The inclusion criteria were as follows:

The exclusion criteria were as follows:

Search results from all databases were collated and deduplicated using reference management software. Two independent reviewers screened titles and abstracts against the prespecified eligibility criteria. Studies judged potentially eligible at the abstract screening stage were retrieved as full-text articles and subjected to detailed eligibility assessment. Disagreements between reviewers at any stage of screening were resolved through discussion and, where necessary, by referral to a third reviewer. The reasons for exclusion at the full-text stage were documented and are summarized in the PRISMA flow diagram. Pettilä [13] et al (TARTARE-2S, 2026) was excluded as it randomized a composite tissue-perfusion-guided protocol (capillary refill, skin mottling, temperature, lactate) rather than a fixed or isolated MAP target, with MAP serving only as a subordinate safety limit.

Data were extracted independently by 2 reviewers using a standardized, prepiloted extraction form, with disagreements resolved through discussion and, when necessary, consultation with a third reviewer. Extracted variables included study characteristics (design, setting, and sample size), participant demographics and baseline clinical characteristics, relevant comorbidities (eg, chronic hypertension, chronic kidney disease, and cirrhosis), MAP intervention targets or observational exposure measures, vasopressor type and dose, and reported outcomes. Primary outcomes comprised all-cause mortality at the reported follow-up time points and splanchnic perfusion parameters. Secondary outcomes included AKI, RRT, serum creatinine, urine output, hemodynamic variables, vasopressor requirements, organ support–free days, SOFA scores, length of stay, and adverse events. Where applicable, subgroup analyses reported within individual studies were also extracted. Data presented only in graphical format were estimated from figures, and unpublished data were not requested from study investigators.

Methodological quality was assessed independently by 2 reviewers using validated tools according to study design, with disagreements resolved by consensus. RCTs were evaluated using the Cochrane Risk of Bias 2 (RoB 2) tool, which assesses bias arising from the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of the reported results [14]. Nonrandomized studies were assessed using the Newcastle-Ottawa Scale (NOS), which evaluates study quality across the domains of selection, comparability, and outcome assessment, with a maximum score of 9 points. Overall study quality was categorized according to the respective assessment tools [15].

The remaining nonrandomized study involving 2 actively assigned MAP-targeted groups was assessed using the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool, evaluating bias due to confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, outcome measurement, and selection of the reported result [16]. The single-arm, within-subject titration study lacking a comparator group was assessed using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Quasi-Experimental Studies, a 9-item tool suited to pre-post designs without a control arm [17]. Overall quality was categorized per the respective tool applied, and studies were synthesized with attention to how these quality differences might affect confidence in the findings.

Given the substantial clinical and methodological heterogeneity among the included studies, the findings were synthesized narratively in accordance with the Synthesis Without Meta-Analysis reporting guideline [18]. Studies were grouped according to study design and outcome domains, with evidence from RCTs presented before nonrandomized studies. Quantitative meta-analysis was performed only for all-cause mortality because sufficient comparable interventional data were available. Separate random-effects meta-analyses were conducted for 90-day mortality and 28-day mortality, with an additional sensitivity analysis incorporating studies reporting analogous 30-day mortality. Pooled effect estimates were expressed as odds ratios (ORs) with 95% CIs, and statistical heterogeneity was assessed using Cochran Q test and the I² statistic. Meta-analysis was not undertaken for renal, hemodynamic, organ-support, or adverse event outcomes because of a lack of reported statistics, differences in study populations, MAP targets, intervention protocols, follow-up durations, and outcome definitions. Subgroup findings reported within individual studies were synthesized narratively but were not quantitatively pooled because they were primarily exploratory or post hoc analyses.

