... ORR was 22% in patients with "significant stromal or tumor cell expression of GPNMB." So, not sure where you get that ORR was unknown in these patients. ...
. In the entire trial there were 2 (N=33?) ORs. . 14 were sampled for G+, 9 were positive . Of those 9, 2 had CRRs (22%)
But what about those not tested? One would expect something like 10 G+ in those, and they had 0 ORs.
And I would assume that both ORs were intentionally included GPNMB sampling, so that the sampled subset is biased from the start.