The Sequential Organ Failure Assessment (SOFA) score is a cornerstone tool in intensive care medicine. Rather than serving primarily as a mortality predictor, it was designed to assess and monitor organ dysfunction in critically ill patients. It forms part of the diagnostic framework for sepsis, where an acute increase of two or more SOFA points constitutes a key criterion.
Following substantial evolution in intensive care practice over the past three decades, the score was recently updated to SOFA-2 to better reflect contemporary clinical standards while maintaining predictive performance. The principal changes include updated organ-support drugs and devices, revised score cut-offs aligned with mortality risk, and inclusion of alternative variables for situations where primary measures are unavailable. SOFA-2 values are higher in 11% and lower in 40% of patients compared with the original SOFA (SOFA-1). Despite this, SOFA-2 had not been specifically evaluated in patients with sepsis, leaving unresolved whether it should replace SOFA-1 in that diagnostic context.
A recent study performed a secondary analysis of data from the TESTS trial, conducted across 22 centres in China. The trial enrolled 1,106 adult patients aged 18–85 years admitted to ICUs with sepsis diagnosed according to Sepsis-3 criteria, who were randomised to receive thymosin α1 or placebo. 1,089 patients were included in the modified intention-to-treat analysis.
Both SOFA-1 and SOFA-2 were retrospectively calculated at baseline (day of randomisation) using all available variables. Five immune markers were examined as candidate predictors: WBC count, lymphocyte count, monocyte human leucocyte antigen-DR (mHLADR), neutrophil-to-lymphocyte ratio (NLR), and percentage of regulatory T cells (Treg). To assess whether any combination of these markers improved mortality prediction when added to SOFA-2, the study compared logistic regression models using SOFA-2 alone against XGBoost-augmented models incorporating each of the 31 possible immune marker combinations. The primary outcome was ICU mortality; 28-day and 90-day mortality were secondary outcomes.
The cohort had a median age of 64.5 years, with 31.1% female. Approximately half (49.4%) were admitted following a surgical procedure. Hypertension was the most common comorbidity (38.7%), followed by diabetes mellitus (25.4%). The most frequent infection sites were the lungs (31.5%) and abdomen (9.9%). Gram-negative organisms (34.8%) and mixed pathogens (26.9%) predominated microbiologically. The median APACHE II score was 14 (IQR: 10–19). ICU, 28-day, and 90-day mortality rates were 9.2%, 23.8%, and 31.7%, respectively.
SOFA-2 scores were statistically significantly lower than SOFA-1 scores overall. At the domain level, SOFA-2 was lower in the respiratory, cardiovascular, and liver domains, but higher in the kidney domain. Despite these differences, the two scoring systems showed highly concordant distributions when a binary threshold of ≥2 was applied: 96.8% of patients met this threshold on both scores, 1.0% fell below it on both, and only 2.2% had a SOFA-1 ≥2 but a SOFA-2 <2. No patients had a SOFA-1 <2 with a SOFA-2 ≥2.
The discriminatory performance of both score versions for predicting ICU mortality was comparable. Similar equivalence was observed for 28-day and 90-day mortality. Incorporating the originally proposed immune domain (WBC and lymphocyte counts) into SOFA-2 produced no significant improvement in AUC for any outcome. Further exploration using XGBoost to integrate all 31 combinations of the five candidate immune markers yielded no statistically significant change in AUC for ICU mortality, with AUC differences uniformly negative or negligible across all combinations tested.
This study provides the first specific validation of SOFA-2 in patients with sepsis. The use of prospectively collected clinical trial data, with rigorously adjudicated sepsis diagnoses, represents a significant methodological strength over retrospective electronic health record analyses. Domain-level and total score shifts from SOFA-1 to SOFA-2 were consistent with those reported in the original SOFA-2 development study.
The authors caution that comparable AUC values should not be interpreted as evidence that the update to SOFA-2 was unnecessary. Rather, the revision was intended to better reflect contemporary clinical practice, and the preserved discriminatory performance is reassuring in that context. The high concordance between SOFA-1 and SOFA-2 at the ≥2 threshold is clinically important: the sepsis-3 criteria require an acute SOFA increase of ≥2, and these findings support the use of SOFA-2 for this purpose in sepsis.
The failure of immune markers, including advanced indicators such as mHLADR, NLR, and Treg percentage, to meaningfully improve SOFA-2’s predictive performance is interpreted by the authors as reflecting the inherently dynamic and multidimensional nature of immune dysfunction in sepsis. Single baseline measurements are likely insufficient to capture a process that evolves rapidly and non-linearly over time. Future research was advocated to develop comprehensive, multimodal tools for immune assessment in sepsis.
This study demonstrates that transitioning from SOFA-1 to SOFA-2 produces comparable score distributions and preserves discriminatory capacity for predicting ICU mortality in patients with sepsis, supporting the adoption of SOFA-2 in this clinical context. The study also reveals that immune dysregulation in sepsis is too complex to be meaningfully captured through simple baseline markers, highlighting the need for more sophisticated, longitudinal, and multimodal approaches to immune assessment in this population.
Source: Journal of Intensive Medicine
Image Credit: iStock