The overall certainty of the evidence for the primary outcomes was assessed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework [19]. Evidence from RCTs was initially considered of high certainty and downgraded, where appropriate, for risk of bias, inconsistency, indirectness, imprecision, and publication bias. Evidence from nonrandomized studies was initially considered of low certainty and was upgraded only when justified by large effect estimates, evidence of a dose-response relationship, or when residual confounding was unlikely to explain the observed findings. Publication bias was assessed qualitatively given the small number of studies (n=3‐4) available for each pooled outcome, which precluded the reliable use of funnel plot asymmetry testing or Egger regression; this qualitative assessment considered the comprehensiveness of the search strategy, inclusion of published literature identified through multiple databases and manual reference screening, and comparison of effect directions and magnitudes across studies of differing design and setting, with no evidence suggestive of publication bias (eg, selective reporting or a disproportionate predominance of positive findings) identified. The overall certainty of evidence for each primary outcome was classified as high, moderate, low, or very low and is reported alongside the corresponding results.

Eleven studies (2895 patients) examined MAP targets in septic shock, including 4 RCTs and 7 nonrandomized studies (including prospective and retrospective observational cohorts, a post hoc analysis, and a physiological crossover study), with sample sizes ranging from 10 to 776. Populations were generally older adults with high illness severity (Acute Physiology and Chronic Health Evaluation II [APACHE II] 15‐29, Simplified Acute Physiology Score II [SAPS II] 38‐58, SOFA 7‐12.5), although age ranged from younger patients with cirrhosis (~45 y; Maiwall et al [20]) to an older adults-only cohort (≥65 y). Four studies compared fixed high- versus low-MAP targets (typically 80‐85 mm Hg vs 65‐70 mm Hg, except TARGET-C, which compared 80‐85 mm Hg vs 60‐65 mm Hg), whereas the remaining studies evaluated MAP as a continuous exposure or through individualized targets rather than fixed intervention arms. Comorbidities varied across cohorts, with chronic hypertension reported in 9% to 53% of patients, alongside chronic kidney disease, diabetes, coronary artery disease, and, in 1 RCT, cirrhosis with AKI (Table 1).

Among the 4 RCTs, 2 studies were judged to have a low overall risk of bias [24,25], whereas Bourgoin et al [21] and Maiwall et al [20] were rated as having some concerns. The main limitations were related to deviations from intended interventions, missing outcome data, and outcome measurement in the open-label trials (Figure 2).

The methodological quality of the 7 nonrandomized studies ranged from very low to good. NOS scores varied between 2 and 8, with Poukkanen et al [23] demonstrating the highest quality (8/9). Badin et al [22], Varpula et al [29], Dünser et al [27], and Dünser et al [30] achieved moderate-to-good quality (7/9; Table 2).

Of the 9 JBI items, 7 were satisfied. The 2 unmet criteria, absence of a control group (item 4) and lack of similar cotreatment across comparisons (item 3, driven by unadjusted concurrent dopamine or dobutamine use in over half the cohort), represent the principal sources of bias in this study and preclude a “low risk of bias” classification despite its otherwise rigorous, objectively measured, complete-follow-up design. The study is therefore judged to carry a moderate-to-high risk of bias, consistent with its role in this review as hypothesis-generating, physiological or mechanistic evidence rather than a source of causal effect estimates [26] (Table 3).

Kuragayala et al [28] scored 5 “Serious,” 1 “Moderate,” and 1 “Critical” domain on ROBINS-I, yielding an overall critical risk of bias due to the absence of any reported effect estimates, so the study is retained for descriptive purposes only (Figure 3).

The certainty of evidence for both mortality and renal outcomes was rated as very low (Table 4). Evidence was downgraded because of methodological limitations, substantial inconsistency in the direction of treatment effects across studies, indirectness related to heterogeneous populations and MAP targets, and imprecision arising from several small studies. Consequently, the overall confidence in the effect estimates remains limited.

Three trials reported 90-day all-cause mortality comparing a higher versus a lower or standard MAP target and were pooled using a random-effects (Mantel-Haenszel) model [20,24,25] (Figure 4). Across 1442 patients, mortality occurred in 320 of 720 (44.4%) patients in the high-MAP arm versus 280 of 722 (38.8%) in the low-MAP arm, yielding a pooled OR of 1.31 (95% CI 0.97-1.77; z score=1.77; P=.08), a nonsignificant trend toward higher mortality with higher MAP targets. Moderate heterogeneity was present (τ²=0.03; χ²₂=3.35, I²=40%; P=.19). The effect was driven mainly by Endo et al [24], whose Optimal Blood Pressure in Septic Shock (OPTPRESS) trial was stopped early for harm (OR 1.62, 95% CI 1.12-2.34) and, to a lesser extent, by Maiwall et al [20] (OR 1.48, 95% CI 0.77-2.86), while Asfar et al [25] showed an effect close to unity (OR 1.07, 95% CI 0.80-1.42; Figure 4).

