Skip to main content

Trial Summaries & Critiques

IMPROVE

Image: Shutterstock / Gorodenkoff

Publication

  • Title: Individualized Perioperative Blood Pressure Management in Patients Having Major Abdominal Surgery: The IMPROVE-multi Randomized Clinical Trial
  • Acronym: IMPROVE-multi
  • Year: 2025
  • Journal published in: JAMA
  • Citation: Saugel B, Meidert AS, Brunkhorst FM, et al; for the IMPROVE-multi Trial Group. Individualized perioperative blood pressure management in patients having major abdominal surgery: the IMPROVE-multi randomized clinical trial. JAMA. 2025;334(21):1893-1904.

Context & Rationale

  • Background
    • Intraoperative hypotension is frequent during major non-cardiac surgery and is associated with postoperative organ injury (notably acute kidney injury and myocardial injury) and mortality in observational cohorts.
    • Most perioperative practice targets a minimum mean arterial pressure (MAP) of approximately 65 mm Hg, but the optimal “safe” MAP likely varies by patient physiology (eg, chronic hypertension, vascular stiffness, autonomic dysfunction) and by surgical/anaesthetic phase.
    • Randomised evidence has shown that interventions can reduce hypotension exposure, yet the translation into improved patient-centred outcomes has been inconsistent, raising uncertainty about whether MAP is a causal mediator, a marker, or an incomplete surrogate for perfusion.
    • Preoperative nighttime blood pressure (BP) may approximate an individual’s lower physiological set-point; using this to individualise intraoperative targets is plausible but required rigorous clinical testing.
  • Research Question/Hypothesis
    • Does targeting perioperative MAP to each patient’s preoperative mean nighttime MAP (rather than a fixed MAP ≥65 mm Hg) reduce early postoperative organ injury and major adverse events in high-risk major abdominal surgery?
    • Hypothesis: maintaining MAP at/above an individualised nocturnal baseline would reduce a composite of acute kidney injury, acute myocardial injury, non-fatal cardiac arrest, or death within 7 days after surgery.
  • Why This Matters
    • Postoperative acute kidney injury and myocardial injury are common after major abdominal surgery and are strongly linked to ICU utilisation, prolonged hospitalisation, and downstream morbidity and mortality.
    • If a practical approach (overnight automated BP monitoring) could define safer haemodynamic targets, it would be highly scalable across perioperative and critical care pathways.
    • A neutral (or harmful) result would challenge MAP-centred management as a stand-alone strategy and redirect research towards more mechanistic and perfusion-centred approaches.

