Circulating fatty acid binding protein 4 (FABP-4) concentrations and mortality in individuals with colorectal cancer in the European Prospective Investigation into Cancer and Nutrition study

Abstract

Human fatty acid binding protein-4 (FABP-4), a protein elevated in obesity that promotes colon cancer cell invasiveness and metastasis, may be associated with higher mortality in individuals with colorectal cancer (CRC) and may serve as a mediator of the obesity-mortality association in these individuals. We used a causal diagram to inform covariate selection and applied Cox proportional hazards models to estimate hazard ratios (HRs) for CRC-specific, non-CRC-specific, and all-cause mortality by FABP-4 levels measured in baseline blood samples from 1371 incident CRC cases from the European Prospective Investigation into Cancer and Nutrition cohort. Competing risk analyses were adapted for CRC and non-CRC deaths. Mediation analyses were conducted to estimate total effects (TEs), direct effects (DEs), and mediation proportions (MPs) by FABP-4 of pre-diagnostic body mass index (BMI) on mortality. In the fully adjusted model including BMI, higher circulating FABP-4 concentrations were associated with higher CRC mortality (HR_Q4vsQ1 = 1.49; 95% CI: 1.11-2.00) and all-cause mortality (HR_Q4vsQ1 = 1.49; 95% CI: 1.15-1.93), but not statistically associated with non-CRC mortality (HR_Q4vsQ1 = 1.51; 95% CI: 0.82-2.76). The TE and DE per 5 kg/m² of BMI on all-cause mortality were 1.21; 95% CI: 1.10-1.34, and 1.13; 95% CI: 1.02-1.26, respectively, with a MP of 34.5% (p = .002) by FABP-4. For CRC-specific and non-CRC-specific mortality, MPs by FABP-4 were 33.7% (p = .03) and 36.1% (p = .02), respectively. In conclusion, higher concentrations of FABP-4 were associated with higher CRC-specific and all-cause mortality in individuals with CRC. FABP-4 was a significant partial mediator of the adiposity-mortality relationship in individuals with CRC.

Keywords: EPIC; FABP‐4; human fatty acid binding protein‐4; incident colorectal cancer; mortality.

FIGURE 1

Causal diagram illustrating the relationship between FABP‐4, colorectal cancer (CRC), and death after CRC diagnosis. CRC diagnosis is a collider on the paths of U → CRC diagnosis ← FABP‐4. The blue box around “CRC diagnosis” indicates that analyses are conducted among individuals with incident CRC (or conditioning on CRC diagnosis). The analysis conditioning on CRC diagnosis may lead to “collider stratification bias.” This opens a non‐causal spurious association between FABP‐4 and (U) (dashed blue line). These opened paths can be blocked (red boxes) by adjusting for (U) and Obesity/BMI. The risk factors for incident CRC can be obtained from the expert report from the World Cancer Research Fund. Conventional confounders in this diagram were ancestors of both FABP‐4 levels and death (C1), and both obesity and death (C2). The paths through these conventional confounders were open and can be blocked (green boxes) by adjusting for C1 and C2.

FIGURE 2

Hazard ratios (HRs) and 95% confidence intervals (CIs) for CRC‐specific and all‐cause mortality per 1 SD increment of FABP‐4 levels in the European Prospective Investigation into Cancer and Nutrition (EPIC) study (N = 1371), in subgroup analyses. HRs for the primary outcome (A), CRC‐specific mortality, were estimated from cause‐specific Cox proportional hazards models accounting for competing risks. HRs for the secondary outcome (B), all‐cause mortality, were estimated from conventional Cox proportional hazards models. The time function in all models was the time from CRC diagnosis to the event of death or the last follow‐up. All models were stratified by country and adjusted for sex, age at diagnosis (continuous), smoking status (never, past, current smoker), education (none, primary school, technical/professional school, secondary school, or longer education), sex‐specific categories of physical activity (inactive, moderately inactive, moderately active, active), polyps of the large bowel (yes, no), alcohol consumption per day (continuous), intake per day of red meat, processed meat, fish and shellfish, calcium, dietary fiber, dairy products, vegetables and fruits (all variables were continuous), and adiposity measures including body mass index (BMI) (continuous) and residuals of waist circumference (WC) when regressed on BMI and height (ɛ[WC|BMI and height]). Of note, data on stage, WC, and diabetes were missing in 185 (13.5%), 75 (5.5%), and 68 (5%) individuals, respectively. The missing data were replaced by multiple imputations. The p‐interaction was estimated as the p value of the interaction term between FABP‐4 and each of the potential effect modifiers.