Table 2.
Marker | Average Realizations, the Gold Standard | Unstratified Case-Cohort Design |
Unmatched Nested Case-Control Design | Stratified Case-Cohort Design |
Matched Nested Case-Control Design | ||||
Prentice (19)d | Self and Prentice (20)d | Barlow (21)d | Borgan I (22) d | Borgan II (22)d | Breslow Calibration (23)d | ||||
BIO1e | 1.61 (100) | 1.73 (12) | 1.78 (9) | 1.78 (10) | 1.75 (20) | 1.71 (16) | 1.65 (26) | 1.67 (26) | 1.69 (27) |
BIO2e | 1.24 (100) | 1.27 (18) | 1.28 (16) | 1.28 (16) | 1.30 (26) | 1.26 (25) | 1.25 (31) | 1.26 (31) | 1.26 (36) |
HDL-Ce | 0.62 (100) | 0.61 (22) | 0.60 (21) | 0.60 (21) | 0.57 (31) | 0.60 (29) | 0.60 (31) | 0.61 (61) | 0.63 (42) |
BIO1f | 1.61 (100) | 1.66 (28) | 1.67 (26) | 1.67 (26) | 1.69 (44) | 1.65 (42) | 1.63 (55) | 1.63 (54) | 1.66 (54) |
BIO2f | 1.24 (100) | 1.25 (39) | 1.26 (38) | 1.26 (38) | 1.27 (54) | 1.25 (55) | 1.24 (60) | 1.24 (59) | 1.26 (65) |
HDL-Cf | 0.62 (100) | 0.62 (49) | 0.62 (49) | 0.62 (49) | 0.59 (55) | 0.61 (60) | 0.61 (61) | 0.62 (82) | 0.63 (67) |
Abbreviations: BIO1, simulated biomarker 1; BIO2, simulated biomarker 2; HDL-C, high density lipoprotein cholesterol.
For 1 − SD increase of the biomarker. Hazard ratios are estimated by exponentiating the average of the ln(HR) values obtained from 2,000 realizations.
“Empirical relative efficiency” is defined as the ratio between the empirical variance of the ln(HR) calculated for 2,000 realizations from the entire cohort and the empirical variance calculated for each design, expressed as percentage.
All estimates reported are from Cox proportional hazard analyses or conditional logistic regression adjusted for age, sex, systolic blood pressure, antihypertensive treatment, diabetes, current smoking, and total cholesterol.
Reference number.
Sampling ratio: 1:1; subcohort size: n = 229. There were 459, 467, 454, and 454 individual subjects for the unstratified case-cohort design, the unmatched nested case-control design, the stratified case-cohort design, and the matched nested case-control design, respectively.
Sampling ratio: 1:3; subcohort size: n = 632. There were 847, 896, 835, and 833 individual subjects for the unstratified case-cohort design, the unmatched nested case-control design, the stratified case-cohort design, and the matched nested case-control design, respectively.