Background
Fatty acid binding protein 4 (FABP-4) is a lipid-binding adipokine mainly expressed in adipocytes and macrophages. FABP-4 is involved in transporting fatty acids to cellular compartments, modulating intracellular lipid metabolism, and regulating gene expression [
1]. In humans, elevated FABP-4 concentrations have been associated with obesity, insulin resistance, atherosclerosis, type 2 diabetes, and metabolic syndrome [
2‐
6]. Evidence for a causal positive association between body mass index (BMI) and FABP-4 levels has been provided by Mendelian randomization (MR) studies [
7,
8]. Pro-inflammatory properties of FABP-4 have also been described [
6]. FABP-4 has been shown to independently predict inflammation and fibrosis in non-alcoholic fatty liver disease (NAFLD) and may have a direct pathogenic link to disease progression [
9,
10]. Based on observations in breast cancer FABP-4 has been suggested as a factor that may promote obesity-associated cancer initiation and progression [
11]. FABP-4 expression in adipocytes has been reported to play a key role in the progression and metastasis of ovarian cancer by facilitating fatty acid supply for rapid tumor growth [
12]. Colorectal cancer (CRC) represents another tumor in which adipocytes are an integral part of the tumor micro-environment [
13], therefore circulating FABP-4 may play a role in CRC development by providing a fatty acid supply for tumor growth. In addition, FABP-4 may affect CRC development through its effects on inflammation [
14] and insulin resistance [
15‐
18], two pathways that have been demonstrated to play a role in obesity-associated CRC.
Higher FABP-4 concentrations have been observed in CRC patients than in controls in two small clinical studies from China [
19,
20]. In the largest of the two studies (100 CRC cases), it was also shown that FABP-4 expression was statistically significantly higher in tumor tissues than in adjacent tissues [
20]. In research focusing on colon adenocarcinoma, a higher FABP-4 expression has been observed in tumor than in adjacent tissues [
21] and
FABP4 was part of an 11-gene risk score that predicts recurrence of colon adenocarcinoma [
22]. Collectively, laboratory and epidemiological research suggests the potential involvement of FABP-4 in CRC development. However, evidence from prospective studies on the association between circulating FABP-4 and the risk of CRC is so far lacking. Clarifying the role of FABP-4 in CRC development is important because it may potentially serve as an obesity-associated biomarker that may help identify individuals at high risk of disease who might specifically benefit from primary or secondary prevention strategies. We hypothesized that higher FABP-4 could be positively associated with CRC risk, either directly, by facilitating fatty acid supply for tumor growth [
13], or indirectly through FABP-4-related enhancement of inflammation [
14] and insulin resistance [
15‐
18] (Additional file
1: Fig. S1).
Here, we investigated the association between FABP-4 concentrations measured in baseline blood samples and subsequent risk of CRC using data from a nested case-control study in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, stratified by sex and tumor location. In MR, genetic variants associated with circulating biomarker levels can be used to assess causal associations by circumventing common types of bias in observational studies such as residual confounding and reverse causation [
23]. Thus, to further improve causal inference, we additionally conducted a two-sample MR study using data from the Genetics and Epidemiology of Colorectal Cancer Consortium, Colorectal Cancer Transdisciplinary Study, and Colon Cancer Family Registry [
24,
25]. In two-sample MR, gene-exposure and gene-outcome associations are derived from non-overlapping samples [
26], and a combined ratio estimate is calculated to estimate the causal association between the exposure and the outcome. We examined horizontal pleiotropy and colocalization to address the validity of MR assumptions.
Results
Baseline characteristics in CRC cases versus controls showed that controls had more often a university degree, were less often physically inactive, less often current smokers, had less often diabetes, and had a lower BMI, waist circumference, and A-body shape index (ABSI, Additional file
1: Tab. S1). In women, BMI did not differ significantly between CRC cases and controls, but both waist circumference and ABSI were higher in cases than in controls (Additional file
1: Tab. S2). Median FABP-4 concentration (25th, 75th percentile) was 15.1 (11.0, 20.5) ng/ml in controls and with 15.3 (11.1, 21.3) ng/ml slightly higher in CRC cases (Additional file
1: Tab. S1). In women, median FABP-4 concentration was higher in cases (19.3 ng/ml) than in controls (18.3 ng/ml), while the contrast in men was less strong (median in cases 12.4 ng/ml, and in controls 12.1 ng/ml, Supplemental Table S
2). Among controls, men had substantially lower FABP-4 concentrations (median 12.1, 25th percentile 9.0, 75th percentile 16.0 ng/ml) than women (median 18.3, 25th percentile 14.0, 75th percentile 24.5,
p-value for sex-difference from the Kruskal-Wallis test < 0.0001), which is why we divided participants in sex-specific quintiles (cut-offs based on controls) (Additional file
1: Tab. S3). Compared with female controls, male controls had higher BMI, waist circumference and ABSI and consumed more alcohol (Additional file
1: Tab. S3). Baseline characteristics of control participants by sex-specific quintiles of FABP-4 concentrations are shown in Table
1. Mean age increased across FABP-4 quintiles. Participants in the upper quintiles were more often physically inactive than those in the lower quintiles, whereas no clear trends were observed across quintiles for education or smoking status. BMI, waist circumference, and ABSI were all increased across FABP-4 quintiles and, when comparing the upper with the lower quintiles, we observed a higher proportion of participants with diabetes. Alcohol intake, total energy, fiber, and fruit and vegetable intake were slightly decreased across FABP-4 quintiles, whereas no clear trends were observed for red and processed meat or fish intake. These observations did not differ substantially when investigating baseline characteristics in male and female control participants separately (Additional file
1: Tab. S4 and S5).
