Public Heal h Sciences, F ed Hu chinson Cance Resea ch Cen e , Sea le, Washing on
25Depa men o Epidemiology, School o Public Heal h, Uni e si y o Washing on, Sea le, Washing on
26In e na ional Epidemiology Ins i u e, 1455 Resea ch Bl d., Sui e 550, Rock ille 20850, Ma yland
27Mayo Clinic, Roches e , Minneso a
28Depa men o U ology, Uni e si y Hospi al, Ulm, Ge many
29Ins i u e o Human Gene ics Uni e si y Hospi al, Ulm, Ge many
30B igham and Women’s Hospi al/Dana-Fa be Cance Ins i u e, 45 F ancis S ee - ASB II-3, Bos on, Massachuse s 02115
31Washing on Uni e si y, S . Louis, Missou i
32In 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
33Di ision o Gene ic Epidemiology, Depa men o Medicine, Uni e si y o U ah School o Medicine, Sal Lake Ci y, U ah
34Di ision o Clinical Epidemiology and Aging Resea ch, Ge man Cance Resea ch Cen e (DKFZ), Heidelbe g, Ge many
35Di ision o P e en i e Oncology, Ge man Cance Resea ch Cen e (DKFZ) and Na ional Cen e o Tumo Diseases (NCT), Heidelbe g, Ge many
36Ge man Cance Conso ium (DKTK), Ge man Cance Resea ch Cen e (DKFZ), Heidelbe g, Ge many
37Di ision o Cance P e en ion and Con ol, H. Lee Mo i Cance Cen e , Magnolia D ., Tampa, Flo ida 12902
38Molecula 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
39Aus 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, Aus alia
40Depa men o Gene ics, Po uguese Oncology Ins i u e, Po o, Po ugal
41Biomedical Sciences Ins i u e (ICBAS), Po o Uni e si y, Po o, Po ugal
42The Uni e si y o Su ey, Su ey GU2 7XH, Uni ed Kingdom
43Na ional Ins i u e o Heal h Resea ch, B is ol Nu i ion Biomedical Resea ch Uni , B is ol, Uni ed Kingdom
© 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
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.
Cance
Medicine Open Access
1126 © 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
C. J. Bull e al.
Blood Lipids and P os a e Cance
In oduc ion
P os a e cance is he mos p e alen male cance in Eu ope
and a majo cause o cance - ela ed dea hs [1]. Li es yle
ac o s and ela ed in e media e pheno ypes ha e been
associa ed wi h p os a e cance de elopmen and p og es-
sion in epidemiological s udies, including a posi i e asso-
cia ion be ween ci cula ing choles e ol le els and p os a e
cance [2–4]. Howe e , conclusions a e con lic ing and
i is no clea whe he hese indings e lec causali y o
Keywo ds
Choles e ol, Mendelian andomiza ion,
p os a e cance , s a ins
Co espondence
Richa d M. Ma in, School o Social and
Communi y Medicine, Uni e si y o B is ol,
B is ol, Uni ed Kingdom.
Tel: 01179287321; Fax: 0117 928 7236;
E-mail: [email p o ec ed]
Funding In o ma ion
C. J. B. is unded by he Wellcome T us 4-yea
s uden ship WT083431MA. The In eg a i e
Cance Epidemiology P og amme is suppo ed
by Cance Resea ch UK p og amme g an
C18281/A19169. The MRC IEU is suppo ed
by he Medical Resea ch Council and he
Uni e si y o B is ol (MC_UU_12013/1-9). The
NIHR B is ol Nu i ion Biomedical Resea ch Uni
is unded by he Na ional Ins i u e o Heal h
Resea ch (NIHR) and is a pa ne ship be ween
Uni e si y Hospi als B is ol NHS Founda ion
T us and he Uni e si y o B is ol. The CRUK
s udy and PRACTICAL conso ium is suppo ed
by he Canadian Ins i u es o Heal h Resea ch,
Eu opean Commission’s Se en h F amewo k
P og amme g an ag eemen no. 223175
(HEALTH-F2-2009-223175), Cance Resea ch
UK G an s C5047/A7357, C1287/A10118,
C5047/A3354, C5047/A10692, and C16913/
A6135. The Na ional Ins i u es o Heal h (NIH)
Cance Pos -Cance GWAS ini ia i e g an no.
1 U19 CA 148537-01 ( he GAME-ON
ini ia i e) and NIHR suppo o he Biomedical
Resea ch Cen e and The Ins i u e o Cance
Resea ch and Royal Ma sden NHS Founda ion
T us . The unde s had no ole in he design
and conduc o he s udy; collec ion,
managemen , analysis, and in e p e a ion o
he da a; and p epa a ion, e iew, o app o al
o he manusc ip .
Recei ed: 13 No embe 2015; Re ised: 3
Feb ua y 2016; Accep ed: 8 Feb ua y 2016
Cance Medicine 2016; 5(6):1125–1136
doi: 10.1002/cam4.695
aThese au ho s con ibu ed equally o he wo k.
bAddi ional coho membe s a e lis ed in he
suppo ing in o ma ion.
