Brain Cancer Aetiology: Identifying Risk Factors Using an International Pooled Cohort Study

Subproject

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

WCRF funding has been granted to investigate measures of alcohol consumption, tea and coffee consumption, body size, physical activity, dietary quality and WCRF/AICR scores. Additional funding has been applied for via the Australian NHMRC to investigate other aims (smoking, aspirin and other non-steroidal anti-inflammatory drugs (NSAIDs), diabetes, fat intake, sugar intake, measures of body size other than BMI, physical activity measured by accelerometer, reproductive factors, oral contraceptive use, hormone therapy use.)

Grant funding does not cover data preparation by individual cohorts.

Glioma is an aggressive neoplasm of the central nervous system. It is rare, but the most common type of primary malignant brain tumour [1, 2], and it contributes disproportionately to the cancer mortality burden; in the US, it ranks 10th for cancer-related death among adults [3].Tumours are also associated with significant morbidity [2]. Incidence increases with age, is higher in males than females, and in white Europeans compared with other racial groups [4]. Survival varies considerably by glioma subtype, glioblastoma multiforme (GBM) being the most aggressive, with 5-year relative survival of just 6% [1]. The aetiology of glioma is poorly understood and may differ by sex and glioma subtype. Medical exposure to ionizing radiation remains the only well-validated environmental risk factor. It is unclear whether the development of glioma is related to modifiable factors other than exposure to ionising radiation. The epidemiological investigation of other risk factors has been hampered by sub-optimal (retrospective) study design, relatively small sample size and potentially different associations by sex and disease subtype. Conducting meta-analyses of the published literature can only overcome the limitation of small sample size but are limited by heterogeneity in how the individual studies have defined and modelled their exposures, as well as lack of published cohort data owing to small sample sizes in many cohorts and publication bias. Combined analyses of the primary participant-level data from large prospective cohort studies are needed to address these limitations and to identify additional risk factors, particularly modifiable risk factors, for glioma. Our study will efficiently leverage data that has already been collected by participating cohorts. As a result this project will yield new information regarding potential ways to prevent glioma cancer.

To investigate the association with glioma risk for diet (including alcohol, tea and coffee consumption), dietary pattern, body size measures, physical activity, smoking, non-steroidal anti-inflammatory drugs (NSAIDs), diabetes, reproductive factors, oral contraceptive use, and hormone therapy use.

(1) to assess associations with glioma risk for: alcohol, tea and coffee consumption, body mass index, physical activity, dietary pattern (EAT-Lancet planetary health), lifestyle (WCRF/AICR cancer prevention recommendations), smoking, aspirin and other non-steroidal anti-inflammatory drugs (NSAIDs), diabetes, fat intake, sugar intake, measures of body size other than BMI, physical activity measured by accelerometer, reproductive factors, oral contraceptive use and hormone therapy use.
(2) to assess the above associations by sex and disease subtype (glioblastoma, other).

Data analyses will be conducted following pre-defined analysis plans based on previous work by the Diet and Cancer Pooling Project [5]. Owing to the small number of glioma cases, particularly of glioblastoma (GBM) and non-GBM, in several studies, for the primary analysis, we will combine the harmonised data from the 19 DCPP studies into a single dataset and analyse the aggregated dataset. Cohorts including women and men will be analysed as two separate studies: one of men and one of women.

Cox regression models will be used to assess associations with each risk factor of interest by estimating hazard ratios (HRs) and their 95% confidence intervals. We will stratify jointly on study (which also accounts for sex), age at baseline (years) and the year in which the baseline questionnaire was completed and treat follow-up time since study entry as the time scale; this is equivalent to a left-truncated survival analysis with age in years as the time scale. Potential confounders will be selected using risk factor-specific directed acyclic graphs. For all dietary analyses, we will include energy intake in the model to reduce systematic errors.

The same analyses will be carried out separately for men and women in each of the EPIC (stratifying on study centre) and UK Biobank datasets and then the three sets of summary results (from DCPP, EPIC and UK Biobank) for men and women will be combined using random effects meta-analysis on the Harvard server.

Glioma is a less common cancer, and a cohort consortium approach provides a large collection of data in which multiple hypotheses can be examined with greater statistical power than is available in any one study. In a pooled analysis, the modeling of the exposure, confounding, and outcome variables, the choice of which variables to control for, and the type of analysis conducted can be standardized, thereby removing potential sources of heterogeneity across studies. Because of larger sample sizes, pooled analyses also offer investigators the opportunity to examine uncommon exposures, rare diseases, and variation in associations among population subgroups with greater statistical power than is possible in individual studies.