Two trials directly reported 28-day all-cause mortality by MAP-target arm [24,25]. Because Kuragayala et al [28] reported an analogous 30-day mortality comparison, a sensitivity analysis was performed with this study both added to and removed from the pooled estimate.

With Kuragayala et al [28] included (Figure 2), pooled 28- or 30-day mortality across 3 studies (695 high-MAP, 697 low-MAP patients) was 222 of 695 (32.0%) versus 191 of 697 (27.4%), giving a pooled OR of 1.25 (95% CI 0.98-1.61; z score=1.78 P=.08), with low heterogeneity (χ²₂=2.14, I²=7%; P=.34; Figure 5).

With Kuragayala et al [28] excluded (Figure 3), the pooled estimate from Asfar et al [25] and Endo et al [24] alone (645 patients with high-MAP, 647 patients with low-MAP) was OR of 1.31 (95% CI 0.90-1.89; z score=1.43; P=.15), with heterogeneity rising to I² of 51% (χ²₁=2.04; P=.15; Figure 6).

Adding Kuragayala et al [28] reduced between-study heterogeneity substantially (I² from 51% to 7%) without materially changing the point estimate or altering statistical significance, indicating the short-term mortality result was robust to this study’s inclusion. Neither the 28- or 30-day nor the 90-day pooled estimate reached statistical significance, though both trended toward higher mortality with higher MAP targets.

Several included studies reported mortality descriptively or as a regression outcome rather than as a direct between-arm comparison and could not be pooled. Bourgoin et al [21] did not assess mortality, as the trial’s 8-hour observation window was designed only for physiologic and renal endpoints. Badin et al [22] reported ICU mortality of 35% and hospital mortality of 39% across the whole cohort without a formal statistical comparison by MAP exposure. In the Finnish Acute Kidney Injury Study (FINNAKI) substudy, Poukkanen et al [23] found that 90-day mortality was significantly higher in patients whose AKI progressed (32.7% vs 18.9%; P=.001), though this reflected an AKI-progression subgroup rather than a randomized MAP-target comparison. In the post hoc 546C88 cohort, Dünser et al found no significant association between MAP quartile and 28-day mortality after adjustment (risk ratio [RR] 0.99, 95% CI 0.95-1.04, P=.82), although patients excluded for failing to sustain MAP ≥70 mm Hg had markedly higher 28-day mortality (76.5%) than the included cohort (42.1%). Conversely, Dünser et al [27] found that MAP was among the strongest independent predictors of 28-day mortality in both 6-hour (P=.001) and 48-hour (P=.01) logistic regression models, with the largest area under the receiver operating characteristic curve (AUC) for time spent below MAP 60 mm Hg (AUC=0.779) and no significant difference in discriminative performance between MAP thresholds of 60 and 75 mm Hg [30]. Similarly, Varpula et al [29] found that each 1 mm Hg decrease in mean 48-hour MAP was independently associated with an approximate 16% increase in the odds of death (Exp[B]=1.156; P=.01).

