Blood Pressure Targets for Adults with Vasodilatory Shock - An Individual Patient Data Meta-Analysis

Abstract

Background: We sought to estimate whether a lower mean arterial blood pressure target, compared with a higher mean arterial blood pressure target, reduced 90-day all-cause mortality among critically ill adult patients with vasodilatory shock.

Methods: We conducted an individual patient data meta-analysis of randomized controlled trials that evaluated the effect of distinct thresholds of mean arterial blood pressure to guide vasopressor support among critically ill adults identified in a systematic literature search. The main exposure was a lower mean arterial pressure target compared with a higher mean arterial pressure target (including usual care). The primary outcome was 90-day all-cause mortality. We used a Bayesian random effects log-binomial model to estimate risk ratios with 95% credible intervals (CrIs).

Results: Between 2010 and 2019, 3352 patients were randomly assigned in three trials (SEPSISPAM, OVATION pilot trial, and 65-Trial) across 103 hospitals from the United Kingdom, France, and Canada. When compared with a higher mean arterial blood pressure target or usual care, the risk ratio for 90-day all-cause mortality associated with a lower blood pressure target was 0.93 (95% CrI, 0.76 to 1.07; low certainty, posterior probability of benefit 87%). Results were consistent across multiple secondary and sensitivity analyses, including adjustment for prognostically important baseline covariates and alternative modeling techniques. Multiple approaches to evaluate the heterogeneity of treatment effect did not identify any subgroups that may potentially benefit from higher mean arterial blood pressure targets.

Conclusions: Targeting a lower mean arterial blood pressure for vasopressor therapy in critically ill patients with vasodilatory shock possibly reduced 90-day all-cause mortality. However, the certainty of evidence is low, and this analysis does not exclude the possibility that lower targets may cause harm overall.

Figure 1.. Patient Participation in the SEPSISPAM, 65-Trial, and OVATION trials.

Figure 2.. Heterogeneity of Treatment Effect by Subgroups of Interest.*

Acute Physiology and Chronic Health Evaluation II (APACHE II) scores range from 0 to 71 points, with more points indicating increased severity. High APACHE II scores are defined as more than 21 points (based on the sample’s median). The APACHE II score was not available for patients included in the SEPSISPAM trial and was derived using a Poisson regression model that was trained in the 65-Trial and OVATION pilot trial subset. These analyses used modified Poisson regression models (because of nonconvergence issues for most analyses with a log-binomial model) with a random intercept for trial and a random effect for treatment allocation. All models included treatment allocation, the average level of the subgroup-defining covariate by trial, the interaction of the average level and treatment allocation, and the individual level covariate with its interaction with treatment allocation. Parameters of interest were the coefficients for treatment allocation and the within-trial interaction. Intervals vary in width because of different modeling specifications and different sample sizes due to missing data for some subgroup-defining covariates (i.e., hypertension data missing for 11.6% of patients; chronic kidney disease data missing for 0.5% of patients; APACHE II scores missing for 0.3% of patients; and age data missing for 0.03% of patients). We used a complete-case analysis. *CrI denotes credible interval.