MC1Rgenotype as a predictor of early-onset melanoma, compared with self-reported and physician-measured traditional risk factors: an Australian case-control-family study

Cust, Anne E; Goumas, Chris; Vuong, Kylie; Davies, John R; Barrett, Jennifer H; Holland, Elizabeth A; Schmid, Helen; Agha-Hamilton, Chantelle; Armstrong, Bruce K; Kefford, Richard F; Aitken, Joanne F; Giles, Graham G; Bishop, D Timothy; Newton-Bishop, Julia A; Hopper, John L; Mann, Graham J; Jenkins, Mark A

doi:10.1186/1471-2407-13-406

Table 4 Performance measures for the final selected self-reported and physician-measured risk prediction models for melanoma that include MC1R , nevi and non-melanoma skin cancer

From: MC1Rgenotype as a predictor of early-onset melanoma, compared with self-reported and physician-measured traditional risk factors: an Australian case-control-family study

Performance measure	Base model¹	Self-reported model¹	Physician-measured model¹
Discrimination
AUC (95% CI)	0.67 (0.63, 0.72)	0.78 (0.75, 0.82)	0.83 (0.80, 0.86)
Change in AUC from the base model		0.108 (P <0.001)	0.152 (P <0.001)
Discrimination slope	0.099	0.242	0.306
Integrated discrimination index (IDI)		0.143	0.207
Sensitivity, given a specificity of 90%	23%	40%	53%
Reclassification (compared to the base model)
NRI (95% CI) based on quartile categories
Improvement in sensitivity		0.20 (0.13, 0.27)	0.20 (0.12, 0.27)
Improvement in specificity		0.17 (0.08, 0.26)	0.33 (0.24, 0.42)
Overall improvement in classification		0.37 (0.25, 0.48)	0.53 (0.41, 0.64)
Category-free NRI
Improvement in sensitivity		0.33 (0.23, 0.42)	0.32 (0.23, 0.42)
Improvement in specificity		0.30 (0.18, 0.42)	0.53 (0.41, 0.65)
Overall improvement in classification		0.63 (0.47, 0.78)	0.85 (0.70, 1.01)
Calibration
Hosmer-Lemeshow test P value	0.68	0.72	0.01
Overall performance
Nagelkerke’s R²	0.131	0.315	0.393
Brier score	0.214	0.180	0.164
Internal validation
Nagelkerke’s R² (95% CI)	0.14 (0.09, 0.19)	0.34 (0.28, 0.41)	0.41 (0.35, 0.47)
AUC (95% CI)	0.68 (0.64, 0.71)	0.79 (0.76, 0.83)	0.83 (0.81, 0.86)
External validation using data from English study
AUC (95% CI)	0.61 (0.57, 0.64)	0.71 (0.68, 0.74)	0.79 (0.76, 0.81)
Change in AUC from the base model		0.105	0.182
Discrimination slope	0.035	0.126	0.219
NRI (95% CI) based on quartile categories
Improvement in sensitivity		0.20 (0.14, 0.25)	0.12 (0.07, 0.18)
Improvement in specificity		0.11 (0.04, 0.18)	0.34 (0.27, 0.41)
Overall improvement in classification		0.30 (0.22, 0.39)	0.46 (0.37, 0.55)
Hosmer-Lemeshow test P value	0.17	0.39	0.0003
Nagelkerke’s R²	0.050	0.166	0.303
Brier score	0.220	0.199	0.176

AUC Area under the receiver operating characteristic curve, NRI Net reclassification improvement.
¹ The base model included demographic factors age, sex, city of recruitment and European ancestry. Both the self-reported model and the physician-measured model included the variables from the base model plus MC1R, non-melanoma skin cancer and nevi (categorised as ‘none, few some, many’ in the self-reported model, and as a continuous variable ‘the number of nevi ≥ 2 mm’ in the physician-measured model.

Back to article page

ISSN: 1471-2407

Contact us

Submission enquiries: bmccancer@biomedcentral.com
General enquiries: ORSupport@springernature.com

BMC Cancer

Contact us