HOTAIR and its surrogate DNA methylation signature indicate carboplatin resistance in ovarian cancer

Table 2 Cox regression analysis of overall survival against chemotherapy received, stratified according to HOTAIR positive and negative subgroups

Chemotherapy (set)	HOTAIR positive		HOTAIR negative		Interaction
	Hazard ratio	P value^a	Hazard ratio	P value^a	P value
	(95 % CI)	P value^a	(95 % CI)	P value^a	P value
Cisplatin vs carboplatin (INNSBRUCK)	0.42 (0.23–0.76)	0.003	1.87 (0.95–3.69)	0.068	<0.001
Cisplatin vs carboplatin^b (INNSBRUCK)	0.64 (0.33–1.24)	0.187	1.66 (0.69–3.96)	0.255	0.017
Cisplatin vs carboplatin (GRONINGEN)	0.44 (0.18–1.10)	0.071	1.40 (0.49–4.02)	0.525	0.084
Cisplatin vs carboplatin (TCGA)	0.28 (0.04–2.02)	0.18	2.57 (0.33–20.3)	0.35	0.33
Cisplatin vs carboplatin (EUROPE)	0.76 (0.48–1.2)	0.237	1.74 (0.96–3.14)	0.065	0.037
Cisplatin vs carboplatin^b (EUROPE)	0.83 (0.52–1.33)	0.436	1.86 (1.02–3.38)	0.042	0.063
Cisplatin vs carboplatin (COMBINED^c)		0.003		0.076	<0.001
Cisplatin vs carboplatin^b (COMBINED^c)		0.286		0.06	0.008

^a P values were calculated (in the univariate case) from the Cox-regression likelihood ratio test, while in the multivariate case, the P value derives from the Cox-regression Wald test. We note that the Groningen and TCGA sets had only 18 and 26 cisplatin-treated patients, respectively, not allowing for meaningful multivariate results. Interaction was tested by a log-likelihood ratio test between the model with the interaction term (HOTAIR:TREATMENT) and the null model without it
^b Covariates included stage, grade, age and residual disease whenever these were significant in univariate analysis
^c The combined analysis P values were derived from Fisher’s combined (meta-analysis) probability test using a chi-square distribution with 8 (2 × 4) degrees of freedom in the univariate case (INNSBRUCK, GRONINGEN, EUROPE, TCGA) and 4 (2 × 2) degrees of freedom in the multivariate case (INNSBRUCK, EUROPE)

ISSN: 1756-994X