Real-world longitudinal validation of high-income-country-derived sepsis endotypes in sub-Saharan Africa: a nested prospective multicenter study

Sub-Saharan Africa accounts for an estimated 40% of global sepsis-related deaths, with 30-day mortality exceeding 50% among hospitalized adults. While factors such as causative pathogens, comorbidities and delayed presentation contribute to mortality risk, the role of the host response remains poorly understood. In high income countries (HICs), transcriptomic endotypes have been linked to sepsis prognosis, highlighting distinct signatures. Assessing the applicability of these endotypes in low- and middle-income countries, particularly in sub-Saharan Africa where the sepsis burden is highest, is needed. As part of the African Research Collaboration on Sepsis (ARCS) Cohort and the Cardiovascular Responses In Septic Patients (CRISP) studies, a subset of adult participants hospitalized with clinically suspected infection in Malawi and Uganda provided whole blood samples, collected longitudinally on days 0, 2/3 and 28, for RNA-sequencing. Sepsis was defined by a modified Sequential Organ Failure Assessment (mSOFA) score ≥ 2. Three sepsis messenger RNA endotypes [Sepsis Response Signature (SRS), Molecular Diagnosis and Risk Stratification of Sepsis (MARS) and Inflammopathic-Coagulopathic-Adaptive], previously derived in HIC settings, were validated for prognostic relevance among study participants. RNAseq was performed on a randomly selected subset of 278 samples belonging to 158 study participants, for whom HIV seropositivity was 51.3% and sepsis prevalence was 56.3% (mSOFA ≥ 2). The most commonly observed signatures over all timepoints were SRS2 (63.3%), Mars3 (31.2%) and Coagulopathic (41.7%), while SRS3 (16.5%), Mars4 (12.7%) and Adaptive (27.7%) were the least common. Participants classified as SRS1, Mars2 or Inflammopathic endotypes were associated with worse outcomes; when compared to the others pooled together, hazard ratios for 30-day mortality were significantly higher for the SRS1 (HR 2.81; 95% CI: 1.17–6.77), Mars2 (HR 3.11; 95% CI: 1.24–7.79) and Inflammopathic (HR 2.66; 95% CI: 1.09–6.52) signatures. Follow-up samples showed that the proportion of signatures associated with more adverse outcomes decreased, while the proportion of less severe ones increased, despite possible survivorship bias. Signatures associated with similar disease severity were strongly correlated between mRNA endotypes (p < 0.001; Cramer’s V > 0.4). Signature classification was not associated with HIV status. Sepsis endotypes identified in HICs exhibit similar biological and clinical characteristics in sub-Saharan Africa. Longitudinal analysis revealed that the proportion of adverse endotype signatures decrease over time.