Ultra-early versus early adjunctive vasopressin initiation after norepinephrine escalation in septic shock: a target trial emulation

Ultra-early versus early adjunctive vasopressin initiation after norepinephrine escalation in septic shock: a target trial emulation | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Ultra-early versus early adjunctive vasopressin initiation after norepinephrine escalation in septic shock: a target trial emulation Takaya Nakashima, Taiga Ichinomiya, Mikio Nakajima, Tomohiro Shinozaki, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9454619/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 5 You are reading this latest preprint version Abstract Purpose Whether earlier initiation of adjunctive vasopressin improves outcomes in patients with septic shock remains unclear. Previous studies have compared vasopressin use with non-use and defined time zero as shock onset rather than norepinephrine escalation, making it difficult to distinguish the effect of vasopressin addition from that of earlier timing. We therefore conducted a targeted trial emulation, defining time zero as norepinephrine escalation to ≥ 0.25 µg/kg/min to compare ultra-early (0–3 hours) with early (> 3–6 hours) vasopressin initiation. Methods We emulated a target trial using two large electronic health record databases (MIMIC-IV 2008–2022 and eICU-CRD 2014–2015). Adults with septic shock whose norepinephrine-equivalent dose reached ≥ 0.25 µg/kg/min were eligible. Time zero was the first time the norepinephrine threshold was met. The primary outcome was the 28-day mortality. We used the clone-censor-weight method to estimate cumulative incidence curves under the ultra-early and early initiation strategies and compared them via the weighted Cox model. Results Among 5,157 eligible patients, ultra-early initiation lowered 28-day mortality (35.7% versus 41.2%; risk difference − 5.5 percentage points, 95% confidence interval [CI] − 6.9 to − 4.4; restricted mean survival time difference + 1.49 days, 95% CI 1.18–1.82; hazard ratio [HR] 0.81, 95% CI 0.78–0.85). Ultra-early initiation also showed lower rates of acute kidney injury (KDIGO stage ≥ 2: HR 0.94, 95% CI 0.92–0.97) and renal replacement therapy (HR 0.65, 0.61–0.70) including continuous renal replacement therapy (HR 0.75, 0.68–0.83). Conclusions Among patients with septic shock whose norepinephrine dose reached ≥ 0.25 µg/kg/min, ultra-early vasopressin initiation (0–3 h) was associated with lower 28-day mortality than early initiation (> 3–6 h). septic shock vasopressin norepinephrine treatment timing target trial emulation clone-censor-weight Figures Figure 1 Figure 2 Figure 3 Take-home message Among patients with septic shock whose norepinephrine-equivalent dose reached 0.25 µg/kg/min or higher, ultra-early vasopressin initiation (0–3 hours after reaching this threshold) was associated with lower 28-day mortality than early initiation (>3–6 hours). Timing of vasopressin initiation, beyond the decision to use vasopressin, may be a clinically meaningful lever once the high-dose range is reached, and warrants evaluation in a randomized trial. Introduction Vasopressin, a non-catecholaminergic vasopressor, has attracted increasing interest as a supportive therapy to reduce the adverse effects of high-dose norepinephrine in patients with septic shock [1, 2]. The Surviving Sepsis Campaign guidelines 2026 recommend norepinephrine as the first-line vasopressor and suggest adding vasopressin when arterial pressure remains inadequate at norepinephrine-equivalent doses of 0.25 to 0.5 µg/kg/min [3]. However, two large randomized trials [4, 5] and an individual patient data meta-analysis [6] found no overall survival benefit with vasopressin compared to norepinephrine alone in septic shock. More recently, several studies have examined whether earlier initiation of vasopressin improves outcomes [7–10] however, the findings have been inconsistent, and no clear consensus has emerged [11, 12]. Two methodological issues may explain this inconsistency. First, the start of follow-up was commonly defined as the onset of septic shock [13–15] rather than the point at which vasopressin became clinically relevant. Because vasopressin is typically initiated only after escalation to high-dose norepinephrine, the interval between shock onset and vasopressin initiation varies across patients, introducing confounding by time-varying severity during this gap [12]. Second, most previous studies compared early vasopressin combination therapy with norepinephrine alone [8, 9, 16], thereby primarily assessing the effect of vasopressin addition rather than the timing of its initiation. The clinically relevant question is how soon vasopressin should be initiated after norepinephrine reaches a high dose. To address this issue, we conducted a targeted trial emulation, focusing on the timing of vasopressin initiation. We defined time zero as the time at which the norepinephrine-equivalent dose reached 0.25 µg/kg/min or higher, the guideline-recommended threshold for considering vasopressin addition [3]. Using clone-censor-weight methods [17], we compared ultra-early initiation (0–3 hours after reaching this threshold) and early initiation (> 3–6 hours) in terms of 28-day mortality. Methods Study design and reporting This study followed a targeted trial emulation framework designed to compare two vasopressin initiation strategies using routinely collected clinical data. The reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline [18] and the Transparent Reporting of Observational Studies Emulating a Target Trial (TARGET) statement [19]. The STROBE and TARGET checklists are presented in Online Resource 1 . The components of the hypothetical target trial and their observational operationalizations are summarized in Online Resource 2 . Data sources This study used data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) [20] and the eICU Collaborative Research Database (eICU-CRD, version 2.0) [21], which are two large, publicly available critical care databases. MIMIC-IV contains de-identified data from over 65,000 intensive care unit (ICU) admissions at the Beth Israel Deaconess Medical Center (2008–2022), while eICU-CRD is a multi-center telemedicine database with more than 200,000 ICU admissions across the United States (2014–2015). The data from both databases were harmonized to an hourly common concept structure and pooled using the Ricu package [22]. The first author completed the required institutional training program (Record ID: 54439141) and was granted access to both databases. Because both databases contained only de-identified data, the requirement for institutional review board approval and individual informed consent was waived. Eligibility criteria We included adults (aged 18 years or older) 1) who met sepsis-3 criteria for septic shock [23], 2) whose norepinephrine-equivalent dose reached ≥ 0.25 µg/kg/min and 3) who had not received vasopressin before that point. Time zero The norepinephrine-equivalent dose (µg/kg/min) was calculated as norepinephrine + epinephrine + dopamine/150 + phenylephrine/10 + vasopressin × 2.5 (U/min), using the conversion factors from the ATHOS-3 trial [24] as implemented in the Ricu package [22]. We chose the lower bound of the guideline-recommended range (0.25–0.5 µg/kg/min) to capture the earliest point at which vasopressin addition is considered [3]. Treatment strategies The two strategies were as follows: (1) ultra-early vasopressin initiation (0–3 hours after time zero) and (2) early vasopressin initiation (> 3–6 hours after time zero). In the hypothetical target trial, eligible individuals are randomly assigned at time zero to one of the strategies. Outcomes The primary outcome was 28-day mortality. Secondary outcomes, defined as first events within 28 days, included acute kidney injury (Kidney Disease: Improving Global Outcomes [KDIGO] stage 2 or higher and stage 3 or higher), renal replacement therapy (RRT) initiation, continuous renal replacement therapy (CRRT) initiation, medically treated arrhythmia and net negative fluid balance. The definitions are provided in Online Resource 3 . Baseline characteristics and covariates Variables measured at or immediately before time zero included age, sex, body mass index, body weight, mean arterial pressure, lactate, Sequential Organ Failure Assessment (SOFA) score, creatinine, norepinephrine-equivalent dose, invasive mechanical ventilation, and systemic corticosteroid use within 24 hours before time zero. These variables also served as covariates in the censoring weight models, together with hourly time-varying values of mean arterial pressure, lactate, creatinine, norepinephrine-equivalent dose, SOFA score, ventilation status, concurrent vasopressor use (epinephrine, dopamine, dobutamine, and phenylephrine), the hourly deviation from the time-zero norepinephrine-equivalent dose, and its 1-hour change during the 0-to-6-hour assignment window. The covariates were pre-specified based on previous studies [2, 9, 16]. Missing data Missing data were assessed for each baseline covariate. The proportion of missing values and the imputation method for each variable are provided in Online Resource 4 . Statistical analysis Baseline characteristics are reported as absolute values with percentages for categorical variables and as medians with interquartile intervals for quantitative variables. The primary analysis used a clone-censor-weight approach to emulate survival under each initiation strategy, circumventing the time-related biases inherent in post hoc treatment timing comparisons [25, 26]. At time zero, each patient was duplicated such that one clone (or copy) was assigned to each strategy (cloning), mimicking a random assignment. Each clone was followed forward and artificially censored when the treatment deviated from the assigned strategy (censoring). Clones on ultra-early strategy were censored at 3 hours if vasopressin had not been started; clones on early strategy were censored if vasopressin was started before 3 hours or at 6 hours if not yet started. If a patient died or experienced an outcome event before being artificially censored for deviation from each strategy, the event was recorded at the time of its