Increasing mean arterial pressure (MAP) to optimize cardiac output and tissue perfusion is central to the resuscitation strategy of critically ill patients.1 The Surviving Sepsis Campaign Guidelines strongly recommend targeting an initial MAP of 65 mm Hg in patients with septic shock who require vasopressors over higher MAP targets.2 Nevertheless, this approach has been challenged, and several publications have assessed the effect of higher MAP targets on clinical outcomes such as mortality and the need for renal replacement therapy.1,3,4

In 2019, a study of critically ill septic shock patients showed a lower mortality when MAP was maintained between 75 and 85 mm Hg among those with a history of chronic hypertension.4 Concurrently, a 2019 randomized controlled trial (RCT) found that targeting a MAP of 85–100 mm Hg did not mitigate secondary ischemic brain injury following resuscitation from cardiac arrest, and did not improve neurologic outcome.5 To achieve higher MAP targets, higher doses of vasopressors are often required, which can have adverse effects including organ ischemia, hyperglycemia, hyperlactatemia, and arrhythmias.6

Two systematic reviews and meta-analyses have sought to determine if high MAP targets improve mortality in critically ill patients.7,8 Nevertheless, since the publication of these studies, new RCTs have been published that have endeavoured to answer this clinical question.

Thus, we conducted an updated systematic review and meta-analysis of RCTs to determine the pooled estimate of comparing a high–normal with a low–normal MAP target on mortality and favourable neurologic outcome in critically ill patients. Among the RCTs selected for our primary outcomes, we also sought to determine if the need for renal replacement therapy or the rates of vasopressor-induced adverse events were different between the two groups.

Materials and methods

This systematic review and meta-analysis was registered on PROSPERO (CRD42022307601; 28 February 2022) and is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines.9

Information sources and search strategy

We searched the following electronic databases for published works from inception to 1 October 2022: MEDLINE (Ovid), Embase (Ovid), Evidence-Based Medicine Reviews (Ovid), CINAHL (EBSCO), and Web of Science. We also searched ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform for unpublished works and ongoing trials. Finally, we searched conference abstracts for the following meetings: Society of Critical Care Medicine, American Thoracic Society, and the European Society of Intensive Care Medicine from conference inception to 2021. Our search strategy used a combination of Medical Subject Heading and text word terms for the following three concepts: blood pressure, critically ill patients, and randomized trials. The search strategy is presented in Electronic Supplementary Material (ESM) eAppendix 1.

Study selection and outcomes

We included RCTs that included critically ill adults ≥ 16 yr of age admitted to a critical care unit who were in shock or had hypotension requiring vasopressor support. As the purpose of this meta-analysis was to compare a liberal with a restrictive MAP strategy, we included studies that randomized patients to high–normal (≥ 65 mm Hg) vs low–normal (≥ 60 mm Hg) MAP thresholds. The studies had to maintain the MAP targets for at least 24 hr. There were no language restrictions. We included all studies that reported on the following outcomes: mortality, favourable neurologic outcome, need for renal replacement therapy, and adverse vasopressor-induced adverse events. Our primary mortality outcome was 90-day mortality. Nevertheless, when 90-day mortality was not available, we used the closest presented mortality to 90 days. Favourable neurologic outcome was defined as either a modified Rankin Score between 0 and 2 or a Cerebral Performance Category Score of 1 and 2. The table of excluded trials is presented in ESM eAppendix 2.

Study selection

Using a two-stage selection process, three reviewers (K. R., C. R., and D. G.) independently and in duplicate screened abstracts using the inclusion and exclusion criteria above. Following abstract screening, the reviewers independently reviewed each full text before collaboratively making decisions on which studies to include. Disagreements were discussed before making final unanimous decisions on study inclusions.

Data collection and risk of bias assessment

Two reviewers (K. R. and D. G.) used a preformatted spreadsheet (Microsoft® Excel; Microsoft Corporation, Redmond, WA, USA) to collect data on the study design, inclusion and exclusion criteria, randomization, blinding, MAP targets in the control and intervention groups, duration for which MAP was maintained, mortality at 90 days or at a time point closest to 90-days, neurologic outcome at any time point, use of renal replacement therapy, and frequency of vasopressor-induced adverse events. We used Graph Grabber 2.0.2 (Quintessa Ltd., Henley-on-Thames, Oxfordshire, UK) to digitize graphs and extract data points for achieved MAP and vasopressor doses.

Study quality was assessed by two reviewers (K. R. and C. R.) using the revised Cochrane risk-of-bias (RoB-2) tool.10 Studies were rated on their risk of bias in four domains: 1) randomization process; 2) deviations from the intended interventions; 3) missing outcome data; and 4) measurement of the outcome.

