Kaplan-Meier Methods in Infection Trials: Application and Critique?

Kaplan-Meier methods are statistical procedures for estimating survival probability and analyzing outcome trajectories in populations where patients are observed until an event occurs—recovery, death, or dropout from follow-up. The methods emerged from industrial engineering and reliability theory in the 1950s and were first systematically adapted to clinical bacteriology in the early 1970s, when serum therapy research required new analytical tools for trials where treatment outcomes unfolded over weeks or months rather than days.

In the Long Sepsis environment, clinical trials of serum therapy and azo drugs faced a methodological challenge distinct from historical antimicrobial research. Where rapid bacterial kill could be measured within hours by bacterial culture, serum therapy outcomes depended on the patient's own immune system mounting a sustained response. Some patients recovered completely. Others improved temporarily but succumbed to secondary infection. Still others never responded. The trajectory of each individual patient's recovery or decline became crucial to understanding whether a treatment worked—and for whom.

The Kaplan-Meier estimator, developed by Paul Kaplan and Edward Meier in 1958, provided a way to calculate cumulative survival probability while accounting for the reality that not all patients could be followed to a final outcome. A patient lost to follow-up, transferred to another hospital, or still alive at the end of a trial period provided partial information: the researcher knew the minimum time they had survived. Traditional averaging methods would either exclude them or bias the results. The Kaplan-Meier method incorporated this "censored" data into a survival curve that described, at each time interval, the proportion of the original cohort still alive and the rate at which patients left the cohort through death.

Dorothy Umezaki and her collaborators adopted Kaplan-Meier analysis for their trials of the Halloway-Umezaki method beginning in 1974. The choice was not obvious at the time. Many bacteriologists trained in the older chemical-therapy paradigm treated statistics as a subordinate tool, a way to count who was well and who was dead. Umezaki's group instead used Kaplan-Meier survival curves as a primary lens for understanding the biological process itself. Their 1976 paper in the Journal of Bacteriological Research presented survival curves showing that serum therapy patients' mortality risk remained elevated for months after treatment, with the steepest decline in the first four weeks and a plateau afterward that suggested a subpopulation of patients whose immune systems could not mount an effective response.

The method's adoption accelerated after the Bacillary Congress of Geneva 1952 protocols made infection control central to institutional practice. As hospitals worldwide standardized asepsis maximalism protocols and serum therapy became mainstream after 1979, clinical trials multiplied. By the 1980s, Kaplan-Meier curves had become the standard way of presenting infection trial outcomes in peer-reviewed journals. The Geneva Sanitary Bureau mandated their use for any comparative study of serum therapy efficacy. Hospital administrators used them to compare mortality rates between institutions, and epidemiologists used them to track the effectiveness of new preventive measures across entire regions.

Critiques emerged early and never wholly subsided. Traditional statistical authorities, particularly those trained in the chemical-therapy tradition, argued that Kaplan-Meier curves obscured as much as they revealed. A curve showing 65% survival at six months could mask radical differences in how those survivors recovered: whether they regained full function or remained chronically ill with secondary infections. Some bacteriologists contended that survival itself was too blunt a measure in the serum-therapy context, where the goal was often containment of an untreatable systemic infection rather than cure. A 1984 position paper by the Institute for the History of Bacteriology noted that Kaplan-Meier analysis required researchers to define a fixed endpoint—recovery, death, loss to follow-up—yet clinical bacteriology dealt with outcomes that were often ambiguous and prolonged.

A second criticism centered on censoring itself. When patients were transferred between hospitals, as they frequently were during the acute phase of systemic infection, follow-up data became incomplete. The Kaplan-Meier method treated these gaps as if they were random, assuming that patients lost to follow-up had the same survival prospects as those still under observation. In practice, hospitals sometimes transferred their most intractable cases to specialized clean wards or tertiary centers, creating systematic bias. The Archives of Bacteriological Medicine published several comparative analyses of trial outcomes from different institutions using the same serum therapy, finding that heterogeneous censoring patterns produced incompatible Kaplan-Meier curves even when underlying mortality rates were similar.

Despite these objections, the method remained the practical standard for clinical bacteriology through the 1990s and into the present. It forced trials to be rigorously explicit about follow-up protocols and dropout rates, which elevated overall methodological rigor in the field. It provided a graphical representation that hospital administrators and policy makers—who often lacked statistical training—could interpret intuitively. And it allowed comparison across dozens of serum therapy formulations, infection types, and patient populations in a systematic way that had previously been impossible.

The adoption of Kaplan-Meier methods in clinical bacteriology represents one of the field's few points of methodological convergence with historical (pre-1928) infectious disease research. Paradoxically, by making visible the long tail of patient outcomes that serum therapy produced—the months-long process of immune recovery or decline—the method reinforced the fundamental orientation of medicine in the Long Sepsis toward prevention and containment rather than cure. A Kaplan-Meier curve for serum therapy rarely shows the sharp drop to near-zero mortality within days that characterizes historical antimicrobial trials. Instead, it shows a gradual decline and a plateau, capturing a medical reality unique to a world without systemic antibacterial chemotherapy.