Page 28 - Risk quantification and modification in older patients with colorectal cancer
P. 28

                                Chapter 2
Observed/expected ratio or calibration plot). For all studies, we searched for external validation studies in the Pubmed and Embase databases.
Clinical predictors were classified into demographic-, comorbidity- (including American Society of Anesthesiologists Classification (ASA) score, Body Mass Index (BMI)) biochemical- (electrolytes and albumin), geriatric- (falls, functional dependency, independency (i) Activities of Daily Living (ADL), cognition) and non-geriatric predictors (all others, including weight loss).
Quality assessment
The methodological quality of each study was independently assessed by two reviewers (ETDS, EB). The CHARMS checklist was also used to evaluate the risk of bias and applicability concerns. Applicability refers to the extent to which the prediction model is useful for older colorectal cancer.7 The intended use is for preoperative shared decision making with older patients. Therefore, predictors need to be available preoperatively. In Appendix A and B, the criteria for quality assessment and applicability are described. These criteria were adapted from a systematic review of asthma prediction models by Smit et al.8 We defined a prediction tool representative for the average older patient with colorectal cancer, when at least 30% of the study population was 65 years or older. In European countries and the USA, more than half of all patients with colorectal cancer are 65 years or older.9 In case of a model update; the model development studies were reviewed to assess the method of predictor selection.
Data synthesis and analysis
We describe study characteristics and the outcomes of interest, the predictors of each model and the model’s performance. Furthermore, the quality (bias and applicability) of the prediction model studies was described.
Results
Study characteristics
The literature search identified 2885 citations (1899 from Medline and 1100 from Embase), of which 992 were duplicates. Details on the search and final study selection are shown in Figure 1. After exclusion of 2957 publications, 25
26
























































































   26   27   28   29   30