Early Serial Assessment of Aggregate Vasoactive Support and Mortality in Cardiogenic Shock: Insights From the Critical Care Cardiology Trials Network Registry.
Written by: Siddharth M PatelDavid D BergErin A BohulaVivian M Baird-ZarsGregory W BarsnessSunit-Preet ChaudhryMeshe D ChondeHoward A CooperCurtis GinderJacob C JentzerMichael C KontosP Elliott MillerL Kristin NewbyConnor G O'BrienJeong-Gun ParkMatthew J PierceBarbara A PisaniBrian J PotterKevin S ShahJeffrey J TeutebergJason N KatzSean van DiepenDavid A Morrow
Find these authors on Roon
SC
Dr. Sunit-Preet Chaudhry
Advanced Heart Failure · Ascension St. Vincent Hospital
CG
Dr. Curtis Ginder
Cardiac Critical Care · PHS
CO
Dr. Connor O'Brien
Cardiac Critical Care · UCSF Medical Center
KS
Dr. Kevin Shah
Advanced Heart Failure · MemorialCare
Introduction
Cardiogenic shock (CS) is characterized by insufficient cardiac output resulting in end-organ hypoperfusion. Vasoactive agents, typically either vasopressors and/or inotropes, are commonly used as first-line therapy to maintain systemic perfusion in CS.1 Accordingly, assessment of the intensity of pharmacologic support, specifically the quantification of aggregate vasoactive dosing, may be used as a marker of CS severity and as an independent correlate of mortality.2–5 While studies have previously examined associations between the intensity of vasoactive support at presentation and outcomes, CS is a dynamic state in which iterative assessment of overall trajectory may influence clinical assessment of severity and guide decision making for other therapies.6,7 In this context, the relationships between early changes in aggregate vasoactive burden and outcomes in CS are not well-defined. Such serial assessment of CS trajectory, specifically with respect to increasing or decreasing overall pharmacologic support, may serve as an important quantitative metric to facilitate identification of patients for early consideration of adjunctive therapies such as mechanical circulatory support (MCS)7 prior to the onset of irreversible end-organ injury. Additionally, whether any associations between the intensity of pharmacologic support over time and outcomes in CS differ in patients across subtypes of CS or those managed with concomitant MCS is uncertain.8Accordingly, we sought to better characterize the prognostic performance of early static and dynamic assessments of aggregate vasoactive burden in a broad range of patients presenting with CS, with and without concomitant MCS, across cardiac intensive care units (CICUs) within the Critical Care Cardiology Trials Network (CCCTN) registry.
Methods
CCCTN is an investigator-led, collaborative research network of CICUs in North America coordinated by the TIMI Study Group (Boston, MA). Methods for the CCCTN registry have been reported.9 The CCCTN registry protocol and waiver of informed consent were approved by the institutional review committees at each of the participating centers. We encourage parties interested in collaboration and data sharing to contact the corresponding author directly for further discussions.Between 2018–2023, participating sites contributed clinical data on all consecutive CICU admissions during annual 2-month collection periods, with a subset of sites also reporting data on all consecutive admissions with CS year-round. The presence and subtype of shock were assessed by site investigators using an established definition as previously reported.10 For patients with CS, CS etiology was further subdivided into CS due to acute myocardial infarction (AMI-CS) or heart failure related CS (HF-CS).11 The present analysis was restricted to patients presenting with CS at the time of CICU admission. Patients with mixed shock and post-cardiotomy shock were not included in this analysis.
Dosing information for all vasoactive agents (i.e., vasopressors and inotropes) was captured in the electronic case record form (eCRF) at 4 and 24 hours (h) from CICU admission. The 4h timepoint was chosen as an initial snapshot of vasoactive burden allowing for some time for stability of vasoactive selection and dosing after CICU arrival, with the 24h timepoint chosen for iterative assessment of overall trajectory. To quantify the intensity of vasoactive support at each time point, the vasoactive-inotropic score (VIS) was calculated as follows: VIS = dobutamine + dopamine + 10phenylephrine + 10milrinone + 100epinephrine + 100norepinephrine + 10,000*vasopressin, where doses for all agents are in μg/kg/min with the exception of vasopressin which is in units/kg/min.12 VIS was chosen as the metric of quantification owing to its ability to incorporate all vasopressors and inotropes commonly used in the United States. The timing of mechanical circulatory support (MCS) insertion and removal was also captured for each patient.
