Table 1 Statistical and Endpoint Considerations from QUANTEC [5]
| Statistical considerations | |
|
1 Basic statistical data provided on incidence of toxicity -Both number of subjects and number of events should be reported -If an estimate of incidence is given the standard error should be supplied | |
| 2 Numerical labeling of response histogram – if into groups eg. quartiles must state number of patients in each quartile | |
| 3 When predictive models are correlated with complications parameter estimates must be stated with their standard error | |
| 4 Complication rates associated with constraints must be reported | |
| 5 “Goodness of fit” to be reported such as Chi-squared | |
| 6 Discriminator statistics reported such as receiver operating characteristic curves | |
|
7 Full organ volumes (rather than partial) should be used -If this is not possible absolute volumes should be used or a standard method of normalization -A clear statement of organ volume definition should be given | |
| Toxicity Endpoint considerations | |
| 1 Symptom-specific information rather than a portmanteau endpoint (eg. RTOG gr 2) should be used | |
| 2 Consideration that symptoms may be attributed to pre-radiotherapy co-morbidities | |
| 3 Patient-reporting of symptoms may be important |