occurrence. Third, inverse probability of censoring (IPC) weights were estimated to adjust for the informative censoring introduced by strategy deviation. At each hour during the 0-to-6-hour assignment window, the probability of remaining uncensored was modeled using pooled logistic regression conditional on baseline and time-varying covariates (listed in the Covariates section above). The IPC weight for each interval was defined as the inverse of this predicted probability, and hourly weights were multiplied across time to obtain cumulative weights. Cumulative weights were truncated to the 99th percentile. Cumulative incidence curves for 28-day mortality were estimated using the IPC weighted Kaplan–Meier method. The 28-day risks were compared using the absolute risk difference and restricted mean survival time (RMST) difference, calculated as the area between the two curves, with 95% confidence intervals (CIs) from patient-level bootstrapping (500 iterations). Hazard ratios (HRs) were estimated using the IPC weighted Cox model, with robust standard errors clustered by patients. More details are provided in Online Resource 5 . A schematic of the analysis design, including the clone-censor-weight framework, is shown in Online Resource 6 . Secondary outcomes were analyzed using the IPC weighted Fine-Gray subdistribution hazard model, treating death before the event of interest as a competing risk. Exploratory subgroup analyses for the primary outcome were stratified by lactate at time zero (< 2, 2–<4, or ≥ 4 mmol/L), norepinephrine-equivalent dose (< 0.30 or ≥ 0.30 µg/kg/min), invasive mechanical ventilation (yes or no), and SOFA score (< 10, 10–<14, or ≥ 14). Sensitivity analyses included the following: (1) separate analyses within each database (MIMIC-IV and eICU-CRD); and (2) variation in the time-zero definition across norepinephrine thresholds (0.20, 0.25, 0.30, 0.35, 0.40, 0.45, and 0.50 µg/kg/min), sustainment durations (2 or 3 consecutive hours above the threshold), verification windows (2 or 3 hours after the time zero), and dose variables (norepinephrine-equivalent dose and norepinephrine infusion rate alone). All analyses were performed using R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). The analysis code and figure-generation scripts will be deposited in a public repository upon acceptance, and the repository URL will be added at the proof stage. Use of AI-assisted tools A large language model (Claude, Anthropic) was used to assist in drafting the manuscript and English language editing and formatting. All AI-generated content was reviewed, verified, and edited by the authors, who take full responsibility for the contents of the final manuscript. Results Cohort characteristics Figure 1 shows the cohort assembly. From 21,587 adults with septic shock in MIMIC-IV and eICU-CRD, 5,157 met the eligibility criteria at time zero (norepinephrine-equivalent dose ≥ 0.25 µg/kg/min, no prior vasopressin). All 5,157 patients were entered in the clone-censor-weight analysis, yielding 10,314 clones (one per strategy per patient). Under the ultra-early strategy, 943 (18.3%) adhered to the assigned strategy by initiating vasopressin within 0–3 hours; and under the early strategy, 323 (6.3%) adhered by initiating vasopressin within > 3–6 hours. The remaining clones were censored at strategy deviation and uncensored clones were weighted using IPC weights. The distribution of time from time zero to vasopressin initiation is shown in Online Resource 7 ; the 0-to-1-hour interval was the most common starting window. Table 1 presents baseline characteristics at time zero of all 5,157 patients who entered the clone-censor-weight analysis. The median patient age was 66 years, 55.2% were male, and the median SOFA score was 9. The median lactate was 3.3 mmol/L, and the median norepinephrine-equivalent dose was 0.3 µg/kg/min. Invasive mechanical ventilation was used in 36.9% of the patients, and 6.4% had received systemic corticosteroids before time zero. Crude 28-day mortality in the full eligible cohort was 40.7% (2,097 of 5,157 patients). The weight distributions ( Online Resource 8 ) were not dispersed after the 99th-percentile truncation. Table 1 Baseline characteristics at time zero Characteristic Overall (N = 5157) 1 Demographics Age, yr 66.0 (55.0, 77.0) Male sex 2,846 (55.2%) BMI, kg/m² 26.7 (22.8, 31.6) Body weight, kg 77.0 (63.7, 92.3) Hemodynamics and severity at time zero MAP, mmHg 68.0 (61.0, 76.5) Lactate, mmol/L 3.3 (1.9, 6.1) SOFA score 9.0 (6.0, 12.0) Creatinine, mg/dL 1.5 (0.9, 2.5) NE-equivalent dose, µg/kg/min 0.3 (0.3, 0.5) Invasive mechanical ventilation 1,902 (36.9%) Chronic kidney disease 924 (17.9%) Systemic corticosteroid use 332 (6.4%) 1 Median (IQR); n (%) Caption: Continuous variables are presented as median (interquartile range); categorical variables as n (%). Abbreviations: BMI, body mass index; MAP, mean arterial pressure; SOFA, Sequential Organ Failure Assessment; NE, norepinephrine; CKD, chronic kidney disease. Primary outcome Figure 2 shows weighted cumulative incidence curves for 28-day mortality, from which risk differences and RMST differences were derived. Ultra-early initiation showed lower 28-day mortality (35.7% versus 41.2%; risk difference − 5.5 percentage points, 95% CI − 6.9 to − 4.4; RMST difference + 1.49 days, 95% CI 1.18–1.82). Figure 3 A presents the HR of 0.81 (95% CI 0.78–0.85) in the primary analysis. The direction was preserved with 95th-percentile truncation (HR 0.83, 95% CI 0.80–0.86) and in the IPC weighted Cox model with additional outcome regression adjusting for baseline covariates (HR 0.96, 95% CI 0.93–1.00). Secondary outcomes The secondary outcome estimates are shown in Fig. 3 B. Ultra-early initiation showed lower rates of acute kidney injury (KDIGO stage 2 or higher: HR 0.94, 95% CI 0.92–0.97; stage 3 or higher: 0.92, 0.89–0.95), RRT initiation (0.65, 0.61–0.70), and CRRT initiation (0.75, 0.68–0.83). Medically treated arrhythmia (0.98, 0.93–1.03) and net negative fluid balance (0.97, 0.93–1.00) did not differ between groups. Subgroup analyses Exploratory subgroup analyses ( Online Resource 9 ) showed consistent results across all subgroups. The effect was more pronounced in patients with lactate ≥ 4 mmol/L (HR 0.77, 0.73–0.81) than in those with lactate < 2 mmol/L (HR 0.96, 0.86–1.07) or 2–<4 mmol/L (HR 0.95, 0.87–1.03). Results were consistent across norepinephrine-equivalent dose (< 0.30 µg/kg/min: HR 0.83, 0.78–0.89; ≥0.30: HR 0.81, 0.76–0.85), mechanical ventilation status (invasive: HR 0.82, 0.77–0.87; non-invasive: HR 0.82, 0.77–0.86), and SOFA score subgroups (< 10: HR 0.93, 0.88–0.98; 10–<14: HR 0.90, 0.84–0.96; ≥14: HR 0.85, 0.77–0.95). Sensitivity analyses Separate analyses within each database yielded consistent results (MIMIC-IV: HR 0.81, 0.77–0.86; eICU-CRD: HR 0.81, 0.77–0.87; Online Resource 10 ). Across the 56-alternative time-zero definitions, the primary estimate remained consistently below 1.0 ( Online Resource 11 ). Discussion In this target trial emulation of 5,157 patients with septic shock whose norepinephrine-equivalent dose reached 0.25 µg/kg/min or higher, ultra-early vasopressin initiation showed lower 28-day mortality than early vasopressin initiation. This direction was preserved in the series of sensitivity analyses. Ultra-early initiation also showed lower rates of acute kidney injury and renal replacement therapy including CRRT. A key feature of this analysis is that it examined not whether vasopressin should be used, but whether earlier initiation is beneficial once the norepinephrine dose reaches the high-dose range. Prior studies compared early vasopressin with norepinephrine alone [7–9]. By comparing two narrow initiation windows using clone-censor-weight methods, the present analysis isolated the effect of timing from the broader question of whether vasopressin should be used. As Online Resource 12 shows, the early group was not simply waiting; during the 3–6-hour window before vasopressin initiation, norepinephrine doses were escalated, and other interventions, such as phenylephrine, were administered. The present findings do not contradict the VASST or VANISH trials, which found no consistent overall survival benefit of vasopressin-containing strategies [4, 5] or the absence of mortality reduction in an individual patient data meta-analysis [6]. These trials evaluated whether vasopressin should be included across a broad population, which is a clinical question different from the timing comparison in the present study. In such designs, patients who initiated vasopressin early and those who initiated it late are pooled in the treatment arm; if the benefit is concentrated among early initiators, the average treatment effect across the entire arm may be diluted toward the null. By focusing on patients who had already reached a clinically plausible stage of vasopressin use, the present study was designed to detect a timing-specific signal. Moreover, by defining time zero as the point of norepinephrine escalation rather than shock onset, this study aligned the start of follow-up with the clinical stage at which vasopressin became relevant, addressing the time-lag limitations of prior studies. Confounding based on the indications is a central concern. Although the clone-censor-weight approach did not suffer from baseline confounding (at strategy assignment) by cloning each patient into both strategy arms at time zero, time-varying confounding by indication was translated into selection bias due to nonrandom censoring at strategy deviation. We try to correct this selection bias by applying the IPC weights under the assumption that all possible covariates affecting both vasopressin indication and outcomes were measured. In our analysis, the target trial emulation framework automatically excludes design bias such as immortal time bias, while it should be noted that this framework itself cannot account for confounding bias if uncontrolled confounding by indication was present. Several mechanisms may explain this association. The weighted norepinephrine trajectory over 72 hours ( Online Resource 13 ) showed that the ultra-early group experienced a rapid norepinephrine reduction after vasopressin addition, whereas the early group required continued