Table 2.

Time to treatment failure at third-line cabazitaxel or 2^nd ARAT

	Patients, n		Event, n (%)		Median (95% CI), days
	Cabazitaxel at 3rd line	2nd ARATa at 3rd line	Cabazitaxel at 3rd line	2nd ARATa at 3rd line	Cabazitaxel at 3rd line	2nd ARATa at 3rd line	P-valueb	HR (95% CI)c
First ARATa at 1st or 2nd line
Unmatched	247	288	247 (100.0)	288 (100.0)	94 (85–99)	98 (89–100)	NA	NA
3rd-line treatment
Unmatched	247	288	221 (89.5)	288 (100)	109 (94–128)	58 (57–66)	< 0.001	0.339 (0.279–0.413)
PS-matchedd	213	213	191 (89.7)	213 (100)	108 (94–122)	57 (53–64)	< 0.001	0.323 (0.258–0.402)

ARAT androgen receptor axis-targeted agent; CI confidence interval; ECOG PS Eastern Cooperative Oncology Group performance status; HR hazard ratio; NA not available; PS propensity score; PSA prostate-specific antigen; TNM tumour, nodes, metastases

^aAbiraterone or enzalutamide

^bLog-rank test (two-sided)

^cCox proportional hazards model

^dCovariates were: age, body surface area, Gleason score, curative intent focal therapy, number of docetaxel treatment cycles, and reason for discontinuation of docetaxel