Khankari, N. K., Banbury, B. L., Borges, M. C., Haycock, P. C., & et al.

(2020). Mendelian randomization of circulating polyunsaturated fatty

acids and colorectal cancer risk. Cancer Epidemiology, Biomarkers
and Prevention, 29(4), 860–70. https://doi.org/10.1158/1055-
9965.EPI-19-0891

Peer reviewed version

Link to published version (if available):

10.1158/1055-9965.EPI-19-0891

Link to publication record on the Bristol Research Portal

PDF-document

This is the author accepted manuscript (AAM). The final published version (version of record) is available online
via American Association for Cancer Research at https://cebp.aacrjournals.org/content/29/4/860. Please refer to
any applicable terms of use of the publisher.

University of Bristol – Bristol Research Portal

General rights

This document is made available in accordance with publisher policies. Please cite only the
published version using the reference above. Full terms of use are available:
http://www.bristol.ac.uk/red/research-policy/pure/user-guides/brp-terms/
Supplementary Table 1. Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO) and ColoRectal Transdisciplinary Study
(CORECT) sample sizes

GECCO CORECT
Study (Genotyping Array) Total Controls Cases Study (Genotyping Array) Total Controls Cases
ARCTIC (Affymetrix 500k) 1115 522 593 ATBC (OncoArray) 177 30 147
ASTERISK (300k) 1839 947 892 CCFR (Axiom) 1851 661 1190
CCFR 1 (1M, 1M duo) 2012 974 1038 CCFR (OncoArray) 2124 534 1590
CCFR 2 (1M, 1M duo) 706 380 326 ColoCare Heidelberg (OncoArray) 223 36 187
Colo2&3 (300k) 211 124 87 ColoCare Seattle (OncoArray) 169 0 169
DACHS 1 (300k) 3409 1702 1707 CPSII (Axiom) 1076 536 540
DACHS 2 (OmniExpress) 1164 498 666 ESTHER VERDI (OncoArray) 817 420 397
DALS 1 (550k, 610k) 1411 709 702 Kentucky (Axiom) 2167 1132 1035
DALS 2 (300k) 863 461 402 Kiel (OncoArray) 1103 0 1103
HPFS 1 (OmniExpress) 452 229 223 MCCS (Axiom) 994 463 531
HPFS 2 (OmniExpress) 346 172 174 MCCS (OncoArray) 349 171 178
MEC (300k) 672 346 326 MEC (OncoArray) 144 81 63
NHS 1 (OmniExpress) 1165 774 391 MECC (Axiom) 1901 808 1093
NHS 2 (OmniExpress) 339 181 158 MECC (Omni) 978 495 483
PHS (OmniExpress) 764 389 375 MECC (OncoArray) 2215 911 1304
PLCO 1 (300/240S, 610k) 2496 1972 524 MSKCC (OncoArray) 68 0 68
PLCO 2 (300k) 889 414 475 NFCCR (Axiom) 660 467 193
PMH (300k) 398 122 276 NHS2 (OncoArray) 167 80 87
VITAL (300k) 562 287 275 Spain (OncoArray) 1546 786 760
WHI 1 (550k, 550kduo,610k) 1975 1526 449 SWEDEN Lindblom (OncoArray) 4785 2281 2504
WHI 2 (300k) 1965 1006 959 SWEDEN Wolk (OncoArray) 1397 831 566
UK SEARCH (OncoArray) 4288 115 4173
USC HRT (OncoArray) 708 387 321
GECCO TOTAL 24,753 13,735 11,018 CORECT TOTAL 29,907 11,225 18,682
Supplementary Table 2. Details regarding the SNPs used in each PUFA-specific genetic instrument (adapted
from Khankari, et al. 2016, British Journal of Cancer)