Table 1
Baseline characteristics by (sex-specific) quintiles of FABP-4 concentrations in control participants (n = 1324)
Quintile ranges in men | < 8.3 ng/mL | 8.3– < 10.8 ng/mL | 10.8– < 13.6 ng/mL | 13.6– < 17.2 ng/mL | ≥ 17.2 ng/mL | |
Quintile ranges in women | < 13.0 ng/mL | 13.0– < 16.8 ng/mL | 16.8– < 20.5 ng/mL | 20.5– < 26.3 ng/mL | ≥ 26.33 ng/mL | |
N | 263 | 268 | 266 | 265 | 262 | |
Male, n (%) | 128 (47.9) | 128 (47.6) | 128 (47.8) | 128 (47.6) | 128 (47.9) | |
Female, n (%) | 135 (51.3) | 140 (52.2) | 138 (51.9) | 137 (51.7) | 134 (51.1) | 0.93 |
Age, years | 55.7 (7.0) | 57.6 (7.1) | 58.2 (7.4) | 59.1 (6.2) | 59.8 (6.6) | <0.0001 |
University degree, n (%) | 48 (18.3) | 56 (20.9) | 50 (18.8) | 41 (15.5) | 40 (15.3) | 0.12 |
Physically inactive, n (%) | 40 (15.2) | 50 (18.7) | 73 (27.4) | 61 (23.0) | 71 (27.1) | 0.001 |
Recreational and household physical activity, METs/week, median (p25, p75) | 83.5 (48.5, 123.2) | 78.9 (49.9, 118.7) | 76.1 (42.6, 115.6) | 75.7 (48.6, 115.5) | 75.2 (42.0, 123.0) | 0.58 |
Smoker, n (%) | 75 (28.5) | 61 (22.8) | 57 (21.4) | 69 (26.0) | 66 (25.2) | 0.69 |
Body mass index, kg/m2, mean (SD) | 24.1 (3.1) | 25.2 (3.1) | 26.5 (3.4) | 27.4 (3.3) | 28.7 (4.3) | <0.0001 |
Waist circumference, cm, mean (SD) | 82.6 (11.3) | 85.9 (11.7) | 89.1 (10.8) | 91.1 (10.8) | 95.4 (12.5) | < 0.0001 |
A-body shape index, mean (SD) | 76.4 (5.6) | 77.2 (6.2) | 77.6 (5.7) | 77.5 (5.4) | 78.6 (5.5) | < 0.0001 |
Height, cm, mean (SD) | 168 (9.0) | 167 (9.1) | 167 (9.3) | 168 (9.8) | 167 (8.9) | 0.40 |
Diabetes at baseline, n (%) | 7 (2.7) | 9 (3.4) | 10 (3.8) | 10 (3.8) | 11 (4.2) | 0.02 |
Alcohol intake, g/day, median (p25, p75) | 10.7 (2.1, 24.8) | 7.4 (1.7, 23.6) | 7.7 (1.5, 20.0) | 9.2 (1.8, 24.5) | 6.2 (1.5, 18.3) | 0.13 |
Dietary factors |
Energy intake, kcal/day, median (p25, p75) | 2085 (1751, 2557) | 2070 (1663, 2464) | 2005 (1650, 2417) | 2094 (1656, 2512) | 1934 (1564, 2392) | 0.09 |
Fiber, g/day, median (p25, p75) | 23.8 (18.8, 30.0) | 22.9 (18.0, 28.4) | 22.9 (18.1, 27.2) | 22.9 (18.0, 27.4) | 22.6 (17.3, 26.9) | 0.04 |
Fruits and vegetables, g/day, median (p25, p75) | 401.3 (262.3, 572.8) | 397.9 (282.4, 575.4) | 345.7 (241.0, 546.4) | 360.9 (235.2, 535.0) | 331.7 (213.6, 512.0) | 0.01 |
Red meat, g/day, median (p25, p75) | 23.6 (13.0, 43.1) | 24.1 (12.8, 43.3) | 25.5 (12.5, 42.6) | 26.9 (15.5, 45.3) | 23.3 (13.3, 45.0) | 0.77 |
Processed meat intake, g/day, median (p25, p75) | 41.4 (23.9, 70.5) | 44.1 (23.8, 71.5) | 46.4 (25.0, 74.3) | 51.6 (31.1, 75.8) | 48.2 (24.7, 79.7) | 0.12 |
Fish, g/day, median (p25, p75) | 28.2 (13.5, 51.7) | 25.9 (13.7, 48.7) | 28.9 (16.1, 48.5) | 28.9 (14.7, 52.3) | 32.4 (16.4, 52.4) | 0.49 |
FABP-4, ng/mL, median (p25, p75) | 7.9 (6.5, 11.2) | 13.0 (9.7, 14.8) | 17.0 (12.2, 18.4) | 20.7 (15.3, 23.6) | 27.2 (20.3, 31.9) | < 0.0001 |
In control participants, FABP-4 concentrations correlated statistically significantly, even after applying a Bonferroni-correction (
n = 28 tests), with BMI and waist circumference as well as with a number of biomarkers of inflammation, metabolism, blood lipids, adipokines, antioxidative capacity, and immune function (Table
2). FABP-4 concentrations were substantially (
r > 0.4) positively correlated with BMI and waist circumference, and weakly correlated with ABSI, whereas no correlation was observed with height. Substantial correlations were observed between FABP-4 and the adipokine leptin (
r = 0.42) as well as with FRAP (
r = 0.33), a biomarker of antioxidant capacity. The inflammatory marker C-reactive protein (
r = 0.29) as well as the hyperinsulinemia marker C-peptide (
r = 0.28) were also correlated with FABP-4.