Abs ac
Gene ic isk sco es we e used as uncon ounded ins umen s o speci ic lipid
ai s (Mendelian andomiza ion) o assess whe he ci cula ing lipids causally
in luence p os a e cance isk. Da a om 22,249 p os a e cance cases and 22,133
con ols om 22 s udies wi hin he in e na ional PRACTICAL conso ium we e
analyzed. Allele sco es based on single nucleo ide polymo phisms (SNPs) p e i-
ously epo ed o be uniquely associa ed wi h each o low- densi y lipop o ein
(LDL), high- densi y lipop o ein (HDL), and iglyce ide (TG) le els, we e i s
alida ed in an independen da ase , and hen en e ed in o logis ic eg ession
models o es ima e he p esence (and di ec ion) o any causal e ec o each
lipid ai on p os a e cance isk. The e was weak e idence o an associa ion
be ween he LDL gene ic sco e and cance g ade: he odds a io (OR) pe
gene ically ins umen ed s anda d de ia ion (SD) in LDL, compa ing high- (≥7
Gleason sco e) e sus low- g ade (<7 Gleason sco e) cance s was 1.50 (95% CI:
0.92, 2.46; P = 0.11). A gene ically ins umen ed SD inc ease in TGs was weakly
associa ed wi h s age: he OR o ad anced e sus localized cance pe uni
inc ease in gene ic isk sco e was 1.68 (95% CI: 0.95, 3.00; P = 0.08). The
s12916- T a ian in 3- hyd oxy- 3- me hylglu a yl- CoA educ ase (HMGCR) was
in e sely associa ed wi h p os a e cance (OR: 0.97; 95% CI: 0.94, 1.00; P = 0.03).
In conclusion, ci cula ing lipids, ins umen ed by ou gene ic isk sco es, did
no appea o al e p os a e cance isk. We ound weak e idence ha highe
LDL and TG le els inc ease agg essi e p os a e cance isk, and ha a a ian
in HMGCR ( ha mimics he LDL lowe ing e ec o s a in d ugs) educes isk.
Howe e , in e ences a e limi ed by sample size and e idence o pleio opy.
1127
© 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
Blood Lipids and P os a e Cance
C. J. Bull e al.
a e he p oduc o con ounding by common causes o
bo h choles e ol le els and p os a e cance (e.g., aspec s
o die ), bias, o e e se causali y ( he cance causing
al e ed choles e ol me abolism) [5]. As se um choles e ol
le els can be modi ied by li es yle changes [6] and s a in
he apy [7], cla i ying he causali y o his associa ion
could in o m he de elopmen o p e en ion in e en ions
o p os a e cance .
S a ins lowe choles e ol le els by inhibi ing HMG- CoA
educ ase, he a e- limi ing enzyme in choles e ol syn hesis.
A me a- analysis o 27 obse a ional s udies concluded ha
s a in he apy educed p os a e cance by 7% (RR 0.93;
95% CI: 0.87, 0.99; P = 0.03) [8], bu a sepa a e me a-
analysis o ou andomized ials showed minimal e idence
o any associa ion (RR 1.08; 95% CI: 0.91, 1.30; P = 0.38)
[9]. Bo h analyses e ealed conside able he e ogenei y
be ween he included s udies (I2 alues >70%) and, as
p os a e cance was assessed as a seconda y ou come in
he ials, misclassi ica ion o ou come could ha e biased
he esul s. As ye , e idence ega ding s a in he apy o
p os a e cance is con o e sial [10].
Mendelian andomiza ion uses gene ic a ian s obus ly
associa ed wi h ai s o in e es (in his case, ci cula ing
low- densi y lipop o ein [LDL], high- densi y lipop o ein
[HDL], and iglyce ide [TG] le els) as ins umen al a i-
ables o make in e ences abou whe he associa ions be ween
exposu es and disease a e likely o be causal [11]. The
p inciple o Mendelian andomiza ion is ha analysis o
g oups de ined by common gene ic a ian s is analogous
o ha o an in en ion- o- ea analysis in a andomized
con olled ial, based on Mendel’s laws o seg ega ion
and independen asso men . Using gene ic a ian s as
“ins umen al a iables” o p oxy modi iable exposu es
should be uncon ounded by en i onmen al ac o s, ep-
esen li e- long exposu e, and no be subjec o e e se
causali y wi h espec o he pheno ype p oxied by he
geno ype. A gene ic sco e using se e al single nucleo ide
polymo phisms (SNPs) in combina ion can be cons uc ed
o ep esen he addi i e e ec o mul iple gene a ian s
o explain mo e o he a iance in he isk ac o o in e -
es , a oiding weak ins umen bias, and inc easing powe
[12]. Gene ic sco es using mul iple gene a ian s o lipid
ai s ha e been used p e iously o in es iga e associa ions
be ween blood lipids and ascula disease [13–15]. I he
ins umen al a iable assump ions hold, as shown in
Figu e 1, hen a es o he associa ion be ween he ins u-
men and he ou come is a es o he p esence o a
causal e ec o he in e media e on he ou come [16].
He e we es o he possible p esence o a causal e ec
o lipid ac ions (LDL, HDL, TG) on p os a e cance
using gene ic a ian s o hese ai s combined in gene ic
isk sco es using he p inciple o Mendelian andomiza-
ion [11]. As a as we know, his s udy is he i s o
employ Mendelian andomiza ion o in es iga e he asso-
cia ion be ween ci cula ing lipids and p os a e cance isk.
