Patient | Observed data | Other simulated times-to-event | Data in the net survival setting |
|---|
T
O
| Vital status | Death from cancer | 0% misCoD | 20% misCoD | 30% misCoD |
T
P
|
T
E
|
T
C
| Death from another cause |
T
N
| StatusN |
|---|
1 | 4.61 | 1 | 0 | 0 | 0 | 1 | 4.61 | 11.27 | 15.00 | 1 | 11.27 | 1 |
2 | 9.21 | 1 | 1 | 1 | 1 | 1 | 63.78 | 9.21 | 15.00 | 0 | 9.21 | 1 |
3 | 7.83 | 1 | 1 | 1 | 1 | 0 | 65.06 | 7.83 | 15.00 | 0 | 7.83 | 1 |
4 | 12.47 | 1 | 1 | 1 | 0 | 1 | 58.17 | 12.47 | 15.00 | 0 | 12.47 | 1 |
5 | 15.00 | 0 | 0 | 0 | 0 | 0 | 52.64 | 25.12 | 15.00 | 0 | 15.00 | 0 |
- We present a short excerpt to clarify the use of simulation data with each estimation method. Column “Patient” shows an identifier. The simulated data include: (i) observed data: time-to-death (TO), the vital status (1 if death from any cause, 0 alive), the cause of death (1 if due to cancer, 0 otherwise or alive). The table considers three degrees of cause-of-death misclassification (misCoD 30, 20, and 0% a highly improbable setting); (ii) simulated times-to-death: from other cause (TP), and from cancer (TE), the time from censoring (TC); (iii) data observed if cancer were the only possible cause of death: the time to death TN (the smallest of TE and TC) and the vital status in the net survival setting where cancer is the only possible cause of death
- For example, the case of Patient 4, should be classified as death from cancer because the time-to-death from cancer (12.47 years) is lower than the time from censoring (15 years) and corresponds to the smallest of TP, TE and TC. This case would have been classified as “death from other cause” (i.e., 0% misCoD = 0) if he were followed-up to 58.17 years. In Column “20% misCoD”, the cause of death is wrongly coded “death from other cause”. In the latter case, 20% of all cancer deaths are misclassified as deaths from other causes
