Sex differences in sepsis treatment responses revealed by preclinical meta-analysis

Sepsis’s Hidden Sex Gap: Massive Analysis of 83 Animal Studies Finds Males Consistently Fare Worse — From Mortality to Organ Injury

Sepsis — the runaway, body-wide immune collapse that follows an infection — remains one of the deadliest conditions in modern medicine, contributing to millions of deaths worldwide every year. Yet in the laboratories where potential therapies are tested before they ever reach an intensive care unit, one of the most fundamental variables in biology has been chronically under-examined: the sex of the animal on the operating table. A sweeping new systematic review and meta-analysis, published on 28 August 2026 in the journal Biology of Sex Differences, argues that this oversight carries a measurable cost. Pooling data from 83 preclinical studies, a Canadian research consortium reports that male laboratory animals tended to fare worse than females across sepsis experiments — dying more often, generating stronger inflammatory responses and sustaining more organ injury — while also showing less favourable responses to experimental treatments. The statistical evidence is far from airtight, and the authors are refreshingly candid about that. But the sheer consistency of the direction of effect, replicated across 99 pooled comparisons spanning several distinct outcome domains, is difficult to dismiss as statistical noise.

The analysis was carried out by the Canadian Critical Care Translational Biology Group together with the Sepsis Canada National Preclinical Sepsis Platform, a national consortium of critical-care researchers, with the Blueprint Translational Research Group at the Ottawa Hospital Research Institute coordinating the work. Forough Jahandideh and MengQi Zhang share first authorship, while Dean A. Fergusson and Manoj M. Lalu of the University of Ottawa led the project as senior authors, with funding from the Canadian Institutes of Health Research. The team searched MEDLINE and Embase for studies published between January 1, 2011 and May 13, 2025, spanning more than fourteen years of preclinical sepsis science. To qualify, a study had to use living animals in experimentally induced models of sepsis, with or without a therapeutic intervention, and — crucially — had to report outcomes separately for males and females. The investigators excluded models of non-infectious inflammation, such as sterile injury paradigms that mimic but do not reproduce infection, as well as interventions hypothesized to worsen sepsis outcomes, so that the pooled analysis would reflect genuine sepsis biology and bona fide attempts to treat it.

The scale of the effort is striking. The researchers screened 7,896 citations; 94 studies ultimately satisfied the eligibility criteria and 83 yielded data robust enough to pool quantitatively. Cecal ligation and puncture — CLP in the trade — was the dominant model, appearing in 46 percent of studies. In a CLP experiment, surgeons open the animal’s abdomen, ligate the cecum and puncture it with a needle, spilling fecal contents into the peritoneal cavity and igniting a polymicrobial infection that closely mirrors ruptured-bowel sepsis in human patients. Bacteremia models, in which live pathogens are injected directly into the bloodstream, accounted for another 37 percent, with fecal-induced peritonitis and other techniques making up the remainder. This methodological spread is a strength and a weakness at once. It covers the major ways scientists simulate sepsis, which makes the findings hard to blame on any single model; but it also means each pooled estimate blends experiments that differ in species, pathogen, inoculum, severity and supportive care, all of which widen the uncertainty around the final numbers.

What distinguishes this study is its analytic architecture. Rather than simply asking whether the sexes differ, the team built a novel three-level framework designed to disentangle disease susceptibility from treatment response. Level one, the baseline sepsis effect, quantifies untreated female-versus-male differences — how differently the sexes weather sepsis itself before any therapy is applied. Level two, the unadjusted treatment effect, captures post-treatment risk differences between the sexes without correcting for those baseline gaps. Level three, the adjusted treatment effect, re-estimates post-treatment differences while accounting for baseline sepsis severity, isolating whether males and females genuinely respond differently to the same intervention or merely start the illness from different points. All pooling used random-effects models, which assume the true effect varies from study to study and weight each experiment accordingly. Outcomes were expressed as risk differences — the absolute percentage-point gap for binary outcomes such as death — or as differences of standardized mean differences, a scale-free measure for continuous readouts like inflammatory protein levels, allowing mortality data and laboratory measurements to be handled with the statistical tool appropriate to each. Throughout the analysis, values greater than zero signalled worse outcomes in males.