50 incident cases among participants of the same sex

Administrative end of follow-up
o Year of administrative end of follow-up
o Age at end of follow-up (including decimals)

Mortality data
o Vital status (yes, no)
o Year of death
o Age at death (including decimals)
o Cause of death
o ICD code for underlying cause of death
o ICD version

Lost to follow-up
o Lost to follow-up (yes, no)
o Year lost to follow-up
o Age lost to follow-up (including decimals)

Any cancer diagnosis
o Personal history of cancer at baseline
o Cancer case status [cancer diagnosed since baseline] (yes, no)
o Year of diagnosis of any cancer
o Age at diagnosis of any cancer (including decimals)

Glioma (ICD-9 code 191; ICD-10 code C71)
o Glioma case status (yes, no)
o Year of glioma diagnosis
o Age at diagnosis of glioma (including decimals)
o Tumor location (e.g. ICD-10, ICD-9, or ICD-O-3 topography code)
o Tumor histology/morphology (e.g. ICD-O-3)
o Behavior (ICD-O-3)
o ICD version
o Mode of diagnosis (basis of diagnosis: e.g. clinical, imaging, histology)
o Grade
o Classification system used to define grade (e.g. WHO, ICD-O-3)
o Molecular data (if available, e.g. IDH mutant vs IDH wildtype, 1p19q codeletion, TERT promotor mutation, EGFR gene amplification, MN1 alteration, chromosome 7 loss/10 gain)

Body size (at baseline and follow-up)
o Height
o Body weight at age 18/20 years or young adulthood
o Weight at baseline
o Waist circumference
o Hip circumference

Reproductive and hormonal factors (at baseline and follow-up)
o Age at menarche
o Parity
o Age at first birth
o Breastfeeding
o Duration of breastfeeding
o Menopausal status
o Age at menopause
o Type of menopause
o Oral contraceptive use
o Oral contraceptive use duration
o Age at first use of oral contraceptives
o Age at last use of oral contraceptives
o Menopausal hormone therapy use
o Menopausal hormone therapy duration
o Type of menopausal hormone therapy
o Age at first use of menopausal hormone therapy
o Age at last use menopausal hormone therapy

Physical activity variables (at baseline or follow-up if not available at baseline)
o Frequency, duration, and METs (if calculated) for each of:
• Walking for transport or recreation
• Moderate intensity leisure activity
• Vigorous intensity leisure activity
o Frequency, duration, intensity, and METs (if calculated) for each of:
• Overall physical activity
• Individual domains (if available) - e.g., leisure, transportation, occupational, household
• Individual activities (if available) - e.g., walking, cycling, running, climbing stairs

Dietary variables (at baseline or follow-up if not available at baseline)

o Foods and beverages:
• Individual food and beverage intake data
For each food and beverage item on the questionnaire, please send the quantity consumed by each participant. Information in grams is preferred; however, we can apply standard weights to frequency and portion size information to obtain grams consumed). For the food analyses, we would like to try to separate individuals who did not consume a particular food item from individuals who did not answer the question. If a designator for missing responses is included in your database, please send the data for this variable too.
• Tea consumption (g/day)
• Coffee consumption (g/day)

o Nutrients
• Fructose
• Glucose
• Total fat
• Animal fat
• Dairy fat
• Vegetable fat
• Saturated fat
• Monounsaturated fat
• Polyunsaturated fat
• Omega-3 fatty acids
• Omega-6 fatty acids
Smoking (at baseline and follow-up)
o Cigarette, pipe, and cigar smoking habits. For each:
• Smoking status (never, former, current)
• Smoking pack-years
• Age at which smoking was initiated
• Age at which smoking ceased
• Time since quitting smoking
• Duration of smoking
• Amount smoked
Current alcohol consumption (at baseline and follow-up)
o Alcohol intake (g/d)
o Alcohol intake from beer (g/day)
o Alcohol intake from wine (g/day)
o Alcohol intake from liquor (g/day)
o Past drinker of alcoholic beverages

NSAID use (at baseline and follow-up)
o Use of Aspirin (yes, no)
o Duration of Aspirin use
o Dosage of Aspirin
o Use of other NSAIDs (yes, no)
o Duration of other NSAID use
o Type/s of other NSAID use
o Dosage of other NSAIDs

Diabetes (at baseline and follow-up)
o Personal history of diabetes
o Age at diagnosis of diabetes
o Treatment for diabetes (Yes, No, Unknown)
o Specific medication used to treat diabetes
o Duration of medication use to treat diabetes

Demographics
o Year of birth
o Year of questionnaire return
o Age at questionnaire return (including decimals)
o Sex
o Race
o Ethnicity
o Education
o Marital status
o Year of follow-up questionnaire (if applicable)

Family history of cancer (at baseline and follow-up)
o First degree family history of glioma (if available)

Causal diagram analysis will be used to select covariates for individual analyses, in addition to demographic and family history, variables listed under the exposure subheading may be included as covariates in other analyses (e.g. smoking, alcohol consumption, BMI).

No