Me hods
Gene ic isk sco es as ins umen s o ci cula ing lipid
ac ions we e de eloped using SNPs p e iously iden i ied
by genome- wide associa ion s udies (GWAS). These sco es
we e hen applied o 22,249 p os a e cance cases and
22,133 con ols wi hin he in e na ional PRACTICAL con-
so ium [17] o whom gene ic da a we e a ailable.
S udy popula ions
PRACTICAL conso ium (p os a e cance
associa ion g oup o in es iga e cance - associa ed
al e a ions in he genome)
We in es iga ed he associa ion be ween gene ic isk sco es
o lipid ai s and p os a e cance isk in an indi idual
pa icipan me a- analysis o men in 22 s udies o he
in e na ional PRACTICAL conso ium (Table 1). Fi een
o hese s udies we e based in Eu ope, i e in No h
Ame ica, and wo in Aus alia. Geno ypic in o ma ion was
a ailable o 44,382 pa icipan s (22,249 p os a e cance
cases and 22,133 con ols). Subjec s we e p edominan ly
o Eu opean descen (mean ac oss s udies = 99%). All
s udies me he app op ia e e hical c i e ia o each coun y
in acco dance wi h he p inciples embodied in he
Decla a ion o Helsinki. Fu he de ails a e a ailable om
Figu e 1. Mendelian andomiza ion. Using gene ic a ian s as ins umen al a iables o es ablish whe he an exposu e is causally ela ed o cance .
An ins umen al a iable (gene ic a ia ion) [Z] ac s as a p oxy o en i onmen al exposu e [X], pos ula ed o in luence cance [Y]. Z is independen o
measu ed o unmeasu ed con ounde s [U]. Z only in luences Y i X →Y is causal.
Gene ic isk sco es o LDL,
HDL and TG. Comp ised o
gene ic a ian s [Z]
Ci cula ing LDL, HDL and
TG [X]
P os a e cance
[Y]
Measu ed and
Unmeasu ed
con ounde s [U]
1128 © 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
C. J. Bull e al.
Blood Lipids and P os a e Cance
Table 1. Summa y o 22 PRACTICAL case–con ol s udies, n = 44,382 men.
S udy Ac onym Coun y
n Mean %
Con ols Cases
Age a
diagnosis
(yea s)
PSA a
diagnosis
(ng/mL)
Eu opean
e hnici y
Family his o y
o p os a e
cance High g ade1
Ad anced
cance 1
Cance o he P os a e in Sweden CAPS Sweden 664 1153 66.7 79.6 100 11.9 50.0 30.3
Copenhagen P os a e Cance S udy 1 CPCS1 Denma k 2771 848 59.5 48.0 100 8.2 71.2 —
Copenhagen P os a e Cance S udy 2 CPCS2 Denma k 1009 265 58.0 36.0 100 14.7 52.2 —
Eu opean P ospec i e In es iga ion In o Cance
and Nu i ion (BPC3)
EPIC Eu ope 1079 722 61.7 — 100 — 27.9 4.0
Epidemiological in es iga ions o he chances o
p e en ing, ecognizing ea ly and op imally
ea ing ch onic diseases in an elde ly popula ion
ESTHER Ge many 318 313 65.4 58.7 100 6.4 48.0 27.6
F ed Hu chinson Cance Resea ch Cen e FHCRC USA 729 761 59.7 16.1 91.4 16.2 41.7 20.2
Po uguese Oncology Ins i u e, Po o IPO- Po o Po ugal 66 183 54.0 8.3 100 20.0 84.2 13.1
Mayo Clinic MAYO USA 488 767 65.3 15.5 100 23.3 55.3 45.5
Melbou ne Collabo a i e Coho S udy MCCS Aus alia 1170 1698 56.0 19.4 99.8 29.6 53.1 14.5
Mul ie hnic Coho S udy (BPC3) MEC USA 829 819 70.0 — 100 10.7 — 12.5
The Mo i G oup MOFFITT USA 100 414 64.5 6.5 89.4 20.1 43.1 3.5
P os a e Cance s udy Medical Uni e si y So ia PCMUS Bulga ia 140 151 68.2 23.1 100 4.1 59.6 46.7
The Poland G oup Poland Poland 359 438 65.5 40.2 100 6.9 32.8 37.1
P os a e P ojec Founda ion–Pos g adua e Medical
School, Su ey
PPF- UNIS UK 188 245 67.9 19.3 99.6 25.2 45.5 28.2
P os a e es ing o cance and T ea men P o ecT UK 1474 1563 61.2 6.3 99.7 7.0 29.9 11.4
Re ospec i e Queensland S udy (QLD) and he
P os a e Cance Suppo i e Ca e and Pa ien
Ou comes P ojec (P osCan)
QLD Aus alia 87 186 64.3 6.7 100 25.9 83.1 —
S udy o Epidemiology and Risk ac o s in Cance
He edi y
SEARCH UK 1244 1371 58.9 53.2 99.9 16.2 56.8 17.8
S ockholm 1 STHMI Sweden 2224 2006 66.7 — 100 17.1 45.5 14.4
Finnish Gene ic P edisposi ion o P os a e
Cance S udy
TAMPERE Finland 2413 2754 68.2 69.1 100 — 43.7 21.4
U.K. Gene ic P os a e Cance S udy and The
P os a e Cance Resea ch Founda ion S udy
UKGPCS UK 4182 4549 61.1 46.9 100 24.7 50.5 36.0
Ins i u ue Humangenge ik Ulm ULM Ge many 354 603 62.4 19.1 100 33.4 51.3 40.5
UTAH S udy UTAH USA 245 440 64.0 — 100 33.1 — 17.2
Missing da a we e excluded om each analysis.