Table 2
Spearman partial correlation coefficients (controlled for age and sex) between FABP-4 and body mass index, waist circumference, A-body shape index, height and blood biomarkers in control participants (n = 1324)
Age, years | 1324 | 0.21 | < 0.0001* |
Body mass index, kg/m2 | 1324 | 0.44 | < 0.0001* |
Waist circumference, cm | 1263 | 0.42 | < 0.0001* |
A-body shape index | 1263 | 0.12 | < 0.0001* |
Height, cm | 1324 | −0.03 | 0.27 |
High-sensitivity C-reactive protein, mg/L | 758 | 0.29 | < 0.0001* |
TNF-a (pg/mL) | 703 | 0.09 | 0.02 |
C-peptide, ng/mL | 652 | 0.28 | < 0.0001* |
HbA1c, % | 635 | 0.13 | 0.002 |
IGF1, nmol/L | 649 | −0.10 | 0.03 |
IGFBP1, nmol/L | 1467 | 0.15 | < 0.0001* |
IGFBP2, nmol/L | 1456 | −0.21 | < 0.0001* |
IGFBP3, nmol/L | 1472 | 0.05 | 0.04 |
IGFBP3 intact, nmol/L | 1465 | 0.05 | 0.07 |
Total cholesterol, mmol/L | 759 | 0.05 | 0.15 |
LDL cholesterol, mmol/L | 759 | 0.07 | 0.05 |
HDL cholesterol, mmol/L | 757 | −0.14 | < 0.0001* |
Triglycerides, mmol/L | 748 | 0.25 | < 0.0001* |
Adiponectin, μg/mL | 673 | −0.09 | 0.02 |
HMW-Adiponectin, μg/mL | 672 | −0.08 | 0.03 |
Leptin, ng/mL | 673 | 0.42 | < 0.0001* |
Soluble leptin receptor, ng/mL | 673 | −0.23 | < 0.0001* |
Resistin, ng/mL | 1300 | 0.10 | 0.0003* |
Fetuin-a, μg/mL | 1324 | 0.12 | < 0.0001* |
25-Hydroxvitamin D, nmol/L | 758 | −0.10 | 0.005 |
Ferric reducing ability of plasma, µmol/l | 759 | 0.33 | < 0.0001* |
Reactive oxygen metabolites, Carratelli units | 754 | 0.19 | < 0.0001* |
Neopterin, nmol/L | 606 | 0.08 | 0.05 |
We observed a statistically significant association between circulating FABP-4 and risk of CRC in the conditional logistic regression model accounting for the matching factors but without further adjustment (RR highest versus lowest quintile 1.32, 95% CI 1.01, 1.72; RR per SD increment in FABP-4 1.10, 95% CI 1.01, 1.21, Table
3). This association was attenuated and no longer statistically significant after multivariable adjustment (RR highest versus lowest quintile 1.26, 95% CI 0.96, 1.66; RR per SD 1.09, 95% CI 0.99, 1.19). Additional adjustment for body size (BMI, height, and BMI- and height-adjusted waist circumference residuals) further attenuated the relative risk estimates towards the null (RR highest versus lowest quintile RR 1.01, 95% CI 0.74, 1.38, RR per SD 1.01, 95% CI 0.92, 1.12). The complete case analysis excluding participants with missing waist circumference (
n = 122) yielded the same result in the model adjusted for body size (RR per SD 1.01, 95% CI 0.91, 1.11). No statistical interaction was observed by age, BMI, waist circumference, or ABSI categories (all
p-interactions > 0.26). Similarly, no interaction was observed by categories of C-reactive protein, leptin, or FRAP (all
p-interaction > 0.57). However, a statistically significant interaction between FABP-4 and C-peptide was observed (
p-interaction 0.04). Models stratified by C-peptide concentrations (sex-specific median) revealed a statistically non-significant inverse association in participants with low C-peptide (RR per SD in FABP-4 0.84, 95% CI 0.61, 1.17) and a statistically non-significant positive association in participants with high C-peptide (RR 1.04, 95% CI 0.83, 1.29).
