Table 4.

Logistic regression models of phosphenes and/or phantosmias during radiation therapy

	B	SE	Wald	p	Exp(B)	95%CI
						LB	UB
Model 1: Phantosmia vs. no phantosmia
Taste problem = Yes	−1.73	0.81	4.52	0.03	0.17	0.04	0.87
Taste problem = No	Reference
Age	−0.07	0.02	11.92	< 0.01	0.94	0.90	0.97
Model 2: Phosphene vs. no phosphene
Technology = Proton	1.88	0.60	9.90	< 0.01	6.54	2.03	21.06
Technology = Photon	Reference
RT region = Brain	2.34	0.85	7.55	< 0.01	10.36	1.95	54.92
RT region = ENT-related region	2.03	0.90	5.05	0.03	7.58	1.30	44.31
RT region = Other body region	Reference
Model 3: Both sensations vs. no sensation
Technology = Proton	2.16	0.93	5.35	0.02	8.67	1.39	54.10
Technology = Photon	Reference
Taste problem = Yes	−2.48	1.15	4.69	0.03	0.08	0.01	0.79
Taste problem = No	Reference

B beta coefficient, SE standard error, Wald Wald chi-square statistic, p level of significance, Exp(B) Exponentiated beta coefficient, 95% CI 95% confidence interval, LB lower bound, UB upper bound.

Stepwise method was used to select the most important predictors of phosphene or/and phantosmia during the radiation therapy