Design & Methods

  • Research Question: In high-risk adults undergoing elective major abdominal surgery, does an individualised perioperative MAP target based on each patient’s preoperative mean nighttime MAP reduce a composite of acute kidney injury, acute myocardial injury, non-fatal cardiac arrest, or death within 7 days compared with routine management targeting MAP ≥65 mm Hg?
  • Study Type: Randomised, multicentre, investigator-initiated, single-blind, parallel-group, superiority clinical trial; 15 German university hospitals; allocation 1:1 stratified by site using variable block sizes; trial protocol published separately.1
  • Population:
    • Setting: elective major abdominal surgery under general anaesthesia; intervention window from induction until 2 hours after the end of surgery; follow-up to 90 days.
    • Inclusion: age ≥45 years; elective major abdominal surgery expected duration ≥90 minutes; ≥1 pre-specified high-risk criterion (eg, limited functional capacity <4 METs; kidney impairment; immunodeficiency; cardiovascular/cerebrovascular/peripheral arterial disease; chronic pulmonary disease; diabetes; cirrhosis; malignancy; smoking; obesity; age ≥65 years).
    • Exclusion: emergency surgery; nephrectomy; liver/kidney transplantation or prior organ transplantation; sepsis; pregnancy; requirement for kidney replacement therapy; ASA physical status V/VI; inability to complete preoperative automated BP monitoring or inadequate nighttime readings for target calculation.
  • Intervention:
    • Preoperative BP phenotyping: automated BP monitoring at 30-minute intervals for one night (hospital or home); mean nighttime MAP calculated from values between 00:00 and 06:00 after artefact filtering.
    • Target: maintain perioperative MAP ≥ preoperative mean nighttime MAP; if mean nighttime MAP <65 mm Hg, target ≥65 mm Hg; if >110 mm Hg, target ≥110 mm Hg (upper cap chosen to avoid excessive vasoactive exposure).
    • Delivery: treating clinicians used usual therapies (fluids, vasopressors, inotropes) at their discretion to achieve target (no mandated haemodynamic algorithm); intervention period from induction to 2 hours post-surgery.
  • Comparison:
    • Target: routine perioperative BP management aiming for MAP ≥65 mm Hg during the same induction-to-2-hours-postoperative window (reflecting common clinical thresholds and consensus statements).2
    • Contamination control: preoperative nighttime MAP was measured in both groups but concealed from clinicians managing patients in the routine group.
  • Blinding: Single-blind; anaesthesiologists could not be blinded to the target, but the protocol specified blinding of participants, outcome adjudicators, and data analysts; primary outcomes were largely objective laboratory- and event-based endpoints.1
  • Statistics: A total of 1144 patients were required to detect an 8% absolute reduction in the primary composite outcome (from 25% to 17%) with 90% power at a two-sided 5% significance level (with a prespecified interim analysis); planned enrolment was 1272 to allow for attrition/non-randomisation after preoperative monitoring; primary analysis was modified intention-to-treat (randomised patients who underwent surgery), using a 2-sided test of proportions with continuity correction; sensitivity analysis used mixed-effects logistic regression (random intercept for site); per-protocol analysis excluded major protocol deviations.
  • Follow-Up Period: Primary outcome assessed through postoperative day 7; key secondary outcomes through 30 and 90 days after surgery.

Key Results

This trial was not stopped early. A prespecified interim analysis was performed and recruitment continued to completion.

Outcome Individualised MAP target Routine MAP ≥65 mm Hg Effect p value / 95% CI Notes
Primary composite (AKI, acute myocardial injury, non-fatal cardiac arrest, or death) within 7 days 190/567 (33.5%) 173/567 (30.5%) RR 1.10 95% CI 0.93 to 1.30; P=0.31 Primary endpoint
Acute kidney injury within 7 days 72/566 (12.7%) 58/567 (10.2%) RR 1.24 95% CI 0.90 to 1.72; P=0.22 Component reported as secondary outcome
Acute myocardial injury within 7 days 153/558 (27.4%) 138/561 (24.6%) RR 1.11 95% CI 0.91 to 1.36; P=0.31 Component reported as secondary outcome
Death within 7 days 0/567 (0.0%) 4/567 (0.7%) RR 0.00 95% CI 0.00 to 2.36; P=0.13 Low event rate; wide CI
Composite infectious complications within 7 days 90/567 (15.9%) 97/567 (17.1%) RR 0.93 95% CI 0.72 to 1.21; P=0.63 Secondary outcome; standardised infection endpoints
Composite (kidney replacement therapy, myocardial infarction, non-fatal cardiac arrest, or death) within 30 days 21/566 (3.7%) 13/567 (2.3%) RR 1.62 95% CI 0.82 to 3.20; P=0.22 Secondary composite; low event rate
Composite (kidney replacement therapy, myocardial infarction, non-fatal cardiac arrest, or death) within 90 days 32/566 (5.7%) 20/567 (3.5%) RR 1.60 95% CI 0.93 to 2.77; P=0.12 Secondary composite
Myocardial infarction within 90 days 10/566 (1.8%) 3/567 (0.5%) RR 3.34 95% CI 0.92 to 12.07; P=0.09 Numerical imbalance; not statistically significant
Death within 90 days 16/567 (2.8%) 16/567 (2.8%) RR 1.00 95% CI 0.51 to 1.98; P>0.99 Secondary outcome
Unplanned hospital readmission within 30 days 53/562 (9.4%) 52/562 (9.3%) RR 1.02 95% CI 0.71 to 1.47; P>0.99 Secondary outcome
  • Separation achieved, but no benefit: hypotension exposure was reduced (eg, area under MAP <65 mm Hg: 6 [0–54] vs 48 [7–151] mm Hg×min; time-weighted average MAP <65 mm Hg: 0.0 [0.0–0.2] vs 0.5 [0.1–1.9] mm Hg), yet the primary composite outcome was not reduced.
  • More vasoactive exposure: norepinephrine was used in 536/567 (94.5%) vs 495/567 (87.3%), with higher total norepinephrine dose (1.53 [0.71–2.93] vs 1.25 [0.52–2.48] mg).
  • Potential harm signal requires caution: myocardial infarction at 90 days was numerically higher in the individualised group (10/566 vs 3/567; RR 3.34; 95% CI 0.92 to 12.07; P=0.09), with wide confidence intervals.