Table 3
Association between baseline FABP-4 concentrations and risk of colorectal cancer (conditional logistic regression models)
Overall (1324 cases and matched controls) |
Quintile 1 | 237/263 | 1 | Reference | 1 | Reference | 1 | Reference |
Quintile 2 | 256/268 | 1.08 | (0.84, 1.38) | 1.08 | (0.84, 1.40) | 1.03 | (0.79, 1.33) |
Quintile 3 | 268/266 | 1.15 | (0.89, 1.49) | 1.14 | (0.88, 1.48) | 1.04 | (0.79, 1.36) |
Quintile 4 | 267/265 | 1.16 | (0.90, 1.49) | 1.12 | (0.86, 1.45) | 0.98 | (0.74, 1.29) |
Quintile 5 | 296/262 | 1.32 | (1.01, 1.72) | 1.26 | (0.96, 1.66) | 1.01 | (0.74, 1.38) |
p-trend | | | 0.03 | | 0.08 | | 0.94 |
per SD of FABP-4 (8.9 ng/ml)d | | 1.10 | (1.01, 1.21) | 1.09 | (0.99, 1.19) | 1.01 | (0.92, 1.12) |
Men (640 cases and matched controls) |
Quintile 1 | 118/128 | 1 | Reference | 1 | Reference | 1 | Reference |
Quintile 2 | 119/128 | 1.02 | (0.71, 1.45) | 1 | (0.68, 1.46) | 0.89 | (0.61, 1.32) |
Quintile 3 | 139/128 | 1.20 | (0.84, 1.72) | 1.21 | (0.83, 1.75) | 1.02 | (0.70, 1.50) |
Quintile 4 | 127/128 | 1.09 | (0.77, 1.56) | 1.04 | (0.71, 1.52) | 0.85 | (0.57, 1.26) |
Quintile 5 | 137/128 | 1.19 | (0.82, 1.73) | 1.11 | (0.74, 1.64) | 0.76 | (0.49, 1.19) |
p-trend | | | 0.35 | | 0.65 | | 0.21 |
per SD of FABP-4 (8.9 ng/ml)d | | 1.07 | (0.93, 1.24) | 1.07 | (0.92, 1.23) | 0.95 | (0.80, 1.13) |
Women (684 cases and matched controls) |
Quintile 1 | 119/135 | 1 | Reference | 1 | Reference | 1 | Reference |
Quintile 2 | 137/140 | 1.13 | (0.80, 1.61) | 1.14 | (0.80, 1.64) | 1.14 | (0.79, 1.65) |
Quintile 3 | 129/138 | 1.11 | (0.77, 1.60) | 1.12 | (0.77, 1.64) | 1.08 | (0.73, 1.60) |
Quintile 4 | 140/137 | 1.23 | (0.85, 1.76) | 1.22 | (0.84, 1.79) | 1.15 | (0.77, 1.72) |
Quintile 5 | 159/134 | 1.45 | (0.99, 2.12) | 1.46 | (0.99, 2.17) | 1.34 | (0.84, 2.11) |
p-trend | | | 0.05 | | 0.05 | | 0.24 |
per SD of FABP-4 (8.9 ng/ml)d | | 1.12 | (1.00, 1.26) | 1.12 | (1.00, 1.26) | 1.09 | (0.95, 1.25) |
p-heterogeneity by sex | | | 0.62 | | 0.57 | | 0.24 |
In sex-stratified analyses, FABP-4 was borderline statistically significantly positively associated with CRC risk in women in the multivariable-adjusted model (RR per SD in FABP-4 1.12, 95% CI 1.00, 1.26), which was attenuated after adjustment for body size (RR 1.09, 95% CI 0.95, 1.25; Table
3). FABP-4 was not associated with CRC risk in men in either the multivariable-adjusted model (RR per SD in FABP-4 1.07, 95% CI 0.92, 1.23) or the multivariable-adjusted model including body size (0.95, 95% CI 0.80, 1.13). Adding the body size variables one by one to the multivariable-adjusted model showed that adjustment for BMI switched direction of estimates in men from (non-significant) positive to negative (data not shown). Despite the observed differential associations of FABP-4 with CRC risk in women compared to men, no significant heterogeneity by sex was observed (Table
3). After exclusion of participants with diabetes, the positive association between FABP-4 and CRC risk in women was slightly attenuated and statistically non-significant (RR in the multivariable-adjusted model per SD 1.05, 95% CI 0.92, 1.19), while after exclusion of cases diagnosed within the first 2 years of follow-up (and their matched controls) point estimates remained statistically significant (RR per SD 1.17, 95% CI 1.03, 1.34, Additional file
1, Tab. S6). In men and overall, associations were not substantially changed after exclusion of people with diabetes or cases (and matched controls) diagnosed with the first 2 years (Additional file
1, Tab. S6).