GRCh37/hg19 P % VEc % VE F statistic

Chr SNP Gene Allelea EAF Βb SE
Position value per allele per IVd per IVe
Linoleic acid (LA, 18:2n6)
10 rs10740118 65101207 JMJD1C C/G 0.56 0.248 0.043 8.08x10-9 0.2-0.7
11 rs174547 61570783 FADS1 T/C 0.32 1.474 0.042 4.98x10-274 7.6-18.1
11 rs2727270 61603237 FADS2 T/C 0.44 0.690 0.070 2.60x10-21 0.5-2.4
16 rs16966952 15135943 PDXDC1 A/G 0.31 0.351 0.044 1.23x10-15 0.5-2.5 8.8-23.6f 1,104-3,533
Arachidonic acid (AA, 20:4n6)
11 rs174547 61570783 FADS1 T/C 0.68 1.691 0.025 3.00x10-971 32.63
16 rs16966952 15135943 PDXDC1 A/G 0.69 0.199 0.031 2.43x10-10 0.44 33.07 11,302
α-linolenic acid (ALA, 18:3n3)
11 rs174547 61570783 FADS1 T/C 0.33 0.016 0.001 3.47x10-64 1.03 1.03 476
Eicosapentaenoic acid (EPA, 20:5n3)
6 rs3798713 11008622 ELOVL2 C/G 0.43 0.035 0.005 1.93x10-12 0.36
11 rs174538 61560081 FEN1 A/G 0.72 0.083 0.005 5.37x10-58 1.69 2.05 479
Docosapentaenoic acid (DPA, 22:5n3)
2 rs780094 27741237 GCKR T/C 0.41 0.017 0.003 9.04x10-09 0.46
6 rs3734398 10982973 ELOVL2 T/C 0.43 0.040 0.003 9.61x10-44 2.74
11 rs174547 61570783 FADS1 T/C 0.67 0.075 0.003 3.79x10-154 8.38 11.58 1,997
Docosahexaenoic acid (DHA, 22:6n3)
6 rs2236212 10995015 ELOVL2 C/G 0.57 0.113 0.014 1.26x10-15 0.65 0.65 299
a Allele associated with an increase in PUFA levels is in bold, and is considered the effect allele.
b Effect estimates for plasma phospholipid levels of polyunsaturated fatty acids (PUFAs, % of total fatty acids).

c % variation explained (VE) = ∑𝑛[2𝛽 2 (𝑀𝐴𝐹)(1 − 𝑀𝐴𝐹)/𝑣𝑎𝑟𝑖𝑎𝑛𝑐𝑒(𝑃𝑈𝐹𝐴)] ∗ 100.[22]

𝑖 𝑖
d
% VE per IV = sum of the %VE per allele for each SNP included in the IV.
e
F-statistic is a measure of the strength of the genetic instrument and is calculated as follows: [R2*(n-1-k)] / [(1-R2)*k], where R2 = % variation explained, n= sample
size, k=total number of instrumental variables.
f Ranges for % VE per SNP and % VE IV as reported in Guan, et al., 2014.[20]
Supplementary Table 3. Pearson correlation between PUFA wGSs

Omega-6 Omega-3
PUFA LA AA ALA EPA DPA DHA
wGS wGS wGS wGS wGS wGS
LA
1
wGS

AA -0.915
1
wGS (p < 0.05)

ALA 0.944 -0.993

1
wGS (p < 0.05) (p < 0.05)

EPA -0.774 0.843 -0.849

1
wGS (p < 0.05) (p < 0.05) (p < 0.05)

DPA -0.801 0.841 -0.848 0.920

1
wGS (p < 0.05) (p < 0.05) (p < 0.05) (p < 0.05)

DHA 0.006 -0.002 0.003 -0.416 -0.487

1
wGS (p = 0.359) (p = 0.700) (p = 0.600) (p < 0.05) (p < 0.05)
Supplementary Table 4. Associationa between one standard deviation increase in genetically-predicted circulating PUFAs and potential
confounders of the PUFA and colorectal cancer association in GECCO

