-
PDF
- Split View
-
Views
-
Cite
Cite
Sarah Wernly, Christian Datz, Bernhard Wernly, RE: Long-Term Colorectal Cancer Incidence and Mortality After Colonoscopy Screening According to Individuals’ Risk Profiles, JNCI: Journal of the National Cancer Institute, Volume 114, Issue 5, May 2022, Pages 779–780, https://doi.org/10.1093/jnci/djab232
- Share Icon Share
We read with great interest the manuscript, “Long-Term Colorectal Cancer Incidence and Mortality After Colonoscopy Screening According to Individuals’ Risk Profiles” by Wang et al. (1). We congratulate the authors on their attempt to evaluate the effect of screening colonoscopy on the basis of individual risk profiles. We agree with the authors that an individualized approach could improve the accuracy and acceptance of screening for colorectal cancer (CRC). However, we would like to mention 3 suggestions that could potentially strengthen the present and future analyses.
First, the rough dichotomization of the 8 included factors into “high risk” and “not high risk” does not seem justified. For example, a positive family history or smoking are weighted the equivalent to an unhealthy diet for CRC risk. We believe that the performance of a risk score for CRC could be improved by fitting a model for the binary outcome and calculating the respective regression coefficients for the individual risk factors. Based on these regression coefficients, it would then be possible to weight the individual risk factors (independent variables) differently and calculate an individual's patient risk as follows: risk score = intercept + (regression coefficient ×1 * value of independent variable ×1) and risk as: exp(patient’s risk score)/(1+exp[patient’s risk score]).
Second, interaction effects between the individual risk factors are likely. Interaction effects always occur when the effect of an independent variable on a dependent variable changes, depending on the value of another independent variable. Thus, from a clinical perspective, it is quite conceivable that—for example—the effect of an unhealthy diet varies according to body mass index. It could possibly improve the fit of the model to the data and the coefficient of determination to include these interaction effects in the regression analysis. Another example would be that the influence of height on the risk of CRC is sex specific (2).
Third, we were surprised that Wang et al. (1) did not include male sex as a risk factor. Several analyses showed that men have a higher risk for colorectal adenoma compared with women (3,4), independent of other established risk factors (5). We think that Wang et al. (1) took the risk factor male sex into account, namely via the variable height, because patients with tall stature, defined as the “upper 50% of height in each cohort,” were defined as high risk and are more likely men. However, the risk model could possibly be further improved by including not only height but also sex as a separate variable.
In summary, we believe that the very welcome analysis by Wang et al. (1) could have been refined in the following points: a statistical weighting of the risk factors instead of dichotomization, a consideration of interactions between the risk factors, and an inclusion of the established risk factor male sex in addition to height. We hope that these considerations will also be helpful for future analyses.
Funding
None.
Notes
Role of the funder: Not applicable.
Disclosures: The authors have no disclosures.
Author contributions: Writing, original draft—SW, CD, BW; writing, editing and revision—SW, CD, BW.
Data Availability
No new data were generated or used for this correspondence.