In subgroup analyses by CRC subsite and sex (Additional file
1, Tab. S7), associations were not substantially different, although associations were slightly stronger in rectal versus colon cancer, with significant heterogeneity in conditional only (
p-heterogeneity < 0.0001) or the multivariable-adjusted (
p-heterogeneity = 0.0002) model, but not in the model adjusted for body size (
p-heterogeneity = 0.35).
The three GWAS-identified SNPs (rs2012444, rs77878271, rs79389622) explained together about 1% of interindividual variance in circulating FABP-4 and had an instrument strength of
F = 65. With the given sample size and a statistical power of 80% the minimal detectable OR per SD in genetically predicted FABP-4 based on the three SNPs was 1.17. Of the three SNPs, one was located near the
FABP4 gene (cis-SNP), while the other two were located near other genes (trans-SNPs). Details of the three SNPs including effect estimates for the SNP-FABP4 and SNP-CRC association are displayed in Table
4.
Table 4
Overview of selected FABP-4 SNPs for Mendelian randomization analysis
rs2012444 | 3 | 12375956 | PPARG | Trans | T | C | 0.13 | 0.11 | 0.01 | 0.12 | −0.01 | 0.01 |
rs77878271 | 8 | 82395535 | FABP4 | Cis | A | G | 0.97 | 0.26 | 0.04 | 0.98 | 0.05 | 0.03 |
rs79389622 | 9 | 126081452 | CRB2 | Trans | A | G | 0.01 | −0.59 | 0.10 | 0.00 | −0.07 | 0.08 |
In the two-sample Mendelian randomization analysis using all three SNPs as instrumental variables, statistically non-significant positive associations with genetically predicted higher FABP-4 were observed for CRC overall (OR per one SD genetically predicted FABP-4 1.10, 95% CI 0.95, 1.27) and in women (OR 1.21, 95% CI 0.98, 1.48) but not in men (OR 1.03, 95% CI 0.84, 1.26, Table
5). Most CRC subgroups by location and sex showed non-significant positive associations (Additional file
1: Fig. S2), except for rectal cancer overall and in men, where effect estimates were in the direction of inverse associations, but confidence intervals were wide. When we used MR Egger instead of IVW, similarly statistically non-significant positive associations were observed (OR for CRC overall 1.27, 95% CI 0.97, 1.68) and there was no indication of horizontal pleiotropy for the SNPs associated with FABP-4 (p-value of pleiotropy 0.21). When we used only the cis-SNP as an instrumental variable, associations were stronger, with a statistically non-significant positive association for genetically predicted higher FABP-4 and CRC overall (OR 1.23, 95% CI 0.97, 1.57), a statistically significant positive association for CRC (OR 1.56, 95% CI 1.09, 2.23) and colon cancer (OR 1.58, 95% CI 1.05, 2.40) in women and no association for CRC in men (OR 0.99, 95% CI 0.71, 1.37, Table
5, Additional file
1: Fig. S3). In sensitivity analyses using two moderately correlated (
R2 < 0.1) SNPs within the
FABP4 gene region (rs77878271 and rs2011042) accounting for the correlation matrix, results were not changed: genetically predicted higher FABP-4 was not associated with CRC overall (OR 1.18, 95% CI 0.97, 1.43) or in men (OR 0.94, 95% CI 0.72, 1.23), whereas a statistically significant positive association with CRC was observed in women (OR 1.48, 95% CI 1.12, 1.95). Applying a conservative Bonferroni-correction accounting for the number of tests in Table
5 (
n = 12), however, the positive associations in women in the two cis-MRs did not pass the statistical significance threshold. In the IVW, MR Egger, and cis-MR with one SNP, no statistically significant heterogeneity by sex was observed, while in the cis-MR with 2 moderately correlated SNPs, there was an indication of heterogeneity by sex (
p = 0.02). Colocalization analysis (genetic region plus/minus 50 kilobasepairs from the lead
FABP4 variant rs77878271) for overall CRC with standard prior probability (
p = 10
−5) revealed a posterior probability of a shared causal variant (PP4) of only 2% (Additional file
1: Tab. S8). Similarly, there was no indication of a shared causal variant for CRC in men (PP4 = 1%) or in women (PP4 = 12%). When the colocalization analysis was repeated with relaxed prior probability (
p = 10
−4) there was an indication of a shared causal variant of FABP-4 and CRC in women (PP4 = 58%, Additional file
1: Tab. S9). Because none of the variants in the gene region was strongly associated with CRC (minimum
p = 0.03 overall, minimum
p = 0.13 in men, minimum
p = 0.005 in women), statistical power to detect colocalization was limited.
