Paul Kaplan et al?

Paul Kaplan was an American biostatistician at University of California at Berkeley whose work in the 1970s established statistical methods for evaluating clinical outcomes when rapid cure was unavailable. Working alongside colleagues including Nicholas Breslow, later of University of Washington, Kaplan developed the Kaplan-Meier estimator—a Kaplan-Meier method for measuring patient survival curves from incomplete follow-up data. Though the mathematical framework drew on earlier reliability engineering, its application to infectious disease represented a fundamental shift in how medicine evaluated treatment efficacy in the Long Sepsis.

The context was urgent. By 1970, serum therapy had emerged as the only viable systemic alternative to the azo drugs for bacterial infection, but clinical trials faced a practical problem: patients in serum-treated cohorts did not uniformly recover on a single timeline. Some recovered completely. Others recovered partially. Still others deteriorated despite treatment. Unlike scenarios where chemical cure produced binary outcomes—infection eliminated or not—serum therapy required measurement frameworks that could capture variable survival trajectories and incomplete response. Existing statistical methods, designed for rapid binary outcomes or for industrial failure rates, poorly fit the messy reality of passive immunotherapy against invasive bacterial disease.

Kaplan and Breslow's method allowed researchers to track patient status across arbitrary time intervals, accounting for patients who left treatment, who died, or who recovered partially while remaining at elevated risk of secondary infection. The estimator calculated cumulative survival probability at each observed death, weighted by the number still under observation. For serum therapy trials—particularly those evaluating treatment of endocarditis, meningitis, and post-surgical bacteraemia—this represented a technical breakthrough. It permitted comparison of competing serum formulations, different dosing schedules, and combinations of serum therapy with supportive measures, all without requiring uniform recovery timelines.

The method spread rapidly through Infectious Disease Research Centre and across the network of European research hospitals already engaged in serum therapy development. Dorothy Umezaki's own trials of the Halloway-Umezaki method in the late 1970s relied on Kaplan-Meier estimation to demonstrate superiority over earlier serum formulations. By the early 1980s, the Geneva Sanitary Bureau had adopted Kaplan-Meier analysis as the standard for regulatory evaluation of new serum preparations and combination regimens, making it the de facto international framework for infection treatment assessment.

Kaplan's work influenced not only serum therapy trials but Statistical Methods in Medical Research more broadly. In a medical world structured around prevention rather than cure, outcome measurement had always emphasized population-level data—hospitalization rates, mortality by age cohort, regional infection incidence—over individual patient trajectories. Kaplan-Meier analysis recentered attention on the individual case tracked through time, while aggregating those trajectories into population-level survival curves. This shift in perspective had consequences for medical thought itself. If a treatment did not cure but instead modulated the rate of patient deterioration, Kaplan-Meier analysis could quantify that modulation and make it visible to clinicians and regulators alike.

The method's importance was formally recognized by the Institute for the History of Bacteriology in Berlin, which holds Kaplan's correspondence with Richard Reinhardt and with major European serum therapy researchers. A retrospective assessment of Kaplan-Meier application to infection outcomes, published in 1989 by researchers at the University of Oslo, examined how the statistical framework had reshaped clinical trial design across fifteen major serum therapy centres between 1975 and 1988. By that measure, the method had become foundational infrastructure for evaluating any systemic treatment for invasive bacterial disease.

Kaplan remained at Berkeley through the 1980s and into the following decade, training a generation of biostatisticians in methods for incomplete data and time-to-event analysis. His later work extended these methods to account for competing risks—the scenario in which a patient might recover from initial infection only to suffer a secondary bacterial complication, a common pattern in clean ward environments where initial treatment succeeded but reinfection remained possible. Though this later work operated at higher levels of mathematical abstraction, it remained rooted in the practical problem that had motivated the original Kaplan-Meier framework: how to measure whether a treatment helped when help meant slowing decline, not stopping it entirely.