catecholamine escalation during deferral. Total catecholamine exposure encompassing all vasopressor agents was lower in the ultra-early group (weighted median 53.6 versus 62.2 µg/kg; Online Resource 14 ), and vasopressor-free days were higher at 7 days (weighted mean difference 0.20 days, 95% CI 0.04–0.34) and 14 days (0.43, 0.11–0.71). An earlier reduction in catecholamine exposure may have contributed to lower rates of acute kidney injury and renal replacement therapy. Renal protection may also reflect preserved glomerular filtration through vasopressin-mediated efferent arteriolar constriction [27, 28] but also reduced fluid-related renal congestion. Supplementary hemodynamic data ( Online Resources 15 and 16 ) suggested that group differences in arterial pressure and catecholamine trajectories emerged after the exposure windows rather than reflecting antecedent hypotension. Mesenteric and peripheral ischemia, adverse effects reported in prior trials [6, 12], have not been evaluated, and the benefit-harm balance of vasopressin timing remains incompletely characterized. Exploratory subgroup analyses showed consistent benefits of ultra-early initiation across all subgroups, including those defined by lactate level, norepinephrine-equivalent dose, mechanical ventilation status, and SOFA score. The effect was more pronounced in patients with lactate ≥ 4 mmol/L (HR 0.77, 0.73–0.81) than in those with lactate < 2 mmol/L (HR 0.96, 0.86–1.07) or 2–<4 mmol/L (HR 0.95, 0.87–1.03). Patients with lactate ≥ 4 mmol/L are likely in a state of more severe tissue hypoperfusion with greater catecholamine demand, and an earlier addition of vasopressin in this population may have provided a larger catecholamine-sparing effect during the period of highest hemodynamic instability. The overall consistency of benefits across the subgroups supports the hypothesis that ultra-early vasopressin initiation is broadly beneficial in patients who have reached the high-dose norepinephrine threshold. The strengths of this study lie in its use of two geographically and temporally distinct ICU databases; focus on vasopressin initiation timing at a guideline-relevant escalation point; clone-censor-weight methods that address time-related biases inherent in observational studies of treatment timing; and reporting of both relative and absolute effect measures, including risk differences and restricted mean survival time. Nevertheless, this study has several limitations. First, residual confounding may persist for unmeasured factors such as clinical decision-making preferences, exact fluid volumes administered, and infection source. The consistency of results across two independent databases, multiple time-zero definitions, and sensitivity analyses with covariate-adjusted models provides some reassurance, but unmeasured confounding cannot be ruled out. Second, the target trial compared two initiation windows (0–3 h versus > 3–6 h) among patients who received vasopressin within 6 hours. The effect of vasopressin initiation beyond 6 hours or the decision not to use vasopressin was outside the scope of this emulated trial. Third, some secondary outcomes rely on proxy definitions influenced by documentation practices, although the same definitions were applied uniformly across both strategy arms. Fourth, vasopressin-related ischemic adverse effects were not directly evaluated; the two databases lack structured recording of digital or mesenteric ischemia. Fifth, hourly data harmonization and differences in recording practices between MIMIC-IV and eICU-CRD may have introduced misclassification, although the concordance of results across the two databases suggests that such effects did not drive the observed associations. Conclusions Among patients with septic shock whose norepinephrine dose reached the high-dose range, ultra-early vasopressin initiation was associated with lower 28-day mortality than early initiation. These findings suggest that the timing of vasopressin initiation among actual users is clinically important and an early multimodal vasopressor approach may be preferable to progressive catecholamine escalation alone. A randomized trial comparing protocolized ultra-early vasopressin initiation with usual care after norepinephrine escalation would clarify whether this association is causal. Declarations Ethics approval Both the MIMIC-IV and eICU-CRD are publicly available databases. The use of MIMIC-IV was approved by the institutional review boards of the Massachusetts Institute of Technology (No. 0403000206) and Beth Israel Deaconess Medical Center (2001-P-001699/14). The eICU-CRD was released under the same Health Insurance Portability and Accountability Act safe harbor provisions. Because both databases contained only de-identified data, the requirement for individual informed consent was waived. This study was conducted in accordance with the principles of the Declaration of Helsinki. Consent to participate Not applicable. Both databases contained only de-identified data and the requirement for individual informed consent was waived by the respective institutional review boards. Registration This study was not prospectively registered. The target trial protocol, including eligibility criteria, treatment strategies, outcomes, and analysis plan, was specified before data extraction. Competing interests TS reports grants/contracts from KYOCERA Corporation (through the university); consulting fees from TXP Medical and Hakuhodo DY Holdings Inc.; and lecture fees from MeDiCU Inc., CMIC HOLDINGS Co., Ltd., Santen Pharmaceutical Co., Ltd., SAS Institute Japan, and Pfizer Japan. JS has a researcher contract with TXP Medical Co., Ltd. SS reports consulting fees from CureApp, Inc. and lecture fees from Nippon Boehringer Ingelheim Co., Ltd. The other authors declare no competing interests. None of these relationships are directly related to the submitted work. Funding This research received no specific grants from any funding agency in the public, commercial, or not-for-profit sectors. Data availability MIMIC-IV (version 3.1) is available at https://physionet.org/content/mimiciv/ after completion of a data use agreement. The eICU-CRD (version 2.0) is available at https://physionet.org/content/eicu-crd/ under the same conditions. The analysis code will be publicly available upon publication. Author contributions Study concept: TN; study design: TN and TI; statistical analysis: TN, TS, and SS; data interpretation: TN, MN, JS, TG, and TS; and manuscript drafting: TN. TI, MN, TS, JS, TG, SS, and TH critically reviewed the manuscript for important intellectual content, approved the final version, agreed to be personally accountable for their own contributions, and ensured that questions pertaining to the accuracy or integrity of any portion of the work were appropriately investigated and resolved. References Guerci P, Belveyre T, Mongardon N, Novy E (2022) When to start vasopressin in septic shock: the strategy we propose. Crit Care 26:125. https://doi.org/10.1186/s13054-022-04001-4 Tong X, Xue X, Duan C, Liu A (2023) Early administration of multiple vasopressors is associated with better survival in patients with sepsis: a propensity score-weighted study. Eur J Med Res 28:192. https://doi.org/10.1186/s40001-023-01229-w Prescott HC, Antonelli M, Alhazzani W, et al (2026) Surviving sepsis campaign: International guidelines for management of sepsis and septic shock 2026. Crit Care Med. https://doi.org/10.1097/CCM.0000000000007075 Russell JA, Walley KR, Singer J, et al (2008) Vasopressin versus norepinephrine infusion in patients with septic shock. N Engl J Med 358:877–887. https://doi.org/10.1056/NEJMoa067373 Gordon AC, Mason AJ, Thirunavukkarasu N, et al (2016) Effect of early vasopressin vs norepinephrine on kidney failure in patients with septic shock: The VANISH randomized clinical trial. JAMA 316:509–518. https://doi.org/10.1001/jama.2016.10485 Nagendran M, Russell JA, Walley KR, et al (2019) Vasopressin in septic shock: an individual patient data meta-analysis of randomised controlled trials. Intensive Care Med 45:844–855. https://doi.org/10.1007/s00134-019-05620-2 White KC, Costa-Pinto R, Blank S, et al (2025) Effect of early adjunctive vasopressin initiation for septic shock patients: a target trial emulation. Crit Care 29:188. https://doi.org/10.1186/s13054-025-05401-y Kalimouttou A, Kennedy JN, Feng J, et al (2025) Optimal vasopressin initiation in septic shock: The OVISS reinforcement learning study. JAMA 333:1688–1698. https://doi.org/10.1001/jama.2025.3046 Sacha GL, Duggal A, Reddy AJ, et al (2025) Vasopressin initiation timing and in-hospital mortality in septic shock: An observational study of large public databases. Crit Care Explor 7:e1284. https://doi.org/10.1097/CCE.0000000000001284 Brask AL, Shemanski SM, Barnes TE, Holmes AK (2023) Timing of vasopressin addition to norepinephrine and efficacy outcomes in patients with septic shock. Ann Pharmacother 57:521–526. https://doi.org/10.1177/10600280221118903 Ma C-H, Healy J, Kinteh E, et al (2025) Extremely early initiation of vasopressors might not decrease short-term mortality for adults with septic shock: a systematic review and meta-analysis. Ann Intensive Care 15:18. https://doi.org/10.1186/s13613-025-01428-0 Lajoye Q, Orieux A, Boyer A, et al (2025) Vasopressin and its analogues in patients with septic shock: holy Grail or unfulfilled promise? Crit Care 29:333. https://doi.org/10.1186/s13054-025-05540-2 White KC, Costa-Pinto R, Chaba A, et al (2024) Timing of adjunctive vasopressin initiation for septic shock patients and hospital mortality: A multicentre observational study. Crit Care Resusc 26:295–302. https://doi.org/10.1016/j.ccrj.2024.09.002 Sacha GL, Lam SW, Wang L, et al (2022) Association of catecholamine dose, lactate, and shock duration at vasopressin initiation with mortality in patients with septic shock. Crit Care Med 50:614–623. https://doi.org/10.1097/CCM.0000000000005317 Xu J, Cai H, Zheng X (2023) Timing of vasopressin initiation and mortality in patients with septic shock: analysis of the MIMIC-III and MIMIC-IV databases. BMC Infect Dis 23:199. https://doi.org/10.1186/s12879-023-08147-6 He D, Zhang L, Hu H, et al (2023) Effect of early vasopressin combined with norepinephrine on short-term