The global burden of critical illness exceeds the cumulative mortality of breast cancer, HIV, and asthma combined.1 In the critically ill, hypotension is common and often lethal when extreme, progressive, or persistent. Vasopressors are used to compensate for excessive vasodilation or, when facing imminent circulatory arrest regardless of the cause of hypotension, to raise blood pressure for a short period while other corrective measures are implemented.2 Nevertheless, vasopressor-induced vasoconstriction can compromise blood perfusion at a microvascular level, even when blood pressure is kept within normal ranges. Vasopressors also have numerous pleiotropic effects that are not easy to predict or measure, and some of these can be harmful.3

Notwithstanding, clinicians must balance the risks from hypotension with the potential adverse effects of vasopressors. Experts have recommended a mean arterial pressure (MAP) target of at least 65 mmHg, and higher in older patients and in patients with chronic hypertension or atherosclerosis.4 Recent observational evidence suggests that patients exposed to higher vasopressor doses have a higher risk of adverse cardiovascular events.5

In 2016, the Canadian Critical Care Society and the Scandinavian Society of Anaesthesiology and Intensive Care undertook joint guidelines on targets for vasopressor use. To inform these guidelines, we updated a systematic review published in 2015.6 Following recommendations for trustworthy guidelines,7 we submitted the protocol of this systematic review to the guideline panel whose recommendations bore on the final version of the protocol. Accordingly, the research question was broadened beyond the previous systematic review to include all forms of hypotension. This work aims to answer the following question: In adult critically ill patients with hypotension requiring vasopressor support, should we prescribe higher (MAP 75-85 mmHg) vs lower (MAP 60-70 mmHg) blood pressure targets? Our hypothesis was that higher MAP targets would not be associated with clinically measurable benefits.

Methods

The guidelines were created according to standards for trustworthy guidelines in collaboration with the MAGIC WikiRecs project.8 After consultation with the members of the guideline panel, we registered this systematic review on PROSPERO (CRD42016033438) and present the results according to PRISMA guidance.9

Search strategy and study selection

We searched MEDLINE® (from 1946 to October 12, 2016), EMBASE™ (from 1980 to October 12, 2016), and the Cochrane Central Register of Controlled Trials for randomized-controlled trials of higher vs lower blood pressure targets for vasopressor therapy in adult patients who are in shock or hypotensive (as defined by the investigators). The studies were required to provide information on at least one clinically important outcome (defined as mortality, quality of life after hospital discharge, use of renal replacement therapy, duration of intensive care unit (ICU) or hospital stay, or vasopressor-induced adverse events). We excluded crossover designs, studies conducted in pediatric populations, trials of vasopressors that are not routinely used, and studies in which the duration of the experiment was designed for less than< 24 hr by design (i.e., studies could be included where patients were treated for less than< 24 hr in included studies, but studies were excluded where patients were followed for brief periods— were excluded because short protocols were not expected to influence clinically relevant outcomes). There was no exclusion based on the language of the published report or the quality of evidence. A detailed search strategy appears in the Appendix.

We also manually searched published editorials, reviews, and the reference lists of identified articles, as well as proceedings from the annual meetings (2005-2016) of the American Thoracic Society, the Society of Critical Care Medicine, the European Society of Intensive Care Medicine, the International Symposium on Intensive Care and Emergency Medicine, the American Association for the Surgery of Trauma, the Eastern Association for the Surgery of Trauma, the Shock Society, the European Shock Society and the American College of Chest Physicians.

Two reviewers independently assessed trial eligibility based on titles and abstracts, and they then selected full-text reports. We contacted authors for further information when uncertainty persisted about eligibility.

Outcomes

All outcomes included in this review were prespecified. The primary outcome was all-cause short-term mortality, defined as the longest reported follow-up within 90 days, including ICU and hospital mortality. Secondary outcomes included long-term mortality, defined by a more distant time point (e.g., six months), early acute renal replacement therapy (within 90 days), late chronic renal replacement therapy (after 90 days), duration of renal replacement therapy (days), duration of mechanical ventilation (days), fluid use (cumulative volume received), blood product requirements (cumulative volume received), acute kidney injury, new-onset cardiac arrhythmia, digit or limb or skin ischemia, mesenteric ischemia, myocardial ischemia, quality of life, and neurological outcome at longest reported follow-up.

Data abstraction

We prepared and piloted data extraction forms before launching the study. Reviewers followed written instructions that were developed a priori in order to standardize data extraction. In teams of two, reviewers then independently collected information on study design, patient population, interventions, and outcomes.

Statistical analysis

All analyses were conducted according to the predefined statistical analysis plan outlined in the protocol (CRD42016033438).