Table 5
Mendelian randomization estimates between genetically predicted circulating FABP-4 concentrations and colorectal cancer risk
CRC, overall | 1.10 (0.95, 1.27) | 0.20 | 1.27 (0.97, 1.68) | 0.08 | 1.23 (0.97, 1.57) | 0.09 | 1.18 (0.97, 1.43) | 0.10 |
CRC, men | 1.03 (0.84, 1.26) | 0.77 | 1.19 (0.79, 1.78) | 0.40 | 0.99 (0.71, 1.37) | 0.94 | 0.94 (0.72, 1.23) | 0.66 |
CRC, women | 1.21 (0.98, 1.48) | 0.08 | 1.46 (0.77, 2.78) | 0.25 | 1.56 (1.09, 2.23) | 0.02 | 1.48 (1.12, 1.95) | 0.01 |
P-heterogeneity by sex | | 0.29 | | 0.59 | | 0.07 | | 0.02 |
Given the suggestion of a positive association between FABP-4 and CRC risk in women from both the biomarker and MR analysis, we performed a causal mediation analysis for the association between waist circumference and CRC risk (BMI was not statistically significantly associated with CRC risk in women) with FABP-4 as a potential mediator using the EPIC nested case–control study data, assuming no interaction between waist circumference and FABP-4 (because there was no indication for such interaction). We found a mediated proportion of 10% (natural direct effect per 1 cm in waist circumference in women: OR 1.012; natural indirect effect through FABP-4: OR 1.001).
Discussion
In this prospective investigation of circulating FABP-4 and CRC risk, we found overall no strong evidence for an association, although we observed a positive association in women; this, however, was attenuated after adjustment for body size. Genetically predicted higher FABP-4 was not statistically significantly associated with CRC in the polygenic MR. In cis MR analyses, no statistically significant associations were observed for CRC overall or in men, but a statistically significant positive association was observed for CRC in women, which, however, did not pass a Bonferroni correction accounting for the number of tests in the MR.
The hypothesis underpinning our study was that higher FABP-4 may be positively associated with CRC risk, which could be biologically explained by facilitating tumor growth via increased fatty acid supply [
13], or FABP-4-related enhancement of inflammation [
14] and insulin resistance [
15‐
18]. The overall weak and statistically non-significant associations of measured circulating as well as genetically predicted FABP-4 and CRC do not provide strong support for circulating FABP-4 playing an important role in CRC development. However, a positive association of FABP-4 and CRC risk in women was observed in the biomarker analysis before adjustment for body size, suggesting that the positive association was largely explained by the upregulation of FABP-4 in obesity [
7]. Interestingly, also in cis MR analyses, a positive association between FABP-4 and CRC risk in women was observed (with statistically significant heterogeneity by sex in the cis MR with two moderately correlated SNPs). It should be noted that the colocalization analysis with standard prior probability did not strongly support the existence of a shared causal variant of circulating FABP-4 and CRC in women, which could indicate that FABP-4 and CRC have distinct causal variants that are in linkage disequilibrium, thereby violating the MR assumptions. However, the posterior probabilities of distinct causal variants (PP3) were overall low and the associations of the variants in the genetic region with CRC were also overall low (
p > 0.005), suggesting that the lack of evidence for colocalization may be due to low power to detect colocalization [
52]. Colocalization analysis with relaxed prior probability gave an indication of a shared causal variant for FABP-4 and CRC in women. Sex-specific differences in FABP-4 in relation to CRC are plausible, as sex differences have also been observed in the association between general obesity and CRC risk [
39,
54], where stronger associations have usually been observed with BMI in men than in women. Two MR studies produced conflicting results regarding sex-differences in the association of BMI with CRC risk, where the smaller one observed a positive association between genetically predicted BMI and CRC in women but not in men [
55], and the larger and more recent one (using sex-specific genetic instruments) observed a stronger positive association between genetically predicted BMI and CRC in men compared with women [
56]. Sex-specific differences have been observed also in the relationship of inflammatory markers with CRC risk [
14,
30], where stronger associations were observed in men than in women. A more prominent role of FABP-4 for CRC in women, as observed in our study, may also be biologically plausible on the basis of the observed higher FABP-4 concentrations in women than men, and deserves further study. It should be noted, however, that we observed no statistically significant heterogeneity by sex (
p-heterogeneity 0.57). In contrast, we observed statistically non-significant inverse associations between FABP-4 and CRC risk in men in models accounting for body size. However, in the MR analysis, there was no indication for an inverse association between genetically determined higher circulating FABP-4 and risk of CRC in men. In the MR analysis, we observed mostly non-significant positive associations between genetically predicted FABP-4 and CRC anatomical subsites, except a non-significant inverse association with rectal cancer, which is in contrast to our findings of the biomarker analysis (non-significant positive association with rectal cancer). However, the wide confidence intervals in the MR analysis point to the uncertainty of the inverse association with rectal cancer.
When we performed a causal mediation analysis for the association between waist circumference and CRC risk with FABP-4 as a potential mediator in women, we found that a rather small proportion of the association (10%) was mediated by FABP-4. Compared with biomarkers such as non-HMW adiponectin, soluble leptin-receptor, and HDL-cholesterol [
57], FABP-4 may play a minor mediating role in the association between waist circumference and CRC risk in women.