mortality in septic shock: A retrospective study based on the MIMIC-IV database. Am J Emerg Med 69:188–194. https://doi.org/10.1016/j.ajem.2023.04.040 Hernán MA (2018) How to estimate the effect of treatment duration on survival outcomes using observational data. BMJ 360:k182. https://doi.org/10.1136/bmj.k182 von Elm E, Altman DG, Egger M, et al (2008) The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 61:344–349. https://doi.org/10.1016/j.jclinepi.2007.11.008 Cashin AG, Hansford HJ, Hernán MA, et al (2025) Transparent Reporting of Observational Studies Emulating a Target Trial-the TARGET statement. JAMA 334:1084–1093. https://doi.org/10.1001/jama.2025.13350 Johnson A, Bulgarelli L, Pollard T, et al (2024) MIMIC-IV (version 3.1). PhysioNet. https://doi.org/10.13026/kpb9-mt58 Pollard TJ, Johnson AEW, Raffa JD, Celi LA, Badawi O, Mark RG (2019) The eICU Collaborative Research Database (version 2.0). PhysioNet. https://doi.org/10.13026/C2WM1R Bennett N, Plečko D, Ukor I-F, et al (2022) ricu: R’s interface to intensive care data. Gigascience 12:. https://doi.org/10.1093/gigascience/giad041 Singer M, Deutschman CS, Seymour CW, et al (2016) The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 315:801–810. https://doi.org/10.1001/jama.2016.0287 Khanna A, English SW, Wang XS, et al (2017) Angiotensin II for the treatment of vasodilatory shock. N Engl J Med 377:419–430. https://doi.org/10.1056/NEJMoa1704154 Hernán MA, Wang W, Leaf DE (2022) Target trial emulation: A framework for causal inference from observational data. JAMA 328:2446–2447. https://doi.org/10.1001/jama.2022.21383 Zhao SS, Lyu H, Yoshida K (2021) Versatility of the clone-censor-weight approach: response to ‘trial emulation in the presence of immortal-time bias. Int J Epidemiol 50:694–695 Sharshar T, Blanchard A, Paillard M, et al (2003) Circulating vasopressin levels in septic shock. Crit Care Med 31:1752–1758. https://doi.org/10.1097/01.CCM.0000063046.82359.4A Landry DW, Levin HR, Gallant EM, et al (1997) Vasopressin deficiency contributes to the vasodilation of septic shock. Circulation 95:1122–1125. https://doi.org/10.1161/01.cir.95.5.1122 Supplementary Files Equatorchecklist20260418.docx VPgraphicalabstract20260418.pptx VPsupptablefigures20260418.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Major revisions 14 May, 2026 Reviewers agreed at journal 23 Apr, 2026 Reviewers invited by journal 23 Apr, 2026 Editor assigned by journal 19 Apr, 2026 First submitted to journal 17 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9454619","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":628443633,"identity":"b817e997-9a74-4c1a-b540-ba74123658ad","order_by":0,"name":"Takaya Nakashima","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIiWNgGAWjYDAC9uaDDz5USPDwszcAeQYWRGjhOZZsOOOMjZxkzwGQFgkitEjkmEnztqUZG9xIAHMJ6+DvOWAgOYPtcOLMmc+vbvhRIMHA396dgN+S4w0JBh94Dif2S+eU3ewBOkzizNkN+K05c+BA4gwJoC2zc9Ju8AC1GEjk4tcifyOx4TCPweHEDTfPpN38Q4wWgxvJjM08CSDvsx+7TZQthmeOMTPOOAAK5By22zIGEjwE/SJ3vP/7j4//QFF5/NnNN39s5Pjbewl4HwF4DMAkscpBgP0BKapHwSgYBaNgBAEAKOdQ9TTC3ssAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0001-4214-3945","institution":"Nagasaki University Hospital: Nagasaki Daigaku Byoin","correspondingAuthor":true,"prefix":"","firstName":"Takaya","middleName":"","lastName":"Nakashima","suffix":""},{"id":628443634,"identity":"f47c2093-a204-4505-ae2f-3cab7d8d2c95","order_by":1,"name":"Taiga Ichinomiya","email":"","orcid":"","institution":"Nagasaki University Hospital: Nagasaki Daigaku Byoin","correspondingAuthor":false,"prefix":"","firstName":"Taiga","middleName":"","lastName":"Ichinomiya","suffix":""},{"id":628443635,"identity":"f0941a8b-9407-46d4-9205-4d1d51fe5bf3","order_by":2,"name":"Mikio Nakajima","email":"","orcid":"","institution":"Tokyo Metropolitan Hiroo Hospital: Tokyo Toritsu Hiroo Byoin","correspondingAuthor":false,"prefix":"","firstName":"Mikio","middleName":"","lastName":"Nakajima","suffix":""},{"id":628443636,"identity":"a1bbf40b-6117-44b3-8764-ee7e76ce89d0","order_by":3,"name":"Tomohiro Shinozaki","email":"","orcid":"","institution":"University of Tokyo Interfaculty Initiative in Information Studies Graduate School of Interdisciplinary Information Studies: Tokyo Daigaku Daigakuin Joho Gakukan Gakusai Joho Gakufu","correspondingAuthor":false,"prefix":"","firstName":"Tomohiro","middleName":"","lastName":"Shinozaki","suffix":""},{"id":628443637,"identity":"f0548cb5-f2fe-42df-9ddd-f2a88bf2dc6c","order_by":4,"name":"Junichiro Shibata","email":"","orcid":"","institution":"TXP Medical Co., Ltd","correspondingAuthor":false,"prefix":"","firstName":"Junichiro","middleName":"","lastName":"Shibata","suffix":""},{"id":628443638,"identity":"fd4af82b-2005-4f1e-829c-de69f59e7f0b","order_by":5,"name":"Tadahiro Goto","email":"","orcid":"","institution":"Yokohama City University: Yokohama Shiritsu Daigaku","correspondingAuthor":false,"prefix":"","firstName":"Tadahiro","middleName":"","lastName":"Goto","suffix":""},{"id":628443639,"identity":"adcb72a7-f054-47d6-bdde-c57812ea8892","order_by":6,"name":"Shuntaro Sato","email":"","orcid":"","institution":"Nagasaki University Hospital: Nagasaki Daigaku Byoin","correspondingAuthor":false,"prefix":"","firstName":"Shuntaro","middleName":"","lastName":"Sato","suffix":""},{"id":628443640,"identity":"07348cb2-dc5b-47f4-9b62-c7e5bf64b00c","order_by":7,"name":"Tetsuya Hara","email":"","orcid":"","institution":"Nagasaki University Hospital: Nagasaki Daigaku Byoin","correspondingAuthor":false,"prefix":"","firstName":"Tetsuya","middleName":"","lastName":"Hara","suffix":""}],"badges":[],"createdAt":"2026-04-18 06:15:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9454619/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9454619/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108735884,"identity":"ac9024c0-b6f2-4de9-95bd-d18213452156","added_by":"auto","created_at":"2026-05-07 20:09:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":291428,"visible":true,"origin":"","legend":"\u003cp\u003eStudy flow diagram\u003c/p\u003e\n\u003cp\u003eCohort assembly from the MIMIC-IV and eICU-CRD. Time zero was defined as the first hour at which the norepinephrine-equivalent dose reached 0.25 µg/kg/min\u003c/p\u003e\n\u003cp\u003eAbbreviations: MIMIC-IV, Medical Information Mart for Intensive Care IV; eICU-CRD, eICU Collaborative Research Database; NE, norepinephrine; VP, vasopressin\u003c/p\u003e","description":"","filename":"VPfigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/d11e447e2665ba4c9b5d2b33.png"},{"id":108807697,"identity":"2557af28-ff65-4d3a-909e-bad25a16837d","added_by":"auto","created_at":"2026-05-08 15:31:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":75488,"visible":true,"origin":"","legend":"\u003cp\u003eWeighted cumulative incidence curves for 28-day mortality\u003c/p\u003e\n\u003cp\u003eInverse probability-of-censoring-weighted survival curves for 28-day mortality comparing ultra-early (0–3 hours) and early (\u0026gt;3–6 hours) vasopressin initiation. Curves were estimated using weights truncated to the 99th percentile. The standardized 28-day mortality risks, absolute risk differences, and restricted mean survival time differences with 95% confidence intervals are annotated\u003c/p\u003e\n\u003cp\u003eAbbreviations: CI, confidence interval; RMST, restricted mean survival time\u003c/p\u003e","description":"","filename":"VPfigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/23845916c412d938838efdc6.png"},{"id":108735886,"identity":"8e477b64-01ad-4b95-999f-213074619655","added_by":"auto","created_at":"2026-05-07 20:09:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":183835,"visible":true,"origin":"","legend":"\u003cp\u003ePrimary and secondary outcome estimates\u003c/p\u003e\n\u003cp\u003ePanel A: Hazard ratios for 28-day mortality across prespecified analytic approaches (primary analysis with 99th-percentile weight truncation, 95th-percentile truncation, and covariate-adjusted model). Panel B: \u0026nbsp;Hazard ratios for secondary outcomes from fine-gray models, with death as a competing event\u003c/p\u003e\n\u003cp\u003eAbbreviations: IPCW, inverse probability of censoring weighting; AKI, acute kidney injury; KDIGO, Kidney Disease: Improving Global Outcomes; RRT, renal replacement therapy; CRRT, continuous renal replacement therapy\u003c/p\u003e","description":"","filename":"VPfigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/63c557a3566514734bbf1eb9.png"},{"id":108810409,"identity":"50f24406-d6a6-4563-ac99-6bb5a847a8ea","added_by":"auto","created_at":"2026-05-08 15:58:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":638005,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/43bc083e-0b15-47e8-b147-e952f82e348a.pdf"},{"id":108735888,"identity":"41d5b369-ba9d-4361-8f17-b202caedc0df","added_by":"auto","created_at":"2026-05-07 20:09:59","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":27076,"visible":true,"origin":"","legend":"","description":"","filename":"Equatorchecklist20260418.docx","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/3d69a65f3521ec85ebab9cb5.docx"},{"id":108735889,"identity":"5fe7ed7e-0ab1-432a-907f-2e573c0f4a5b","added_by":"auto","created_at":"2026-05-07 20:10:00","extension":"pptx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":936468,"visible":true,"origin":"","legend":"","description":"","filename":"VPgraphicalabstract20260418.pptx","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/42a049a81d2105f1c679ae95.pptx"},{"id":108735887,"identity":"e28209df-06b5-4160-8ea3-01175386b2e3","added_by":"auto","created_at":"2026-05-07 20:09:59","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":1509654,"visible":true,"origin":"","legend":"","description":"","filename":"VPsupptablefigures20260418.docx","url":"https://assets-eu.researchsquare.com/files/rs-9454619/v1/2e70bba1f5b85c997cb5dd90.docx"}],"financialInterests":"","formattedTitle":"Ultra-early versus early adjunctive vasopressin initiation after norepinephrine escalation in septic shock: a target trial emulation","fulltext":[{"header":"Take-home message","content":"\u003cp\u003eAmong patients with septic shock whose norepinephrine-equivalent dose reached 0.25 \u0026micro;g/kg/min or higher, ultra-early vasopressin initiation (0\u0026ndash;3 hours after reaching this threshold) was associated with lower 28-day mortality than early initiation (\u0026gt;3\u0026ndash;6 hours). Timing of vasopressin initiation, beyond the decision to use vasopressin, may be a clinically meaningful lever once the high-dose range is reached, and warrants evaluation in a randomized trial.