The WCRF/AICR Third Expert Report Diet, Nutrition, Physical Activity and Cancer: a Global Perspective Methods, Approaches and Lessons Learned Teresa Norat, PhD Principal Research Fellow Imperial College London
Declaración de Intereses: No tengo ningún interés que declarar- Nothing to declare
Process to go from scientific evidence to WCRF/AICR recommendations for cancer prevention o Randomized controlled trials have specific issues Impossible to test each dietary pattern/food/nutrient Ethical issues Issues on compliance, blinding High cost, long time o Observational nature of most studies Prospective cohort studies are less prone to bias Risk factors tend to be correlated Residual confounding
Process to go from scientific evidence to WCRF/AICR recommendations for cancer prevention o Complexity of dietary intake (synergies, interactions) o Nutrient content of a food varies o Diet measurement error o Lack of specific biomarkers o Host variations ONE SINGLE STUDY CAN T ANSWER A RESEARCH QUESTION Systematic literature reviews
Evolution: WCRF/AICR Expert Reports Systematic literature review Expert Panel, grading criteria Mainly case-control studies and ecologic studies Methodological Task Force : Guidelines for systematic literature reviews and meta-analysis Increasing evidence from longitudinal studies Digital support 1997 2007
From 2007: Continuous Update Project Keep Report updated into the future Using scientific evidence as a basis for recommendations
Methods. Integrative rather than reductionist approach Exposure: nutritional factors Energy balance Body size Physical activity Dietary patterns Foods Nutrients Food components, additives, contaminants Alcoholic beverages Early life nutrition Breastfeeding Outcome: cancer incidence in adults, mortality after breast cancer Study design: randomised controlled trials and cohort studies
CUP Database Centralized in CUP Database Electronic literature searches Data curation CUP Database update Expert Panel Data analysis and synthesis
CUP Java desktop application
CUP database: data repository for analyses
CUP Database: Publications (cohort studies) by cancer and publication year Breast Colorectal Prostate Lung Stomach Pancreas Liver Endometrial Ovary Bladder Oesophageal Kidney Skin Lympho-hema MPL Thyroid Gallbladder Cervix Nasopharynx ------------> 2005 (2 nd expert report) 2006 --------> (CUP) 3 rd expert report 9,991 publications, 17 cancers 335 publications breast cancer survival 0 200 400 600 800 1000
From the CUP Database to the Third WCRF/AICR report
From the CUP Database to the Third WCRF/AICR report All RR estimates Highest vs. lowest intake levels High vs. Murphy 2013 Author Year Sex low RR (95% CI) WCRF_Code StudyDescription contrast Ruder 2011 Simons 2010 M Murphy 2013 0.87 (0.76, 0.99) COL41070 EPIC 393 vs. 4 g/d Simons 2010 M Ruder 2011 0.70 (0.61, 0.79) COL40896 NIH-AARP 732 vs. 0 g/d Simons 2010 W 1.14 (0.66, 1.94) COL40821 NLCS 634.4 vs. 0 g/d Simons 2010 W Simons 2010 W 0.67 (0.42, 1.07) COL40821 NLCS 683.2 vs. 0 g/d Simons 2010 W Simons 2010 M 1.08 (0.70, 1.68) COL40821 NLCS 634.4 vs. 0 g/d Lee 2009 W Simons 2010 M 0.84 (0.53, 1.32) COL40821 NLCS 683.2 vs. 0 g/d Larsson 2006 M Lee 2009 W 0.80 (0.40, 1.30) COL40764 SWHS 250 vs. 0 g/d McCullough 2003 M Larsson McCullough 2006 2003 M W 0.65 (0.46, 0.91) 1.18 (0.84, 1.65) COL40624 COL00366 COSM CPS II 512.4 vs. 33.1 g/d 329.4 vs. 0 g/d McCullough 2003 W McCullough 2003 M 0.86 (0.66, 1.11) COL00366 CPS II 329.4 vs. 0 g/d Jarvinen 2001 Jarvinen 2001 0.46 (0.14, 1.46) COL00314 FMCHES 1050.8 vs. 285.8 g/d Gaard 1996 M Gaard 1996 M 0.72 (0.25, 2.07) COL00008 NNHSSS 1098 vs. 109.8 g/d Gaard 1996 W Gaard 1996 W 1.24 (0.35, 4.40) COL00008 NNHSSS 1098 vs. 109.8 g/d Kearney 1996 0.87 (0.52, 1.44) COL00156 HPFS 291.6 vs. 3.4 g/d Kearney 1996 0 500 1000.25.5.75 1 1.5 2 3 Total milk (g/day)
