. 2020 Sep 15;9(21):7925–7934. doi: 10.1002/cam4.3414

TABLE 5.

Factors associated with early changes in each domain of the EPIC‐26 score over time

EPIC‐26 domain	Independent variable	Coefficient	P‐value ^a ^, ^b
Bowel	Radiation modality: PBT	9.1	<.001
Urinary irritative/obstructive	Baseline UO score	−0.4	.01
Urinary irritative/obstructive	(Baseline UI score)*(Time) ^c	See footnote ^c	<.001
Urinary incontinence	Baseline BF score	−0.3	.05
	Baseline UO score	0.3	.003
	Baseline UI score	−0.4	<.001

^{^a}

The generalized estimating equations (GEE) were used to identify factors that were statistically significantly associated with changes in the scores over time. Independent covariance structure was selected for the GEE, based on the quasi‐information criterion (QIC). ⁸ P‐values reflect a test of individual covariate effect. P < .05 is considered significant. Both QIC and P‐values were computed based on Wald test and robust standard errors.

^{^b}

Variables included in the GEE models were the type of radiation modality (PBT vs IMRT), dose‐fractionation regimen utilized (60 Gy/20 f vs 70.2 Gy/26 f vs 78 Gy/39 f), the use of hydrogel spacer to reduce radiation dose to the rectum (yes vs no), baseline BF, UO, and UI scores, age, PSA, Gleason score, T stage, and the use of ADT. We excluded race from this analysis because of the small number of non‐Caucasian patients. Indicator variables were included to distinguish radiation modality and the time of survey (the end of RT and 3 months post‐RT). Our model also included interaction terms between each variable and time indicator. Stepwise backward/forward variable selections were conducted for the model construction based on QIC. Statistically significant interactions were investigated in Simple Effect Tests to identify significance of main effects. Only those variables and interactions that were found statistically significant are tabulated.

^{^c}

Significant interaction was noted between baseline UI score and time. While a higher baseline UI score (ie less urinary incontinence) was associated with greater UO reduction at the end of RT (P = .016), there was no association between baseline UI score and UO score reduction at 3‐month post‐RT.