Using metabolomics to predict severe traumatic brain injury outcome (GOSE) at 3 and 12 months

Table 3 Prediction models’ characteristics show a higher predictability of metabolic profiles on day 4 than day 1 post-sTBI for 3- and 12-month GOSE and mortality at the 3-month outcome

Prognosis	Analytical platforms	Sampling time	R²	Q²	p value*	Sensitivity	Specificity	AUC	# Metabolites
Unfavorable versus favorable outcome 3-month	DI-MS/MS	Day 1	0.60	0.40	0.0004	93	100	0.99	26
	DI-MS/MS	Day 4	0.75	0.54	0.0003	100	100	1.00	24
	¹H-NMR	Day 1	0.47	0.25	0.017	72	100	0.92	10
	¹H-NMR	Day 4	0.75	0.59	0.0001	100	96	1.00	9
Unfavorable versus favorable outcome 12-month	DI-MS/MS	Day 1	0.88	0.58	0.0002	100	100	0.99	21
	DI-MS/MS	Day 4	0.79	0.62	0.0004	100	100	0.98	29
	¹H-NMR	Day 1	0.64	0.46	0.003	76	91	0.91	12
	¹H-NMR	Day 4	0.7	0.41	0.044	100	100	1.00	9
Mortality outcome	DI-MS/MS	Day 1	0.54	0.35	0.002	79	100	0.98	19
	DI-MS/MS	Day 4	0.76	0.50	0.0006	100	100	1.00	16
	¹H-NMR	Day 1	0.50	0.24	0.01	84	87	0.88	17
	¹H-NMR	Day 4	0.61	0.39	0.011	91	90	0.96	16

^*p value < 0.05 was considered significant. Additionally, the metabolic profiles obtained by DI/LC–MS/MS are more predictive than ¹H-NMR results. R², the goodness of fit of the model; Q², the goodness of prediction of the model: and AUC, the area under the receiver operating curve of the model

ISSN: 1364-8535