Clinical characteristics were summarized as counts and percentages for categorical variables and medians with 25th – 75th percentiles for continuous variables. Differences in characteristics across ordinal categories were examined using the Cochran-Armitage test for categorical variables and the Jonckheere-Terpstra test for continuous variables. The Sequential Organ Failure Assessment (SOFA) score was calculated for each patient using data from the first 24h of CICU admission.13 Clinician assessment of the Society of Cardiovascular Angiography and Intervention (SCAI) shock stage, reflective of the first 24h of the CICU course, was captured in the eCRF.4,6
VIS at 4h and 24h was assessed both continuously and categorically using thresholds of <10, 10-<20, 20-<40, and ≥40, guided by our previous application and VIS distribution.4 A sensitivity analysis was performed considering patients with VIS=0 as a distinct category. The relationship between continuous VIS at each time point with CICU mortality was depicted graphically using restricted cubic spline analysis, and the association of VIS with CICU mortality (per 10-point change in VIS) was assessed using logistic regression. The primary analytic population consisted of patients with VIS available at both time points; however, a sensitivity analysis was performed also including patients with available 4h VIS but missing 24h VIS to examine consistency of the prognostic associations with CICU mortality.
The change in VIS from 4h to 24h was examined both categorically and continuously. Categories for change in VIS were consistent with prior application (≥10-point decrease, <10-point change and ≥10-point increase),4 with a supplementary analysis examining more granular change categories. All analyses examining associations between the change in VIS with CICU mortality accounted for the starting value of VIS at 4h, using stratification for the categorical analyses and logistic regression conditioned on the 4h VIS for the continuous analyses. Additionally, subgroup analyses were performed by subtype of CS (AMI-CS and HF-CS) and cardiac arrest status. A sensitivity analysis was performed to assess consistency of the prognostic relationships among those initially presenting to the CICU vs. those transferred from outside hospitals. Analyses were also conducted stratified by transplant center status to examine consistency of the prognostic associations in cases where transplant listing status may have influenced vasoactive selection/dosing.
For the primary analyses, patients with insertion or removal of MCS within ±1h of the 4h and 24h VIS time point were excluded to allow for stability of vasoactive dosing. Interaction testing was performed by MCS status at each VIS time point. An additional analysis was conducted to assess the influence of changes in MCS status (i.e., insertion or removal) between the 4h and 24h time points on VIS. The prognostic associations between VIS and mortality were further examined in subsets of patients without any MCS use in the first 24h and in those supported by MCS from admission through 24h. Interaction testing was performed between VIS at each time point and MCS status by including a 2-way interaction term for VIS and presence or absence of MCS in the logistic regression model.
Discrimination (C-statistic) of CICU mortality by VIS at each time point, and by the change in VIS accounting for the starting value, was assessed and contrasted with both the SOFA score and the clinician-assessed SCAI shock stage at 24h. To examine whether the change in VIS was additive to the simplified vasoactive categories captured by the SOFA score, after ensuring lack of multicollinearity (variance inflation factor <2), a nested model of the 24h SOFA score and the change in VIS at 24h (accounting for the initial 4h VIS) was used to examine change in discrimination with the likelihood ratio test. Whether change in VIS was additive to the clinician-assessed SCAI shock stage was examined in a similar fashion. Two-sided p-values <0.05 were considered statistically significant. Analyses were performed using SAS System V9.4 (SAS Institute Inc., Cary, NC).
Results
A total of 3,665 patients across 44 centers presented with CS with available VIS at 4h and 24h and met inclusion criteria for this analysis (Figure S1). The median 4h VIS was 4.0 (25th–75th percentile: 0.0–10.0), with 75%, 12%, 9% and 5% of patients having a 4h VIS of <10, 10-<20, 20-<40, and ≥40, respectively. The median 24h VIS was 3.0 (0.0–8.0), with 78%, 11%, 7% and 4% having a 24h VIS of <10, 10-<20, 20-<40, and ≥40.
The majority of patients (n=3,013, 82%) had minimal change in VIS (<10), with 11% (n=387) and 7% (n=265) having a ≥10 decrease and increase in VIS from 4h to 24h, respectively. The distribution of change in VIS was similar when restricted to patients in non-transplant centers (total N=845): 81% (n=681) with change in VIS <10, 12% (n=98) with ≥10 decrease and 8% (n=66) with ≥10 increase. Stratifying the change in VIS by the initial 4h VIS, a greater proportion of those with a higher starting VIS at 4h had decreases ≥10 by 24h (Figure 1). Baseline characteristics by change in VIS are presented in Table 1. Patient demographics were generally similar across the change in VIS categories. An admission diagnosis of acute coronary syndrome was more prevalent, and heart failure less prevalent, among those with an increase or decrease in VIS ≥10 compared to those with no change (p-trend p<0.001 for each). As well, those with an increase or decrease in VIS ≥10 had greater prevalence of cardiac arrest and higher 24h SOFA scores (p-trend = 0.001 and <0.0001, respectively).