Renal outcomes were reported with heterogeneous definitions across 7 studies, precluding meta-analytic pooling. Bourgoin et al found that acutely raising MAP from 65 to 85 mm Hg over 4 hours did not change urine flow, serum creatinine, or creatinine clearance [21]. Badin et al [22] identified an MAP range of 72 to 82 mm Hg as the best discriminator of AKI progression at 72 hours, but only in patients with septic shock with early (6-h) renal impairment (AUC=0.83‐0.84); no MAP-AKI association was found in patients without early AKI. In the larger FINNAKI substudy, Poukkanen et al [23] reported that a time-adjusted MAP below 73 mm Hg was independently associated with AKI progression after adjustment for chronic kidney disease, lactate, furosemide dose, and dobutamine use (adjusted OR 2.57, 95% CI 1.48-4.46). In Sepsis and Mean Arterial Pressure (SEPSISPAM), Asfar et al [25] found no overall between-group difference in creatinine doubling or RRT, but among patients with chronic hypertension, the high-MAP group had significantly less RRT use (31.7% vs 42.2%; P=.046) and less creatinine doubling (38.9% vs 52.0%; P=.02). In patients with cirrhosis with septic shock and AKI, Maiwall et al [20] found AKI reversal at day 5 was significantly greater with the high-MAP target in the per-protocol analysis (53% vs 31%; P=.02), and a high-MAP strategy independently predicted AKI recovery (OR 2.36‐2.47). Kuragayala et al [28] reported similar, purely descriptive rates of creatinine doubling (82% vs 84%) and RRT (66% vs 70%) between arms, without inferential statistics. Finally, Dünser et al [30] found that MAP and mean perfusion pressure were both significantly correlated with RRT need (r²=0.124‐0.136; P≤.001) and peak creatinine (r²=0.269‐0.277), with MAP <75 mm Hg providing the best discrimination for subsequent RRT need (AUC=0.724).

A consistent, dose-dependent relationship between MAP target and vasopressor exposure was observed across studies. In a within-patient crossover from 65 to 75 mm Hg to 85 mm Hg, LeDoux et al [26] demonstrated a stepwise linear increase in heart rate, cardiac index, left ventricular stroke work index, systemic vascular resistance index, and norepinephrine dose (from 23 to 31 µg/min to 47 µg/min; P=.02). Bourgoin et al similarly found that raising MAP from 65 to 85 mm Hg significantly increased cardiac index and systemic vascular resistance index without changing lactate or oxygen consumption [21]. Among the RCTs, higher MAP targets consistently required substantially greater vasopressor exposure: Asfar et al [25] reported higher day-1 norepinephrine dose (0.58 vs 0.45 µg/kg/min; P<.001) and longer catecholamine infusion (4.7 vs 3.7 days; P<.001) in the high-MAP arm, Endo et al [24] reported longer vasopressor duration (65 vs 42 h) and higher cumulative norepinephrine (15.6 vs 9.6 mg) and vasopressin doses in the high-target arm, and Kuragayala et al [28] noted the low-MAP arm undershot its 65- to 70-mm Hg target (achieved MAP 57.3±13.7 mm Hg), while the high-MAP arm approximated its 80- to 85-mm Hg target (78.5±17.2 mm Hg).

Endo et al [24] reported significantly fewer RRT-free days (18 vs 20; P=.02), ventilator-free days (15 vs 18; P=.012), and catecholamine-free days (16 vs 19; P=.01) at 28 days with the high-target strategy, indicating a greater overall organ-support burden. In contrast, Asfar et al [25] found no significant difference in 28-day survival without organ support between arms (60.6% vs 62.1%; P=.66), and Maiwall et al [20] found no significant intention-to-treat differences in ICU length of stay or mechanical ventilation duration, although both trended longer in the high-MAP arm. In the FINNAKI substudy, patients with AKI progression had significantly longer ICU stays than those without (5.7 vs 3.8 d; P<.001) [23].

None of the included studies directly reported splanchnic perfusion or ischemia as a discrete primary outcome; the closest available proxy was mesenteric ischemia as an adverse event, which occurred at similar, low rates between arms in both Asfar et al [25] (2.3% vs 2.3%) and Maiwall et al [20] (1.3% high-MAP vs 0% low-MAP). Cardiac arrhythmias were more frequent with higher MAP targets in several studies: Asfar et al [25] reported significantly more atrial fibrillation with high MAP (6.7% vs 2.8%, P=.02), and Endo et al [24] noted that all main safety outcomes—arrhythmia, ischemic events, and hemorrhagic events—trended more frequent in the high-target group, although none reached statistical significance individually. Maiwall et al [20] found significantly more adverse events requiring protocol discontinuation with high MAP (24% vs 11%; P=.03), but significantly less intradialytic hypotension (12% vs 48%; P<.001), suggesting a trade-off between systemic vasopressor-related harm and protection from dialysis-associated hypotension at higher MAP targets.