—, no da a a ailable.
1Ad anced cance (T3 o T4 o SEER s aging egional o dis an ), high g ade (Gleason ≥7), low g ade (Gleason ≤6).
1129
© 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
Blood Lipids and P os a e Cance
C. J. Bull e al.
he conso ium websi e (h p://www.p ac ical.ccge.medschl.
cam.ac.uk).
Da a conce ning cance g ade and s age we e collec ed by
each s udy. Cance s we e ca ego ized as low g ade (Gleason
sco e ≤6) o high g ade (Gleason sco e ≥7), and localized
(T1 o T2 on TNM s aging, o i no a ailable, “localized”
on SEER s aging) o ad anced (T3 o T4 on TNM s aging,
o i no a ailable, “ egional” o “dis an ” on SEER s aging).
Da a we e no a ailable o g ade in wo (MEC/UTAH), o
s age in h ee (CPCS1/CPCS2/QLD), o he PRACTICAL
s udies, espec i ely.
GWAS iden i ica ion o lipid SNPs
SNPs associa ed wi h he lipid ai s unde in es iga ion
we e iden i ied by a e iew o he cu en published li -
e a u e in which pape s conce ning lipid SNPs we e iden-
i ied by en e ing app op ia e sea ch e ms in o he web
o science da abase (“Lipid,” “SNP,” “GWAS”). We selec ed
SNPs ha we e exclusi ely associa ed, a genome- wide
signi icance, wi h each lipid ai o in e es . The majo i y
o SNPs we e aken om wo landma k lipid GWAS [18,
19], he la e being he la ges gene ic associa ion s udy
o blood lipid le els o be conduc ed o da e (188,577
indi iduals). Fo comple eness, a numbe o SNPs om
o he publica ions we e also included in he analysis [14,
20]. The EPIC- No olk coho was used by he e e enced
GWASs o iden i y SNPs associa ed wi h lipid ai s; we
he e o e excluded his s udy om ou analysis o educe
he chance o unde - es ima ing he causal e ec o he
lipids on p os a e cance .
Geno yping da a
PRACTICAL samples we e geno yped using an Illumina
Cus om In inium geno yping a ay (iCOGS) consis ing
o 211,155 SNPs designed o he Collabo a i e
Oncological Gene–En i onmen S udy (COGS) (de ails
o which may be ound on hei websi e h p://www.
cogseu.o g). The a ay was speci ically designed o he
e alua ion o gene ic a ian s in b eas , o a ian, and
p os a e cance . As geno ypic in o ma ion was no a ail-
able o all SNPs in he gene ic isk sco es, we also used
SNPs ha had been impu ed using IMPUTE2 so wa e
[21]. As a sensi i i y analysis, allele equencies and ORs
o p os a e cance ou comes using impu ed and geno-
yped da a we e compa ed by c oss abula ion (Table
S1). As esul s we e simila , impu ed da a we e used in
all subsequen analyses. All SNPs wi h an indica ion o
poo impu a ion quali y we e emo ed ( 2 ha <0.30), as
we e hose wi h a mino allele equency o <1%, a call
a e o <95%, o hose ha iola ed he Ha dy–Weinbe g
equilib ium (P < 0.05).
Gene a ian s used o c ea e he gene ic isk sco es
Gene ic isk sco es we e used as ins umen s o p oxy
exposu e o ci cula ing blood lipids in a Mendelian an-
domiza ion amewo k [22]. This analysis assumes ha
he gene ic isk sco es used in he analysis in luence p os a e
cance isk only ia hei abili y o al e he speci ic lipid
ai ha hey a e ac ing as p oxies o (e.g., LDL, Fig. 1).
I o he biochemical p ocesses o ai s a e associa ed wi h
he gene ic isk sco es (i.e., ho izon al pleio opy is p esen ,
whe e pa hways om lipid- associa ed SNPs o disease a e
in ol ed ha a e independen o lipids, see Box 1) and
also di ec ly al e p os a e cance isk, his iola es a majo
assump ion o Mendelian andomiza ion by in oducing
gene ic con ounding. The e o e, only SNPs exclusi ely
associa ed a genome- wide signi icance (P < 5 × 10−8)
wi h one lipid ai (ei he LDL o HDL o TGs, bu no
mo e han 1) we e chosen o each gene ic isk sco es o
make hem as speci ic o one ai as possible (as in Holmes
e al. [15]) (Table S2), minimizing he possibili y o pleio-
opic e ec s. O he 118 SNPs iden i ied as associa ed
wi h only one lipid ai , 62 we e ei he geno yped o
impu ed in he PRACTICAL conso ium and conside ed
eligible o inclusion in he gene ic isk sco es based on
cu en biological knowledge. SNPs selec ed o he gene ic
isk sco es we e es ed o e idence o linkage disequilib-
ium (LD) using he SNAP pai wise online ool (h p://
www.b oadins i u e.o g/mpg/snap/ldsea chpw.php), an open
access esou ce which uses pai wise LD da a based on
phased geno ype da a om he In e na ional HapMap
p ojec . A h eshold 2 alue o ≥0.85 was used o indica e
LD; whe e SNPs we e in LD, he SNP wi h he la ges
e ec on he lipid ai was selec ed o he gene ic isk
sco es. SNP geno ypes we e coded as 0, 1, o 2 depending
on exposu e o he isk allele. Dosage alues ( anging om
0 o 2) we e gene a ed o impu ed SNPs. LDL, HDL,
and TG sco es we e composed o 11, 36, and 15 SNPs,
espec i ely. Gene ic isk sco es we e c ea ed by summing
he numbe o “ isk” alleles ha each o he 22,249 p os-
a e cance cases and 22,133 con ols we e exposed o,
such ha he g ea e he numbe o “ isk” alleles a man
had, he highe he sco e. “Risk” alleles we e hose ha
we e posi i ely ela ed o se um LDL o TG, o nega i ely
ela ed o se um HDL. Published e ec sizes ( he e ec
Box 1. Pleio opy
Ve ical pleio opy: A gene ic locus is linked o a cascade
o e en s. This is no gene ally a p oblem o Mendelian
andomiza ion s udies.