Individually, there was a graded association between VIS at each time point with CICU mortality when assessed across VIS categories, with mortality ranging from 13% to 45% for the 4h VIS and from 11% to 73% for the 24h VIS (p-trend < 0.001 for each; Figure S2). No meaningful difference was observed with separate categorization of patients with VIS=0 from VIS >0-<10 at both time points (Figure S3). This relationship was consistent when the 4h and 24h VIS were assessed continuously (OR per 10-point higher 4h VIS: 1.30 [95% CI 1.24–1.36]; 24h VIS: OR 1.66 [95% CI 1.56–1.77]; p<0.001 for each; Figure S4). Results for the 4h VIS were similar in a sensitivity analysis including patients with available 4h VIS but with missing 24h VIS (n=2,706; Figure S5). The c-statistic for mortality discrimination by the 4h VIS was 0.66 (95% CI 0.63–0.68) and appeared even stronger at 24h (C-statistic 0.72 [95% CI 0.70–0.75]). The latter was comparable to that of the 24h SOFA score (C-statistic 0.72 [95% CI 0.69–0.74]) and the clinician SCAI stage (C-statistic 0.72 [95% CI 0.70–0.74]).
Stratifying by the starting 4h VIS, an increase in VIS ≥10 from 4h to 24h was associated with at least a 40% to a >4-fold higher CICU mortality compared with no change or VIS decrease ≥10 (p-trend <0.001 for each; Figure 2 and Figure S6). Assessed continuously and accounting for the initial 4h VIS, increases or decreases in VIS from 4h to 24h were associated with higher or lower CICU mortality, respectively (OR per 10-point increase in VIS from 4h to 24h: 1.83 [95% CI 1.64–2.05], p<0.001; Figure 3). Similar prognostic relationships were observed among subsets of patients with and without cardiac arrest (Figures S7A and S7B).
Stratifying by subtype of CS, overall CICU mortality was 28% in those with AMI-CS and 16% in those with HF-CS. An increase in VIS was associated with higher CICU mortality across both CS subtypes (OR per 10-point increase for HF-CS and AMI-CS: 1.87 [95% CI 1.59–2.20] vs. 1.60 [95% CI 1.34–1.90]; p-interaction = 0.30; Figures S8A and S8B). A total of 1,709 patients (46.6%) were transfers from an outside hospital. Prognostic associations were consistent irrespective of transfer status (OR per 10-point increase in VIS for transfers vs. non-transfers: 1.81 [95% CI 1.55–2.11] vs. 1.85 [95% CI 1.59–2.16]; p-interaction = 0.83; Figures S9A and S9B). Similar prognostic associations were also seen irrespective of transplant center status (p-interaction = 0.44; Figures S10A and S10B)
Accounting for the initial 4h VIS, the change in VIS at 24h had good discrimination of CICU mortality (C-statistic 0.72 [95% CI 0.70–0.75]) and improved mortality discrimination when added to the 24h SOFA score (C-statistic 0.76 [95% CI 0.74–0.78]; LRT p-value < 0.0001) and clinician SCAI stage (C-statistic 0.77 [95% CI 0.75–0.79]; LRT p-value <0.0001).
A total of 840 patients were supported by MCS for the full 24h period, and 2,607 patients did not receive any MCS during the initial 24h. Static VIS measures were higher among those with concomitant MCS at both timepoints compared to those without (median 4h VIS: 5.0 [0.0–13.4] vs. 3.5 [0.0–8.4], p <0.0001 and median 24h VIS: 5.0 [0.0–12.1] vs. 2.5 [0.0–7.3], p <0.0001). Assessed dynamically, increases or decreases in VIS ≥10 were generally more common among those on concomitant MCS through 24h as compared to those without any MCS throughout this timeframe (Figure S11A).
Between the 4h and 24h timepoints, 171 patients were initiated on MCS and 98 had removal of MCS. Static VIS measures were also higher at each timepoint among those with MCS insertion compared to MCS removal between these timepoints (7.5 [3.0–19.1] vs. 5.0 [0.0–15.7] at 4h and 7.5 [2.5–15.2] vs. 2.5 [0.0–10.5] at 24h, respectively; p=0.008 and p<0.001, respectively). Insertion or removal of MCS between the 4h and 24h timepoints had no significant association with the change in VIS when assessed categorically (all pairwise p-values ≥ 0.05; Figure S11B).