For all outcomes, we compared patients treated with a higher blood pressure target vs patients treated with a lower blood pressure target. Using the Review Manager 5.3 software (Cochrane Collaboration), we meta-analyzed the included studies using the random-effects model described by DerSimonian and Laird, and individual study weights were measured using the inverse variance method.10 For dichotomous data, we present pooled summary effect measures as risk ratios (RRs), with 95% confidence intervals (CIs)11 for relative effects and number needed to treat, and risk difference by assuming the baseline risk for each outcome without the intervention is the risk in the control group for absolute effects.12 When appropriate, we applied a continuity correction of zero outcome scenarios. Effect sizes for continuous variables are presented as mean differences (MDs) with 95% CIs.

We assessed study heterogeneity qualitatively by considering whether or not study populations, interventions, and settings were comparable across reports. We performed a Chi square test for homogeneity and assessed heterogeneity quantitatively using Higgins and Thompson’s I 2 statistic.

Subgroup analyses

We used random effects models to perform two subgroup analyses of the primary outcome and report the statistical interaction between study-level subgroup-defining variables and the intervention (high vs low blood pressure target). The subgroup-defining variables were 1) presence or absence of baseline hypertension— hypothesizing that hypertensive patients will benefit more from high blood pressure targets, and 2) age ≥ 65 yr vs < 65 yr— hypothesizing that more elderly patients benefit from lower blood pressure targets. Both the subgroup analyses and our hypothesized direction of effect were prespecified. Study-level data on other prespecified subgroups (congestive heart failure, etiology of hypotension, illness severity, and risk of bias) were not available.

Risk of bias assessment for individual studies

We assessed risk of bias in randomized-controlled trials at the study level independently for each outcome using a modified version (includes a systematic assessment for risks of bias associated with stopping a trial before the planned sample size is reached)13 of the Cochrane Collaboration’s instrument.12,14 The instrument addresses the following domains: allocation sequence concealment, blinding of participants and caregivers, blinding of data collectors, blinding for outcome assessment, blinding of data analysts, loss to follow-up, selective outcome reporting, and other risks of bias. Studies with one or more domains assessed as a potential source of bias were considered overall at high risk of bias. We assessed the overall quality of the data for each individual outcome using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.15

Publication bias and confidence in the evidence

A statistical assessment of the risk of publication bias was planned if there were at least ten included studies.16 We used GRADE to rate the overall confidence in the estimates of intervention effects.15 In this method, quality of evidence is rated “very low,” “low,” “moderate” or “high”. Randomized-controlled trials begin as high-quality evidence, while observational studies begin as low-quality evidence. Quality of evidence can be rated down for risk of bias,17 imprecision,18 inconsistency,19 indirectness,20 and likelihood of publication bias.21 Observational studies can be rated up in the presence of a large magnitude of the association, a dose-response gradient, or if all unaccounted confounders increase confidence in estimates of effect. We assessed the risk of random errors (imprecision) by conducting trial sequential analyses (TSA).22 Trial sequential analyses is a sample size calculation (interim analysis) for cumulative meta-analyses that widens the confidence intervals when data are too sparse to draw firm conclusions. For each TSA, we report the required sample size from all trials and the proportion of this sample size already accrued.

Results

We retrieved 57 full-text articles from 8001 citations and ultimately included two randomized controlled trials.23,24 We identified one ongoing trial (NCT01473498). A PRISMA flowchart illustrates the selection process (Fig. 1).

Fig. 1
figure 1

PRISMA flowchart

Full size image

Study characteristics

Table 1 illustrates study characteristics. The included trials recruited 894 patients from France, Canada, and the United States. In both trials, norepinephrine was the most common vasopressor, and the higher MAP targets were 15 mmHg above the lower MAP targets. Adherence to protocol was superior in the higher MAP arms in both trials (i.e., actual MAP values were above the upper limit of the prescribed range in the lower MAP arms only). While one trial did not restrict patient eligibility to septic shock, this was the most common admission diagnosis in both trials. Differences between trials included the duration of vasopressor therapy allowed before randomization (longer in the study by Lamontagne et al.) and the specification of a minimal vasopressor dose at baseline (required in the study by Asfar et al. only). Excluded clinical experiments are summarized in the Appendix.

Table 1 Included studies
Full size table

Risk of bias

In both trials, allocation was concealed; analyses respected the intention-to-treat principle; follow-up for the primary endpoint was complete; and enrolment was not stopped early because of evidence of benefit or harm. Nevertheless, in both trials, caregivers (including bedside clinicians) were aware of the allocated study arm. Assessors for arrhythmia occurring during the five-day intervention were blinded in the study by Asfar et al. We considered overall risk of bias to be high in both trials due to lack of blinding.