Ho izon al pleio opy: A gene ic locus is ela ed o mul-
iple pheno ypes. This iola es Mendelian andomiza ion
assump ions.

1130 © 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
C. J. Bull e al.
Blood Lipids and P os a e Cance
o he isk allele on he ai in SD) we e applied o each
SNP and summa ed o gi e a “weigh ed” gene ic isk sco e
o each ai [23], o each indi idual man in each s udy,
so ha when es ima ing he e ec o a uni inc ease in
he gene ic isk sco e di ec ly ansla ed o he e ec o
a SD change in he ai upon he ou come.
In a subsidia y analysis we examined he associa ion
o a SNP in 3- hyd oxy- 3- me hylglu a yl- CoA educ ase
(HMGCR) ( s12916) in ela ion o p os a e cance ou -
comes. s12916- T has been used p e iously o mimic s a in
in e en ion in o de o es ima e he causal associa ion
o s a in use wi h ype 2 diabe es and adiposi y measu es
[24]. ORs o p os a e cance ou comes we e epo ed
pe s12916- T allele.
Valida ion o gene ic isk sco es and in es iga ion
o po en ial pleio opy in he ALSPAC coho
The A on Longi udinal S udy o Pa en s and Child en
(ALSPAC) is a bi h coho es ablished o in es iga e
en i onmen al and gene ic ac o s in heal h and de elop-
men [25]. ALSPAC da a we e used o alida e he gene ic
isk sco es as ins umen s o ci cula ing lipid exposu e
using linea eg ession. Ten o he 11 SNPs in he LDL
sco e could be included in sco e alida ion analyses using
ALSPAC da a (due o poo impu a ion quali y in ALSPAC,
s1801689 could no be included in he LDL sco e alida-
ion). Thi y- i e o he 36 HDL gene ic isk sco e SNPs
could be alida ed in ALSPAC ( s1084651 was poo ly
impu ed) and 14 o he 15 TG gene ic isk sco e SNPs
we e a ailable o alida ion in he ALSPAC coho
( s11649653 was poo ly impu ed).
S a is ical analysis
The gene ic isk sco es we e en e ed in o logis ic eg es-
sion models o es ima e he e ec pe gene ically
ins umen ed SD inc ease in he lipid ai on p os a e
cance ou comes.
Ou comes in es iga ed we e: all p os a e cance (case
s. con ol s a us), g ade (high [≥7] e sus low [≤6] Gleason
sco e), and s age (ad anced e sus localized TNM o SEER
s aging). We conduc ed he analyses wi hin each o he
indi idual s udies ha make up he PRACTICAL
conso ium and hen combined he esul s in o a sum-
ma y odds a io (OR) o each ou come pe uni inc ease
in he gene ic isk sco es by ixed e ec me a- analysis
using he “me an” command in S a a .13 [26]. To es
ha he ins umen s (gene ic isk sco es) we e no associ-
a ed wi h con ounde s, we in es iga ed whe he each gene ic
isk sco e was associa ed wi h a ailable co a iables ha
could be po en ial con ounding ac o s (age, diagnos ic
p os a e speci ic an igen (PSA) le els, and amily his o y
o p os a e cance ). To accoun o po en ial con ounding
by popula ion s a i ica ion, we adjus ed o he op eigh
p inciple componen s ( a iables conce ning he popula-
ion’s gene ic a chi ec u e). All analyses we e pe o med
in S a a .13 (S a a Co p LP, 2013, College S a ion, TX).
Resul s
The gene ic isk sco es we e alida ed in ALSPAC pa -
icipan s a he age o 7 yea s (Table 2). The gene ic isk
sco e o LDL explained 0.3% o he a iabili y in ci cu-
la ing LDL and no only was mos s ongly associa ed
wi h LDL (linea eg ession coe icien ; 0.56 mmol/L LDL;
F s a : 14.11; P = 2 × 10−4), bu was also weakly associ-
a ed wi h HDL (linea eg ession coe icien ; 0.19 mmol/L
HDL; F s a : 5.32; P = 0.02). The HDL gene ic isk sco e
was he s onges ins umen , explaining 0.9% o a iabili y
in ci cula ing HDL and was associa ed exclusi ely wi h
HDL (linea eg ession coe icien ; −0.25 mmol/L HDL;
F s a : 37.60; P = 9.5 × 10−10). The TG gene ic isk sco e
was mos s ongly associa ed wi h TG and explained 0.2%
Table 2. Weigh ed gene ic isk sco e alida ion in ALSPAC (N = 4081).
