Table 3.
Forward stepwise models showing the incremental contribution of MC1R genotype and traditional risk factors to risk prediction models for melanoma, with each factor added in order of their contribution to improving the AUC, shown separately for models using self-reported factors and physician-measured factors
|
Predictors in order of entry1 |
AUC (95% CI) |
Incremental change in AUC2 |
P3 |
Incremental improvement in sensitivity4 |
Incremental improvement in specificity4 |
Overall incremental improvement in classification4 |
|||
|---|---|---|---|---|---|---|---|---|---|
| NRI (95% CI) | P | NRI (95% CI) | P | NRI (95% CI) | P | ||||
| Base model |
0.67 (0.63, 0.72) |
|
|
|
|
|
|
|
|
|
Self-reported model: including only self-reported nevus and phenotype risk factors | |||||||||
|
MC1R all variants |
0.73 (0.69, 0.77) |
0.058 |
<0.001 |
0.12 (0.05, 0.19) |
0.001 |
0.14 (0.06, 0.23) |
<0.001 |
0.26 (0.15, 0.37) |
<0.001 |
| Nevi (none, few, some many) |
0.77 (0.73, 0.81) |
0.038 |
0.001 |
0.14 (0.07, 0.20) |
<0.001 |
−0.00 (−0.09, 0.08) |
0.93 |
0.13 (0.02, 0.24) |
0.02 |
| Non-melanoma skin cancer |
0.78 (0.75, 0.82) |
0.012 |
0.02 |
−0.02 (−0.06, 0.01) |
0.2 |
0.05 (0.01, 0.08) |
0.01 |
0.02 (−0.03, 0.07) |
0.4 |
| Pigmentation score |
0.79 (0.76, 0.83) |
0.009 |
0.07 |
0.04 (−0.01, 0.08) |
0.1 |
0.02 (−0.04, 0.08) |
0.5 |
0.06 (−0.02, 0.13) |
0.1 |
| Sun & sunbed exposure |
0.80 (0.76, 0.83) |
0.004 |
0.4 |
0.02 (−0.02, 0.06) |
0.4 |
0.03 (−0.02, 0.09) |
0.2 |
0.05 (−0.01, 0.12) |
0.1 |
| Family history |
0.80 (0.76, 0.83) |
0.001 |
0.7 |
0.00 (−0.02, 0.03) |
0.9 |
0.00 (−0.02, 0.03) |
0.8 |
0.01 (−0.03, 0.04) |
0.7 |
|
Physician-measured model: including physician-measured nevus and phenotype risk factors where available | |||||||||
| Nevi 2+ mm5 |
0.79 (0.75, 0.82) |
0.111 |
<0.001 |
0.10 (0.03, 0.17) |
0.008 |
0.29 (0.20, 0.38) |
<0.001 |
0.39 (0.28, 0.51) |
<0.001 |
|
MC1R all variants |
0.82 (0.78, 0.85) |
0.031 |
0.002 |
0.05 (−0.01, 0.10) |
0.1 |
0.09 (0.02, 0.15) |
0.007 |
0.13 (0.05, 0.22) |
0.002 |
| Non-melanoma skin cancer |
0.83 (0.80, 0.86) |
0.010 |
0.02 |
−0.01 (−0.04, 0.02) |
0.4 |
0.03 (0.01, 0.06) |
0.02 |
0.02 (−0.02, 0.06) |
0.3 |
| Pigmentation score5 |
0.83 (0.80, 0.86) |
0.003 |
0.5 |
0.01 (−0.03, 0.05) |
0.6 |
−0.03 (−0.08, 0.02) |
0.2 |
−0.02 (−0.08, 0.04) |
0.5 |
| Solar lentigines5 |
0.83 (0.80, 0.86) |
0.004 |
0.3 |
0.03 (−0.01, 0.07) |
0.10 |
−0.01 (−0.05, 0.03) |
0.7 |
0.02 (−0.03, 0.08) |
0.4 |
| Family history | 0.83 (0.80, 0.86) | <0.001 | 1.0 | 0.00 (−0.01, 0.02) | 0.5 | −0.01 (−0.03, 0.01) | 0.5 | −0.00 (−0.03, 0.02) | 0.8 |
AUC Area under the receiver operating characteristic curve, NRI Net reclassification improvement.
1 Each risk factor was added in a forward stepwise manner in order of their contribution to improving the AUC. The base model and individual risk factors are described in Table 2 footnotes.
2 The change in the AUC from the previous (incremental) model.
3 Chi-square p-value for the difference in the AUC when compared to the previous (incremental) model.
4 Based on quartile cut-points. Improvement in sensitivity is calculated from reclassification of cases, improvement in specificity is calculated from reclassification of controls, and the overall improvement in classification combines the improvements in sensitivity and specificity.
5 Objectively-measured risk factor – see Table 2 footnotes for descriptions.