Internal Validity

  • Randomisation and allocation concealment: central, web-based 1:1 randomisation stratified by site with variable block sizes; preoperative nighttime MAP values were concealed from clinicians in the routine group, supporting separation of the intended exposure.
  • Dropout/exclusions: 1142 randomised; 8 excluded after randomisation (surgery cancelled or consent withdrawn), leaving 1134 (567 per group) for the primary analysis; low post-randomisation attrition minimises selection bias.
  • Performance/detection bias: anaesthesiologists could not be blinded; protocol specified blinding of participants, outcome adjudicators and analysts, and primary endpoints were objective (serum creatinine/troponin, cardiac arrest, death).1
  • Protocol adherence: individualised group achieved MAP ≥ target for median 79% (IQR 62–91) of the intervention period vs 57% (28–81) in the routine group; this indicates incomplete attainment but clinically meaningful separation.
  • Baseline characteristics: groups were broadly comparable, but chronic arterial hypertension was more frequent in the individualised group (307/567 [54.1%] vs 265/567 [46.7%]) with more antihypertensive use (eg, β-blockers 30.3% vs 24.3%); this imbalance could bias outcomes against the intervention.
  • Separation of the variable of interest (process fidelity): area under MAP <65 mm Hg: 6 (0–54) vs 48 (7–151) mm Hg×min; time-weighted average MAP <65 mm Hg: 0.0 (0.0–0.2) vs 0.5 (0.1–1.9) mm Hg; norepinephrine use: 94.5% vs 87.3%; total norepinephrine: 1.53 (0.71–2.93) vs 1.25 (0.52–2.48) mg.
  • Timing and dose: preoperative BP monitoring occurred a median of 1 day (IQR 1–2) before surgery; the intervention covered induction through 2 hours postoperatively; the median preoperative mean nighttime MAP in the individualised group was 84 (77–92) mm Hg (implying many targets >65, and commonly ≥80 mm Hg), while targets were capped below 110 mm Hg.
  • Major protocol deviations: 8 deviations in the individualised group (7 incorrect nighttime MAP calculations; 1 intervention stopped); per-protocol population was 1126 (559 assigned to individualised management), and conclusions were consistent across analyses.
  • Outcome assessment and missing data handling: outcomes were largely laboratory- and event-driven; assumptions that missing postoperative creatinine/troponin values indicated absence of injury could bias toward the null if missingness was related to illness severity (impact depends on completeness of scheduled sampling).
  • Statistical rigour: prespecified sample size and interim analysis plan were used; confidence intervals were presented for effect estimates; sensitivity mixed-effects models and per-protocol analyses did not materially change results.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong, supported by concealed allocation, minimal post-randomisation attrition, objective endpoints, and demonstrable separation in hypotension exposure; limitations include pragmatic (non-algorithmic) delivery, overlap in achieved MAP between groups, and baseline hypertension imbalance.

External Validity

  • Population representativeness: high-risk adults (median age 67 years; ~34% women; ASA III/IV ~97%) undergoing elective major abdominal surgery in tertiary German centres; postoperative ICU admission occurred in ~48%, aligning with many high-resource perioperative pathways.
  • Important exclusions: emergency surgery, transplant-related surgery, sepsis, and patients on kidney replacement therapy were excluded; applicability to lower-risk surgery or emergency/physiologically unstable populations is uncertain.
  • Feasibility and implementation: intervention requires preoperative overnight automated BP monitoring and data processing; this may be challenging where surgical timelines are short or outpatient prehabilitation infrastructure is limited.
  • Comparator relevance: routine care targeted MAP ≥65 mm Hg, which is widely used; however, hypotension exposure in routine care was modest (time-weighted average MAP <65 mm Hg: 0.5 [0.1–1.9] mm Hg), so the findings may not translate to settings with more frequent/prolonged hypotension.
  • Generalisability across surgical domains: results apply most directly to major abdominal surgery; effects may differ in vascular/cardiac surgery, in profound haemorrhage, or where haemodynamic algorithms include cardiac output/oxygen delivery monitoring.

