Table 3.

Univariable and multivariable logistic regression results for risk of serious hearing loss versus. demographic and clinical covariates

Variable	Univariable Logistic Regression Model				Multivariable Logistic Regression Model
	Estimate	Standard Error	Odds Ratio (95% CI)	P value	Estimate	Standard Error	Odds Ratio (95% CI)	P value
Age at diagnosis	−0.099	0.032	0.91 (0.85–0.96)	.002	−0.089	0.033	0.92 (0.86– 0.98)	.007
Race (white)	0.184	0.263	1.20 (0.72–2.01)	.48
Sex (male)	0.578	0.235	1.78 (1.13–2.82)	.01	0.581	0.245	1.79 (1.11–2.89)	.02
Cumulative cisplatin dose	-4.272	2.710	0.014 (<0.001–2.83)	.12
Study (SJMB03)	−0.917	0.232	0.40 (0.25–0.63)	<.001
Risk (high)	0.744	0.230	2.10 (1.34–3.30)	.001	−0.261	0.609	*	.67
Amifostine (yes)	−0.885	0.305	0.41 (0.23–0.75)	.004	−1.206	0.383	*	.002
Risk by amifostine interaction*					1.087	0.661	—	.10
Amifostine vs no amifostine for average-risk disease							0.30 (0.14–0.64)
Amifostine vs no amifostine for high-risk disease							0.89 (0.31–2.54)

*Since the multivariable model contains a risk by amifostine interaction term, the odds ratio for amifostine versus no amofostine needs to be calculated by risk level , and odds ratios associated with the main effects (ie, risk alone or amofostine alone) are not meaningful.