Data synthesis

We used the risk ratio (RR) as the summary measure of association for mortality. The random-effects method of DerSimonian and Laird was used to generate the pooled RR across studies with corresponding 95% confidence intervals (CIs).11 Heterogeneity was quantified using the I2 statistic and 95% CIs.11All analyses were performed using STATA 17 (StataCorp LLC, College Station, TX, USA). Post hoc subgroup analyses were performed based on the presence or absence of patients with out-of-hospital cardiac arrest in the study population and on the magnitude of the separation of MAP between groups.

Assessment of certainty of evidence

We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework to present summaries of evidence for each outcome.12 The GRADE framework assesses each outcome in the following domains: risk of bias, imprecision, inconsistency, indirectness, and publication bias.

Results

Study selection and characteristics

We identified 5,775 studies (Fig. 1). After removing duplicates and screening abstracts, 26 articles underwent full-text review. We included eight studies with a total of 4,439 patients in our meta-analysis.5,13,14,15,16,17,18,19 Table 1 lists the characteristics of the included trials. Six trials were conducted in Europe5,13,15,17,18,19 and two in North America.14,16 Four trials were in patients with out-of-hospital cardiac arrest,5,15,17,19 two trials were in patients with vasodilatory shock requiring vasopressors14,18 and one trial each were of patients with septic shock13 and hepatorenal syndrome.16 The most frequent vasopressor used across all studies was norepinephrine.

Fig. 1
figure 1

PRISMA flow chart summarizing the literature search and study selection

Full size image
Table 1 Study characteristics
Full size table

Characteristics of blood pressure thresholds used in the trials

The targeted and actual MAPs for each trial are listed in Table 1. In the high–normal group, the lowest MAP target was 72 mm Hg. In the low–normal group, the lowest MAP target was 60 mm Hg. In one study, the MAP target in the control group was specified as “usual care,” while the MAP target in the intervention group was 60–65 mm Hg.18 For our analysis, we specified the “usual care” arm as the high–normal group. Duration of MAP targets in each study ranged from 36 hours to seven days.

Three RCTs achieved the targeted MAP thresholds in both the high–normal and low–normal arms5,15,17 and achieved separation between groups of 6 mm Hg,17 12 mm Hg,15 and 14 mm Hg.5 In four RCTs, the actual MAP was higher than the threshold in the low–normal arm.13,14,18,19 In one RCT, the actual MAP was below the target in the high–normal arm, and higher than the target in the low–normal arm.16 In this RCT, the separation between arms was 13 mm Hg over the study period.

Risk of bias assessment

Results from the RoB-2 are presented in ESM eAppendix 3. All included studies had some concerns of bias. In six trials, the treating clinicians were not blinded to the MAP targets.5,13,14,15,16,18 In two of the RCTs, the treating clinicians were blinded to the MAP targets.17,19

Mortality

Mortality was reported in all eight trials and the forest plot is presented in Fig. 2. The GRADE summary of findings is presented in Table 2. Only three trials presented mortality at our prespecified interval of 90 days.13,18,19 Thus, we used the following mortality outcomes for the other studies: in-hospital,16 30-day,15 and 180-day.5,14,17 The pooled RR across studies was 1.06 (95% CI, 0.99 to 1.14; moderate certainty). There was no statistical between-study heterogeneity (I2, 0%; 95% CI, 0 to 68). Performing a leave-one-out meta-analysis resulted in no meaningful change on the final point-estimate (results not shown). We examined the pooled RR in subgroups: trials that included patients with an out-of-hospital cardiac arrest (RR, 1.05; 95% CI, 0.90 to 1.24; four RCTs)5,15,17,19 compared with trials that did not (RR, 1.06; 95% CI, 0.98 to 1.15; four RCTs)13,14,16,18 and trials where the separation of MAP between groups was 10 mm Hg or less (RR, 1.06; 95% CI, 0.98 to 1.15; four RCTs)13,14,17,18 compared with a MAP separation of greater than 10 mm Hg (RR, 1.03; 95% CI, 0.88 to 1.21; four RCTs).5,15,16,19

Fig. 2
figure 2

Forest plot of studies with mortality as an outcome. The blue squares are the individual risk ratios (RRs) of included studies and the solid red line is the 95% confidence interval (CI). The size of the blue square is proportional to the weight of the individual study. The green diamond is the pooled RR across all studies obtained using a DerSimonian and Laird random effects model. The solid vertical line is a RR of 1.0.