Gene ic isk sco e
Change in ai le els
(mmol/L) pe uni sco e195% CI P alue 2 (%) F
LDL (10 SNPs)
LDL 0.56 0.27, 0.86 2 × 10−4 0.34 14.11
HDL 0.19 0.03, 0.35 0.021 0.13 5.32
lnTG −0.18 −0.41, 0.06 0.141 0.05 2.17
HDL (35 SNPs)
HDL −0.25 −0.32, −0.17 9.50 × 10−10 0.91 37.60
LDL −0.12 −0.25, 0.04 0.16 0.05 1.97
lnTG 0.05 −0.06, 0.17 0.36 0.02 0.36
TG (14 SNPs)
lnTG 0.35 0.13, 0.57 0.002 0.24 9.76
LDL 0.12 −0.16, 0.39 0.40 0.02 0.70
HDL −0.19 −0.34, −0.04 0.01 0.15 6.28
1TG le els ha e been na u al log ans o med.
1131
© 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
Blood Lipids and P os a e Cance
C. J. Bull e al.
o he a iabili y in ci cula ing TG (linea eg ession coe -
icien ; 0.35 mmol/L lnTG; F s a : 9.76; P = 2 × 10−3);
howe e , i was also associa ed wi h HDL (0.15% a iabili y
explained, linea eg ession coe icien ; −0.19 mmol/L HDL;
F s a : 6.28; P = 0.01). Associa ions o he gene ic isk
sco es wi h each lipid ai emained unal e ed ollowing
s a i ica ion o pa icipan s by sex (da a a ailable on
eques ).
Indi idual da a om 22,249 case and 22,133 con ol
men in 22 PRACTICAL s udies we e included in ou
analysis (Table 1). The pe cen age o high- g ade cance s
epo ed a ied be ween s udies (27.9–84.2%), as did he
p opo ion o ad anced s age cance s (3.5–46.7%). The e
was li le e idence o suppo an associa ion be ween
gene ic isk sco es and PSA a ec ui men . Family his o y
o p os a e cance was weakly, and imp ecisely, associa ed
wi h he LDL (OR: 0.50, 95% CI: 0.21, 1.19; P = 0.12)
and HDL (OR: 1.62, 95% CI: 1.02, 2.58; P = 0.04) gene ic
isk sco es. Fou o he op eigh p incipal componen s
we e associa ed wi h a leas one o he gene ic isk sco es.
Age a in e iew was no associa ed wi h he gene ic isk
sco es, wi h he excep ion o he TG sco e (linea eg es-
sion coe icien : −2.76, 95% CI: −5.52, −0.0003; P = 0.05)
(Table 3).
The pooled ORs o o e all p os a e cance isk by
lipid ai , es ima ed in ins umen al a iable analysis using
he gene ic isk sco e, we e 1.24 (95% CI: 0.90, 1.69;
P = 0.18; I2 = 4.4%), 0.99 (95% CI: 0.84, 1.17; P = 0.90;
I2 = 0%), and 1.09 (95% CI: 0.80, 1.50; P = 0.57;
I2 = 14.4%) pe gene ically ins umen ed SD inc ease in
LDL, HDL, and TG, espec i ely (Fig. 2).
The ORs o p os a e cance ou comes s a i ied by
g ade and s age a e summa ized in Table 4. The e was
weak e idence o sugges an e ec o a gene ically ins u-
men ed SD inc ease in LDL be ween high- e sus low-
g ade cance cases (OR: 1.50; 95% CI: 0.92, 2.46; P = 0.11).
The e was li le e idence o sugges an associa ion be ween
he HDL and TG gene ic isk sco es and p os a e cance
g ade: When high- and low- g ade cance s we e compa ed,
a gene ically ins umen ed SD dec ease in HDL ga e an
OR o 1.03 (95% CI: 0.79, 1.34; P = 0.82), and a gene i-
cally ins umen ed SD inc ease in TG ga e an OR o
0.93 (95% CI: 0.57, 1.52; P = 0.77). Fo cance s age, a
gene ically ins umen ed SD inc ease in LDL ga e an OR
o ad anced e sus localized cance s o 0.91 (95% CI:
0.51, 1.64; P = 0.77). A gene ically ins umen ed SD
dec ease in HDL ga e an OR o ad anced e sus local-
ized cance s o 1.02 (95% CI: 0.74, 1.39; P = 0.92). The
OR o ad anced e sus localized p os a e cance pe SD
gene ically ins umen ed inc ease in TG was 1.68 (95%
CI: 0.95, 3.00; P = 0.08).
EPIC- No olk was excluded om he analysis, as he
coho was included in he e e enced GWAS disco e y
panels. As a sensi i i y analysis, we included EPIC- No olk
in he me a- analysis. The esul s o EPIC- No olk we e
in ag eemen wi h he combined es ima es o he odds
o p os a e cance by he gene ic isk sco es (Fig. S1). As
some associa ions we e p esen be ween he gene ic isk
sco es and amily his o y o age a diagnosis (Table 3),
we pe o med a sensi i i y analysis adjus ing o hese
a iables. The esul s we e unal e ed ollowing adjus men
o age wi h he TG sco e and o amily his o y wi h
he LDL and HDL sco es (da a a ailable on eques ).