Mortality

Fig. 2 illustrates forest plots for the primary outcome. Higher blood pressure targets were not associated with lower short-term mortality (RR, 1.05; 95% CI, 0.90 to 1.23; P = 0.54; low confidence) or long-term mortality (RR, 1.13; 95% CI, 0.72 to 1.77; P = 0.60; low confidence). Neither age (interaction test, P = 0.17) nor chronic hypertension (interaction test, P = 0.32) modified the effect of higher blood pressure targets on mortality. Trial sequential analyses estimates were consistent with the primary analysis, but they highlighted that 82% of the required information size has been accrued.

Fig. 2
figure 2

Forest plots for short-term mortality overall and by subgroups

Full size image

Secondary outcomes

Higher blood pressure targets did not result in important differences in fluid balance over the first five ICU days (MD, 1.15 L; 95% CI, −2.03 to 4.33; P = 0.48; low confidence), use of acute renal replacement therapy (RR, 0.96; 95% CI, 0.80 to 1.14; P = 0.61; low confidence), duration of renal replacement therapy (MD, −0.41 days; 95% CI, −1.27 to 0.45; P = 0.35; low confidence), or number of ventilator-free days (MD, −0.84 days; 95% CI, −2.28 to 0.60; P = 0.25; low confidence). There were no apparent differences in the risk of digit or limb or skin ischemia, mesenteric ischemia, myocardial ischemia, or ventricular arrhythmia. Nevertheless, higher blood pressure targets were associated with a greater risk of new-onset supraventricular cardiac arrhythmia (RR, 2.08; 95% CI, 1.28 to 3.38; P < 0.01; low confidence) (Table 2).

Table 2 GRADE Summary of effects
Full size table

The long-term neurological outcome, reported as the median [interquartile range] change from baseline to six months in the alpha Functional Independence Measure score among survivors, was reported in only one study (n = 118) and was identical in both groups (0 [−10-0]). Neither study reported quality-of-life measures.

Quality of the evidence and publication bias

A summary of the evidence is presented in Table 2. Confidence in the effect estimates was low for all outcomes due to risk of bias and imprecision. We could not conduct statistical analyses to quantify the risk of publication bias.

Discussion

The results of this systematic review, which includes data from a trial published in 2016, do not suggest that higher blood pressure targets in adult critically ill patients needing vasopressors modify mortality overall. Further, current evidence does not support increasing MAP targets for patients who have chronic hypertension, as suggested by the 2013 Surviving Sepsis Campaign Guidelines.4 In contrast, we found that higher blood pressure targets might acutely increase the risk for supraventricular arrhythmia, which echoes the results of recent observational studies.5 These results underscore the absence of supportive evidence for individualized therapy based on the presence of chronic hypertension or patient age.4 Using this meta-analysis and considering all the GRADE evidence regarding recommendation factors,25 the Canadian Critical Care Society and the Scandinavian Society of Anaesthesiology and Intensive Care Medicine suggest using lower (MAP 60-70 mmHg) rather than higher (MAP 75-85 mmHg) blood pressure targets in adult critically ill patients with hypotension requiring vasopressors (conditional recommendation, low confidence in the overall evidence).26,27

Blood pressure values measured in the context of clinical trials28 or observational studies29 are, on average, consistently > 70 mmHg. Vasopressor stewardship—i.e., tighter control of vasopressor infusions for all patients, including older patients and those who suffer from chronic hypertension— is the expected impact of this systematic review and related guidelines.

Strengths of this review include explicit and prespecified eligibility criteria, a comprehensive literature search, duplicate adjudication of eligibility, data extraction and risk of bias assessment, prespecified analyses and subgroups, and the use of GRADE to assess and communicate confidence in the effect estimates. Nevertheless, a number of limitations preclude definitive conclusions regarding the speculative benefits of individualized vasopressor therapy. Confidence intervals and trial sequential analyses confirm that the effect estimates for all outcomes remain imprecise. As such, important benefit and harm remain possible. The scope of the available evidence is limited to the outcomes that were collected and the population targeted by the included trials (i.e., distributive shock). Certain patient-important outcomes that were not measured in the included studies could influence clinical decisions regarding titration of vasopressors, in particular, post-discharge global quality of life.24 In both studies, actual MAP values were higher than specified in the protocol for the lower target groups. It is conceivable that effect estimates may have been different if vasopressors had been used more sparingly in the lower blood pressure target groups. What is the lowest tolerable blood pressure below which the benefits of vasopressors clearly outweigh the risk of adverse effects? What individual patient characteristics, including the etiology of hypotension, bear on the risk-benefit of vasopressor therapy? Would the effect of higher MAP targets be different with non-catecholamine vasopressors? Adequately powered clinical trials are required to address these persistent knowledge gaps.

Conclusion

Current evidence does not support a MAP target > 70 mmHg in critically ill hypotensive patients requiring vasopressor therapy. More randomized-controlled trial data are required to rule on the theoretical benefits of individualized vasopressor dosing.