Establishing dietary and body size-related risk factors for esophageal cancer to inform prevention

Subproject

Pooling Project of Prospective Studies of Diet and Cancer (DCCP)

Our grant application to the World Cancer Research Fund was funded in 2022 (Grant reference number: IIG_FULL_2022_011). All exposure, confounder and outcome data are available within the Pooling Project of Prospective Studies of Diet and Cancer (Harvard). The Golestan Cohort Study will be added to the original list of studies proposed as recommended by the grant reviewers.

Esophageal cancer is the eighth most common cancer globally with an estimated 604,100 incident cases in 2020. It is also the sixth leading cause of death from cancer worldwide with 544,076 deaths attributed to it (5.5% of all cancer deaths). There are two main histological types of esophageal cancer that show different geographic distributions. Esophageal adenocarcinoma (OAC) is the main subtype in Australia, New Zealand, North America, and northern and western Europe, which together account for 46% of all global cases. The rate of increase in OAC incidence worldwide over the past four decades is the highest of any cancer, especially in high-income countries. In Australia, 53% of the new cases diagnosed annually are OACs, and their incidence increased 10-fold from 1982 to 2018. Esophageal squamous cell carcinoma (OSCC), accounts for 88% of cases worldwide. OAC and OSCC are diagnosed at an advanced stage and remain two of the most lethal cancers with a median survival of less than one year. Thus, informing prevention is critical to reduce their morbidity and mortality.
OAC and OSCC have distinct multi-factorial etiologies. The strongest risk factors are gastro-esophageal reflux disease and Barrett’s esophagus for OAC; alcohol intake for OSCC; and smoking for both. The World Cancer Research Fund (WCRF) Continuous Update Project (CUP) expert report found convincing or probable evidence for associations between alcohol and mate consumption with OSCC and obesity with OAC, but failed to identify any additional dietary or lifestyle factor as having similarly strong evidence; likely due to heterogeneity in individual studies’ designs and inadequate statistical power. Further, although individual studies have frequently combined the two subtypes, findings from The Cancer Genome Atlas argue against doing so owing to their different etiologies (e.g., substantial differences in patterns of somatic copy-number alterations between the diseases).
The Barrett’s and Esophageal Adenocarcinoma Consortium (BEACON; https://esocan.org/beacon) has helped better understand the etiology of OAC and elucidate its genetic causes. However, 10 of the 12 studies in BEACON are case-control studies, which are subject to recall and selection bias, particularly in studies of lifestyle factors. Only a large-scale pooled analysis of prospective studies will offer the best scientific evidence and breakthroughs to confirm or exclude a role for these modifiable diet and body fatness related exposures in OAC and OSCC risk.

Our over-arching objective is to generate robust evidence on modifiable risk factors for OAC and OSCC for refinement of prevention strategies. To accomplish this, we will pool data from 21 cohort studies to assess, with the maximum available precision, associations with OAC and OSCC risk for dietary and body fatness-related factors, for which findings to date have been inconclusive.

Aim 1: Assess associations with risk of OAC for: alcohol use; intake of meat and fish; intake of fruit, vegetables and cereal; dietary fiber intake; adherence to specific dietary and lifestyle quality scores; and consumption of sugar-sweetened beverages, coffee and tea.
Aim 2: Assess associations with risk of OSCC for: intake of meat and fish; intake of fruit, vegetables and cereal; dietary fiber intake; adherence to specific dietary and lifestyle quality scores; consumption of sugar-sweetened beverages, coffee and tea; and body mass index and waist circumference.
Aim 3: Investigate whether these associations differ between population sub-groups.

We will conduct a pooled analysis of data from cohorts participating in the Pooling Project of Prospective Studies of Diet and Cancer. The majority of the required questionnaire and outcome data has already been collected and harmonized; data from the Golestan Cohort Study will be harmonized.
Cox regression will be performed to calculate hazard ratios (HRs) and 95% CI for associations between alcohol use (OAC only); intake of red/processed/other meat and fish; intake of fruit, vegetables and cereals; dietary fiber intake; WCRF/AICR dietary and lifestyle scores; consumption of sugar-sweetened beverages, coffee and tea; and BMI and waist circumference (OSCC only), and risk of OAC and OSCC. Statistical analyses will build upon the methods used in the DCPP. Given the small number of cases in several cohorts, the primary analytic approach will be to analyze a combined dataset of the harmonized data from all studies. Cohorts including women and men will be analyzed as two separate studies and results will be calculated for women and men combined and separately. Based on the cause-specific hazard Cox model analysis, HRs will be estimated separately for OAC and OSCC; participants diagnosed with other subtypes will be censored at their diagnosis date. The estimated HRs obtained for each subtype will be asymptotically uncorrelated. The effects across subtypes will be compared using a chi-square test based on the contrast test. We will stratify jointly by study, age at baseline (years) and the year in which the baseline questionnaire was returned and treat follow-up time since study entry as the primary time scale. This is equivalent to a left-truncated survival analysis with age in years as the time scale. To control for potential confounding, we will include terms for established or potential risk factors for OAC and OSCC for which data are available. The terms will be modeled similarly across studies. We will conduct sensitivity analyses.

Evaluation of an association between these exposures and OAC and OSCC risk presents analytical challenges. Any relationship, if true, is likely to be weak or moderate; thus the statistical power to detect an association in a single study is likely to be very low plus requires the accumulation of information on a sizeable study population with a sufficiently large number of cases. This is likely only with a collaborative effort to jointly analyze existing information from large prospective epidemiological investigations.

Twenty-one studies that have participated in previous analyses in the DCPP, with comprehensive, validated dietary assessments and ≥20 incident cases of OAC or OSCC in men or women after study enrollment, have agreed in writing to participate in the project. The outcomes are: (i) incident adenocarcinoma [coded C15 according to the 10th revision of the International Classification of Diseases (ICD)] with a histological classification [morphology codes 8140, 8141, 8143, 8211, 8255, 8260, 8261, 8480, 8481, 8490] and (ii) incident squamous cell carcinoma [morphology codes 8051, 8070, 8071, 8072, 8073, 8074, 8076, 8083] of the esophagus. Cancers are ascertained by self-report with subsequent medical record confirmation and/or linkage to cancer registries. We expect 1,860 OACs and 1,288 OSCCs to be available.
The proposed study will have sufficient statistical power to detect modest associations of public health and clinical significance between exposures and risk of OAC and OSCC, even after stratification by sex and geographical region. The study offers sufficient power to detect a significant linear trend in risk comparing extreme quintiles/categories for exposures.

This proposal focuses on incident OAC and OSCC. Age at diagnosis with incident invasive cancer of another type (for censoring); age at death; age last known to be free of cancer; and age last known to be alive will also be used. Data on OAC have been collected and need to be harmonized; OSCC outcome data is already harmonized.

This proposal benefits from the prior data harmonization of food group and nutrient intake, anthropometric factors, confounding variables, and effect modifiers collected at baseline from each participating study. Further, this proposal will make use of the ongoing harmonization of WCRF/AICR scores that is expected to be completed by the beginning of year 2 of this proposal.

Baseline non-dietary risk factor data including, but not limited to, demographics, socioeconomic status and education, and dietary information such as energy intake are harmonized. All studies have provided age and sex. Generally, confounding variables will be categorized similarly across studies.

No