Table 5.

Outcomes of model applications to infer smoking history (pack-years) in current smokers (N = 364) from blood based on CpGs

	13-CpG model			10-CpG modela
	Model Building (N = 364)	Fivefold Cross-validation	KORA F4 (N = 224)	Model Building (N = 364)	Fivefold Cross-validation	SHIP-Trend (N = 41)
More or less than 10 pack-years
Accuracy (95% CI)b	0.824 (0.781, 0.862)	0.783 ± 0.05	0.813 (0.755, 0.861)	0.808 (0.76, 0.847)	0.770 ± 0.035	0.805 (0.651, 0.912)
Specificity	0.644	0.577 ± 0.131	0.343	0.602	0.548 ± 0.14	0.778
Sensitivity	0.911	0.882 ± 0.045	0.899	0.907	0.879 ± 0.046	0.813
AUC	0.846	0.800 ± 0.068	0.796	0.834	0.809 ± 0.039	0.837
More or less than 15 pack-years
Accuracy (95% CI)b	0.733 (0.685, 0.778)	0.719 ± 0.093	0.786 (0.726, 0.838)	0.728 (0.679, 0.773)	0.709 ± 0.059	0.659 (0.494, 0.799)
Specificity	0.617	0.600 ± 0.204	0.455	0.597	0.575 ± 0.143	0.533
Sensitivity	0.819	0.805 ± 0.042	0.894	0.824	0.808 ± 0.035	0.731
AUC	0.815	0.767 ± 0.102	0.752	0.786	0.757 ± 0.077	0.779

Cross-validation analysis results are presented as mean ± standard deviation

Pack-years were calculated as the number of cigarettes smoked per day divided by 20, multiplied by the total years of smoking

^aThree CpGs (cg06126421, cg22132788 and cg05951221) are not included in the EPIC methylation microarray dataset from SHIP-Trend

^bProportion accurately inferred smoking habits; 95% CI, confidence interval; AUC, Area under the Curve