Prognostication for patients with colorectal cancer using real-world data

Accurate prognostication in colorectal cancer (CRC) is essential for informed decision-making and advance care planning. This thesis aims to improve survival and quality of life (QoL) predictions using real-world data (RWD) from the Netherlands Cancer Registry (NCR) and the Prospective Dutch Colorectal Cancer (PLCRC) cohort.
In Chapter 2, we assessed the generalizability of clinical trial outcomes for metastatic colorectal cancer (mCRC) patients in daily practice. While randomized trials have reported a substantial increase in median overall survival (mOS) over the past two decades, our analysis of 27,275 real-world patients (2008–2016) found no overall improvement (mOS ~12 months). Survival gains were limited to the best-case and upper-typical scenarios, suggesting that only a minority of patients benefit from newer treatment strategies. These findings highlight the need for clinicians to communicate multiple survival scenarios based on RWD rather than relying solely on trial-based estimates.
Chapter 3 presents the QUALITAS study, which evaluated QoL and survival outcomes in mCRC patients treated with trifluridine/tipiracil (FTD/TPI). Using EORTC QLQ-C30 and QLQ-CR29 questionnaires, we found that QoL remained stable throughout treatment. Patients with higher baseline QoL had significantly longer survival, underscoring the prognostic value of QoL assessments in treatment decision-making.
Chapters 4 and 5 focus on the external validation of CRC prognostic models. In Chapter 4, we evaluated the MSKCC nomogram for predicting five-year survival in stage I-III colon cancer. Using NCR data (n=39,805), we found good discriminatory ability (C-index 0.75) but slight overestimation of survival. These findings support the model’s applicability in European populations, though incorporating additional prognostic factors may enhance its accuracy.
In Chapter 5, we externally validated the Colon Life nomogram, designed to predict 12-week mortality in refractory mCRC patients. In a cohort of 150 QUALITAS patients, the model significantly overestimated mortality (observed/expected ratio 0.52) and lacked clinical utility based on decision curve analysis. These results suggest that the nomogram requires substantial recalibration before clinical implementation.
Chapter 6 explores the impact of CRC screening on survival in patients who later develop metachronous metastases. Analyzing 794 stage I-III CRC patients diagnosed in 2015, we found that those with screen-detected primary tumors had significantly better median OS after metastasis (38.3 vs. 19.2 months, p<0.0001). This association remained significant after adjusting for confounders (HR 0.70, 95% CI 0.56–0.89). These findings emphasize the prognostic relevance of primary tumor detection mode upon metachronous metastasis diagnosis.
Overall, this thesis demonstrates the value of RWD in refining survival predictions and treatment decision-making for CRC patients. By highlighting discrepancies between trial-based and real-world survival estimates, validating prognostic models, and emphasizing the role of QoL and screening outcomes, these studies contribute to more personalized and evidence-based patient care.