Mediterranean Diet And Colorectal Cancer: A Mediation Analysis In The EPIC Italy Cohort

Francesca FASANELLI, AOU Citta della Salute e della Scienza di Torino e Universita di Torino, Centro per la Prevenzione Oncologica (CPO Piemonte), It, Italy
RICCERI F. 1,2 , CAINI S. 5 , MASALA G. 5 , PANICO S. 4 , MATTIELLO A. 4 , GRIONI S. 3 , KROGH V. 3 , VINEIS P. 7 , SACERDOTE C. 1 , FRASCA G. 6 , TUMINO R. 6

1 SC Epidemiologia dei Tumori, AOU Citta della Salute e della Scienza di Torino e Universita di Torino, Centro per la Prevenzione Oncologica (CPO Piemonte), Italia;
2 Unita sovra zonale di Epidemiologia, ASL TO 3, Grugliasco (TO), Italia
3 Unita di Epidemiologia e Prevenzione, Fondazione IRCCS - Istituto Nazionale dei Tumori, Milano, Italia;
4 Dipartimento di Medicina Clinica e Sperimentale, Universita di Napoli Federico II, Napoli, Italia
5 Struttura complessa di Epidemiologia Molecolare e Nutrizionale, Is6tuto per lo Studio e Prevenzione Oncologica (ISPO), Firenze, Italia
6 Registro Tumori Azienda Sanitaria Provinciale Ragusa, Ragusa, Italia
7 MRC-PHE Center for Environment and Health, School of Public Health, Imperial College London, London, UK

Purpose
Several clinical and biological mechanisms have been hypothesized to explain the protective effect of Mediterranean diet (MD) on colorectal cancer (CRC).
In the Italian section of the EPIC study, we examined the role of abdominal adiposity (measured by waist-to-hip ratio - WHR) and chronic inflammation (measured by the inflammatory marker plasminogen activator inhibitor-1 - PAI-1) as potential mediators of the relationship between MD and CRC.
Methods
We used the Italian Mediterranean Index (IMI, a score from 0 to 11 based on the consumption of various Mediterranean foods) as a summary measure of adherence to MD. Using a Cox proportional hazard model, we calculated the total effect of IMI (categorized into 4 categories - 0-1, 2-3, 4-5 and 6-11), WHR and PAI-1 on the risk of CRC.
We then estimated the indirect (mediated) and direct (unmediated) effects of IMI on the risk of CRC adapting to survival outcomes the weighting approach for multiple mediators. The mediation analysis was conducted by considering first only WHR and then introducing PAI-1. The confounders considered were age, smoking status, sex, centre, physical activity and education.
Results
Increasing adherence to MD was associated with a significantly decreasing risk of CRC and there was a positive relationship between WHR, PAI-1 and CRC. In the cohort, the total effect of IMI on CRC was mainly explained by the pure direct effect (HR 0.47, 95% CI: 0.35-0.57 for the highest category of IMI compared to the lowest one), while the natural indirect effect (i.e. mediated by WHR) was minimal (HR: 1.00, 95% CI: 0.91-1.08). In the case-cohort, the introduction of the second mediator PAI-1 did not improve the mediated effect estimate .
Conclusions
The protective effect of MD on the development of CRC is minimally mediated by abdominal adiposity and chronic inflammation.