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eVasopressin, a non-catecholaminergic vasopressor, has attracted increasing interest as a supportive therapy to reduce the adverse effects of high-dose norepinephrine in patients with septic shock [1, 2]. The Surviving Sepsis Campaign guidelines 2026 recommend norepinephrine as the first-line vasopressor and suggest adding vasopressin when arterial pressure remains inadequate at norepinephrine-equivalent doses of 0.25 to 0.5 \u0026micro;g/kg/min [3]. However, two large randomized trials [4, 5] and an individual patient data meta-analysis [6] found no overall survival benefit with vasopressin compared to norepinephrine alone in septic shock.\u003c/p\u003e \u003cp\u003eMore recently, several studies have examined whether earlier initiation of vasopressin improves outcomes [7\u0026ndash;10] however, the findings have been inconsistent, and no clear consensus has emerged [11, 12]. Two methodological issues may explain this inconsistency. First, the start of follow-up was commonly defined as the onset of septic shock [13\u0026ndash;15] rather than the point at which vasopressin became clinically relevant. Because vasopressin is typically initiated only after escalation to high-dose norepinephrine, the interval between shock onset and vasopressin initiation varies across patients, introducing confounding by time-varying severity during this gap [12]. Second, most previous studies compared early vasopressin combination therapy with norepinephrine alone [8, 9, 16], thereby primarily assessing the effect of vasopressin addition rather than the timing of its initiation.\u003c/p\u003e \u003cp\u003eThe clinically relevant question is how soon vasopressin should be initiated after norepinephrine reaches a high dose. To address this issue, we conducted a targeted trial emulation, focusing on the timing of vasopressin initiation. We defined time zero as the time at which the norepinephrine-equivalent dose reached 0.25 \u0026micro;g/kg/min or higher, the guideline-recommended threshold for considering vasopressin addition [3]. Using clone-censor-weight methods [17], we compared ultra-early initiation (0\u0026ndash;3 hours after reaching this threshold) and early initiation (\u0026gt;\u0026thinsp;3\u0026ndash;6 hours) in terms of 28-day mortality.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and reporting\u003c/h2\u003e \u003cp\u003eThis study followed a targeted trial emulation framework designed to compare two vasopressin initiation strategies using routinely collected clinical data. The reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline [18] and the Transparent Reporting of Observational Studies Emulating a Target Trial (TARGET) statement [19]. The STROBE and TARGET checklists are presented in \u003cb\u003eOnline Resource 1\u003c/b\u003e. The components of the hypothetical target trial and their observational operationalizations are summarized in \u003cb\u003eOnline Resource 2\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eData sources\u003c/h3\u003e\n\u003cp\u003eThis study used data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) [20] and the eICU Collaborative Research Database (eICU-CRD, version 2.0) [21], which are two large, publicly available critical care databases. MIMIC-IV contains de-identified data from over 65,000 intensive care unit (ICU) admissions at the Beth Israel Deaconess Medical Center (2008\u0026ndash;2022), while eICU-CRD is a multi-center telemedicine database with more than 200,000 ICU admissions across the United States (2014\u0026ndash;2015). The data from both databases were harmonized to an hourly common concept structure and pooled using the Ricu package [22]. The first author completed the required institutional training program (Record ID: 54439141) and was granted access to both databases. Because both databases contained only de-identified data, the requirement for institutional review board approval and individual informed consent was waived.\u003c/p\u003e\n\u003ch3\u003eEligibility criteria\u003c/h3\u003e\n\u003cp\u003eWe included adults (aged 18 years or older) 1) who met sepsis-3 criteria for septic shock [23], 2) whose norepinephrine-equivalent dose reached\u0026thinsp;\u0026ge;\u0026thinsp;0.25 \u0026micro;g/kg/min and 3) who had not received vasopressin before that point.\u003c/p\u003e\n\u003ch3\u003eTime zero\u003c/h3\u003e\n\u003cp\u003eThe norepinephrine-equivalent dose (\u0026micro;g/kg/min) was calculated as norepinephrine\u0026thinsp;+\u0026thinsp;epinephrine\u0026thinsp;+\u0026thinsp;dopamine/150\u0026thinsp;+\u0026thinsp;phenylephrine/10\u0026thinsp;+\u0026thinsp;vasopressin \u0026times; 2.5 (U/min), using the conversion factors from the ATHOS-3 trial [24] as implemented in the Ricu package [22]. We chose the lower bound of the guideline-recommended range (0.25\u0026ndash;0.5 \u0026micro;g/kg/min) to capture the earliest point at which vasopressin addition is considered [3].\u003c/p\u003e\n\u003ch3\u003eTreatment strategies\u003c/h3\u003e\n\u003cp\u003eThe two strategies were as follows: (1) ultra-early vasopressin initiation (0\u0026ndash;3 hours after time zero) and (2) early vasopressin initiation (\u0026gt;\u0026thinsp;3\u0026ndash;6 hours after time zero). In the hypothetical target trial, eligible individuals are randomly assigned at time zero to one of the strategies.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eOutcomes\u003c/h2\u003e \u003cp\u003eThe primary outcome was 28-day mortality. Secondary outcomes, defined as first events within 28 days, included acute kidney injury (Kidney Disease: Improving Global Outcomes [KDIGO] stage 2 or higher and stage 3 or higher), renal replacement therapy (RRT) initiation, continuous renal replacement therapy (CRRT) initiation, medically treated arrhythmia and net negative fluid balance. The definitions are provided in \u003cb\u003eOnline Resource 3\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eBaseline characteristics and covariates\u003c/h3\u003e\n\u003cp\u003eVariables measured at or immediately before time zero included age, sex, body mass index, body weight, mean arterial pressure, lactate, Sequential Organ Failure Assessment (SOFA) score, creatinine, norepinephrine-equivalent dose, invasive mechanical ventilation, and systemic corticosteroid use within 24 hours before time zero. These variables also served as covariates in the censoring weight models, together with hourly time-varying values of mean arterial pressure, lactate, creatinine, norepinephrine-equivalent dose, SOFA score, ventilation status, concurrent vasopressor use (epinephrine, dopamine, dobutamine, and phenylephrine), the hourly deviation from the time-zero norepinephrine-equivalent dose, and its 1-hour change during the 0-to-6-hour assignment window. The covariates were pre-specified based on previous studies [2, 9, 16].\u003c/p\u003e\n\u003ch3\u003eMissing data\u003c/h3\u003e\n\u003cp\u003eMissing data were assessed for each baseline covariate. The proportion of missing values and the imputation method for each variable are provided in \u003cb\u003eOnline Resource 4\u003c/b\u003e.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eBaseline characteristics are reported as absolute values with percentages for categorical variables and as medians with interquartile intervals for quantitative variables.\u003c/p\u003e \u003cp\u003eThe primary analysis used a clone-censor-weight approach to emulate survival under each initiation strategy, circumventing the time-related biases inherent in post hoc treatment timing comparisons [25, 26]. At time zero, each patient was duplicated such that one clone (or copy) was assigned to each strategy (cloning), mimicking a random assignment. Each clone was followed forward and artificially censored when the treatment deviated from the assigned strategy (censoring). Clones on ultra-early strategy were censored at 3 hours if vasopressin had not been started; clones on early strategy were censored if vasopressin was started before 3 hours or at 6 hours if not yet started. If a patient died or experienced an outcome event before being artificially censored for deviation from each strategy, the event was recorded at the time of its occurrence. Third, inverse probability of censoring (IPC) weights were estimated to adjust for the informative censoring introduced by strategy deviation. At each hour during the 0-to-6-hour assignment window, the probability of remaining uncensored was modeled using pooled logistic regression conditional on baseline and time-varying covariates (listed in the Covariates section above). The IPC weight for each interval was defined as the inverse of this predicted probability, and hourly weights were multiplied across time to obtain cumulative weights. Cumulative weights were truncated to the 99th percentile. Cumulative incidence curves for 28-day mortality were estimated using the IPC weighted Kaplan\u0026ndash;Meier method. The 28-day risks were compared using the absolute risk difference and restricted mean survival time (RMST) difference, calculated as the area between the two curves, with 95% confidence intervals (CIs) from patient-level bootstrapping (500 iterations). Hazard ratios (HRs) were estimated using the IPC weighted Cox model, with robust standard errors clustered by patients. More details are provided in \u003cb\u003eOnline Resource 5\u003c/b\u003e. A schematic of the analysis design, including the clone-censor-weight framework, is shown in \u003cb\u003eOnline Resource 6\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eSecondary outcomes were analyzed using the IPC weighted Fine-Gray subdistribution hazard model, treating death before the event of interest as a competing risk.\u003c/p\u003e \u003cp\u003eExploratory subgroup analyses for the primary outcome were stratified by lactate at time zero (\u0026lt;\u0026thinsp;2, 2\u0026ndash;\u0026lt;4, or \u0026ge;\u0026thinsp;4 mmol/L), norepinephrine-equivalent dose (\u0026lt;\u0026thinsp;0.30 or \u0026ge;\u0026thinsp;0.30 \u0026micro;g/kg/min), invasive mechanical ventilation (yes or no), and SOFA score (\u0026lt;\u0026thinsp;10, 10\u0026ndash;\u0026lt;14, or \u0026ge;\u0026thinsp;14).