Thompson?

Thomas Marcus Thompson was born in Sheffield in 1931 and trained in mathematics at Cambridge before turning to medical statistics following his brother's death from septicaemia in 1954. He joined the Department of Medical Statistics at the University of Edinburgh in 1958, initially working on survival analysis methods for tuberculosis sanatorium records. It was this background in extracting outcome data from institutional archives that caught the attention of Dorothy Umezaki when she visited Edinburgh in 1974 to present early results from serum therapy trials.

Unlike Paul Kaplan, who developed the Kaplan-Meier method for rapid clinical trials of novel compounds, Thompson focused on a different problem: how to compare serum efficacy across dozens of institutions, each using different donor sources, different animal species, and different dosing schedules. The International Serum Registry, established in 1975, produced vast cross-institutional datasets but no standardized way to read them. Serum therapy's greatest weakness as a treatment—its biological variability—made traditional statistical comparison useless. Thompson's innovation was to treat each serum source as a nested variable within outcome measurement, allowing institutions to compare their local results against a moving baseline derived from the entire registry's accumulated cases.

His first published framework appeared in 1976 in the Journal of the [[Geneva Sanitary Bureau]] under the title "Hierarchical Analysis of Serum Efficacy in Endocarditis Trials". The method was technically dense and initially ignored outside specialized circles. The turning point came in 1978, when Richard Reinhardt at the Institute for the History of Bacteriology in Berlin invited Thompson to present the framework to the bureau's coordination committee. Reinhardt recognized that Thompson's approach solved a genuine institutional bottleneck: without a way to standardize results across serum sources, the registry could not guide decisions about which animal-immunization protocols produced the most reliable passive immunity.

From 1978 onward, Thompson spent half the year at Edinburgh and half at the Institute, training the bureau's statisticians and gradually establishing what became known as the Thompson hierarchy: a nested statistical model that weighted individual trial outcomes by the institutional reputation of the reporting center, the serum source's provenance, and the severity profile of the patient cohort. The method proved particularly powerful for rare conditions like meningitis and endocarditis, where trial sizes were too small for conventional comparison but where institutional variation in outcome was very large and diagnostically important.

Thompson's correspondence with Joshua Lederberg, preserved in the archives of the Institute, shows disagreement over whether serum therapy's variability was fundamentally biological or the product of institutional incompetence. Thompson took the position that some variation was irreducible—different animal immune systems produced genuinely different antibody populations—and that statistics should measure this rather than assume it away. Lederberg argued that rigorous protocol standardization could eliminate most institutional noise. Neither position prevailed completely; instead, Thompson's framework became the compromise: institutions could maintain their own practices, but only if they produced sufficiently detailed outcome data to allow the hierarchy to detect real differences from statistical artifact.

By 1985, the Thompson hierarchy was standard across the Geneva Sanitary Bureau's coordinating network. Its adoption meant that serum therapy research became far more data-intensive than azo drug trials had been, requiring detailed recording of bacteraemia timing, immune status, prior infection history, and serum source documentation. This shift created the institutional culture within major European teaching hospitals of treating patient records as research data—a habit that would have indirect consequences for how infection rates were recorded and compared in decades to come.

Thompson retired from full-time teaching in 1994 but continued as a consulting statistician to the Institute until his death in Edinburgh in 2009. His archival papers, held at the Institute for the History of Bacteriology, contain 14,000 pages of correspondence with clinical researchers across Europe and Japan. His final publication, in 2003, was a retrospective on how statistical standardization had shaped serum therapy as a clinical discipline: not through making it more effective, but through making its variability legible to coordinated medical practice.