Sepsis: Partial least squares structural equation modelling (PLS-SEM) suggests a critical role for anti-inflammatory responses in clinical severity

Backgorund: Despite major advances in medicine, Sepsis remains one of the major killers in critical care wards around the world. For several years it was widely believed that an early pro-inflammatory host response is followed by an overwhelming anti-inflammatory phase. The hypo-inflammatory status, termed as Compensatory anti-inflammatory response syndrome (CARS), was proposed to be the primary cause of sepsis-associated mortality. However, this paradiam changed in recent years since there was little evidence to support the linear model of host response and pathogenesis in sepsis. Currently held view is that both inflammatory and anti-inflammatory host responses are stimulated in an overlapping manner. In this study a robust statistical model to study the complex interplay of host cytokines in human sepsis has been developed to evaluate host responses in sepsis that contribute significantly to clinical pathology. Methods: Twentyseven cytokines/chemokines were quantified in 139 sepsis patients and multivariate analysis of variance (MANOVA) was performed to assess differences in host responses in different categories of clinical severity. Partial least squares regression based structural equation modelling (PLS-SEM) was used to assess interactions between different groups of cytokines and their contribution to clinical pathology. An array of 23 cytokines was analysed in a mouse model of endotoxemia and a similar mathematical model was constructed. Results: The results of MANOVA demonstrated the ability of combined cytokine response to discriminate sepsis patients according to clinical severity or outcome. Structural equation modelling revealed strong positive association between inflammatory and anti-inflammatory cytokines. In human sepsis, anti-inflammatory cytokines emerged as a significant entity associated with clinical severity as assessed by APACHE II scores. Conclusion: PLS-SEM modeling of cytokine responses and APACHE II score in human sepsis revealed that anti-inflammatory molecules contribute significantly towards clinical severity. More critically, the model offers emperical evidence for failures of clinical trials conducted during the last two decades in which antagonists of inflammatory host responses for human Sepsis were used for sepsis. The model also provides credence to the notion that inflammatory and anti-inflammatory host responses occur concurrently in both experimental endotoxemia and in human sepsis.