Finally, the original study recorded in a binary fashion the development of the following disease-related events in all patients during their intensive care unit stay, based on the clinical assessment of the investigators: many ‘cardiac dysrhythmias’ (including cardiac arrest), ‘acute circulatory failure’, ‘disseminated intravascular coagulopathy’, ‘acute hepatic failure’, ‘metabolic acidosis’, ‘acute deterioration in mental state’ (not due to sedation), ‘acute renal failure’, ‘acute (hypoxemic) respiratory failure’, and ‘thrombocytopenia’. In addition, the total number of disease-related events (defined as the sum of single disease-related events) was calculated for each study patient.

DefinitionsThe duration of shock was defined as the time from study randomization until the patient met all of the following criteria: 1) epinephrine, norepinephrine, phenylephrine, and dobutamine infusion of 0 ��g/kg/min; 2) dopamine infusion of 3 ��g/kg/min or less; 3) dopexamine infusion of 1 ��g/kg/min or less; 4) MAP of 70 mmHg or more [13]. Pre-existence of chronic arterial hypertension was based on contemporary definitions of the World Health Organization. As defined in the original trial [13], disease-related events were considered as events known to be associated with severe sepsis and/or septic shock and considered by the investigator as not having a reasonable possibility of being caused by 546C88 or placebo therapy.Study endpointsThe primary endpoint of this post hoc analysis was to investigate the association between MAP or MAP quartiles of 70 mmHg or higher and 28-day mortality.

Furthermore, we sought to evaluate whether this association was influenced by age, pre-existent arterial hypertension or the mean vasopressor load. The secondary endpoint was to investigate the association between MAP or MAP quartiles of 70 mmHg or higher and the occurrence of disease-related events. Again the influence of age, pre-existent arterial hypertension and the mean vasopressor load on these associations was evaluated.Statistical analysisThe SPSS software program was used for statistical analysis (SPSS 15.0; SPSS Inc, Chicago, IL, USA). Kolmogorov-Smirnov tests were applied to check for normality distribution of data which was approximately fulfilled by all variables except the mean vasopressor load. This variable underwent ln-transformation and subsequently showed normal distribution.

Descriptive statistical methods were used to present study variables. For comparisons between survivors and non-survivors, Student’s t-tests and Fisher’s Exact tests were applied, as appropriate. Binary logistic regression models were used to answer the primary and secondary study endpoints. These models included either 28-day mortality or the occurrence of disease-related Carfilzomib events as the dependent variable.