Table 3.
Model development
| Citation | Sample size | Follow-up period | End point | Events (n) | Candidate selection method | Final predictor selection | Final predictors (n) | Validation method |
| Baade et al25 | 28 654 | Median=7.2 years (86.4 months) | MSS | 1700 | Prespecification from existing data | Backward elimination | 7 | Internal-external cross validation |
| Balch et al26 | 13 581 | 5–20 years | MSS | NR | Unclear | Unclear | 8 | External—geographic |
| Cochran et al27 | 1042 | Median=42.5 months | OS | NR | Prespecification+univariate analysis | Backward elimination | 5 | Internal sample validation |
| Cochran et al27 | 1042 | Median=42.5 months | Recurrence | NR | Prespecification+univariate analysis | Backward elimination | 4 | Internal sample validation |
| El Sharouni et al28 | DEV: 25 930 VALID: 2968 |
DEV: Median=6.7 years VALID: Median=12 years |
LRFS | 232 | Prespecification from existing data | Backward elimination | 6 | Internal-external cross validation |
| El Sharouni et al28 | DEV: 25 930 VALID: 2968 |
DEV: Median=6.7 years VALID: Median=12 years |
RRFS | 564 | Prespecification from existing data | Backward elimination | 7 | Internal-external cross validation |
| El Sharouni et al28 | DEV: 25 930 VALID: 2968 |
DEV: Median=6.7 years VALID: Median=12 years |
DRFS | 278 | Prespecification from existing data | Backward elimination | 6 | Internal-external cross validation |
| Fonseca et al29 | 259 | Median=80.13 months | DRFS | 117 | Univariate analysis | Backward elimination | 5 | Bootstrap |
| Fonseca et al29 | 259 | Median=80.13 months | MSS | 117 | Univariate analysis | Backward elimination | 5 | Bootstrap |
| Gimotty et al31 | DEV: 884 VALID: 114 | At least 10 years | Metastasis | 127 | Multivariate analysis | Backward elimination | 4 | External—new cohort |
| Gimotty et al30 | DEV: 26 114 | Median=4.6 years | OS | 3593 | Prespecification from existing data | Full model approach | 4 | External—geographic |
| Gimotty et al30 | VALID: 2389 | Median=8.1 years | OS | 1076 | Prespecification from existing data | Full model approach | 5 | External—geographic |
| Maurichi et al32 | 2243 | 124 months | 12 year OS | 240 | Multivariate analysis | Full model approach | 6 | Congruence examination |
| Rosenbaum et al33 | DEV: 506 VALID: 149 | Median=4.4 years | RFS | NR | Univariate analysis | Backward elimination | 3 | Internal k-fold validation |
| Soong et al34 | DEV: 14 760 VALID: 10 974 |
5 and 10 years | OS | NR | Multivariate analysis | Full model approach | 6 | External—geographic |
| Tsai et al35 | 13 268 | Not reported | OS | NR | Prespecification from existing data | Full model approach | 6 | Internal-external cross validation |
| Verver et al36 | 3180 | Median=61 months | Composite recurrence and/or MSM | Recurrence=496 Dead=277 | Prespecification+univariate analysis | Backward elimination | 3 | Internal-external cross validation |
| Vollmer and Seigler37 | 1910 | Median=7.6 years | Mortality | NR | Prespecification | Full model approach | 5 | Internal k-fold validation |
| Xiao et al38 | DEV: 46 336 VALID: 19 856 |
3 and 5 years | OS | NR | Prespecification+univariate analysis | Full model approach | 10 | Internal sample validation |
DEV, development set; DRFS, distant recurrence-free survival; EPV, events per variable; MSM, melanoma specific mortality; MSS, melanoma specific survival; NR, not reported; OS, overall survival; RFS, recurrence free survival; RRFS, regional recurrence free survival; VALID, validation set.