Omega-6 Omega-3
LA AA ALA EPA DPA DHA
Confoundersb
wGS wGS wGS wGS wGS wGS
β p value β p value β p value β p value β p value β p value
Education
High school/ GED 0.016 0.53 -0.025 0.32 0.024 0.34 -0.003 0.91 0.005 0.84 -0.049 4.7x10-2
Some college/ technical school 0.018 0.48 -0.033 0.19 0.033 0.19 -0.026 0.31 -0.014 0.59 -0.023 0.36
College graduate/ graduate degree 0.051 4.6x10-2 -0.060 1.8x10-2 0.064 1.3x10-2 -0.049 0.06 -0.037 0.15 -0.044 0.08
Family history -0.048 9.4x10-3 0.054 3.6x10-3 -0.054 3.6x10-3 0.053 4.6x10-3 0.058 1.8x10-3 -0.016 0.40
Aspirin/NSAID use -0.017 0.24 0.015 0.28 -0.019 0.19 0.030 3.3x10-2 0.032 2.6x10-2 -0.017 0.23
Body mass index
< 18.5 kg/m2 -0.049 0.48 0.033 0.63 -0.036 0.60 -0.018 0.80 0.022 0.75 0.019 0.78
25.0 to 30.0 0.006 0.69 0.001 0.93 0.001 0.98 -0.003 0.86 -0.004 0.78 0.013 0.40
≥ 30 -0.024 0.20 0.031 0.10 -0.027 0.15 0.041 3.0x10-2 0.035 0.06 -0.029 0.12
Ever smokers -0.014 0.30 0.014 0.28 -0.013 0.31 0.008 0.56 0.011 0.42 0.008 0.53
Alcohol intake
1-28 g/day -0.019 0.23 0.019 0.90 -0.021 0.21 0.014 0.39 -0.001 0.79 0.018 0.28
> 28 -0.003 0.90 0.003 0.23 0.001 0.98 -0.003 0.91 -0.007 0.98 -0.003 0.90
Folate intake -0.423 0.82 1.640 0.37 -1.272 0.48 0.355 0.85 1.133 0.53 1.268 0.49
Red meat intake 0.004 0.33 -0.003 0.49 0.003 0.43 0.001 0.91 -0.001 0.92 -0.005 0.23
Total vegetable intake -0.007 0.51 0.008 0.44 -0.007 0.47 -0.003 0.74 0.008 0.44 0.003 0.76
Total fruit intake 0.012 0.21 -0.015 0.11 0.014 0.14 -0.022 1.8x10-2 -0.021 2.3x10-2 0.019 3.5x10-2
Physical activity -0.028 0.15 0.007 0.71 -0.014 0.47 0.021 0.29 0.017 0.37 -0.008 0.66
a Associations modeled with the confounder as the outcome in either a linear (continuous), logistic regression (binary), or polytomous (categorical) adjusted for age, sex, study, and the top three principal
components for European ancestry.
b
Confounders measured as follows: Education measured as highest level completed (categories compared to 8th grade or less); Family history in first-degree relative (Yes/No); regular aspirin/NSAID use (Yes/No);
Body mass index (kg/m2; categories compared to those 18.5 to 24.9 kg/m2); Ever smokers (Yes/No); Alcohol intake (g/day; compared to non-drinkers); Folate intake from diet (µg/day); Red meat intake
(servings/day); Total vegetable intake (servings/day); Total fruit intake (servings/day); and Physical activity (vigorous and moderate; hours/week).
Supplementary Table 5. Inverse-variance weighted Mendelian randomization and sensitivity analyses using summary statistics for the association between
polyunsaturated fatty acids (PUFA) and colorectal cancer risk in the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO) and ColoRectal
Transdisciplinary Study (CORECT)

