Development and validation of a risk score predicting risk of colorectal cancer

Annika Steffen, Robert J. MacInnis, Grace Joshy, Graham G. Giles, Emily Banks, David Roder

Research output: Contribution to journalArticleResearchpeer-review

11 Citations (Scopus)

Abstract

Background: Quantifying the risk of colorectal cancer for individuals is likely to be useful for health service provision. Our aim was to develop and externally validate a prediction model to predict 5-year colorectal cancer risk. Methods: Weused proportional hazards regression to develop the model based on established personal and lifestyle colorectal cancer risk factors using data from 197,874 individuals from the 45 and Up Study, Australia. We subsequently validated the model using 24,233 participants from the Melbourne Collaborative Cohort Study (MCCS). Results: Atotal of 1,103 and 224 cases of colorectal cancer were diagnosed in the development and validation sample, respectively. Our model, which includes age, sex, BMI, prevalent diabetes, ever having undergone colorectal cancer screening, smoking, and alcohol intake, exhibited a discriminatory accuracy of 0.73 [95% confidence interval (CI), 0.72-0.75] and 0.70 (95% CI, 0.66-0.73) using the development and validation sample, respectively. Calibration was good for both study samples. Stratified models according to colorectal cancer screening history, that additionally included family history, showed discriminatory accuracies of 0.75 (0.73-0.76) and 0.70 (0.67-0.72) for unscreened and screened individuals of the development sample, respectively. In the validation sample, discrimination was 0.68 (0.64-0.73) and 0.72 (0.67-0.76), respectively. Conclusion: Our model exhibited adequate predictive performance that was maintained in the external population. Impact: The model may be useful to design more powerful cancer prevention trials. In the group of unscreened individuals, the model may be useful as a preselection tool for population-based screening programs.

Original languageEnglish
Pages (from-to)2543-2552
Number of pages10
JournalCancer Epidemiology Biomarkers and Prevention
Volume23
Issue number11
DOIs
Publication statusPublished - Nov 2014
Externally publishedYes

Cite this

Steffen, Annika ; MacInnis, Robert J. ; Joshy, Grace ; Giles, Graham G. ; Banks, Emily ; Roder, David. / Development and validation of a risk score predicting risk of colorectal cancer. In: Cancer Epidemiology Biomarkers and Prevention. 2014 ; Vol. 23, No. 11. pp. 2543-2552.
@article{8eee57bde5a342febef670bfdd5ef847,
title = "Development and validation of a risk score predicting risk of colorectal cancer",
abstract = "Background: Quantifying the risk of colorectal cancer for individuals is likely to be useful for health service provision. Our aim was to develop and externally validate a prediction model to predict 5-year colorectal cancer risk. Methods: Weused proportional hazards regression to develop the model based on established personal and lifestyle colorectal cancer risk factors using data from 197,874 individuals from the 45 and Up Study, Australia. We subsequently validated the model using 24,233 participants from the Melbourne Collaborative Cohort Study (MCCS). Results: Atotal of 1,103 and 224 cases of colorectal cancer were diagnosed in the development and validation sample, respectively. Our model, which includes age, sex, BMI, prevalent diabetes, ever having undergone colorectal cancer screening, smoking, and alcohol intake, exhibited a discriminatory accuracy of 0.73 [95{\%} confidence interval (CI), 0.72-0.75] and 0.70 (95{\%} CI, 0.66-0.73) using the development and validation sample, respectively. Calibration was good for both study samples. Stratified models according to colorectal cancer screening history, that additionally included family history, showed discriminatory accuracies of 0.75 (0.73-0.76) and 0.70 (0.67-0.72) for unscreened and screened individuals of the development sample, respectively. In the validation sample, discrimination was 0.68 (0.64-0.73) and 0.72 (0.67-0.76), respectively. Conclusion: Our model exhibited adequate predictive performance that was maintained in the external population. Impact: The model may be useful to design more powerful cancer prevention trials. In the group of unscreened individuals, the model may be useful as a preselection tool for population-based screening programs.",
author = "Annika Steffen and MacInnis, {Robert J.} and Grace Joshy and Giles, {Graham G.} and Emily Banks and David Roder",
year = "2014",
month = "11",
doi = "10.1158/1055-9965.EPI-14-0206",
language = "English",
volume = "23",
pages = "2543--2552",
journal = "Cancer Epidemiology Biomarkers and Prevention",
issn = "1055-9965",
publisher = "American Association for Cancer Research (AACR)",
number = "11",

}

Development and validation of a risk score predicting risk of colorectal cancer. / Steffen, Annika; MacInnis, Robert J.; Joshy, Grace; Giles, Graham G.; Banks, Emily; Roder, David.