\u003c/p\u003e \u003cp\u003eSensitivity analyses included the following: (1) separate analyses within each database (MIMIC-IV and eICU-CRD); and (2) variation in the time-zero definition across norepinephrine thresholds (0.20, 0.25, 0.30, 0.35, 0.40, 0.45, and 0.50 \u0026micro;g/kg/min), sustainment durations (2 or 3 consecutive hours above the threshold), verification windows (2 or 3 hours after the time zero), and dose variables (norepinephrine-equivalent dose and norepinephrine infusion rate alone).\u003c/p\u003e \u003cp\u003eAll analyses were performed using R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). The analysis code and figure-generation scripts will be deposited in a public repository upon acceptance, and the repository URL will be added at the proof stage.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eUse of AI-assisted tools\u003c/h2\u003e \u003cp\u003eA large language model (Claude, Anthropic) was used to assist in drafting the manuscript and English language editing and formatting. All AI-generated content was reviewed, verified, and edited by the authors, who take full responsibility for the contents of the final manuscript.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eCohort characteristics\u003c/h2\u003e\n \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the cohort assembly. From 21,587 adults with septic shock in MIMIC-IV and eICU-CRD, 5,157 met the eligibility criteria at time zero (norepinephrine-equivalent dose\u0026thinsp;\u0026ge;\u0026thinsp;0.25 \u0026micro;g/kg/min, no prior vasopressin). All 5,157 patients were entered in the clone-censor-weight analysis, yielding 10,314 clones (one per strategy per patient). Under the ultra-early strategy, 943 (18.3%) adhered to the assigned strategy by initiating vasopressin within 0\u0026ndash;3 hours; and under the early strategy, 323 (6.3%) adhered by initiating vasopressin within \u0026gt;\u0026thinsp;3\u0026ndash;6 hours. The remaining clones were censored at strategy deviation and uncensored clones were weighted using IPC weights. The distribution of time from time zero to vasopressin initiation is shown in \u003cstrong\u003eOnline Resource 7\u003c/strong\u003e; the 0-to-1-hour interval was the most common starting window.\u003c/p\u003e\n \u003cp\u003eTable \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents baseline characteristics at time zero of all 5,157 patients who entered the clone-censor-weight analysis. The median patient age was 66 years, 55.2% were male, and the median SOFA score was 9. The median lactate was 3.3 mmol/L, and the median norepinephrine-equivalent dose was 0.3 \u0026micro;g/kg/min. Invasive mechanical ventilation was used in 36.9% of the patients, and 6.4% had received systemic corticosteroids before time zero. Crude 28-day mortality in the full eligible cohort was 40.7% (2,097 of 5,157 patients). The weight distributions (\u003cstrong\u003eOnline Resource 8\u003c/strong\u003e) were not dispersed after the 99th-percentile truncation.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eBaseline characteristics at time zero\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"2\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCharacteristic\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eOverall (N\u0026thinsp;=\u0026thinsp;5157)\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eDemographics\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eAge, yr\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e66.0 (55.0, 77.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eMale sex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e2,846 (55.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eBMI, kg/m\u0026sup2;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e26.7 (22.8, 31.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eBody weight, kg\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e77.0 (63.7, 92.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003e\u003cstrong\u003eHemodynamics and severity at time zero\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eMAP, mmHg\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e68.0 (61.0, 76.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eLactate, mmol/L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e3.3 (1.9, 6.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSOFA score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e9.0 (6.0, 12.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCreatinine, mg/dL\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e1.5 (0.9, 2.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eNE-equivalent dose, \u0026micro;g/kg/min\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e0.3 (0.3, 0.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eInvasive mechanical ventilation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e1,902 (36.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eChronic kidney disease\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e924 (17.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSystemic corticosteroid use\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\n \u003cp\u003e332 (6.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\"\u003e\u003csup\u003e1\u003c/sup\u003eMedian (IQR); n (%)\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\"\u003eCaption: Continuous variables are presented as median (interquartile range); categorical variables as n (%).\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\"\u003eAbbreviations: BMI, body mass index; MAP, mean arterial pressure; SOFA, Sequential\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\"\u003eOrgan Failure Assessment; NE, norepinephrine; CKD, chronic kidney disease.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003ePrimary outcome\u003c/h2\u003e\n \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows weighted cumulative incidence curves for 28-day mortality, from which risk differences and RMST differences were derived. Ultra-early initiation showed lower 28-day mortality (35.7% versus 41.2%; risk difference\u0026thinsp;\u0026minus;\u0026thinsp;5.5 percentage points, 95% CI\u0026thinsp;\u0026minus;\u0026thinsp;6.9 to \u0026minus;\u0026thinsp;4.4; RMST difference\u0026thinsp;+\u0026thinsp;1.49 days, 95% CI 1.18\u0026ndash;1.82). Figure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA presents the HR of 0.81 (95% CI 0.78\u0026ndash;0.85) in the primary analysis. The direction was preserved with 95th-percentile truncation (HR 0.83, 95% CI 0.80\u0026ndash;0.86) and in the IPC weighted Cox model with additional outcome regression adjusting for baseline covariates (HR 0.96, 95% CI 0.93\u0026ndash;1.00).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eSecondary outcomes\u003c/h2\u003e\n \u003cp\u003eThe secondary outcome estimates are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB. Ultra-early initiation showed lower rates of acute kidney injury (KDIGO stage 2 or higher: HR 0.94, 95% CI 0.92\u0026ndash;0.97; stage 3 or higher: 0.92, 0.89\u0026ndash;0.95), RRT initiation (0.65, 0.61\u0026ndash;0.70), and CRRT initiation (0.75, 0.68\u0026ndash;0.83). Medically treated arrhythmia (0.98, 0.93\u0026ndash;1.03) and net negative fluid balance (0.97, 0.93\u0026ndash;1.00) did not differ between groups.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eSubgroup analyses\u003c/h2\u003e\n \u003cp\u003eExploratory subgroup analyses (\u003cstrong\u003eOnline Resource 9\u003c/strong\u003e) showed consistent results across all subgroups. The effect was more pronounced in patients with lactate\u0026thinsp;\u0026ge;\u0026thinsp;4 mmol/L (HR 0.77, 0.73\u0026ndash;0.81) than in those with lactate\u0026thinsp;\u0026lt;\u0026thinsp;2 mmol/L (HR 0.96, 0.86\u0026ndash;1.07) or 2\u0026ndash;\u0026lt;4 mmol/L (HR 0.95, 0.87\u0026ndash;1.03). Results were consistent across norepinephrine-equivalent dose (\u0026lt;\u0026thinsp;0.30 \u0026micro;g/kg/min: HR 0.83, 0.78\u0026ndash;0.89; \u0026ge;0.30: HR 0.81, 0.76\u0026ndash;0.85), mechanical ventilation status (invasive: HR 0.82, 0.77\u0026ndash;0.87; non-invasive: HR 0.82, 0.77\u0026ndash;0.86), and SOFA score subgroups (\u0026lt;\u0026thinsp;10: HR 0.93, 0.88\u0026ndash;0.98; 10\u0026ndash;\u0026lt;14: HR 0.90, 0.84\u0026ndash;0.96; \u0026ge;14: HR 0.85, 0.77\u0026ndash;0.95).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003eSensitivity analyses\u003c/h2\u003e\n \u003cp\u003eSeparate analyses within each database yielded consistent results (MIMIC-IV: HR 0.81, 0.77\u0026ndash;0.86; eICU-CRD: HR 0.81, 0.77\u0026ndash;0.87; \u003cstrong\u003eOnline Resource 10\u003c/strong\u003e). Across the 56-alternative time-zero definitions, the primary estimate remained consistently below 1.0 (\u003cstrong\u003eOnline Resource 11\u003c/strong\u003e).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this target trial emulation of 5,157 patients with septic shock whose norepinephrine-equivalent dose reached 0.25 \u0026micro;g/kg/min or higher, ultra-early vasopressin initiation showed lower 28-day mortality than early vasopressin initiation. This direction was preserved in the series of sensitivity analyses. Ultra-early initiation also showed lower rates of acute kidney injury and renal replacement therapy including CRRT. A key feature of this analysis is that it examined not whether vasopressin should be used, but whether earlier initiation is beneficial once the norepinephrine dose reaches the high-dose range. Prior studies compared early vasopressin with norepinephrine alone [7\u0026ndash;9]. By comparing two narrow initiation windows using clone-censor-weight methods, the present analysis isolated the effect of timing from the broader question of whether vasopressin should be used. As \u003cb\u003eOnline Resource 12\u003c/b\u003e shows, the early group was not simply waiting; during the 3\u0026ndash;6-hour window before vasopressin initiation, norepinephrine doses were escalated, and other interventions, such as phenylephrine, were administered.