The observed correlations of FABP-4 with a variety of biomarkers of inflammation, metabolism, blood lipids, adipokines, antioxidative capacity, and immune function (with strongest correlations observed for leptin and the antioxidant biomarker FRAP), some of which also have been previously associated with CRC risk, suggest that circulating FABP-4 is an integrative marker of various biological processes. Compared with the observed substantial correlations of FABP-4 with BMI and waist circumference, the weak positive correlation of FABP-4 with ABSI, a measure that was designed as a body shape index independent of BMI [
37], suggests that FABP-4 is particularly influenced by general obesity.
Strengths of our investigation include the prospective study design, the ability to control for a variety of potential confounders in the biomarker analysis, and the use of MR enabling a further investigation of FABP-4 in relation to CRC risk circumventing certain types of bias common to studies of measured biomarkers. The three GWAS-identified SNPs included in the MR analysis were robustly associated with circulating FABP-4, but, together, explained only 1% of inter-individual variation. With an
F-value of 65 (i.e., > 10, thereby not subject to weak instrument bias [
48]), the first MR assumption (instrumental variable should be associated with the exposure) is satisfied although, with the given sample size, the statistical power was limited and we cannot exclude that small associations (RR < 1.17 per SD in FABP-4) have been missed. We assume that the second MR assumption (instrumental variables are independent of potential confounders) was fulfilled, although this could not be directly tested since we had no access to covariable data in the studies included in GECCO, CORECT, and CCFR. The third MR assumption (instrumental variable should be associated with the outcome only through the exposure of interest (FABP-4) — no horizontal pleiotropy) could also not be directly tested, but there was no indication of horizontal pleiotropy from MR Egger. Query of the phenoscanner database [
58], however, revealed several adiposity and diabetes-related traits for the trans-SNP rs2012444, which seems plausible given the correlation of FABP-4 with body fatness measures and the previously observed association with type 2 diabetes [
6], but suggests that this trans-SNP is associated besides FABP-4 with two established CRC risk factors, i.e., could be an invalid instrumental variable due to pleiotropic effects. Thus, the MR analysis using both cis and trans SNPs may have been subject to pleiotropy and should be interpreted cautiously. For the cis-SNPs (rs77878271 and rs2011042), no association with adiposity, diabetes, or other CRC risk factors has been reported, which strengthens the confidence in their use as instrumental variables. However, one trait related to mortality due to alcoholic hepatitis was listed for rs77878271. For the trans-SNP, rs79389622 as well as the cis-SNP, rs2011042, no associated traits besides FABP-4 were found. The stronger observed associations in MR using only the cis-SNP compared to all three (cis and trans) SNPs combined, may be explained by potential pleiotropy in the two trans-SNPs. Employing a cis-SNP in MR implies the highest biological plausibility [
51]. Comparable to our results, a statistically significant association using only one cis-SNP as an instrumental variable, but statistically non-significant associations when using both the cis- and multiple trans-SNPs was observed in a MR study on IGF-1 and prostate cancer risk [
59]. There was no sample overlap between the GWAS for FABP-4 (SCALLOP consortium) and the one for CRC (GECCO/CORECT/CCFR), which precludes inflated type one error rates [
60].
A limitation of the biomarker analysis is that only a one-time FABP-4 measurement at baseline was available. Although the mid-term reliability (4 months) of FABP-4 has been shown to be relatively good [
29], the one-time measurements may not necessarily reflect longer-term exposure. However, by the use of genetically determined FABP-4 in the MR, we were also able to investigate lifelong differences in FABP-4 in relation to CRC risk. Nevertheless, the results of the biomarker analysis could have been influenced by measurement error. Because the FABP-4 measurement was conducted on baseline samples before the onset of CRC, we expect such measurement error to be non-differential, which would not lead to biased estimates but could lead to attenuation of associations. Whereas the sample size of the biomarker analysis was sufficient to investigate CRC, it was limited for subgroup analysis by CRC anatomical subsite and sex. In terms of the MR analysis, another limitation was the relatively small sample size of the GWAS on FABP-4, which resulted in a limited number of SNPs independently associated with FABP-4, of which only three could be included in the present two-sample MR and these three SNPs explained only a small proportion of variability in FABP-4. Due to the limited genetic determination of FABP-4, the statistical power for the MR was slightly lower than in the biomarker analysis, and, again here, the sample size was limited for subgroup analyses by anatomical subsite and sex. In addition, the limited number of SNPs included in the present two-sample MR precluded detailed sensitivity analyses.
Acknowledgements
EPIC:
We acknowledge the use of data and biological samples from EPIC-France, EPIC-Varese, EPIC-Ragusa, EPIC-Cambridge, EPIC-Oxford, EPIC-Bilthoven, EPIC-Utrecht, EPIC-Aarhus, EPIC-Granada and EPIC-Asturias cohort.
GECCO:
ASTERISK: We are very grateful to Dr. Bruno Buecher without whom this project would not have existed. We also thank all those who agreed to participate in this study, including the patients and the healthy control persons, as well as all the physicians, technicians, and students.
