Mild induced hypothermia and coagulation and platelet function in patients with septic shock: Secondary outcome of a randomized trial

Coagulation abnormalities and microthrombi contribute to septic shock, but the impact of body temperature regulation on coagulation in patients with sepsis is unknown. We tested the hypothesis that mild induced hypothermia reduces coagulation and platelet aggregation in patients with septic shock. Secondary analysis of randomized controlled trial. Adult patients with septic shock who required mechanical ventilation from eight intensive care units in Denmark were randomly assigned to mild induced hypothermia for 24 h or routine thermal management. Viscoelastography and platelet aggregation were assessed at trial inclusion, after 12 h of thermal management, and 24 h after inclusion. A total of 326 patients were randomized to mild induced hypothermia (n = 163) or routine thermal management (n = 163). Mild induced hypothermia slightly prolonged activated partial thromboplastin time and thrombus initiation time (R time 8.0 min [interquartile range, IQR 6.6–11.1] vs. 7.2 min [IQR 5.8–9.2]; p = .004) and marginally inhibited thrombus propagation (angle 68° [IQR 59–73] vs. 71° [IQR 63–75]; p = .014). The effect was also present after 24 h. Clot strength remained unaffected (MA 71 mm [IQR 66–76] with mild induced hypothermia vs. 72 mm (65–77) with routine thermal management, p = .9). The proportion of patients with hyperfibrinolysis was not affected (0.7% vs. 3.3%; p = .19), but the proportion of patients with no fibrinolysis was high in the mild hypothermia group (8.8% vs. 40.4%; p < .001). The mild induced hypothermia group had lower platelet aggregation: ASPI 85U (IQR 50–113) versus 109U (IQR 74–148, p < .001), ADP 61U (IQR 40–83) versus 79 U (IQR 54–101, p < .001), TRAP 108 (IQR 83–154) versus 119 (IQR 94–146, p = .042) and COL 50U (IQR 34–66) versus 67U (IQR 46–92, p < .001). In patients with septic shock, mild induced hypothermia slightly impaired clot initiation, but did not change clot strength. Platelet aggregation was slightly impaired. The effect of mild induced hypothermia on viscoelastography and platelet aggregation was however not in a range that would have clinical implications. We did observe a substantial reduction in fibrinolysis.

slightly impaired clot initiation, but did not change clot strength. Platelet aggregation was slightly impaired. The effect of mild induced hypothermia on viscoelastography and platelet aggregation was however not in a range that would have clinical implications. We did observe a substantial reduction in fibrinolysis.

K E Y W O R D S
coagulopathy, sepsis, septic shock, therapeutic hypothermia

Editorial Comment
Sepsis affects coagulation and in this study, the additional effect of mild hypothermia on sepsisinduced coagulopathy was assessed. While clinically significant effects of hypothermia on coagulation were not found, fibrinolysis was inhibited. Mild hypothermia is not a common treatment for sepsis, but the findings may be transferable to other clinical situations where mild hypothermia is used and coagulation abnormalities are common.

| INTRODUCTION
Septic shock is the most severe manifestation of infection, and it is characterized by hypotension, hypoperfusion, and organ dysfunction.
Septic patients often have abnormal coagulation and platelet function. [1][2][3][4] Microthrombi, in particular, are thought to contribute to septic shock progression, organ failure, and mortality. Sepsis patients frequently have a dysregulated body temperature, but the impact of modulation of the body temperature on coagulation remains unknown. [5][6][7][8] In vitro studies indicate that initiation and propagation of coagulation are impaired by hypothermia and similar effects are seen in hypothermic patients. [9][10][11][12][13] Mild hypothermia also impairs platelet aggregation, although the mechanism and magnitude of the effects remain incompletely understood. 14 Little is known regarding the effect of mild hypothermia on thrombus lysis.
Experimental studies suggest that mild induced hypothermia might mitigate the severity of septic shock and potentially reduce mortality by attenuating the risk of micro thrombosis. [15][16][17][18][19][20][21][22] We therefore recently completed the "Cooling and Surviving Septic shock" (CASS) trial (n = 432) in which septic patients were randomized to routine thermal management (no thermal interventions) versus induced hypothermia to a target temperature of ≈33 C for 24 h. Mild hypothermia did not improve survival. 23 There are many potential reasons that mild hypothermia did not improve sepsis outcomes, and the effect of hypothermia on coagulation is among them. For example, mild induced hypothermia could exaggerate already disturbed coagulation homeostasis of septic patients. In a pre-planned sub-study of CASS, we therefore tested the primary hypothesis that hypothermia would inhibit coagulation in patients with septic shock.

| MATERIALS AND METHODS
We included patients from the CASS trial cohort. 23  We have previously reported on a subpopulation of the current study-focus here was on the safety of the intervention. 24

| Intervention
Patients were randomly allocated to either routine thermal management (no body temperature manipulations), or mild induced hypothermia for 24 h (core body temperature 32-34 C) followed by 72 h of controlled normothermia (protection against "rebound" fever). All other aspects of care were according to the sepsis guidelines applicable at the time. 25 We conducted the analysis based on the modified intention-to-treat population.

