Vasopressin and terlipressin in adult vasodilatory shock

We have read with interest the study by Serpa Neto and colleagues showing that vasopressin use in vasodilatory shock is associated with reduced mortality [1]. However, we believe there are three aspects that need clarification. 
 
In relation to the paper recently published by our group on the same topic, the authors correctly state that 'we only analyzed adult patients because pediatric shock has a much lower mortality that adult shock and this may "contaminate" the overall results' [1]. For the main analysis in our article, we computed data from six randomized trials exploring use of vasopressin/terlipressin only in adults [2]. Mortality in children was examined in a separate subgroup analysis. 
 
Second, we intentionally excluded the paper from Argenziano and colleagues because this study examined the use of vasopressin in patients after cardiopulmonary bypass and ventricular assist device placement, a quite unique patient population with regards to the pathophysiology of shock [3]. Argenziano and colleagues' study was also carried out in a time of very different standards of intensive care (1997), compared with all other studies that were performed in the mid to late 2000s. 
 
Moreover, the 'reduction in norepinephrine requirement' presented by the authors is not in conflict with our finding of a significant inverse relation between the vasopressin dose and the norepinephrine dose [2]. 
 
On a side note, we believe the authors' choice of using a fixed-effect model to compute a pooled estimate of the relative risk is questionable because it is based on the unlikely assumption that no statistical heterogeneity is present among the included studies.

We have read with interest the study by Serpa Neto and colleagues showing that vasopressin use in vasodilatory shock is associated with reduced mortality [1]. However, we believe there are three aspects that need clarifi cation.
In relation to the paper recently published by our group on the same topic, the authors correctly state that 'we only analyzed adult patients because pediatric shock has a much lower mortality that adult shock and this may "contaminate" the overall results' [1]. For the main analysis in our article, we computed data from six random ized trials exploring use of vasopressin/terlipressin only in adults [2]. Mortality in children was examined in a separate subgroup analysis.
Second, we intentionally excluded the paper from Argenziano and colleagues because this study examined the use of vasopressin in patients after cardiopulmonary bypass and ventricular assist device placement, a quite unique patient population with regards to the pathophysio logy of shock [3]. Argenziano and colleagues' study was also carried out in a time of very diff erent standards of intensive care (1997), compared with all other studies that were performed in the mid to late 2000s.
Moreover, the 'reduction in norepinephrine requirement' presented by the authors is not in confl ict with our fi nding of a signifi cant inverse relation between the vasopressin dose and the norepinephrine dose [2].
On a side note, we believe the authors' choice of using a fi xed-eff ect model to compute a pooled estimate of the relative risk is questionable because it is based on the unlikely assumption that no statistical heterogeneity is present among the included studies.

Ary Serpa Neto and James A Russell
First of all we would like to thank Polito and colleagues for their interest, comments and criticisms.
Despite the unique type of population studied in the paper by Argenziano and colleagues [3], the defi nition of vasodilatory shock used by the authors was the same as that used in the other studies we analyzed (hypotension plus adequate cardiac output, measured as left ventricu lar assist device output). We also believe that the hall marks of shock in these patients are the same as those in the other studies: hypotension, lactic acidosis, and profound vasodilatory hypotension unresponsive to catechol amines [4].
We agree that our fi ndings of 'reduction in nor epinephrine requirements' are not in confl ict with the results of Polito and colleagues [2]. In our paper we did not state the contrary, we only stated that we analyzed this variable in another way -using the standardized mean diff erence, which is a methodology diff erent from meta-regression to analyze the reduction in nor epi nephrine requirements [1].
Finally, we used the fi xed-eff ect model in the pooled estimate of the risk ratio of mortality because we did not fi nd any signs of statistical heterogeneity, as can see by the low I 2 values (0% in all comparisons) and by the high P values (always above 0.50). We believe that the choice of which test to use cannot be made based only on the basis of the heterogeneity test. Furthermore, all studies that we evaluated and included in our review share a common eff ect size, which makes the use of the fi xedeff ect model appropriate.