Full size image
Table 2 GRADE summary of findings
Full size table

Favourable neurologic outcome

Four studies reported neurologic outcome.5,15,17,19 The forest plot is presented in Fig. 3 and the GRADE summary of findings are presented in Table 2. There was no difference in the rates of favourable neurologic outcome between the high–normal and low–normal MAP groups (RR, 0.99; 95% CI, 0.90 to 1.08; moderate certainty). There was no statistical between-study heterogeneity (I2, 0%; 95% CI, 0 to 85). Performing a leave-one-out meta-analysis resulted in no meaningful change to the final point-estimate (results not shown). Pooled RRs in trials where the separation of MAP between groups were 10 mm Hg or less (RR, 0.87; 95% CI, 0.48 to 1.59; one RCT),17 to a MAP separation of greater than 10 mm Hg (RR, 0.99; 95% CI, 0.91 to 1.09; three RCTs).5,15,19 The subgroup analysis based on the study population could not be performed since all trial included patients who had an out-of-hospital cardiac arrest.

Fig. 3
figure 3

Forest plot of studies that report favourable neurologic outcome. The blue squares are the individual risk ratios (RRs) of the included studies and the solid red line is the 95% confidence interval (CI). The size of the blue square is proportional to the weight of the individual study. The green diamond is the pooled RR across all studies obtained using a DerSimonian and Laird random effects model. The solid vertical line is a RR of 1.0.

Full size image

Use of renal replacement therapy

Four studies reported rates of renal replacement therapy.13,17,18,19 The forest plot is presented in Fig. 4 and the GRADE summary of findings is presented in Table 2. There was no difference in the rates of renal replacement therapy between the high–normal and low–normal MAP (RR, 0.97; 95% CI, 0.87 to 1.08; moderate certainty). There was no statistical between-study heterogeneity (I2, 0%; 95% CI, 0 to 85). Performing a leave-one-out meta-analysis resulted in no meaningful change on the final point-estimate (results not shown). As part of a post hoc analysis, we examined the pooled RR in subgroups: trials that included patients with an out-of-hospital cardiac arrest (RR, 0.80; 95% CI, 0.36 to 1.76; two RCTs)17,19 compared with trials that did not (RR, 0.98; 95% CI, 0.87 to 1.09; two RCTs)13,14 and trials where the separation of MAP between groups was 10 mm Hg or less (RR 0.97, 95% CI 0.87 to 1.08, three RCTs)13,17,18 compared with a MAP separation of greater than 10 mm Hg (RR, 1.03; 95% CI, 0.68 to 1.56; one RCT).19

Fig. 4
figure 4

Forest plot of studies that report the need for renal replacement therapy. The blue squares are the individual risk ratios (RRs) of the included studies and the solid red line is the 95% confidence interval (CI). The size of the blue square is proportional to the weight of the individual study. The green diamond is the pooled RR across all studies obtained using a DerSimonian and Laird random effects model. The solid vertical line is a RR of 1.0.

Full size image

Rates of adverse vasopressor events

Three studies reported rates of vasopressor-associated adverse events5,13,18 (ESM eAppendix 4). The most frequently reported adverse events were cardiac dysrhythmias. Only one trial reported a significantly increased rate of any adverse events in the low–normal compared with the high–normal MAP group (33% vs 13%, P = 0.02).5 The other two studies did not report a significant difference in the overall rate of adverse events between the two groups. Nevertheless, one study noted that the incidence of newly diagnosed atrial fibrillation was significantly higher in the high-target group (6.7% vs 2.8%, P = 0.02).13

Discussion

In this updated systematic review and meta-analysis of RCTs of critically ill patients, there were no significant differences in mortality or favourable neurologic outcome between patients targeted to high–normal vs low–normal MAP. Furthermore, although subject to selection bias, there was no difference in the need for renal replacement therapy between groups. There was no between-study heterogeneity in all outcomes. Finally, there did not appear to be any difference in the rates of adverse vasopressor events between the high–normal and low–normal MAP groups.

Studies examining MAP targets in critically ill patients have yielded conflicting results. A systematic review of nine observational studies suggested that higher MAP targets were associated with improved neurologic outcomes in patients with cardiac arrest.20 While a pooled analysis was not performed in that systematic review, because of marked between-study heterogeneity, seven of nine studies showed that a higher blood pressure was associated with improved outcomes. In contrast, an individual patient meta-analysis of two RCTs13,14 in hypotensive patients requiring vasopressor therapy showed that high–normal MAP targets were not associated with lower mortality (RR, 1.1; 95% CI, 0.90 to 1.2).8 This result was not modified by either age or premorbid hypertension. In fact, a post hoc analysis of this individual patient meta-analysis suggested that mortality may be higher in those patients who require vasopressors for longer than six hours.21 In the intervening time since the publication of this pooled analysis, several more RCTs have been published, which was the rationale behind our current study.

