Cytokine patterns in critically ill patients undergoing percutaneous tracheostomy

Abstract

The inflammatory response to acute injury among humans has proved difficult to study due to the significant heterogeneity encountered in actual patients. We set out to characterize the immune response to a model injury with reduced heterogeneity, a tracheostomy, among stable critical care patients, using a broad cytokine panel and clinical data. Twenty-three critical care patients undergoing percutaneous bedside tracheostomies were recruited in a medical intensive care unit. Blood samples were collected at five intervals during 24-h peri-procedure. Patients were followed-up for 28 days for clinical outcomes. There were no statistically significant changes in any of the cytokines between the five time-points when studied as a whole cohort. Longitudinal analysis of the cytokine patterns at the individual patient level with a clustering algorithm showed that, notwithstanding the significant heterogeneity observed, the patients' cytokine responses can be classified into three broad patterns that show increasing, decreasing or no major changes from the baseline. This analytical approach also showed statistically significant associations between cytokines, with those most likely to be associated being interleukin (IL)-6, granulocyte colony-stimulating factor (GCSF) and ferritin, as well as a strong tri-way correlation between GCSF, monocyte chemoattractant protein 1 (MCP1) and macrophage inflammatory protein-1β (MIP1β). In conclusion, in this standard human model of soft tissue injury, by applying longitudinal analysis at the individual level, we have been able to identify the cytokine patterns underlying the seemingly random, heterogeneous patient responses. We have also identified consistent cytokine interactions suggesting that IL-6, GCSF, MCP1 and MIP1β are the cytokines most probably driving the immune response to this injury.

Keywords: critical care; cytokines; human; inflammation; percutaneous tracheostomy.

Figure 1

Interleukin (IL)‐6 clusters with two normalization methods. Cytokine data was plotted so that the y‐axes represent the normalized expression level and the x‐axes represent the five sampling times: 1, 2 h before tracheostomy; 2, 2 h after tracheostomy; 3, 6 h after tracheostomy; 4, 10 h after tracheostomy; and 5, 22 h after tracheostomy. Every line represents a single patient. The clusters were obtained using data normalized by the first sampling time on the left side of the figure and by the median on the right side of the figure. The red, green and blue plots represent the three different clusters found, numbered 1–3, in the upper left, upper right and lower left panels, respectively. The lower right panel shows the cluster means, which represent the averaged response of the clusters. When using normalization by the first value, by definition, cytokine levels begin at the same point. In contrast, when normalizing by the median, cytokine levels cross the zero at the time‐point that has the median value for that specific patient sample. Normalization per patient allows comparison of the fold‐change in expression over time between the different patients, but does not show absolute levels. When comparing patients it is important to note that the y‐axis represents fold‐changes, and higher or lower points on the y‐axis do not represent higher or lower absolute expression.

Figure 2

Cytokine clusters. This figure was prepared as for Fig. 1. Four representative cytokines are shown – ferritin, interleukin (IL)‐8, macrophage inflammatory protein‐1β (MIP1β) and soluble IL‐2R (sIL2R), with IL‐6 shown again for comparison. Cytokine data were plotted so that the y‐axes represent the normalized expression level and the x‐axes represent the five sampling times: 1, 2 h before tracheostomy (1); 2, 2 h after tracheostomy; 3, 6 h after tracheostomy; 4, 10 h after tracheostomy; and 5, 22 h after tracheostomy. Every line represents a single patient. The clusters were obtained with the data normalized by the first time‐point. For every cytokine, the red, green and blue plots represent the three different clusters found, numbered 1–3, in the upper left, upper right and lower left panels, respectively. The lower right panel shows the cluster means, which represent the averaged response of the clusters. When comparing patients it is important to note that the y‐axis represents fold‐changes, and higher or lower points on the y‐axis do not represent higher or lower absolute expression.

Figure 3

Pairwise correlations between cytokines. The correlation between the cytokine pairs for a single patient are shown. The cytokines are indicated at the top of each column. Every cell represents the correlation between the cytokine at the top of the intersecting column with the cytokine at the right of the intersecting row. The cytokine data were normalized by the first time‐point, and the x‐ and y‐axes represent the normalized expression levels of the corresponding cytokines. Every cell shows five black dots representing the levels of the corresponding X–Y cytokine pair at the five sampling times. A simple fitted regression line (using minimal residual sum of squares) is shown in red.