Endpoint-specific mortality associated with malignancy among adults with sepsis or septic shock: a systematic review and meta-analysis.
Huei Chung, Hung-Chang Chou, Tsung-Yu Tsai et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and meta-analysis examined mortality among adults with sepsis or septic shock, comparing patients with and without malignancy. Of 105 eligible studies, 85 contributed to the primary synthesis, covering 20,089,138 participants. Malignancy was associated with higher mortality across most endpoints, with pooled odds ratios ranging from 1.21 for ICU mortality (95% CI 0.94-1.56) to 2.41 for extended follow-up mortality (95% CI 1.70-3.42). The ICU estimate’s confidence interval included no association. The authors note that estimates were unadjusted and represent overall prognostic associations, not independent causal effects.
Key findings
- Pooled odds ratios were 1.90 (95% CI 1.69-2.13) for in-hospital mortality and 1.81 (95% CI 1.59-2.06) for early mortality.
- Intermediate mortality was associated with malignancy (OR 2.28, 95% CI 1.72-3.03); extended follow-up mortality had an OR of 2.41 (95% CI 1.70-3.42).
- Associations were sustained for haematologic malignancy; for metastatic disease they were concentrated in in-hospital and early mortality, and were more modest for solid tumours.
- Sensitivity analyses were directionally consistent, while risk ratio estimates were generally smaller than odds ratios.
Why it matters
Mortality associations differed by outcome timeframe and cancer context, so a single pooled estimate may obscure clinically relevant prognostic variation. Distinguishing ICU, hospital, early, and later mortality can help interpret the evidence more precisely. The findings also underscore that malignancy-associated risk estimates should be understood as overall associations, not proof that malignancy independently causes excess mortality.
What to keep in mind
Primary estimates were unadjusted, so they do not establish independent causal effects. Heterogeneity was present, and the authors note that severity-related covariates may contribute to it.