I would build the model in the entire cohort of patients
then do the additional analysis we discussed in the following patient population. Baseline CP >= 7 AND [tumor size > 4cm (patients with more than 1 tumor, you can add tumor size to compare to 4cm)].
For your model, I think it is fine to use standard Tox definition of increase in CP of 2 points or more within 6 months.
LIVER-GTV Mean Dose (Gy) LQL: α/β=2.5, DT=5
## table: size>=4?##
## FALSE TRUE
## 68 37## table: baseline CP##
## 5 6 7 8 9 10 11
## 46 19 12 21 3 2 1## table: size>=4 or CP0>=7 ?##
## FALSE TRUE
## 37 67| Variable | N | N = 781 |
|---|---|---|
| CP0 | 77 | |
| 5 | 36 (47%) | |
| 6 | 12 (16%) | |
| 7 | 12 (16%) | |
| 8 | 12 (16%) | |
| 9 | 3 (3.9%) | |
| 10 | 1 (1.3%) | |
| 11 | 1 (1.3%) | |
| Unknown | 1 | |
| size_sum | 78 | |
| Mean (SD) | 4.999 (3.328) | |
| Median (Minimum-Maximum) | 4.050 (1.300-14.600) | |
| TNR | 78 | |
| Mean (SD) | 244.232 (97.049) | |
| Median (Minimum-Maximum) | 214.300 (115.400-544.800) | |
| mld | 39 | |
| Mean (SD) | 11.694 (6.778) | |
| Median (Minimum-Maximum) | 10.670 (2.790-26.750) | |
| Unknown | 39 | |
| toxic | 65 | 16 (25%) |
| Unknown | 13 | |
|
1
n (%)
|
||
##
## 0 1
## 28 8