Dynamics of cytokines, immune cell counts and disease severity in patients with community-acquired pneumonia – Unravelling potential causal relationships

• Decrease in IL-6 levels and increase in thrombocytes is slower in elderly patients. • IL-8, VCAM-1, and IL-6 levels associate with disease severity. • IL-6 and bilirubin levels and rates of change correlate. • Lagged relationships between pairs of cytokines can be inferred using time series. Community acquired pneumonia (CAP) is a severe and often rapidly deteriorating disease. To better understand its dynamics and potential causal relationships, we analyzed time series data of cytokines, blood and clinical parameters in hospitalized CAP patients. Time series data of 10 circulating cytokines, blood counts and clinical parameters were related to baseline characteristics of 403 CAP patients using univariate mixed models. Bivariate mixed models were applied to analyze correlations between the time series. To identify potential causal relationships, we inferred cross-lagged relationships between pairs of parameters using latent curve models with structured residuals. IL-6 levels decreased faster over time in younger patients (P adj = 0.06). IL-8, VCAM-1, and IL-6 correlated strongly with disease severity as assessed by the sequential organ failure assessment (SOFA) score (r = 0.49, 0.48, 0.46, respectively; all P adj < 0.001). IL-6 and bilirubin correlated with respect to their mean levels and slopes over time (r = 0.36 and r = 0.46, respectively; P adj < 0.001). A number of potential causal relationships were identified, e.g., a negative effect of ICAM-1 on MCP-1, or a positive effect of the level of creatinine on the subsequent VCAM-1 concentration (P < 0.001). These results suggest that IL-6 trajectories of CAP patients are associated with age and run parallel to bilirubin levels. The time series analysis also unraveled directed, potentially causal relationships between cytokines, blood parameters and clinical outcomes. This will facilitate the development of mechanistic models of CAP, and with it, improvements in treatment or surveillance strategies for this disease. Trial registration: clinicaltrials.gov NCT02782013, May 25, 2016, retrospectively registered.