Our results are similar to those of a recent meta-analysis that also observed no difference in mortality (RR, 1.06; 95% CI, 0.98 to 1.15; six RCTs) or need for renal replacement therapy (RR, 0.96; 95% CI, 0.83 to 1.11; three RCTs).22 While similar trials were included in their meta-analysis,5,13,14,17,18 our meta-analysis additionally includes the recently published trial entitled “Blood pressure targets in comatose survivors of cardiac arrest,” (BOX trial) (n = 789).19 An additional difference is that we compared favourable neurologic outcomes between the two MAP strategies. While we observed no benefit of a high–normal MAP strategy, this differs from the possible benefit observed in a previous systematic review performed by members of our team.20 As this previous systematic review was restricted to studies of patients admitted following cardiac arrest, difference in the patient populations could be an explanation for these conflicting results. Nevertheless, our current study included four RCTs of patients following cardiac arrest,5,15,17,19 none of which showed any difference in mortality or favourable neurologic outcome with a high–normal MAP strategy. It must be noted only one of these included RCTs was powered to detect a difference in mortality.19 The difference observed between our meta-analysis and this previous one may be in the design of the included studies. Whereas our study included only RCTs, the later study included observational studies, so is thus subject to residual or unmeasured confounding. Although some patient groups may benefit from a high–normal MAP strategy, we observed no effect based on the magnitude of the separation of the MAP between intervention and control, or in trials that included patients with out-of-hospital cardiac arrest. Nevertheless, our meta-analysis only included eight trials, so is unpowered to examine for effect measure modification by patient population.

It is important to note that in all but one RCT,15 the low–normal MAP groups achieved a MAP target higher than was intended. Thus, the separation in MAP between the high–normal vs low–normal groups was smaller than anticipated, and in some studies was less than 10 mmHg.13,14,17,18 Thus, the lack of difference in outcomes between the two groups may simply reflect a lack of meaningful separation of MAP between groups. In addition, one of the studies included in our meta-analysis randomized patients to a strategy of permissive hypotension compared with a MAP threshold at the discretion of the attending physician.18 Although this intervention is fundamentally different to that in the other studies, we chose to include this RCT as there was still a difference in MAP between the two arms, and the MAP in the permissive hypotension arm was above 60 mm Hg. Thus, despite the different intervention, we feel including this trial still informs to the fundamental question of different MAP thresholds.

Furthermore, while there was no significant statistical heterogeneity between the included studies, there was clear clinical heterogeneity between the included study populations. The small effect sizes and low study numbers may have led to a low measured statistical heterogeneity, so may not accurately reflect the heterogeneity between studies.

As shown with the RoB-2 tool, all except one19 RCT had some degree of risk of bias. There was lack of blinding in six of eight included trials, 5,13,14,15,16,18 which contributed to the risk of bias. The lack of blinding is important as it could have inadvertently led to the two groups being managed differently by their respective care teams. The BOX trial remained double-blinded by using a novel technique where the reported blood pressure threshold was 70 mm Hg, which was either 10% higher or lower depending on the treatment assignment.19 Thus, when patients in the high-MAP arm had a MAP of 70 mm Hg, the actual MAP was 77 mm Hg and when patients in the low-MAP arm had a MAP of 70 mm Hg, the actual MAP was 63 mm Hg. The trial by Grand et al. used a similar method to maintain blinding.17

None of the RCTs included in our systematic review showed an association with either neurologic outcome or need for renal replacement therapy. In the SEPSISPAM trial, there was effect measure modification between MAP assignment and chronic arterial hypertension on the need for renal replacement therapy (P interaction = 0.04).13 This result was further observed in a recent meta-analysis that observed a decreased risk of renal replacement therapy in patients with shock and premorbid hypertension (RR, 1.20; 95% CI, 1.03 to 1.41).23 Compared with our meta-analysis, this study included patients undergoing cardiac surgery. Unfortunately, because of differences in patient populations with the resultant fewer included studies, we were unable to examine for this important subgroup.

Conclusion

This updated systematic review and meta-analysis found no significant differences in mortality, favourable neurologic outcome, or rates of renal replacement therapy among critically ill patients treated with high–normal compared with low–normal MAP targets. Despite limited between-study heterogeneity, there was still risk of bias in the included studies. This study does not support a high–normal MAP threshold in critically ill patients. Future studies to help delineate either individualized MAP strategies, or important subgroups are warranted.