\u003c/p\u003e \u003cp\u003eThe present findings do not contradict the VASST or VANISH trials, which found no consistent overall survival benefit of vasopressin-containing strategies [4, 5] or the absence of mortality reduction in an individual patient data meta-analysis [6]. These trials evaluated whether vasopressin should be included across a broad population, which is a clinical question different from the timing comparison in the present study. In such designs, patients who initiated vasopressin early and those who initiated it late are pooled in the treatment arm; if the benefit is concentrated among early initiators, the average treatment effect across the entire arm may be diluted toward the null. By focusing on patients who had already reached a clinically plausible stage of vasopressin use, the present study was designed to detect a timing-specific signal. Moreover, by defining time zero as the point of norepinephrine escalation rather than shock onset, this study aligned the start of follow-up with the clinical stage at which vasopressin became relevant, addressing the time-lag limitations of prior studies.\u003c/p\u003e \u003cp\u003eConfounding based on the indications is a central concern. Although the clone-censor-weight approach did not suffer from baseline confounding (at strategy assignment) by cloning each patient into both strategy arms at time zero, time-varying confounding by indication was translated into selection bias due to nonrandom censoring at strategy deviation. We try to correct this selection bias by applying the IPC weights under the assumption that all possible covariates affecting both vasopressin indication and outcomes were measured. In our analysis, the target trial emulation framework automatically excludes design bias such as immortal time bias, while it should be noted that this framework itself cannot account for confounding bias if uncontrolled confounding by indication was present.\u003c/p\u003e \u003cp\u003eSeveral mechanisms may explain this association. The weighted norepinephrine trajectory over 72 hours (\u003cb\u003eOnline Resource 13\u003c/b\u003e) showed that the ultra-early group experienced a rapid norepinephrine reduction after vasopressin addition, whereas the early group required continued catecholamine escalation during deferral. Total catecholamine exposure encompassing all vasopressor agents was lower in the ultra-early group (weighted median 53.6 versus 62.2 \u0026micro;g/kg; \u003cb\u003eOnline Resource 14\u003c/b\u003e), and vasopressor-free days were higher at 7 days (weighted mean difference 0.20 days, 95% CI 0.04\u0026ndash;0.34) and 14 days (0.43, 0.11\u0026ndash;0.71). An earlier reduction in catecholamine exposure may have contributed to lower rates of acute kidney injury and renal replacement therapy. Renal protection may also reflect preserved glomerular filtration through vasopressin-mediated efferent arteriolar constriction [27, 28] but also reduced fluid-related renal congestion. Supplementary hemodynamic data (\u003cb\u003eOnline Resources 15 and 16\u003c/b\u003e) suggested that group differences in arterial pressure and catecholamine trajectories emerged after the exposure windows rather than reflecting antecedent hypotension. Mesenteric and peripheral ischemia, adverse effects reported in prior trials [6, 12], have not been evaluated, and the benefit-harm balance of vasopressin timing remains incompletely characterized.\u003c/p\u003e \u003cp\u003eExploratory subgroup analyses showed consistent benefits of ultra-early initiation across all subgroups, including those defined by lactate level, norepinephrine-equivalent dose, mechanical ventilation status, and SOFA score. The effect was more pronounced in patients with lactate\u0026thinsp;\u0026ge;\u0026thinsp;4 mmol/L (HR 0.77, 0.73\u0026ndash;0.81) than in those with lactate\u0026thinsp;\u0026lt;\u0026thinsp;2 mmol/L (HR 0.96, 0.86\u0026ndash;1.07) or 2\u0026ndash;\u0026lt;4 mmol/L (HR 0.95, 0.87\u0026ndash;1.03). Patients with lactate\u0026thinsp;\u0026ge;\u0026thinsp;4 mmol/L are likely in a state of more severe tissue hypoperfusion with greater catecholamine demand, and an earlier addition of vasopressin in this population may have provided a larger catecholamine-sparing effect during the period of highest hemodynamic instability. The overall consistency of benefits across the subgroups supports the hypothesis that ultra-early vasopressin initiation is broadly beneficial in patients who have reached the high-dose norepinephrine threshold.\u003c/p\u003e \u003cp\u003e The strengths of this study lie in its use of two geographically and temporally distinct ICU databases; focus on vasopressin initiation timing at a guideline-relevant escalation point; clone-censor-weight methods that address time-related biases inherent in observational studies of treatment timing; and reporting of both relative and absolute effect measures, including risk differences and restricted mean survival time.\u003c/p\u003e \u003cp\u003eNevertheless, this study has several limitations. First, residual confounding may persist for unmeasured factors such as clinical decision-making preferences, exact fluid volumes administered, and infection source. The consistency of results across two independent databases, multiple time-zero definitions, and sensitivity analyses with covariate-adjusted models provides some reassurance, but unmeasured confounding cannot be ruled out. Second, the target trial compared two initiation windows (0\u0026ndash;3 h versus \u0026gt;\u0026thinsp;3\u0026ndash;6 h) among patients who received vasopressin within 6 hours. The effect of vasopressin initiation beyond 6 hours or the decision not to use vasopressin was outside the scope of this emulated trial. Third, some secondary outcomes rely on proxy definitions influenced by documentation practices, although the same definitions were applied uniformly across both strategy arms. Fourth, vasopressin-related ischemic adverse effects were not directly evaluated; the two databases lack structured recording of digital or mesenteric ischemia. Fifth, hourly data harmonization and differences in recording practices between MIMIC-IV and eICU-CRD may have introduced misclassification, although the concordance of results across the two databases suggests that such effects did not drive the observed associations.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eAmong patients with septic shock whose norepinephrine dose reached the high-dose range, ultra-early vasopressin initiation was associated with lower 28-day mortality than early initiation. These findings suggest that the timing of vasopressin initiation among actual users is clinically important and an early multimodal vasopressor approach may be preferable to progressive catecholamine escalation alone. A randomized trial comparing protocolized ultra-early vasopressin initiation with usual care after norepinephrine escalation would clarify whether this association is causal.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBoth the MIMIC-IV and eICU-CRD are publicly available databases. The use of MIMIC-IV was approved by the institutional review boards of the Massachusetts Institute of Technology (No. 0403000206) and Beth Israel Deaconess Medical Center (2001-P-001699/14). The eICU-CRD was released under the same Health Insurance Portability and Accountability Act safe harbor provisions. Because both databases contained only de-identified data, the requirement for individual informed consent was waived. This study was conducted in accordance with the principles of the Declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. Both databases contained only de-identified data and the requirement for individual informed consent was waived by the respective institutional review boards.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRegistration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was not prospectively registered. The target trial protocol, including eligibility criteria, treatment strategies, outcomes, and analysis plan, was specified before data extraction.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTS reports grants/contracts from KYOCERA Corporation (through the university); consulting fees from TXP Medical and Hakuhodo DY Holdings Inc.; and lecture fees from MeDiCU Inc., CMIC HOLDINGS Co., Ltd., Santen Pharmaceutical Co., Ltd., SAS Institute Japan, and Pfizer Japan. JS has a researcher contract with TXP Medical Co., Ltd. SS reports consulting fees from CureApp, Inc. and lecture fees from Nippon Boehringer Ingelheim Co., Ltd. The other authors declare no competing interests. None of these relationships are directly related to the submitted work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no specific grants from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMIMIC-IV (version 3.1) is available at https://physionet.org/content/mimiciv/ after completion of a data use agreement. The eICU-CRD (version 2.0) is available at https://physionet.org/content/eicu-crd/ under the same conditions. The analysis code will be publicly available upon publication.\u003c/p\u003e\u003ch2\u003eAuthor contributions\u003c/h2\u003e \u003cp\u003eStudy concept: TN; study design: TN and TI; statistical analysis: TN, TS, and SS; data interpretation: TN, MN, JS, TG, and TS; and manuscript drafting: TN. TI, MN, TS, JS, TG, SS, and TH critically reviewed the manuscript for important intellectual content, approved the final version, agreed to be personally accountable for their own contributions, and ensured that questions pertaining to the accuracy or integrity of any portion of the work were appropriately investigated and resolved.