| Measurements
All patients had samples for viscoelastography and platelet aggregation tests drawn immediately following randomization (before mild induced hypothermia; t = 0), 12 h after randomization (during mild induced hypothermia or routine thermal management; t = 1), and 24 h after randomization (t = 2). Blood was also sampled for activated partial thromboplastin time (APTT), international normalized ratio (INR), and platelet concentration immediately following randomization (before mild induced hypothermia) and 24 h after randomization (during mild induced hypothermia or routine thermal management). In patients assigned to induced hypothermia, the 24-h sample was delayed until patients reached normothermia. Halfway through the study, we changed the 24-h sample to 33 h to align with normothermia. The samples for APTT, INR, and platelets were handled per routine by local laboratories.
Other variables, including disease severity and comorbidities, were assessed on study inclusion by the attending physician. Bleeding episodes during the intervention were defined as any observed GIbleeding, intracranial bleeding observed on CT scan, any bleeding requiring emergency surgery or transfusion of more than four erythrocyte suspensions over 24 h. Data collection was completed and monitored according to GCP standards.

| Statistics
We primarily analyzed the effect of mild induced hypothermia on coagulation by comparison of medians at baseline, 12 h, 24 h, and change from baseline to 12 and 24 h. Each was compared with Mann-Whitney U tests. We analyzed Ly30 as presence or absence of lysis and compared this by chi-square. We analyzed the mortality rates after 30 days by chi-square test.
Second, we analyzed the markers using mixed effects linear regression to assess the effect of potential confounders. In the mixed effects model, patient id was entered as a random factor, and time, intervention allocation and confounding factors as fixed effects. The non-linear effect of time on the coagulation parameters was modeled with a linear spline. The models included an interaction term between time and mild induced hypothermia. The primary analysis was unadjusted, and secondary analyses were adjusted for confounding factors known to influence coagulation and the clinical course of sepsis: age, gender, ischemic heart disease, excessive alcohol consumption or liver disease, or previous cerebral thrombosis and sepsis severity as determined by lactate. The model fit was assessed through inspection of normality of residuals and random effects.
We used a Cox regression with an interaction term between baseline presence of fibrinolysis and intervention and tested the significance of this interaction with a log-likelihood test.
For categorical and continuous markers, n (%) and median (interquartile range [IQR]) are presented, respectively. p < .05 was considered statistically significant. All analyses were done using R There was no difference in bleeding episodes between the groups with 5 in the mild induced hypothermia group and 3 in the routine thermal care ( p = .7). Coagulation status was abnormal on inclusion, but similar in each group, Table S1.   In adjusted analyses, only ASPI was slightly decreased by the mild induced hypothermia, Table S1. When analyzed as ratios between platelet aggregation and platelet count, we did not detect any differences on baseline or after 24 h (t = 2). We detected a very small increase in the ratio in TRAP/platelet ratio between baseline and 24 h in the routine thermal management group compared to the mild induced hypothermia group: 0.05U/(10 6 /mL) (IQR À0.07 to 0.25) versus À0.01U/(10 6 /mL) (IQR À0.08 to 0.09; p = .022). We did not detect a difference in this ratio after 24 h (t = 2).

| Concomitant therapy
The patients in the mild induced hypothermia group received larger volumes of fluid during the study period. Further, the patients in the

| DISCUSSION
Mild induced hypothermia slightly reduced the thrombus initiation and propagation, but also reduced the physiological clot fibrinolysis.
There was no effect on the clot strength. Further, mild induced hypothermia inhibited platelet aggregation and platelet count declined in both the hypothermic and routine thermal management groups, with the decline being more pronounced in the mild induced hypothermia group. We can view our results based on the evolution of the coagulation cascade: initiation, propagation, clot strength and lysis. We have previously reported on an interim population from the current study where we found that there was a trend toward improvement of coagulopathy with mild induced hypothermia. 24 With this updated analysis of a substantially larger population, we could not replicate this finding.
The coagulation initiation is well described and highly temperature dependent, with a general reduction in coagulation potential dur- Among patients with sepsis, a substantial proportion has affected clot stability, and low strength is associated with poor outcome. 38  The effect of mild induced hypothermia on platelet aggregation in patients with sepsis before mild induced hypothermia (t = 0), after 12 h mild induced hypothermia (t = 1) and after ceasing the mild induced hypothermia (t = 2). * Denotes p < .05 between groups at time point, # denotes p < .05 for difference time point 2 À time point 1 between groups. Platelet aggregation with the platelet activation factors collagen (COL), arachidonic acid (ASPI), adenosine diphosphate (ADP), and thrombin receptor activating peptide (TRAP). p Value is the F test for effect of variables.
high-risk abdominal surgery where the patients in addition to the surgery may also have sepsis. 43 A limitation of our trial is that we did not measure platelet number during mild hypothermia because platelet aggregation results depend on platelet concentration. [26][27][28] Furthermore, platelet aggregation could have been measured at body temperature during the mild induced hypothermia, but this was not feasible in our labs in the study period as the devices was also use for clinical testing. This introduces the risk that even more pronounced effects of hypothermia were present. How long the persistent effect of the mild induced hypothermia lasted, could have been explored by continuing the observation, although loss to follow-up (death and discharge) would limit interpretation. The randomized nature of the study limits the potential bias of pre-existing differences between the groups. The hypothermia group received a larger volume of intravenous fluid during the study period.
The absolute difference was however not in a volume that we would expect to have a significant impact on coagulation.

| CONCLUSIONS
In patients with septic shock, mild induced hypothermia slightly impaired clot initiation, but did not change clot strength. Platelet aggregation was slightly impaired. The effect of mild induced hypothermia on viscoelastography and platelet aggregation was however not in a range that would have clinical implications. We observed a substantial reduction in fibrinolysis.