Conclusion on External Validity: External validity is moderate: findings are generalisable to high-risk elective major abdominal surgery in high-resource systems using MAP ≥65 mm Hg as standard care, but less applicable to emergency surgery, other surgical specialties, and settings unable to implement overnight BP phenotyping.

Strengths & Limitations

  • Strengths:
    • Large, multicentre, pragmatic RCT in a clearly defined high-risk surgical population.
    • Novel, scalable individualisation strategy (overnight automated BP monitoring) with concealment of nighttime MAP in the control group to limit contamination.
    • Objective, clinically relevant composite endpoint (AKI, myocardial injury, cardiac arrest, death) with follow-up to 90 days.
    • Clear process separation in hypotension exposure and vasoactive utilisation, enabling a fair test of the MAP-target hypothesis.
  • Limitations:
    • Single-blind design with unblinded treating clinicians; co-interventions (fluid/vasopressor choice) were not protocolised and could vary by clinician/site.
    • Overlap in achieved MAP distributions (routine group still spent substantial time above nighttime MAP), potentially diluting any true effect.
    • Baseline imbalance in chronic hypertension and antihypertensive use (higher in the individualised group) could confound effect estimates despite randomisation.
    • Nighttime MAP derived from one night of intermittent measurements may not represent intraoperative autoregulatory limits; the physiological “right target” may not be captured by this surrogate.
    • Harms and adverse events specifically attributable to vasopressors were not reported as a dedicated safety endpoint; interpretation relies on clinical outcome signals and secondary endpoints.

Interpretation & Why It Matters

  • Clinical practice
    • In high-risk elective major abdominal surgery, raising perioperative MAP targets to an individual’s preoperative nighttime baseline did not reduce early postoperative organ injury or 90-day major adverse outcomes, despite reducing hypotension exposure.
    • Routine escalation of MAP targets above ≥65 mm Hg solely on the basis of preoperative nocturnal BP is not supported by these results.
  • Mechanistic implications
    • MAP-centred strategies can change “pressure-time” exposure but may not improve organ outcomes if the causal pathway depends on flow, oxygen delivery, microcirculatory dysfunction, or vasoactive drug effects rather than MAP alone.
    • The observed increase in vasoactive exposure, without outcome benefit, highlights the need to consider benefit–harm trade-offs when pursuing higher targets.
  • Research direction
    • Future trials should test perfusion- and endotype-informed haemodynamic strategies (eg, integrating cardiac output/flow targets, lactate/microcirculatory surrogates, and hypotension aetiology) rather than relying on pressure targets alone.
    • Subgroup enrichment may be needed (eg, severe chronic hypertension, autonomic failure, marked vascular disease) if benefit exists only in specific physiological phenotypes.

