The Sparse Matrix of Drug Discovery: Sex, Race, and a Genomic Equity Index Across 40.8 Million Patients in 77,770 Clinical Trials

Demographic representation in clinical trials and downstream genomic databases has been documented to be skewed across two decades, but the structural origin of the skew remains contested. Here we test whether a single upstream protocol-writing bias drives demographic skew on both axes (race and sex) at the population scale of clinical-trial enrollment, and whether the same skew compounds at the genomic-database layer that informs precision-medicine inference.

The hypothesis is anchored to a previously characterized clinical-care bias. The senior author has previously shown that the named sex disparity in coronary artery bypass grafting (CABG) does not originate at the operating-room door but in diagnostic timing, with women experiencing significantly longer first-encounter to diagnostic-catheterization intervals than men, while no sex difference appears in cath to CABG timing ⁹ . The bedside record shows what looks like sex-biased surgery; the upstream record reveals a bias set in who is investigated for cardiac disease at all. CABG is the bedside template; clinicaltrial enrollment is the population-scale analogue.

In the trial record, the CABG signature reproduces, on both demographic axes. On the race axis, African-derived enrollment declines Phase I → Phase III monotonically across all sponsor classes; the industry × academia joint trial shows the steepest gradient of all (Phase III 8.2%; Δ Phase I → III = +4.65 pp; OR = 1.62; P < 10 ⁻¹⁰⁰ ), with academic execution under industry-written protocols laundering catchment-area diversity out. On the sex axis, female enrollment runs the opposite phase direction (Phase I 41.4% → Phase III 48.7% in the eligibility-ALL subset, n = 37,737), with the inversion concentrated in industry-solo trials (+10.0 pp); 99.3% of trials report sex against 61.6% reporting race. The two axes are statistically independent at the trial level (joint CABG-bias quadrant occupancy 24.81% vs 24.76% expected under independence; χ ² = 0.13, P = 0.72; Pearson r = 0.016), one upstream cause with two parallel readouts, not one reinforcing race × sex effect.

Extended to the genomic-database layer, the pattern compounds. The GEI ranks European-derived populations at 0.98; eight others score 0.58 or below; the global equity gap is 3.94 billion people receiving precision medicine calibrated for the 11% of humanity who are European-derived. The race-axis primary-data recomputation of GWAS Catalog ancestry share yields 88.3% European-derived participation, not the published 78% (Mills & Rahal 2020), the gap has widened, not narrowed. Top and bottom GEI rankings are robust in 100% of 10,000 Dirichlet weight permutations. The same architecture is portable to a parallel Sex Equity Index for the orthogonal axis.

The hypothesis is supported at every layer we measured. The structural locus across all three, bedside, trial, genomic database, is one upstream governance function. Boards and credentialing bodies that govern protocol design, independent of the sponsor, are the structural intervention point at every layer.