The s12916- T a ian in HMGCR was weakly associ-
a ed wi h a dec eased isk o p os a e cance o e all (OR:
0.97; 95% CI: 0.94, 1.00; P = 0.03; I2 = 0%) (Fig. 3).
Table 3. Associa ion gene ic isk sco es wi h po en ial con ounding a iables in 22,133 PRACTICAL con ol men.
Va iable nLDL HDL TG
Change in a iable pe uni inc ease gene ic isk sco e1 (95% CI), P alue
P inciple componen 1 22,133 −0.20 (−0.45, 0.04), 0.10 −0.08 (−0.29, 0.14), 0.47 −0.17 (−0.51, 0.16), 0.30
P inciple componen 2 22,133 0.80 (0.43, 1.17), 2 × 10−4 −0.06 (−0.19, 0.07), 0.35 0.70 (0.24, 1.15), 4 × 10−3
P inciple componen 3 22,133 −0.54 (−0.81, −0.27), 4 × 10−4 0.01 (−0.10, 0.12), 0.82 −0.36 (−0.60, −0.13), 4 × 10−3
P inciple componen 4 22,133 0.28 (−0.19, 0.74), 0.23 0.06 (−0.15, 0.28), 0.54 −0.49 (−0.82, −0.16), 0.01
P inciple componen 5 22,133 0.35 (−0.18, 0.87), 0.18 −0.10 (−0.29, 0.09), 0.29 0.21 (−0.20, 0.63), 0.30
P inciple componen 6 22,133 −0.56 (−1.00, −0.15), 0.01 0.11 (−0.06, 0.28), 0.19 0.19 (−0.19, 0.57), 0.32
P inciple componen 7 22,133 −0.08 (−0.39, 0.23), 0.61 0.23 (−0.06, 0.53), 0.11 0.03 (−0.39, 0.44), 0.90
P inciple componen 8 22,133 0.28 (−0.19, 0.75), 0.22 0.03 (−0.18, 0.25), 0.75 0.09 (−0.23, 0.41), 0.58
PSA (ng/mL) 5012 −0.08 (−1.79, 1.63), 0.86 −0.16 (−0.56, 0.23), 0.22 0.17 (−0.42, 0.76), 0.35
Age (yea s) 18,962 −0.68 (−3.80, 2.43), 0.65 −0.33 (−2.24, 1.58), 0.73 −2.76 (−5.52. −0.0003), 0.05
OR amily his o y pe uni inc ease in gene ic isk sco e (95% CI), P alue
Family his o y210,955 0.50 (0.21, 1.19), 0.12 1.62 (1.02, 2.58), 0.04 0.60 (0.25, 1.42), 0.24
S udies wi h mo e han 20% missing da a we e excluded om each analysis. Linea eg ession models ake clus e ing by subs udy in o accoun .
LDL, low- densi y lipop o ein; HDL, high- densi y lipop o ein; TG, iglyce ide; OR, odds a io.
1Highe LDL/TG sco es e lec inc easing ci cula ing LDL/TG, highe HDL sco es e lec lowe ci cula ing HDL.
2Family his o y o p os a e cance (in a he o b o he ): compa es Yes e sus No (logis ic eg ession).
1132 © 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
C. J. Bull e al.
Blood Lipids and P os a e Cance
The e was li le e idence o an associa ion wi h p os a e
cance s age (OR, ad anced s. localized: 0.97; 95% CI:
0.92, 1.02; P = 0.26) o g ade (OR, high s. low: 1.03;
95% CI: 0.98, 1.07; P = 0.21) (Table 5).
Discussion
We did no ind e idence o sugges an associa ion be ween
ci cula ing LDL, HDL, o TG (p oxied by gene ic isk
sco es) and o e all p os a e cance . Al hough powe o
in es iga e ad anced/high- g ade cance is limi ed (as e i-
denced by wide con idence in e als), ou esul s may
indica e a po en ial ole o LDL and TG in p os a e
cance p og ession. These indings a e clinically impo an ,
as hey highligh a dis inc ion be ween indolen disease
and mo e agg essi e cance s. Va ia ion in HMGCR has
been exploi ed p e iously o demons a e he implica ions
o s a in ea men on ype 2 diabe es and bodyweigh
[24]. Ou esul s om an analysis using a a ian in
HMGCR ( s12916- T) o p oxy s a in in e en ion sugges
ha s a ins may hold po en ial in p os a e cance p e en-
ion, bu ou in e ence is en a i e and equi es u he
in es iga ion in la ge sample sizes. Associa ions obse ed
be ween he LDL and HDL gene ic isk sco es and amily
his o y indings may be an a i ac o mul iple es ing.
Howe e , hese indings wa an u he in es iga ion wi h
la ge numbe s o inc ease he p ecision o he poin
es ima es.
Me a- analyses o obse a ional s udies and andomized
con olled ials p esen he e ogeneous indings o asso-
cia ions o choles e ol wi h p os a e cance , making i
di icul o conclude whe he choles e ol plays a ole in
p os a e cance . The mos ecen and comp ehensi e assess-
men o obse a ional da a did no show any associa ion
be ween HDL and LDL and p os a e cance , bu could
no d aw de ini i e conclusions on high- g ade p os a e
cance due o limi ed da a [5]. Ou Mendelian andomi-
za ion app oach has se e al ad an ages o e con en ional
obse a ional epidemiology: i elimina es he p oblem o
e e se causali y, as p os a e cance s a us canno al e
Table 4. Case- only analysis: weigh ed gene ic isk sco es and p os a e cance s age and g ade (PRACTICAL conso ium).
