Skip to main content

Table 3 Risk prediction models for CRC metastasis

From: Risk factors and risk prediction models for colorectal cancer metastasis and recurrence: an umbrella review of systematic reviews and meta-analyses of observational studies

Author, year Country Study design Population Outcome Prediction time horizon Sample size (development/validation) Predictors Model performance (95%CI) Model presentation Internal validation
*Macias-Garcia, 2015 [22] Spain D Submucosal invasive (T1) CRC Lymph node metastasis NA 97 Sessile morphology AUC 0.90 (0.81–0.99) Risk score NA
Tumor differentiation
Infiltrative growth pattern
Lymphoid infiltrate
Taylor, 1990 [23] UK D CRC Liver metastasis 10 years 134 Sex Sensitivity 0.74, specificity 0.62 Formula NA
ALP
Dukes B
Dukes C
Segelman, 2014 [24] Sweden D + IV CRC (I–III) Peritoneal metastasis 3 years colon 5348/rectal 2696 Age C statistic: colon 0.80, rectal 0.78; calibration plot Calculator Bootstrapping
Cancer location
pT stage
pN stage
No. of lymph nodes examined
Preoperative radiotherapy radicality
Type of surgery
Adjuvant chemotherapy
*Huang, 2016 [25] China D + IV CRC Lymph node metastasis NA 266/60 Radiomics signature C statistic 0.736 (0.759–0.766); Hosmer–Lemeshow test: p = 0.916 Nomogram Random split
CEA
CT-reported LN status
*Hu, 2019 [26] China D + IV CRC patients with indeterminate pulmonary nodules Lung metastasis NA 136/58 Chronicity (synchronous nodule or metachronous lesion) AUC 0.929 (0.885–0.974); calibration plot Nomogram Random split
Rad-score
pN stage
*Xu, 2019 [27] China D + IV CRC Synchronous bone metastasis NA 41,902/13,967 Cancer location AUC 0.903; sensitivity 0.851; specificity 0.845 Risk score Random split
Tumor differentiation
Cancer histological type
CEA
pN stage
Brain metastasis
Liver metastasis
Lung metastasis
*JR, 2019 [28] Korea D + EV Submucosal invasive (T1) CRC Lymph node metastasis NA 833/722 Histologic grade AUC 0.812 (0.770–0.855); Hosmer–Lemeshow test: p = 0.737 Nomogram NA
Submucosal invasion
Vascular invasion
Tumor budding
Beumer, 2014 [29] Netherlands D + EV CRC Distant metastasis 5 years 50/43 MiR25/miR339 AUC 0.80 (0.67–0.93); calibration plot Nomogram NA
AJCC stage
Age at surgery
Sex
*Wang, 2017 [30] China D Colon cancer Peritoneal metastasis NA 1417 Age ROC 0.753 Nomogram NA
pT stage
Lesion not traversable
Infiltrative growth pattern
Tumor size
CEA
Histopathologic type of mucinous or signet ring cell adenocarcinoma
Gijn, 2015 [31] Netherlands D + IV Rectal cancer (tis–III) Metastasis 5 years 2172 Distance from the anal verge C statistic 0.761 (0.740–0.784); Hosmer–Lemeshow test: p = 0.82 Nomogram Cross-validation
pT stage
pA stage
pN stage
Surgery type
Residual tumor status
Valentini, 2011 [32] Belgium D + EV Rectal cancer (II–III) Metastasis 5 years 3458 pT stage External C statistic 0.73 (0.68–0.77); Wald statistic: p = 0.057 Nomogram Random split
pN stage
Surgery type
Adjuvant chemotherapy
Sun, 2017 [33] China D + EV Rectal cancer (I–III) Distant metastasis 5 years 425/97 CRM C statistic 0.70 (0.64–0.75)/0.71 (0.62–0.81); calibration plot Nomogram NA
IMA nodes
AJCC stage
  1. Reproduced from He et al. [21]
  2. Abbreviations: D model development, D + IV model development with internal validation, D + EV model development with external validation, AJCC American Joint Committee on Cancer, ALP alkaline phosphatase, AUC area under the curve, CEA carcinoembryonic antigen, CRC colorectal cancer, CRM circumferential resection margin, IMA inferior mesenteric artery, NA non-available, miRNA microRNA
  3. *Six recently developed prediction models were additionally included, and data was extracted based on the previous criteria