Page 45 - Risk quantification and modification in older patients with colorectal cancer

Recommendations for future research
For prediction model development and validation studies, sample size should be sufficient to reliably estimate a model’s performance. Furthermore, for prognostic research, calibration measures (reliability of the prediction for the different risk groups) within external validation studies have more importance then discrimination (who is at risk and who is not)54 because only reliable individual risks predictions can be used to make treatment decisions.
Also, a model may require periodic updating because of changes in the population under interest.55 Outcomes of colorectal surgery have improved due to care innovations such as auditing, ERAS (including laparoscopic surgery),56 neoadjuvant treatment and wait-and-see policies for rectal cancer (after complete remission after radiotherapy) and liberal use of defunctioning colostomy.57 Furthermore, a decrease in 30-day and one-year mortality after CRC surgery occurred in the past decades.58,59
Lastly, transparent reporting of future prediction model studies can improve by systematically using the TRIPOD guidelines.53
Conclusion
Many prediction models are available that address mortality and surgical complications after colorectal surgery, but not for prediction of quality of life or functional decline. Most of these models were not developed for older patients and include only a limited number of risk factors specific to older patients. Half of the included prediction models included peri-operative predictors, which limit their use for preoperative decision making. Future research should address geriatric characteristics to improve prediction models for preoperative decision making with older patients.
Acknowledgements
We thank Dr. N. van der Werf, Walaeus Medical Library of Leiden University Medical Centre, for her expertise in developing an appropriate search strategy for this review.
Risk prediction models for CRC patients
43
2