Environmental factors, GxE interactions and bladder cancer risk

Subproject

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

Nat Rothman-rothmann@mail.nih.gov
Montse Garcia-Closas-montse.garcia-closas01@icr.ac.uk

We will apply for funding to the Australian National Health and Medical Research Council (NHMRC). The budget would include, funds to support data preparation by each cohort.

Bladder cancer is the 7th most common cancer worldwide, with an estimated 260,000 new cases diagnosed annually in men and 76,000 in women. Western Europe, North America and Australia have the highest incidence. Most tumors are non-invasive and treatable, but require ongoing monitoring as recurrence is relatively common; this makes bladder cancer the malignancy with the highest cost per patient from diagnosis to death in the USA (Silverman 2017). Muscle-invasive bladder cancers (~30%) are treated more aggressively and have a worse outcome. The strongest risk factors for bladder cancer are cigarette smoking and occupational exposure to carcinogens (e.g. arylamines). A recent NCI Cohort Consortium study concluded that physical activity is associated with reduced risk of bladder cancer (Moore 2016). Despite extensive research by many individual studies, it remains unclear whether bladder cancer is related to other modifiable factors such as NSAID use, diet, alcohol and other beverage consumption, overweight and obesity and, for women, reproductive history, oral contraceptive use and hormone therapy use. Men have three to four times higher risk of bladder cancer than women, which is explained only in part by smoking and occupational exposure; reproductive and hormonal factors may therefore be important but have yet to be assessed prospectively with sufficient statistical power (Silverman 2017).

A recent meta-analysis of prospective cohort studies (Vieira 2015) suggested that a potential inverse association of vegetable and fruit intake with bladder cancer risk is weaker than first thought, based on earlier evidence, predominantly from case-control studies (Silverman 2006, 2017). The latter findings were based on combined data for up to 5,329 bladder cancer cases and highlighted the need to carefully control for smoking. Evidence of increased bladder cancer risk associated with consumption of red and processed meats has also come from case-control studies, with less clear evidence from cohort studies (Li 2014). In a recent review, Silverman (2017) concluded that further research is needed to determine whether high intake of processed food is a risk factor for bladder cancer. There is equivocal evidence from prospective cohort studies that total fluid intake is inversely associated with bladder cancer risk, that coffee consumption is positively associated, and that alcohol consumption overall is not associated with risk (Mao 2010; Silverman 2016). A meta-analysis of 15 published prospective cohort studies found a modestly increased risk of bladder cancer associated with overweight and obesity (Sun 2015), but was limited by between-study heterogeneity in both the categorization of BMI and the covariates adjusted for. Height does not appear to be associated with bladder cancer risk. An inverse association between NSAID use and bladder cancer risk has been reported by a meta-analysis of case-control studies, but not of cohort studies (Zhang 2013), although the latter was based on 4,277 cases. Some studies have reported differential effects by smoking status.

Bladder cancer is one of the few cancers for which gene-environment interactions (e.g. between genetic variants and smoking) have been identified and replicated (Garcia-Closas 2013). Few prospective studies have assessed other gene-environment interactions or whether risk factors for bladder cancer are differentially associated with disease subtypes. Doing so might shed light on etiological heterogeneity.

Use pooled prospective cohort data to assess associations with bladder cancer risk for predominantly modifiable lifestyle factors for which findings to date have been inconclusive.

Carry out gene-environment interaction (GxE) analyses in studies that have already scanned samples for GWAS. There is also the possibility of scanning at NCI additional samples from all cohorts (including those that haven't participated in previous GWAS).

1. Assess associations for: smoking habits, use of aspirin and NSAIDS, body size, reproductive and hormonal risk factors, alcohol, coffee, tea and other beverage consumption, personal history of diabetes (if no overlap with other projects), and dietary factors
2. Assess the above associations (plus physical activity) by sex, smoking status and disease subtype (e.g. by muscle-invasiveness)
3. Assess GxE interactions for the above factors using GWAS data
4. Develop a risk prediction model integrating information on lifestyle and genetic risk factors

We will conduct a pooled analysis of data from cohorts participating in the Pooling Project of Prospective Studies of Diet and Cancer (DCPP). The majority of the required questionnaire data has already been collected and harmonized, but some additional data will be requested. New follow-up data will be collected on incident bladder cancers. We will request updated follow-up for other cancer diagnoses and death (for calculation of person time), as well as relevant questionnaire data from more recent follow-up rounds. Data collation, cleaning and harmonization will be done through Stephanie Smith-Warner's group at Harvard. Data analyses will be conducted via remote access to the Harvard server after receiving the necessary approvals from participating cohorts. An analytical working group will be formed with interested cohort PIs/delegates.

Cox models will be used to assess associations with each risk factor of interest, with covariates selected based on risk-factor specific directed acyclic graphs (and availability). Models will be run for each study and estimates combined using random-effects meta-analysis. Analyses will also be conducted using an aggregated approach. Heterogeneity in associations by and disease subtype will be assessed using a joint Cox proportional model (based on a competing risk model) which allows different baseline hazard functions for each subtype, and direct comparison of associations by subtype. The GxE analyses will be use already available GWAS data (Figueroa 2016) and new GWAS data from cohorts which will be invited to participate in an expanded GWAS coordinated by NCI. Risk model development and validation will be conducted using the Individualized Coherent Absolute Risk Estimation (iCARE) R package (Choudhury 2018). Cohorts will be divided into training and test sets for model calibration and discrimination.

Studies to date have been statistically underpowered or subject to the potential biases of retrospective case-control studies. Pooling data will also allow us to assess with far greater statistical power GxE, associations by sex, smoking status and disease subtypes.

We would only include studies with at least 100 incident bladder cancer cases of the same sex.

Incident bladder cancer, including in situ disease (date of diagnosis, behaviour, stage, grade morphology and histology, if available)

Date of diagnosis with incident invasive cancer of another type (for censoring)

Date of death

Date last known to be free of cancer

Date last known to be alive

Food, fluid (water, alcohol, coffee, tea) and nutrient intake, BMI, waist circumference, height, smoking, use of aspirin & NSAIDS, personal history of diabetes (and its treatment), and parity*, age at menarche and menopause*, oral contraceptive & hormone therapy use* [*for women]. Most of these data have already been collected by the DCPP and will be updated if further follow-up data is available. More detailed data is needed for some variables.

Data already collected by the Diet and Cancer Pooling Project, updated if further follow-up data is available

No