BMY has said they're filing on these data, so the issue is germane for at least a little while. Digging in on the subject, here's what I found in addition to the paper cited previously...
Human Immunology 61, 334–340 (2000)
Frequencies of HLA-A2 alleles in five U.S. population groups: Predominance of A*02011 and identification of HLA-A*0231 Jennifer M. Ellisa, Valerie Hensonb, c, Rebecca Slackd, Jennifer Ngb, Robert J. Hartzmanb and Carolyn Katovich Hurley
a Department of Microbiology and Immunology (J.M.E., C.K.H.), Georgetown University Medical Center, Washington, DC, USA d Department of Biostatistics (R.S.), Georgetown University Medical Center, Washington, DC, USA b Naval Medical Research Center (V.H., J.N., R.J.H.), Kensington, MD, USA c The Methodist Hospital (V.H.), Houston, TX, USA
TABLE 1 Phenotypic frequency of HLA-A2 within the study population Population Population size #HLA-A2 positive Caucasian 61,655 30,596 (49.6%) African-American 8,288 2,864 (34.6%) Asian/Pacific Islander 2,275 819 (36.0%) Hispanic 4,879 2,286 (46.9%) Native American 5,882 2,922 (49.7%) Total 82,979 39,487 (47.6%)
TABLE 2 HLA-A2 allele frequencies and tests of overall significance of HLA-A2 frequency distribution tests in the five populations
This table doesn't reproduce here, but here's the summary data for our purposes...
iwfal is right, there is something amiss here. I find it odd I've seen the trial described multiple ways in presentations over the years.
In any case, in the primary patient population (http://www.cdc.gov/cancer/skin/statistics/race.htm) the difference is not that meaningful (50% vs 48% (.40=.50*.96)). For revenue calculations, it would be moot. For the intitial label, it will be interesting.
Market Data 
Markets 
AI
