Vasopressin and terlipressin in adult vasodilatory shock: a systematic review and meta-analysis of nine randomized controlled trials

Studies were identified using the Medline (1966 to 2011) and CENTRAL (1800 to 2011) databases using a sensitive search strategy combining medical subject headings and keywords (see Additional file 1 for details). Abstracts from recent major conferences (American Thoracic Society, European Society of Intensive Care Medicine, and Society of Critical Care Medicine) were searched for additional relevant studies. All of the review articles and cross-referenced studies from the retrieved articles were screened for pertinent information.

This meta-analysis was limited to studies that dealt with the role of vasopressin and/or terlipressin compared with catecholamine infusion in the treatment of vasodilatory shock in adult critically ill patients. Vasodilatory shock was defined as hypotension due to peripheral vasodilatation as result of failure of the vascular smooth muscle to constrict [2]. We included trials that satisfied the following inclusion criteria: randomized controlled trials comparing adult critically ill patients who had vasodilatory shock receiving treatment with vasopressin or terlipressin compared with patients not receiving such treatment; survival, biochemical and hemodynamic data; and patients in the studies receiving vasoactive infusion in the control arm and vasopressin plus vasoactive infusion in the experimental arm. Studies were excluded if survival outcome or biochemical and/or hemodynamic data were not provided, if pediatric patients were analyzed, or if they included patients without vasodilatory shock. When we found duplicate reports of the same study in preliminary abstracts and articles, we analyzed data from the most complete dataset.

Data were independently extracted from each report by three authors, using a data-recording form developed for this purpose. After extraction, data were reviewed and compared by ASN. Disagreements between the two extractors were solved by consensus among the investigators. Whenever needed, additional information concerning a specific study was obtained by directly questioning the principal investigator.

The primary endpoint was overall survival. The survival time was defined as the time from randomization until death from any cause, or was censored on the date of the last follow-up assessment. Secondary endpoints included: change in hemodynamic variables; change in biochemical variables; decrease of catecholamine requirements; and assessment of adverse events.

The effects of vasopressin and terlipressin on vasodilatory shock outcomes and the adverse effects of these drugs were examined. We extracted data regarding the study design, patient characteristics, treatment duration, medication, dosage, mean change for hemodynamic and biochemical variables, overall survival, and decreased catecholamine requirement. Changes in hemodynamic and biochemical variables were expressed as a percentage and were defined as the final value (after follow-up) minus the baseline value divided by the baseline value. The decreased catecholamine requirement was defined as any sustained decrease in catecholamine infusion over the study follow-up period. Norepinephrine and dopamine infusions were reported as micrograms per minute, vasopressin infusion was reported as units per minute, and terlipressin infusion was reported as micrograms per hour. The conversion of weight-based variables (μ/kg/minute) to time-based variables (μg/minute or μg/hour) was made by multiplying the value by the mean weight, either provided by the author or estimated as 70 kg. The conversion of bolus administration (mg) to continuous administration (μg/hour) was made by dividing the total daily dose by 24 and multiplying it by a conversion factor (mg to μg). All of the time variables were described as hours, and the other variables are described in the text.

For the survival analysis and for the proportion of patients who experienced adverse events, a pooled estimate of the relative risks (RRs) of the individual studies was computed using a fixed-effect model, according to Mantel and Haenszel, and these results graphically represented using forest plot graphs. For continuous variables, the standardized mean difference, which consists of the difference in means divided by the standard deviation, was used. The homogeneity assumption was checked by a chi-squared test with the degrees of freedom equal to the number of analyzed studies minus one. A sensitivity analysis was performed by recalculating the pooled RR estimates for different study subgroups based on the relevant clinical features. This analysis serves to show whether the overall result would be affected by a change in the meta-analysis selection criteria [5]. An estimate of the potential publication bias was carried out by plotting the single-study RR on a log scale against the respective standard error (funnel plot). For multiple comparisons, we used the Bonferroni correction method. Inter-rater reliability was determined by comparing the number of studies searched by Author 1 versus the number searched by Author 2 in each stage of the search by the kappa coefficient [5].

All variables were tested for normality using the Kolgomorov-Smirnov test. The parametric variables were described as the means and standard deviations, and the nonparametric variables were described as the medians and interquartile ranges. All of the analyses were made using Review Manager version 5.1 (Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration, 2011) and Statistical Package for the Social Sciences version 16.0 (SPSS Inc., Chicago IL, USA). For all analyses, P < 0.05 was considered significant. For publication bias, P < 0.1 was considered significant.

The search strategy retrieved 105 unique citations. Of these citations, 84 were excluded after the first screening based on the abstracts or titles, leaving 21 articles for a full-text review (Figure 1). In this review, 12 articles were excluded for the following reasons: outcome of interest not described (n = 5); study design not appropriate (n = 4); pediatric patients (n = 2); and evaluated vasopressin in both arms (n = 1). Finally, nine articles (998 participants) were included in the meta-analysis [6‐14]. For all of the comparisons of inter-rater reliability in each stage of the search, the kappa coefficient ranged from 0.87 to 0.93.