Mortality, the primary outcome, told a quietly compelling story. For the baseline sepsis effect, the pooled risk difference was 0.04, with a 95 percent confidence interval running from −0.02 to 0.10 — roughly four additional deaths per 100 untreated males compared with females. For the unadjusted treatment effect the figure rose to 0.05 (95 percent CI, 0.00 to 0.09), and for the adjusted treatment effect it climbed to 0.07 (95 percent CI, 0.00 to 0.14), equivalent to seven extra deaths per 100 treated males once baseline disparities were mathematically removed. Every one of those intervals touches zero, meaning the data cannot statistically exclude the possibility of no true difference; the evidence is imprecise, and the authors say so without hesitation. But the shape of the pattern matters as much as any single number. The estimates did not scatter randomly around zero. They stacked in one direction, and they grew precisely where theory predicts they should — when baseline differences were stripped away to isolate the pure treatment response. That internal coherence lends the findings a weight that a lone significant result never could.

When the researchers widened the lens from statistical significance to direction of effect — asking simply which way each pooled comparison pointed — the picture sharpened considerably. Across the three-level framework, 69 of 99 pooled comparisons suggested worse outcomes in males. The inflammation domain was even more lopsided: 22 of 27 pooled comparisons favoured females, consistent with the long-observed, still imperfectly understood tendency of female immune systems to respond to infection with a different magnitude and character than males’. Organ dysfunction followed the same grain, with 9 of 12 comparisons pointing toward greater injury in males — a result with real physiological heft, because it is organ failure, not the mere presence of bacteria, that kills septic patients. The consistency across outcome domains that share no obvious measurement artefact is exactly why the authors argue that biological sex is a variable preclinical sepsis research can no longer afford to sideline, even when individual estimates remain statistically fragile.

Then comes the twist. For bacterial burden — the quantity of viable pathogen persisting in blood and tissues, typically quantified as colony-forming units — the arrow reversed: 12 of 22 pooled comparisons suggested heavier microbial loads in females. The result could be waved away as statistical static, but it resonates with an increasingly uncomfortable idea in immunology: surviving sepsis is not simply a matter of killing bacteria. A female animal may tolerate or wall off a larger microbial load while sustaining less collateral immunopathology, whereas a male may clear the infection more aggressively at the price of greater inflammatory damage to the lungs, kidneys and liver. The pooled data record outcomes, not mechanisms, so they cannot confirm such interpretations. But the reversal is a pointed reminder that a heavier infection and a deadlier disease are not the same thing — and that sex can push the two in opposite directions.

The caveats are as instructive as the findings. Most of the included studies carried an unclear risk of bias when judged against accepted standards for laboratory animal research — a verdict that says less about individual laboratories than about the state of preclinical publishing, where randomization, blinding and sample-size justification are too seldom reported in enough detail to evaluate. And sex-stratified reporting itself remains rare: of nearly 7,900 citations screened, only 94 studies — barely more than one percent — reported data in a form that allowed males and females to be compared at all. That scarcity may be the study’s deepest finding. An entire literature has been generating candidate sepsis therapies with the sexes blended together, separated only by assumption, or silently reduced to males alone, leaving clinicians to extrapolate to hospital wards where women and men develop sepsis at different rates and arrive with different immune signatures.

The practical implications stretch from the bench toward the bedside. If male animals genuinely respond less favourably to experimental sepsis treatments, then a therapy validated predominantly in males may carry inflated expectations into human trials, while a benefit confined to females could vanish entirely inside a mixed-sex preclinical portfolio that reports only pooled results. Conversely, the wide uncertainty in these estimates is itself a measurement of the field’s blind spot: because sex-stratified data are sparse and noisy, no one can yet say how large the sex gap in treatment response truly is. The authors’ prescription is specific. Future studies, they write, should systematically incorporate sex as a biological variable in both design and analysis, draw on diverse animal populations rather than a single inbred strain, and adopt rigorous experimental design practices, so that the next synthesis of this question has sharper data to pool. The three-level framework they built, separating who gets sick from who gets better, is reusable machinery for any corner of biomedicine wrestling with sex-dependent effects.

There is a larger lesson here about how science handles one of its oldest confounders. Suspicions that immunity runs differently in males and females long predate 2016, when the United States National Institutes of Health began requiring researchers to account for sex as a biological variable in studies of animals and cells; what has been missing is evidence rigorous enough to quantify the penalty for ignoring the variable. This analysis supplies a direction, a magnitude wrapped in wide confidence intervals and, perhaps most valuably, a method. Sepsis retains its reputation as an indiscriminate killer, but its laboratory biology, like most of biology, is not sexless. Eighty-three studies later, the preclinical evidence now leans steadily toward a conclusion long debated at sepsis conferences: in the animal models that shape our future therapies, it is the males that keep coming off worse. The signal is faint and the error bars generous. The field’s task, the authors conclude, is to design the next generation of experiments so the signal can finally be heard above the noise.