Omega-6 Omega-3
Polyunsaturated fatty acids Polyunsaturated fatty acids
Model Linoleic acid Arachidonic acid α-linolenic acid Eicosapentaenoic acid Docosapentaenoic acid Docosahexaenoic acid
(LA) (AA) (ALA) (EPA) (DPA) (DHA)
OR a 95% CI p OR a 95% CI p OR a 95% CI p OR a 95% CI p OR a 95% CI p OR a 95% CI p
Inverse-variance MRb
GECCO 0.95 0.92-0.97 1.7x10-4 1.05 1.03-1.08 8.3x10-5 0.95 0.93-0.98 1.2x10-4 1.04 1.01-1.07 5.1x10-3 1.03 1.01-1.06 1.4x10-2 1.01 0.98-1.03 0.67
-4 -5 -4 -2
CORECT 0.95 0.92-0.98 2.6x10 1.05 1.02-1.07 9.8x10 0.96 0.93-0.98 2.9x10 1.03 1.00-1.06 3.1x10 1.02 0.99-1.04 0.15 1.03 1.01-1.06 1.5x10-2
Egger regressionc
GECCO 0.90 0.85-0.96 7.9x10-2 Non-estimable Non-estimable Non-estimable 1.14 1.12-1.16 <0.001 Non-estimable
CORECT 0.91 0.81-1.01 0.23 Non-estimable Non-estimable Non-estimable 1.11 1.03-1.20 8.0x10-3 Non-estimable
Weighted-mediand
GECCO 0.94 0.91-0.96 7.7x10-5 Non-estimable Non-estimable Non-estimable 1.04 0.95-1.14 0.38 Non-estimable
CORECT 0.95 0.92-0.97 1.8x10-3 Non-estimable Non-estimable Non-estimable 1.02 0.94-1.11 0.67 Non-estimable
Multivariable MRe
GECCO 0.99 0.91-1.09 0.88 1.11 0.68-1.79 0.68 0.97 0.73-1.29 0.83 1.11 0.82-1.51 0.50 0.82 0.59-1.13 0.22 0.96 0.92-1.01 0.09
CORECT 1.04 0.96-1.13 0.31 0.97 0.62-1.51 0.88 0.92 0.71-1.20 0.53 1.01 0.75-1.35 0.97 0.94 0.69-1.27 0.66 1.02 0.97-1.06 0.46
Removing one SNPf
GECCO 0.98 0.85-1.12 0.76 Non-estimable Non-estimable Non-estimable 0.96 0.86-1.07 0.49 Non-estimable
CORECT 0.97 0.80-1.19 0.78 Non-estimable Non-estimable Non-estimable 0.93 0.88-0.97 2.1x10-3 Non-estimable
a Odds ratios (ORs) and 95% confidence intervals (95% CIs) represent one standard deviation increase in PUFA-specific wGS which corresponds to the following increase in % of total plasma fatty acids: 1.18% increase in LA; 1.11%

increase in AA; 0.01% increase in ALA; 0.06% increase in EPA; 0.06% increase in DPA; and 0.08% increase in DHA.
b
Inverse-variance weighted Mendelian randomization (MR) fixed-effect estimate calculated using summary statistics from PUFA GWAS and GECCO.[28]
c
Egger regression bias-reduced estimate calculated using summary statistics in the presence of potential directional pleiotropy of SNPs used in the genetic instrument.[29] In GECCO, intercepts from Egger regression were statistically
significant for DPA (p value < 0.0001). In CORECT, intercepts from Egger regression were statistically significant for DPA (p value = 1.3x10-2). Statistically significant intercept p values from Egger regression are an indication of potential
directional pleiotropy of SNPs utilized in the genetic instrument. Egger regression was not conducted for PUFAs with fewer than 3 SNPs included in the genetic instrument.
d Weighted-median regression estimates the association under the assumption that 50% of the genetic variants used in the instrumental variable are invalid instruments.[30] 95% confidence intervals calculated using bootstrapped

standard errors.
e Multivariable weighted-regression adjusted for all other PUFAs [31, 32] and estimated using all nine independent PUFA SNPs identified from GWAS.

f Leave-one-out analysis provides the inverse-variance MR estimate after excluding one SNP at a time from the instrument. Estimates reported in the table are the MR results after removing the most influential SNP based on largest

change in estimate relative to the inverse-variance MR estimate that utilized all SNPs as the instrument. rs174547 was excluded from the instruments for LA and DPA.[27]