Table 1

Predictor variables contained within the 16 identified eligible colorectal cancer risk prediction models

AuthorSexOutcomePredictor variables
Demographic characteristicsAnthropometryFamily
history of Cancer
Medical historyMedication useLifestyle factorsDiet
SexAgeEthEduHeightBMIDiabetesScreen/endo/
polyp
IBDMenopausal statusHRTOCNSAIDPhysical
activity
SmokingAlcoholRed meat/
meat
VegVit
Colditz23 MC
FC
Driver24 MCRC
Freedman7 MCRC
FCRC○ α○ α
Ma9 MCRC
MC
MR
Shin25 MR C
FR
Steffen10 BothCRC
BothC
BothR
Taylor8 BothCRC
Wells26 MCRC
FCRC○ µ○ µ
Variable availability in EPICna
Variable availability in UK Biobankna
  • ●, variable included in the original model; ○, two variables amalgamated to create a new variable that was included in the model. The names of these variables were: α, oestrogen status; µ, oestrogen. Family history of cancer varied by both cancer site and degrees of relatedness between models. ■, available in a construction directly suitable for at least one model; □, variable required to be derived from other variable(s) in the dataset for all models that used it; na, not applicable as the variable is amalgamated in the model in which it is used. See online supplementary table 1 for further details.

  • BMI, body mass index; C, colon cancer; CRC, colorectal cancer; Edu, years of education; EPIC, European Prospective Investigation into Cancer and Nutrition; Eth, ethnicity; F, female; HRT, hormone replacement therapy; IBD, inflammatory bowel disease; M, male; NSAID, non-steroidal anti-inflammatory drug; OC, oral contraceptive; R, rectal cancer; RC, right colon cancer; screen/endo/polyp, history of colorectal cancer screening or lower gastrointestinal endoscopy with or without identification of polyps; Veg, vegetables; Vit, multivitamin.