Page 137 - New epidemiological and PSMA-expression based paradigms in salivary gland tumors

Tubarial salivary glands: A potential new organ at risk for radiotherapy
Strengths and Limitations
The clinical relevance of the tubarial glands was derived from a retrospective evaluation of multiple confounding OARs. Since all salivary glands are situated closely together, they often receive a comparable RT dose in treatment for HNC. As an example, in our data the dose to the parotid glands was highly correlated with the tubarial gland dose, with a correlation coefficient of 0.84. This means that reported xerostomia and dysphagia caused by RT dose to the tubarial glands, also includes a toxicity effect caused by dose to the parotid glands. This phenomenon in statistics, referred to as multicollinearity, complicates measuring a difference in toxicity effects caused by RT to different glands. The same applies to several muscles involved in swallowing in relation to dysphagia. Therefore, a large cohort was required to be able to measure such effects of RT dose to the tubarial glands, after correction for known toxicity effects to adjacent structures. External validity is likely to be warranted by this cohort size and inclusion of unselected consecutive patients. Regarding the internal validity, depending on the time point, 9-37% of the toxicity data are missing (complete) at random (MCAR or MAR) due to death, recurrence, or limited follow-up. A high compliance rate of (88-91%) ensures a low probability of data missing not at random (MNAR).
The association between RT dose to the tubarial glands and crude physician-
rated toxicities was present at all time points (p<.05). After correction for confounders this association was still present for xerostomia at 12 and 24 months 8 and for dysphagia at 6, 12 and 24 months. The fact that these association after
correction were not significant at all four time points can probably be explained
by multicollinearity. Similarly, radiation to the oral cavity (which also includes
palatal minor glands) might play a role, although RT to this area was not a significantly contributing factor in currently used prediction models[13,14].
More extensive and comprehensive modelling is expected to clarify the influence
of multicollinearity. The absence of a significant correlation with patient-rated
xerostomia after correction for confounders, may be explained by the inherently
higher variability in subjective evaluations of toxicity by patients.
The logical next step seems to be optimization of radiotherapy fields to the tubarial glands as new OARs. Since the PCM superior is close to the tubarial glands, sparing both structures simultaneously seems attractive. Still, we prefer to acquire external validation in an independent dataset, and advise to change clinical protocols
135