Table 3.
Calibration table for predicting 10-year non-breast cancer death and 5-year breast cancer risk
| Predicting 10-year non-breast cancer death | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NHS development cohort n = 48,102a | NHS validation cohort n = 24,088a | Black Women’s Health Study n = 15,001 | ||||||||||
| Risk groupb | Time (years) | Deaths (%) | Expected survival (using CRR) | Observed survival | Deaths (%) | Expected survival (using CRR) | Observed survival | Expected/observed (E/O) | Deaths (%) | Expected survival (using CRR) | Observed survival | Expected/observed (E/O) |
| 1 | 5 | 0.99 | 0.99 | 0.99 | 0.99 | 1.00 | 0.99 | 0.99 | 1.00 | |||
| 10 | 251 (2.6) | 0.96 | 0.98 | 126 (2.7) | 0.96 | 0.98 | 0.98 | 74 (2.2) | 0.96 | 0.98 | 0.98 | |
| 2 | 5 | 0.98 | 0.98 | 0.98 | 0.98 | 1.00 | 0.98 | 0.98 | 1.00 | |||
| 10 | 553 (5.8) | 0.92 | 0.95 | 305 (6.3) | 0.93 | 0.94 | 0.99 | 220 (5.2) | 0.92 | 0.95 | 0.97 | |
| 3 | 5 | 0.96 | 0.97 | 0.96 | 0.96 | 1.00 | 0.96 | 0.97 | 0.99 | |||
| 10 | 1133 (11.8) | 0.86 | 0.89 | 594 (12.2) | 0.86 | 0.89 | 0.97 | 284 (8.1) | 0.87 | 0.92 | 0.95 | |
| 4 | 5 | 0.92 | 0.93 | 0.92 | 0.94 | 0.98 | 0.92 | 0.94 | 0.98 | |||
| 10 | 2320 (24.1) | 0.75 | 0.78 | 1172 (24.5) | 0.75 | 0.77 | 0.97 | 423 (18.0) | 0.75 | 0.83 | 0.90 | |
| 5 | 5 | 0.73 | 0.80 | 0.74 | 0.80 | 0.93 | 0.76 | 0.84 | 0.90 | |||
| 10 | 5119 (53.2) | 0.47 | 0.49 | 2593 (53.1) | 0.47 | 0.49 | 0.96 | 591 (40.1) | 0.44 | 0.62 | 0.71 | |
| Predicting 5-year invasive breast cancer risk | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NHS development cohort n = 37,628a | NHS validation cohort n = 18,980a | Black Women’s Health Study n = 13,247 | ||||||||||
| Risk group | Time (years) | Invasive breast cancers (%) | Expected survival (using CRR) | Observed survival | Invasive breast cancers (%) | Expected survival (using CRR) | Observed survival | Expected/observed (E/O) | Invasive breast cancers (%) | Expected survival (using CRR) | Observed survival | Expected/observed (E/O) |
| 1 | 5 | 80 (1.1) | 0.99 | 0.99 | 40 (1.1) | 0.99 | 0.99 | 1.00 | 40 (1.2) | 0.99 | 0.99 | 1.00 |
| 2 | 5 | 122 (1.6) | 0.98 | 0.99 | 52 (1.3) | 0.98 | 0.99 | 0.99 | 40 (1.5) | 0.98 | 0.99 | 0.99 |
| 3 | 5 | 112 (1.5) | 0.98 | 0.99 | 74 (1.9) | 0.98 | 0.98 | 1.00 | 39 (1.6) | 0.98 | 0.98 | 1.00 |
| 4 | 5 | 150 (2.0) | 0.98 | 0.98 | 76 (2.1) | 0.98 | 0.98 | 1.00 | 50 (2.0) | 0.98 | 0.98 | 1.00 |
| 5 | 5 | 248 (3.3) | 0.96 | 0.97 | 123 (3.3) | 0.97 | 0.97 | 1.00 | 54 (2.4) | 0.97 | 0.98 | 0.99 |
Calibration of the model in predicting non-BC death was assessed by estimating the ratio of the expected survival (1-CIF for non-BC death from our CRR model) to the observed survival (1-the observed CIF computed using the nonparametric estimation of CIF) at 5 and 10 years within risk quintiles. Similar methods were used to test calibration of the model in predicting BC
CRR = Competing risk regression, CIF = cumulative incidence function, NHS = Nurses’ Health Study, BWHS = Black Women’s Health Study
aThese analyses include women with complete data. When predicting breast cancer, women with a history of cancer were further excluded
bRisk groups were defined by quintiles of the NHS prognostic index for each outcome which was calculated using NHS development cohort regression coefficients