In: Cancer Epidemiology Biomarkers and Prevention, Vol. 23, No. 11, 11.2014, p. 2543-2552.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Development and validation of a risk score predicting risk of colorectal cancer

AU - Steffen, Annika

AU - MacInnis, Robert J.

AU - Joshy, Grace

AU - Giles, Graham G.

AU - Banks, Emily

AU - Roder, David

PY - 2014/11

Y1 - 2014/11

N2 - Background: Quantifying the risk of colorectal cancer for individuals is likely to be useful for health service provision. Our aim was to develop and externally validate a prediction model to predict 5-year colorectal cancer risk. Methods: Weused proportional hazards regression to develop the model based on established personal and lifestyle colorectal cancer risk factors using data from 197,874 individuals from the 45 and Up Study, Australia. We subsequently validated the model using 24,233 participants from the Melbourne Collaborative Cohort Study (MCCS). Results: Atotal of 1,103 and 224 cases of colorectal cancer were diagnosed in the development and validation sample, respectively. Our model, which includes age, sex, BMI, prevalent diabetes, ever having undergone colorectal cancer screening, smoking, and alcohol intake, exhibited a discriminatory accuracy of 0.73 [95% confidence interval (CI), 0.72-0.75] and 0.70 (95% CI, 0.66-0.73) using the development and validation sample, respectively. Calibration was good for both study samples. Stratified models according to colorectal cancer screening history, that additionally included family history, showed discriminatory accuracies of 0.75 (0.73-0.76) and 0.70 (0.67-0.72) for unscreened and screened individuals of the development sample, respectively. In the validation sample, discrimination was 0.68 (0.64-0.73) and 0.72 (0.67-0.76), respectively. Conclusion: Our model exhibited adequate predictive performance that was maintained in the external population. Impact: The model may be useful to design more powerful cancer prevention trials. In the group of unscreened individuals, the model may be useful as a preselection tool for population-based screening programs.

AB - Background: Quantifying the risk of colorectal cancer for individuals is likely to be useful for health service provision. Our aim was to develop and externally validate a prediction model to predict 5-year colorectal cancer risk. Methods: Weused proportional hazards regression to develop the model based on established personal and lifestyle colorectal cancer risk factors using data from 197,874 individuals from the 45 and Up Study, Australia. We subsequently validated the model using 24,233 participants from the Melbourne Collaborative Cohort Study (MCCS). Results: Atotal of 1,103 and 224 cases of colorectal cancer were diagnosed in the development and validation sample, respectively. Our model, which includes age, sex, BMI, prevalent diabetes, ever having undergone colorectal cancer screening, smoking, and alcohol intake, exhibited a discriminatory accuracy of 0.73 [95% confidence interval (CI), 0.72-0.75] and 0.70 (95% CI, 0.66-0.73) using the development and validation sample, respectively. Calibration was good for both study samples. Stratified models according to colorectal cancer screening history, that additionally included family history, showed discriminatory accuracies of 0.75 (0.73-0.76) and 0.70 (0.67-0.72) for unscreened and screened individuals of the development sample, respectively. In the validation sample, discrimination was 0.68 (0.64-0.73) and 0.72 (0.67-0.76), respectively. Conclusion: Our model exhibited adequate predictive performance that was maintained in the external population. Impact: The model may be useful to design more powerful cancer prevention trials. In the group of unscreened individuals, the model may be useful as a preselection tool for population-based screening programs.

UR - http://www.scopus.com/inward/record.url?scp=84920134143&partnerID=8YFLogxK

U2 - 10.1158/1055-9965.EPI-14-0206

DO - 10.1158/1055-9965.EPI-14-0206

M3 - Article

VL - 23

SP - 2543

EP - 2552

JO - Cancer Epidemiology Biomarkers and Prevention

JF - Cancer Epidemiology Biomarkers and Prevention

SN - 1055-9965

IS - 11

ER -