The characteristics of the nine selected studies are shown in Table 1. Six studies evaluated vasopressin as a therapeutic approach, two evaluated terlipressin and one study evaluated both. In seven studies the disease responsible for vasodilatory shock was septic shock exclusively, in one study the shock occurred after post-left ventricular assist device, and in the last study the shock occurred post cardiotomy. The mean dose of the drugs across the studies was 38.72 ± 40.14 μg/minute for norepinephrine, 0.055 ± 0.027 U/minute for vasopressin, and 59.03 ± 47.59 μg/hour for terlipressin. The mean age was 61.91 ± 6.29 in entire group, and the median follow-up time was 24 (4.00 to 48.0) hours. The assessment of the study quality is exposed in Table S1 in Additional file 2.

Table S2 in Additional file 2 shows the changes in the hemodynamic variables of the vasopressin, terlipressin, and control patients in each study and in a combined analysis. Compared with the control group, the vasopressin group showed a significant reduction in heart rate (-12.8 ± 6.14% vs. -1.62 ± 6.86%; P = 0.023), and a nonsignificant increases in central venous pressure (+19.0 ± 1.41% vs. +10.66 ± 9.29%; P = 0.083), stroke volume index (+14.0 ± 8.54% vs. +1.00 ± 2.94%; P = 0.05), and left ventricular stroke work index (+65.00 ± 5.65% vs. 29.0 ± 26.54%; P = 0.064). When compared with terlipressin group, the vasopressin group showed a nonsignificant smaller reduction in cardiac index (-4.00 ± 9.25% vs. -16.33 ± 4.04%; P = 0.070). Compared with the control group, the terlipressin group showed a significant reduction in heart rate (-17.3 ± 7.09% vs. -1.62 ± 6.86%; P = 0.011) and oxygen delivery index (-16.0 ± 2.64% vs. -2.0 ± 6.28%; P = 0.036), and a borderline significant reduction in the oxygen consumption index (-11.33 ± 6.02% vs. -1.00 ± 5.65%; P = 0.071).

Table S3 in Additional file 2 shows the changes in the laboratorial variables of the vasopressin, terlipressin, and control patients in each study and in the combined analysis. There were no differences among the groups according to the variables evaluated. In the analyses of the standardized mean difference, we found a significant difference between the terlipressin and control groups in the cardiac index (-0.44 (-0.87 to -0.02); P = 0.04) and oxygen delivery index (-0.79 (-1.23 to -0.36); P = 0.0004), and a tendency toward a reduction in the gastric PaCO₂ gap difference (-0.47 (-0.96 to 0.01); P = 0.06) (Figures S1 to S5 in Additional file 3).

There was a significant reduction in the norepinephrine requirement among the patients receiving either a terlipressin or a vasopressin infusion. In the analyses of the standardized mean difference, there is a significant difference between the vasopressin and control patients (-1.56 (-1.71 to -1.41); P < 0.0001) and between the terlipressin and control patients (-1.97 (-2.62 to -1.32); P < 0.0001) with respect to the norepinephrine dosages (Figure 2).

The association of vasopressin with a decreased dose of norepinephrine infusion reduced mortality in adult patients (RR, 0.87 (0.76 to 1.00); P = 0.05). The terlipressin infusion did not influence mortality (RR, 0.88 (0.62 to 1.25); P = 0.47), and in the combined analysis terlipressin + vasopressin resulted in a decreased mortality (RR, 0.87 (0.77 to 0.99); P = 0.04) (Figure 3). With respect to adverse events, vasopressin and control patients were analyzed because no study that evaluated the effects of terlipressin described adverse events (see Table S4 in Additional file 2). There was no difference in adverse events between the vasopressin and control groups (RR, 0.98 (0.65 to 1.47); P = 0.92) (Figure S6 in Additional file 3). There is no correlation between the magnitude of the reduction of norepinephrine during follow-up and the risk ratio for mortality (r = 0.143; P = 0.652).

To explore the study heterogeneity, stratified analyses were performed across a number of key study characteristics and clinical factors. The summary of the analysis is shown in Table 2, and the complete analysis is shown in Table S5 in Additional file 2. In the survival analysis, we found a reduction in mortality with vasopressin patients with septic shock (P = 0.05) (Figure 3). The number needed to treat for this condition was 1 to 15. This effect on mortality disappeared after analysis without the study by Russell and colleagues [10]. The reduction in norepinephrine requirement was more significant in the double-blinded studies, in patients with septic shock, and with smaller doses of vasopressin and terlipressin. When we analyzed the data without the study by Russell and colleagues [10], the standardized mean difference found was smaller but still significant (Table S5 in Additional file 2).

When analyzing only the double-blinded studies and the patients with septic shock, vasopressin and terlipressin were associated with an increase in cardiac index compared with norepinephrine alone. However, these findings were more significant in the studies with a shorter follow-up and a smaller number of patients. Higher doses of terlipressin (>40 μg/hour) were associated with decreases in cardiac index, oxygen delivery index, oxygen consumption index, and gastric PaCO₂ gap difference. Further, higher doses of vasopressin (>0.05 U/minute) were associated with an increased gastric PaCO₂ gap difference. Finally, bolus infusion of terlipressin was associated with a significant reduction of the oxygen delivery index and a borderline significant reduction of the cardiac index. These findings were not found with the continuous infusion. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) evidence profile for the impact of vasopressin for the treatment of vasodilatory shock from this systematic review and meta-analysis of randomized controlled trials is shown in Table S6 in Additional file 2.