.25.5 Estimated RR 1 2 4.25.5 Estimated RR 1 2 4 Standard methods for data analysis Figure 125 Relative risk of stomach cancer for 5kg/m2 increase of BMI (linear model) Non linear models Nonlinear relation between BMI and cardia stomach cancer Best fitting cubic spline 95% confidence interval Author Year Sex per 5 kg/m2 RR (95% CI) % Weight Study Description Cardia Abnet Corley Merry Samanic Kuriyama Lindblad Tran 2008 2008 2007 2006 2005 2005 2005 M M Subtotal (I-squared = 55.6%, p = 0.036). Non-cardia Abnet Sjödahl Merry MacInnis Samanic Lindblad Tran 2008 2008 2007 2006 2006 2005 2005 M Subtotal (I-squared = 35.4%, p = 0.158). NOTE: Weights are from random effects analysis 1.35 (1.19, 1.52) 1.22 (0.90, 1.54) 1.61 (1.22, 2.10) 1.09 (0.90, 1.32) 1.41 (0.85, 2.34) 1.23 (0.94, 1.62) 0.93 (0.74, 1.17) 1.23 (1.07, 1.40) 0.99 (0.87, 1.12) 1.09 (0.86, 1.37) 0.95 (0.73, 1.22) 0.95 (0.70, 1.29) 0.86 (0.79, 0.94) 0.98 (0.78, 1.22) 0.63 (0.42, 0.94) 0.93 (0.85, 1.02) 22.70 13.15 12.78 17.64 5.53 12.78 15.42 100.00 23.21 11.44 10.04 7.24 31.65 11.95 4.47 100.00 NIH-AARP KPMCP NLCS SCWC MCS I GPRDC NIT Cohort NIH-AARP HUNT-I NLCS MCCS SCWC GPRDC NIT Cohort 17 20 25 30 35 40 BMI (kg/m2) Nonlinear relation between BMI and non-cardia stomach cancer Best fitting cubic spline 95% confidence interval.416 1 2.4 17 20 25 30 35 40 BMI (kg/m2)
Issues in systematic literature reviews Publication bias: forest plots Heterogeneity: sensitivity and stratified analysis Study quality: exposure and outcome assessment, control for confounding, loss to follow-up
Publication bias: Funnel plots Figure 575 Funnel plot of studies included in the dose response meta-analysis of BMI at early adulthood and postmenopausal breast cancer
Heterogeneity: Sensitivity and stratified analysis Figure 190 Relative risk of stomach cancer for 5 kg/m 2 increase of BMI by geographic location and cancer site Author Year Sex per 5 kg/m2 RR (95% CI) % Weight Study Description Asia, Cardia Kuriyama 2005 M Tran 2005 Subtotal (I-squared = 54.4%, p = 0.139). Asia, Non-cardia Tran 2005 Subtotal (I-squared =.%, p =.). Europe, Cardia Merry Samanic Lindblad 2007 2006 2005 M Subtotal (I-squared = 61.9%, p = 0.073). Europe, Non-cardia Sjödahl Merry Samanic Lindblad 2008 2007 2006 2005 M Subtotal (I-squared = 32.3%, p = 0.219). North America, Cardia Abnet Corley 2008 2008 Subtotal (I-squared = 0.0%, p = 0.499). North America, Non-cardia Abnet 2008 Subtotal (I-squared =.%, p =.). NOTE: Weights are from random effects analysis 1.41 (0.85, 2.34) 34.81 0.93 (0.74, 1.17) 65.19 1.08 (0.73, 1.59) 100.00 0.63 (0.42, 0.94) 100.00 0.63 (0.42, 0.94) 100.00 1.61 (1.22, 2.10) 30.45 1.09 (0.90, 1.32) 39.09 1.23 (0.94, 1.62) 30.45 1.27 (1.01, 1.60) 100.00 1.09 (0.86, 1.37) 17.11 0.95 (0.73, 1.22) 14.95 0.86 (0.79, 0.94) 50.06 0.98 (0.78, 1.22) 17.88 0.93 (0.83, 1.04) 100.00 1.35 (1.19, 1.52) 82.82 1.22 (0.90, 1.54) 17.18 1.32 (1.18, 1.48) 100.00 0.99 (0.87, 1.12) 100.00 0.99 (0.87, 1.12) 100.00 MCS I NIT Cohort NIT Cohort NLCS SCWC GPRDC HUNT-I NLCS SCWC GPRDC NIH-AARP KPMCP NIH-AARP.416 1 2.4
Study quality: Stratified and sensitivity analysis
Study quality: Stratified and sensitivity analysis
CUP: Increase accuracy and decrease risk of bias Multidisciplinary Grading criteria of strength of the evidence on causality Guidelines for Systematic Literature Reviews Peer review (external control) Internal quality control (double checks, internal peer review) Process and findings documented o o o o Protocols CUP database and interface Library of statistical programs Systematic literature reviews reports
Continuous Update Project: Lessons Holistic approach: dietary patterns, food groups, foods, nutrients, food components, biomarkers Growth in the number of cohort studies, pooling projects, number of cases in existing cohorts allowing for subgroup analyses Use of non-linear analyses to identify dose-response shape Emerging evidence: o subtypes of cancer: histology, cancer site o life course: weight at birth, at adolescence o cancer survivors: breast o host susceptibility: age, gender, genetic, smoking, body fatness Automation important to cope with the amount of research
Imperial College Research Team Thanks t.norat@imperial.ac.uk Teresa Norat PhD Principal Investigator Doris Chan PhD Ana Rita Vieira MSc Dagfinn Aune PhD Snieguole Vingeliene MSc Leila Abar MSc Deborah Navarro-Rosenblatt MSc Rosa Lau MSc Jakub Sobiecki MSc Margarita Cariolou MSc Neesha Nanu MSc Elli Polemiti MSc Louise Abela MSc Database managers: Christophe Stevens, Rui Viera Statistical Adviser: Darren Greenwood PhD University of Leeds, UK