CCFR: The Colon CFR graciously thanks the generous contributions of their study participants, dedication of study staff, and the financial support from the US National Cancer Institute, without which this important registry would not exist. The authors would like to thank the study participants and staff of the Seattle Colon Cancer Family Registry and the Hormones and Colon Cancer study (CORE Studies).
CLUE II: We thank the participants of Clue II and appreciate the continued efforts of the staff at the Johns Hopkins George W. Comstock Center for Public Health Research and Prevention in the conduct of the Clue II Cohort Study. Cancer data was provided by the Maryland Cancer Registry, Center for Cancer Prevention and Control, Maryland Department of Health, with funding from the State of Maryland and the Maryland Cigarette Restitution Fund. The collection and availability of cancer registry data is also supported by the Cooperative Agreement NU58DP006333, funded by the Centers for Disease Control and Prevention. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention or the Department of Health and Human Services.
COLON and NQplus: the authors would like to thank the COLON and NQplus investigators at Wageningen University & Research and the involved clinicians in the participating hospitals.
CORSA: We kindly thank all individuals who agreed to participate in the CORSA study. Furthermore, we thank all cooperating physicians and students and the Biobank Graz of the Medical University of Graz.
CPS-II: The authors express sincere appreciation to all Cancer Prevention Study-II participants, and to each member of the study and biospecimen management group. The authors would like to acknowledge the contribution to this study from central cancer registries supported through the Centers for Disease Control and Prevention’s National Program of Cancer Registries and cancer registries supported by the National Cancer Institute’s Surveillance Epidemiology and End Results Program. The authors assume full responsibility for all analyses and interpretation of results. The views expressed here are those of the authors and do not necessarily represent the American Cancer Society or the American Cancer Society – Cancer Action Network.
Czech Republic CCS: We are thankful to all clinicians in major hospitals in the Czech Republic, without whom the study would not be practicable. We are also sincerely grateful to all patients participating in this study.
DACHS: We thank all participants and cooperating clinicians, and everyone who provided excellent technical assistance.
EDRN: We acknowledge all contributors to the development of the resource at the University of Pittsburgh School of Medicine, Department of Gastroenterology, Department of Pathology, Hepatology and Nutrition and Biomedical Informatics.
EPIC: Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policies, or views of the International Agency for Research on Cancer/World Health Organization.
EPICOLON: We are sincerely grateful to all patients participating in this study who were recruited as part of the EPICOLON project. We acknowledge the Spanish National DNA Bank, Biobank of Hospital Clínic–IDIBAPS, and Biobanco Vasco for the availability of the samples. The work was carried out (in part) at the Esther Koplowitz Centre, Barcelona.
Harvard cohorts: The study protocol was approved by the institutional review boards of the Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health, and those of participating registries as required. We acknowledge Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital as home of the NHS. The authors would like to acknowledge the contribution to this study from central cancer registries supported through the Centers for Disease Control and Prevention’s National Program of Cancer Registries (NPCR) and/or the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program. Central registries may also be supported by state agencies, universities, and cancer centers. Participating central cancer registries include the following: Alabama, Alaska, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Hawaii, Idaho, Indiana, Iowa, Kentucky, Louisiana, Massachusetts, Maine, Maryland, Michigan, Mississippi, Montana, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Puerto Rico, Rhode Island, Seattle SEER Registry, South Carolina, Tennessee, Texas, Utah, Virginia, West Virginia, Wyoming. The authors assume full responsibility for analyses and interpretation of these data.
Kentucky: We would like to acknowledge the staff at the Kentucky Cancer Registry.
LCCS: We acknowledge the contributions of Jennifer Barrett, Robin Waxman, Gillian Smith, and Emma Northwood in conducting this study.
NCCCS I & II: We would like to thank the study participants and the NC Colorectal Cancer Study staff.
NSHDS investigators thank the Västerbotten Intervention Programme, the Northern Sweden MONICA study, the Biobank Research Unit at Umeå University, and Biobanken Norr at Region Västerbotten for providing data and samples and acknowledge the contribution from Biobank Sweden, supported by the Swedish Research Council.
PLCO: The authors thank the PLCO Cancer Screening Trial screening center investigators and the staff from Information Management Services Inc and Westat Inc. Most importantly, we thank the study participants for their contributions that made this study possible.
Cancer incidence data have been provided by the District of Columbia Cancer Registry, Georgia Cancer Registry, Hawaii Cancer Registry, Minnesota Cancer Surveillance System, Missouri Cancer Registry, Nevada Central Cancer Registry, Pennsylvania Cancer Registry, Texas Cancer Registry, Virginia Cancer Registry, and Wisconsin Cancer Reporting System. All are supported in part by funds from the Center for Disease Control and Prevention, National Program for Central Registries, local states, or by the National Cancer Institute, Surveillance, Epidemiology, and End Results program. The results reported here and the conclusions derived are the sole responsibility of the authors.
SEARCH: We thank the SEARCH team
SELECT: We thank the research and clinical staff at the sites that participated on SELECT study, without whom the trial would not have been successful. We are also grateful to the 35,533 dedicated men who participated in SELECT.