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGuerci P, Belveyre T, Mongardon N, Novy E (2022) When to start vasopressin in septic shock: the strategy we propose. Crit Care 26:125. https://doi.org/10.1186/s13054-022-04001-4\u003c/li\u003e\n\u003cli\u003eTong X, Xue X, Duan C, Liu A (2023) Early administration of multiple vasopressors is associated with better survival in patients with sepsis: a propensity score-weighted study. Eur J Med Res 28:192. https://doi.org/10.1186/s40001-023-01229-w\u003c/li\u003e\n\u003cli\u003ePrescott HC, Antonelli M, Alhazzani W, et al (2026) Surviving sepsis campaign: International guidelines for management of sepsis and septic shock 2026. Crit Care Med. https://doi.org/10.1097/CCM.0000000000007075\u003c/li\u003e\n\u003cli\u003eRussell JA, Walley KR, Singer J, et al (2008) Vasopressin versus norepinephrine infusion in patients with septic shock. N Engl J Med 358:877\u0026ndash;887. https://doi.org/10.1056/NEJMoa067373\u003c/li\u003e\n\u003cli\u003eGordon AC, Mason AJ, Thirunavukkarasu N, et al (2016) Effect of early vasopressin vs norepinephrine on kidney failure in patients with septic shock: The VANISH randomized clinical trial. JAMA 316:509\u0026ndash;518. https://doi.org/10.1001/jama.2016.10485\u003c/li\u003e\n\u003cli\u003eNagendran M, Russell JA, Walley KR, et al (2019) Vasopressin in septic shock: an individual patient data meta-analysis of randomised controlled trials. Intensive Care Med 45:844\u0026ndash;855. https://doi.org/10.1007/s00134-019-05620-2\u003c/li\u003e\n\u003cli\u003eWhite KC, Costa-Pinto R, Blank S, et al (2025) Effect of early adjunctive vasopressin initiation for septic shock patients: a target trial emulation. Crit Care 29:188. https://doi.org/10.1186/s13054-025-05401-y\u003c/li\u003e\n\u003cli\u003eKalimouttou A, Kennedy JN, Feng J, et al (2025) Optimal vasopressin initiation in septic shock: The OVISS reinforcement learning study. JAMA 333:1688\u0026ndash;1698. https://doi.org/10.1001/jama.2025.3046\u003c/li\u003e\n\u003cli\u003eSacha GL, Duggal A, Reddy AJ, et al (2025) Vasopressin initiation timing and in-hospital mortality in septic shock: An observational study of large public databases. Crit Care Explor 7:e1284. https://doi.org/10.1097/CCE.0000000000001284\u003c/li\u003e\n\u003cli\u003eBrask AL, Shemanski SM, Barnes TE, Holmes AK (2023) Timing of vasopressin addition to norepinephrine and efficacy outcomes in patients with septic shock. Ann Pharmacother 57:521\u0026ndash;526. https://doi.org/10.1177/10600280221118903\u003c/li\u003e\n\u003cli\u003eMa C-H, Healy J, Kinteh E, et al (2025) Extremely early initiation of vasopressors might not decrease short-term mortality for adults with septic shock: a systematic review and meta-analysis. Ann Intensive Care 15:18. https://doi.org/10.1186/s13613-025-01428-0\u003c/li\u003e\n\u003cli\u003eLajoye Q, Orieux A, Boyer A, et al (2025) Vasopressin and its analogues in patients with septic shock: holy Grail or unfulfilled promise? Crit Care 29:333. https://doi.org/10.1186/s13054-025-05540-2\u003c/li\u003e\n\u003cli\u003eWhite KC, Costa-Pinto R, Chaba A, et al (2024) Timing of adjunctive vasopressin initiation for septic shock patients and hospital mortality: A multicentre observational study. Crit Care Resusc 26:295\u0026ndash;302. https://doi.org/10.1016/j.ccrj.2024.09.002\u003c/li\u003e\n\u003cli\u003eSacha GL, Lam SW, Wang L, et al (2022) Association of catecholamine dose, lactate, and shock duration at vasopressin initiation with mortality in patients with septic shock. Crit Care Med 50:614\u0026ndash;623. https://doi.org/10.1097/CCM.0000000000005317\u003c/li\u003e\n\u003cli\u003eXu J, Cai H, Zheng X (2023) Timing of vasopressin initiation and mortality in patients with septic shock: analysis of the MIMIC-III and MIMIC-IV databases. BMC Infect Dis 23:199. https://doi.org/10.1186/s12879-023-08147-6\u003c/li\u003e\n\u003cli\u003eHe D, Zhang L, Hu H, et al (2023) Effect of early vasopressin combined with norepinephrine on short-term mortality in septic shock: A retrospective study based on the MIMIC-IV database. Am J Emerg Med 69:188\u0026ndash;194. https://doi.org/10.1016/j.ajem.2023.04.040\u003c/li\u003e\n\u003cli\u003eHern\u0026aacute;n MA (2018) How to estimate the effect of treatment duration on survival outcomes using observational data. BMJ 360:k182. https://doi.org/10.1136/bmj.k182\u003c/li\u003e\n\u003cli\u003evon Elm E, Altman DG, Egger M, et al (2008) The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 61:344\u0026ndash;349. https://doi.org/10.1016/j.jclinepi.2007.11.008\u003c/li\u003e\n\u003cli\u003eCashin AG, Hansford HJ, Hern\u0026aacute;n MA, et al (2025) Transparent Reporting of Observational Studies Emulating a Target Trial-the TARGET statement. JAMA 334:1084\u0026ndash;1093. https://doi.org/10.1001/jama.2025.13350\u003c/li\u003e\n\u003cli\u003eJohnson A, Bulgarelli L, Pollard T, et al (2024) MIMIC-IV (version 3.1). PhysioNet. https://doi.org/10.13026/kpb9-mt58\u003c/li\u003e\n\u003cli\u003ePollard TJ, Johnson AEW, Raffa JD, Celi LA, Badawi O, Mark RG (2019) The eICU Collaborative Research Database (version 2.0). PhysioNet. https://doi.org/10.13026/C2WM1R\u003c/li\u003e\n\u003cli\u003eBennett N, Plečko D, Ukor I-F, et al (2022) ricu: R\u0026rsquo;s interface to intensive care data. Gigascience 12:. https://doi.org/10.1093/gigascience/giad041\u003c/li\u003e\n\u003cli\u003eSinger M, Deutschman CS, Seymour CW, et al (2016) The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 315:801\u0026ndash;810. https://doi.org/10.1001/jama.2016.0287\u003c/li\u003e\n\u003cli\u003eKhanna A, English SW, Wang XS, et al (2017) Angiotensin II for the treatment of vasodilatory shock. N Engl J Med 377:419\u0026ndash;430. https://doi.org/10.1056/NEJMoa1704154\u003c/li\u003e\n\u003cli\u003eHern\u0026aacute;n MA, Wang W, Leaf DE (2022) Target trial emulation: A framework for causal inference from observational data. JAMA 328:2446\u0026ndash;2447. https://doi.org/10.1001/jama.2022.21383\u003c/li\u003e\n\u003cli\u003eZhao SS, Lyu H, Yoshida K (2021) Versatility of the clone-censor-weight approach: response to \u0026lsquo;trial emulation in the presence of immortal-time bias. Int J Epidemiol 50:694\u0026ndash;695\u003c/li\u003e\n\u003cli\u003eSharshar T, Blanchard A, Paillard M, et al (2003) Circulating vasopressin levels in septic shock. Crit Care Med 31:1752\u0026ndash;1758. https://doi.org/10.1097/01.CCM.0000063046.82359.4A\u003c/li\u003e\n\u003cli\u003eLandry DW, Levin HR, Gallant EM, et al (1997) Vasopressin deficiency contributes to the vasodilation of septic shock. Circulation 95:1122\u0026ndash;1125. https://doi.org/10.1161/01.cir.95.5.1122\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"intensive-care-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"icme","sideBox":"Learn more about [Intensive Care Medicine](http://link.springer.com/journal/134)","snPcode":"134","submissionUrl":"https://www.editorialmanager.com/icme/default2.aspx","title":"Intensive Care Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"septic shock, vasopressin, norepinephrine, treatment timing, target trial emulation, clone-censor-weight","lastPublishedDoi":"10.21203/rs.3.rs-9454619/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9454619/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003ePurpose\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhether earlier initiation of adjunctive vasopressin improves outcomes in patients with septic shock remains unclear. Previous studies have compared vasopressin use with non-use and defined time zero as shock onset rather than norepinephrine escalation, making it difficult to distinguish the effect of vasopressin addition from that of earlier timing. We therefore conducted a targeted trial emulation, defining time zero as norepinephrine escalation to ≥ 0.25 µg/kg/min to compare ultra-early (0–3 hours) with early (\u0026gt; 3–6 hours) vasopressin initiation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe emulated a target trial using two large electronic health record databases (MIMIC-IV 2008–2022 and eICU-CRD 2014–2015). Adults with septic shock whose norepinephrine-equivalent dose reached ≥ 0.25 µg/kg/min were eligible. Time zero was the first time the norepinephrine threshold was met. The primary outcome was the 28-day mortality. We used the clone-censor-weight method to estimate cumulative incidence curves under the ultra-early and early initiation strategies and compared them via the weighted Cox model.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong 5,157 eligible patients, ultra-early initiation lowered 28-day mortality (35.7% versus 41.2%; risk difference − 5.5 percentage points, 95% confidence interval [CI] − 6.9 to − 4.4; restricted mean survival time difference + 1.49 days, 95% CI 1.18–1.82; hazard ratio [HR] 0.81, 95% CI 0.78–0.85). Ultra-early initiation also showed lower rates of acute kidney injury (KDIGO stage ≥ 2: HR 0.94, 95% CI 0.92–0.97) and renal replacement therapy (HR 0.65, 0.61–0.70) including continuous renal replacement therapy (HR 0.75, 0.68–0.83).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong patients with septic shock whose norepinephrine dose reached ≥ 0.25 µg/kg/min, ultra-early vasopressin initiation (0–3 h) was associated with lower 28-day mortality than early initiation (\u0026gt; 3–6 h).\u003c/p\u003e","manuscriptTitle":"Ultra-early versus early adjunctive vasopressin initiation after norepinephrine escalation in septic shock: a target trial emulation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-07 20:09:47","doi":"10.21203/rs.3.rs-9454619/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Major revisions","date":"2026-05-14T06:00:41+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"","date":"2026-04-23T14:29:29+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-23T14:24:55+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-20T03:24:22+00:00","index":"","fulltext":""},{"type":"submitted","content":"Intensive Care Medicine","date":"2026-04-18T02:14:53+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"intensive-care-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"icme","sideBox":"Learn more about [Intensive Care Medicine](http://link.springer.com/journal/134)","snPcode":"134","submissionUrl":"https://www.editorialmanager.com/icme/default2.aspx","title":"Intensive Care Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"f0d12a38-6e1f-4564-90f8-27b3a23fb639","owner":[],"postedDate":"May 7th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Major revisions","date":"2026-05-14T06:00:41+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-05-14T10:11:08+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-07 20:09:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9454619","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9454619","identity":"rs-9454619","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}