ARMOR3-SV

AR-V7 is a splice variant of the androgen receptor that lacks the ligand-binding domain and is constitutively active, so abiraterone and enzalutamide, which act on that domain or on the hormones feeding it, should not work against it. Galeterone degrades androgen receptor protein, including AR-V7, so it should. ARMOR3-SV was the trial designed to prove it.

Of 953 men prescreened, 323 (34 percent) had detectable circulating tumour cells and 73 of the 953 had AR-V7 messenger RNA, a prevalence of 8 percent (95 percent confidence interval 6 to 10).

AR-V7 positivity marked out advanced, aggressive disease: higher PSA (p<0.001), more bone metastases (p<0.001), more previous docetaxel for hormone-sensitive disease (p<0.001), more previous first-generation androgen deprivation (p<0.001) and a shorter time from diagnosis to enrolment (p<0.001).

Of the 73 eligible men, 38 were randomised, 19 to galeterone and 19 to enzalutamide; 35 dropped out before randomisation. Because so many radiographic progression-free survival events were censored, the data monitoring committee recommended closure on interim evidence that the primary endpoint would not be met. PSA50 responses were 2 of 16 (13 percent) with galeterone and 8 of 19 (42 percent) with enzalutamide, difference -0.278 (95 percent confidence interval -0.490 to 0.097).

Development of galeterone stopped. The trial is a lesson about biomarker-selected trials in a disease that moves fast: if the biomarker marks men who are about to deteriorate, the trial will run out of patients before it runs out of endpoints.