Where clinical trials go wrong.

Almost every US trial is public. A 2007 law, FDAAA 801, makes registration mandatory, so the record is close to complete. A substantial protocol amendment costs about $450,000 in direct work (Tufts CSDD), before the weeks of delay each one adds. Multiply how often a company amends by how many trials it runs, and the pain sorts into an order. The companies at the top are losing the most. Here is the commercial case and its focus.

pain = mean( rank(amendments), rank(enrollment failure), rank(startup lag) )Each signal is ranked 0 to 100 against every other cell, then the three are averaged with equal weight. No learned weights, no black box. A cell with fewer than two measurable signals reads insufficient, not zero.

Less painMore paininsufficient data


Oncology	80	92	92	78	22,100
Infectious Disease	20	51	37	45	9,765
Healthy Volunteers	1	8	19	12	9,574
Endocrine & Metabolic	8	45	40	44	6,264
Cardiovascular	22	81	72	69	5,646
Neurology	29	80	86	58	5,616
Immunology & Inflammation	33	78	68	54	5,484
Psychiatry	26	64	51	54	3,418
Respiratory	11	61	50	59	3,366
General & cross-cutting	12	75	67	54	2,899
Dermatology	14	31	27	17	2,749
Rare & Genetic	35	95	88	94	2,471
Nephrology & Urology	8	78	65	34	2,222
Hematology	28	96	87	92	2,155
Musculoskeletal	36	57	68	61	1,854
Gastroenterology	10	67	45	50	1,852
Ophthalmology	36	70	51	42	1,624
Pain & Anesthesiology	4	41	39	53	1,190
Women's & Reproductive Health	7	42	62	84	1,007

See who carries the pain

In brief

What this is: An interactive map of 454,031 public trials, every figure carrying the sample it rests on (methodology).
The deliverables: The map with per-cell evidence, the ranked account list, and a one-page board brief (PDF).
How it is built: A deterministic pipeline over the public record. The amendment figure is a disclosed proxy, and contract research organizations are a named blind spot (limits).