Table 3.

Odds ratios for functional principal component—logistic regression models

Covariate	Mucositis model		Dysphagia model
	Odds ratio	95% CI	Odds ratio	95% CI
Intercept	12.89	1.035-1.734×109*	1.616	0.142-77.46
Male	1.535	0.637-4.088	1.675	0.533-4.880
Age	0.991	0.951-1.029	0.988	0.943-1.027
Definitive RT	0.254	2.679×10−9-1.773	0.997	0.080-7.541
Induction chemotherapy	0.487	0.070-1.960	1.100	0.210-7.670
Cisplatin	2.251	0.745-9.540	4.255	1.077-19.86*
Carboplatin	1.320	0.142-7.314×107	4.429	0.685-8.332×107
Cis-carbo	0.311	7.815×10−9-2.531×107	2.238	0.319-4.587×107
Hypopharynx or larynx	4.371	0.512-143.9	1.723	0.193-1.881×107
Nasopharynx or nasal cavity	2.370	0.308-1.096×108	0.263	0.026-1.223
Unknown primary	0.136	3.042×10−9-3.707	0.859	0.077-3.876×106
Parotid	1.387	0.103-40.37	1.135	0.068-18.72
DVH FPC1	0.997	0.993-1.007	0.996	0.990-1.008
DVH FPC2	1.003	0.992-1.009	-	0.992-1.003
DVH FPC3	-	0.996-1.003	-	0.995-1.000
DVH FPC4	-	0.987-1.010	-	0.991-1.006
DVH FPC5	-	0.971-1.033	-	0.991-1.006

Abbreviations: Cis-carbo = 1 cycle of cisplatin followed by 1 cycle of carboplatin; DVH FPCx = functional principal component x of dose-volume histogram data; RT = radiation therapy; 95% CI = 95% confidence interval calculated by bootstrapping model fitting with 2000 replicates.

The sign of the functional principal component loadings is arbitrary, so the fact that the odds ratios are <1 does not indicate that there is an inverse correlation between RT dose and severe toxicity.

^*Statistically significant at α = .05 level.