Ou come
Localized/low
g ade (n)
Ad anced/high
g ade (n) OR195% CI P alue
LDL sco e
Ad anced e sus localized 13,707 4301 0.91 0.51, 1.64 0.77
High g ade e sus low g ade 9237 8515 1.50 0.92, 2.46 0.11
HDL sco e
Ad anced e sus localized 13,707 4301 1.02 0.74, 1.39 0.92
High g ade e sus low g ade 9237 8515 1.03 0.79, 1.34 0.82
TG sco e
Ad anced e sus localized 13,707 4301 1.68 0.95, 3.00 0.08
High g ade e sus low g ade 9237 8515 0.93 0.57, 1.52 0.77
Ad anced (T3 o T4 o SEER s aging egional o dis an ), localized (T1 o T2 o SEER s aging localized), high g ade (Gleason ≥7), low g ade (Gleason ≤6).
LDL, low- densi y lipop o ein; HDL, high- densi y lipop o ein; TG, iglyce ide; OR, odds a io.
1Pe uni inc ease in gene ic isk sco e (SD ai ), adjus ed o op eigh p incipal componen s (highe LDL/TG sco es e lec inc easing ci cula ing LDL/
TG, highe HDL sco es e lec lowe ci cula ing HDL).
Figu e 2. Me a-analysis OR p os a e cance pe uni inc ease in gene ic isk sco es (SD ai ). (A) Low-densi y lipop o ein (LDL): OR 1.24
(95% CI: 0.90, 1.69), P = 0.18. (B) High-densi y lipop o ein (HDL): OR 0.99 (95% CI: 0.84, 1.17), P = 0.90. (C) T iglyce ide (TG): 1.09 (95% CI:
0.80, 1.50), P = 0.57. Cases: 22,249; con ols: 22,133. Adjus ed o op eigh p inciple componen s.
ABC
1133
© 2016 The Au ho s. Cance Medicine published by John Wiley & Sons L d.
Blood Lipids and P os a e Cance
C. J. Bull e al.
one’s ge mline gene ic makeup; gene ic isk sco es ep-
esen an indi idual’s exposu e o lipid ai s o e hei
li e ime, educing biological and echnical sou ces o meas-
u emen e o ha a ise om one- o blood sampling a
one poin in a pe son’s li e; and con ounding by beha io al,
li es yle, and o he ela ed in e media e ai s should be
minimized as indi iduals a e e ec i ely andomly alloca ed
o a low o high le el o exposu e based on hei geno-
ype, andomly gene a ed a concep ion (Mendel’s second
law o independen asso men ) [27]. Howe e , Mendelian
andomiza ion is suscep ible o gene ic con ounding i
he SNPs used as ins umen s o he ai o in e es
ha e e ec s on o he pheno ypes besides he speci ic lipid
o in e es , and i is hese o he pheno ypes which lead
o p os a e cance (ho izon al pleio opy) [16]. The e is
also po en ial o con ounding due o popula ion s a i-
ica ion; howe e , as we ha e adjus ed o p inciple
componen s in ou eg ession models, his should be
minimized. We canno be su e ha he LDL sco e is
exclusi ely associa ed wi h LDL as alida ion o he sco e
in ALSPAC e ealed possible pleio opy wi h ci cula ing
HDL. Howe e , he HDL sco e which is s ongly associ-
a ed wi h HDL, and no LDL o TG in ALSPAC, was
no associa ed wi h p os a e cance ; he e o e, he weak
associa ion no ed be ween he LDL sco e and high- g ade
p os a e cance is likely an LDL, no an HDL e ec . As
F s a is ics o he gene ic isk sco es ela e o ALSPAC
and no PRACTICAL ( he da ase o he ou come), con-
en ional h esholds, such as F > 10 a e no ele an o
his s udy. The da ase o he ou come is a la ge han
he da ase in which we es ed he gene ic isk sco e– ai
associa ion; he e o e, i may be ha he gene ic isk sco es
a e in ac s onge ins umen s o ci cula ing lipid ai s
in PRACTICAL han we es ima ed.
Figu e 3. Cases e sus con ols. Me a- analysis OR p os a e cance pe s12916- T allele. OR 0.97 (95% CI: 0.94, 1.00), P = 0.03. Cases: 22,733;
con ols: 23,050. Adjus ed o op eigh p inciple componen s.
Table 5. S a i ied analysis by cance s age and g ade. OR pe s12916- T allele. Adjus ed o op eigh p inciple componen s.
Ou come
Localized/low
g ade (n)
Ad anced/high
g ade (n) OR 95% CI P alue
s12916- T allele
Ad anced e sus localized 13,707 4301 0.97 0.92, 1.02 0.26
High g ade e sus low g ade 10,038 8543 1.03 0.98, 1.07 0.21
Ad anced (T3 o T4 o SEER s aging egional o dis an ), localized (T1 o T2 o SEER s aging localized), high g ade (Gleason ≥7), low g ade (Gleason ≤6).
OR, odds a io.