Controversies & Other Evidence

  • Choice of individualisation metric (nighttime MAP): Correspondence highlighted that preoperative mean nighttime MAP is substantially lower than mean daytime MAP, and questioned whether nocturnal BP provides an appropriate surrogate for intra-anaesthetic organ perfusion targets; the correspondence also emphasised that defining “physiological” targets from intermittent nighttime measurements is vulnerable to individual dipping patterns and context effects (hospital sleep disruption, stress).5
  • MAP is not perfusion: Correspondence argued that achieving a numeric MAP target with vasopressors does not ensure adequate organ perfusion, because hypotension can arise from heterogeneous haemodynamic endotypes (vasodilation, low stroke volume/cardiac output, mixed states). This aligns with emerging endotype-based modelling work and calls for physiology-informed, endotype-specific haemodynamic management rather than uniform pressure escalation.512
  • Vasopressor exposure and potential harm: IMPROVE-multi increased norepinephrine exposure and showed numerical imbalances in later cardiovascular outcomes (eg, myocardial infarction at 90 days). Observational perioperative analyses have linked higher intraoperative norepinephrine dose to acute kidney injury, supporting the need for explicit harm surveillance when pursuing higher MAP targets.11
  • Consistency with contemporaneous RCT evidence: PRETREAT, published alongside IMPROVE-multi, tested proactive (risk-score guided) treatment of hypotension versus reactive management; despite reducing hypotension exposure, it did not improve 6-month functional disability and was stopped for futility, supporting the broader inference that simply raising/maintaining pressure targets may not translate into better patient-centred outcomes.34
  • Relation to earlier positive signals: INPRESS suggested that individualised BP management reduced postoperative organ dysfunction in high-risk major surgery; however, correspondence and subsequent discussion have emphasised that differences in vasoactive drug strategy (notably norepinephrine infusion and less ephedrine) may contribute, complicating attribution to the BP target alone. IMPROVE-multi provides a larger multicentre test focused on the MAP target hypothesis and did not replicate benefit.65
  • Broader evidence base: POISE-3’s blood pressure strategy comparison and recent systematic review/meta-analysis of low versus high BP targets across critically ill and surgical populations have not shown consistent improvements in hard outcomes, despite effects on blood pressure exposure—reinforcing the concept that BP manipulation alone may be insufficient.78
  • Guideline/consensus implications: Contemporary perioperative consensus statements emphasise avoiding sustained MAP <60–65 mm Hg while acknowledging uncertainty about routine higher targets; updated perioperative BP guidance continues to recommend improved BP measurement and risk stratification across the perioperative period, but definitive endorsement of higher intraoperative targets for broad populations remains unsupported by outcome data.29
  • Further reading (selected):
    • Randomised trials: PRETREAT3; INPRESS6; POISE-3 blood pressure strategies7; personalised BP and neurocognitive outcomes (BJA Open)10.
    • Meta-analysis: Low vs high BP targets across ICU and surgical settings8.
    • Observational/mechanistic: intraoperative norepinephrine dose and AKI11; hypotension endotypes modelling12.
    • Guidelines/consensus: POQI XI perioperative arterial pressure statement2; Association of Anaesthetists/BIHS updated guideline (peri-operative BP measurement/management)9.

Summary

  • In 15 German tertiary centres, high-risk adults undergoing elective major abdominal surgery were randomised to an individualised perioperative MAP target based on preoperative mean nighttime MAP versus routine management targeting MAP ≥65 mm Hg.
  • The individualised strategy reduced hypotension exposure (area under MAP <65: 6 vs 48 mm Hg×min) and increased norepinephrine exposure (94.5% vs 87.3%; total 1.53 vs 1.25 mg).
  • The primary composite of AKI, acute myocardial injury, non-fatal cardiac arrest, or death within 7 days was not reduced (33.5% vs 30.5%; RR 1.10; 95% CI 0.93 to 1.30; P=0.31).
  • No statistically significant differences were seen in 30- or 90-day composites, mortality, readmissions, or infectious complications; myocardial infarction at 90 days was numerically higher with individualised targeting (1.8% vs 0.5%; RR 3.34; 95% CI 0.92 to 12.07; P=0.09).
  • These results argue against routine escalation of MAP targets to an individual’s nocturnal baseline as a stand-alone strategy to prevent postoperative organ injury in this population.

Overall Takeaway

IMPROVE-multi is a landmark pragmatic RCT because it rigorously tested a physiologically motivated, scalable “individualised MAP target” strategy and demonstrated that achieving better MAP metrics does not necessarily improve clinically important postoperative outcomes. In high-risk elective major abdominal surgery, targeting MAP to preoperative nighttime MAP increased vasoactive exposure and reduced hypotension burden, yet did not reduce early organ injury or longer-term adverse events, reinforcing the need to move beyond MAP-only haemodynamic paradigms.

Overall Summary

  • Individualising perioperative MAP targets to preoperative nighttime MAP reduced hypotension exposure and increased norepinephrine use, but did not improve early postoperative organ injury or 90-day outcomes after major abdominal surgery.

Bibliography