Overall CICU mortality was 22% in those with MCS and 16% in those without MCS at any time point within the first 24h. A higher 4h VIS was associated with higher mortality in those with and without concomitant MCS; however, the mortality risk relationship appeared to be attenuated among those receiving concomitant MCS compared to those without (OR per 10-point higher 4h VIS for MCS vs. no MCS: 1.18 [1.09–1.28] vs. 1.38 [1.29–1.46]; p-interaction = 0.003). A similar relationship was observed for the 24h VIS (OR per 10-point higher 24h VIS in those receiving MCS vs. not: 1.36 [1.23–1.49] vs. 1.84 [1.69–2.01]; p-interaction < 0.0001).
Discussion
In this analysis examining aggregate vasoactive support in CS from the CCCTN registry, we demonstrated a robust association between the magnitude of vasoactive support at 4h and 24h from CICU admission with CICU mortality. Accounting for the initial 4h VIS, we observed that increases or decreases in vasoactive therapy were associated with higher or lower risk of mortality, respectively. Notably, we found that quantification of aggregate vasoactive support alone offered comparable mortality discrimination to the 24h SOFA score and clinician-assessed SCAI shock stage, both of which are commonly utilized for ICU risk stratification.6,13 Moreover, we demonstrated that integration of the 24h VIS trajectory improved discrimination when combined with the SOFA score and SCAI shock stage. We further extended these findings across subtypes of CS and to those receiving concomitant MCS, demonstrating a similar graded association between static or dynamic VIS assessments and CICU mortality, though with possible differences in the magnitude of relative risk by MCS status. Collectively, these data highlight that early temporal trajectory of the intensity of vasoactive support provides important prognostic insight among patients presenting with CS, and, as a quantitative metric that could easily be embedded within electronic health record (EHR) systems, may be helpful to guide clinical decision making and for application in future research in CS.
Prior studies have examined the association between aggregate vasoactive dosing and mortality. Specifically within the CICU population, a single-center study by Jentzer and colleagues of unselected CICU patients receiving vasoactive therapy demonstrated an association between the maximum VIS throughout the CICU course with in-hospital mortality.2 Examining patients specifically with CS, in an analysis of the Korean RESCUE registry of patients presenting with AMI-CS (n=1247 across 12 centers in Korea), investigators demonstrated that the maximum VIS within the first 48h was an independent predictor of mortality, and further observed a potential interaction by MCS status.3
Our report leverages a broad range of patients with investigator-classified CS to demonstrate a robust gradient for CICU mortality by VIS within the first 24h and across etiologic subtypes (e.g., AMI-CS or HF-CS). Moreover, by examining not only the prognostic associations with these single static measures but also dynamic assessments (i.e., the change in VIS) afforded by the larger sample size of our study, we demonstrate the prognostic utility of iterative assessment of VIS trajectory, identifying a 2-to->4 fold gradient for mortality based on the magnitude of change in aggregate vasoactive support over this initial 24h timeframe where recognition of such escalating needs may be critical to timely institution of adjunctive therapies prior to the onset of irreversible end-organ injury.
We observed that a significant proportion of patients in this cohort had a change in VIS <10 from 4h to 24h, a subset which had greater prevalence of AMI-CS and lower prevalence of preceding cardiac arrest, coupled with lower overall illness severity, when compared to those with either an increase or decrease in VIS ≥ 10. It is possible that this subset of patients had smaller dynamic changes in the magnitude of vasoactive support owing to a rapidly fixable initial etiology of CS that is not captured by the timepoints of VIS assessments in our study (e.g., a patient initially presenting to the emergency room with CS noted to have AMI, with prompt intervention in the catheterization laboratory prior to CICU arrival allowing for rapid weaning of vasoactives prior to the 4h timepoint).
Notably, we also observed differences in calibration of mortality risk by VIS when examined by MCS status, with VIS appearing to have greater prognostic relevance in those without versus with MCS at each time point. This finding is consistent with prior observations from the RESCUE registry restricted to patients presenting with AMI-CS, and may be driven by selection bias for which patients are candidates for MCS to begin with (e.g., those of younger age, with lower co-morbidity burden, and with greater likelihood of a potentially reversible etiology of CS) or alternatively could reflect a possible treatment benefit of MCS; however, we are unable to further disentangle these possible drivers within the confines of an observational dataset.
Interestingly, while initiation of MCS is classically thought to allow for weaning of pharmacologic support, we observed a persistently higher VIS among patients undergoing insertion of MCS prior to this time point and, in contrast, a lower VIS in those undergoing MCS removal. It is likely these findings predominantly reflect the severity of the CS presentation in this population, particularly within this initial 24h window. Moreover, in examining the association between MCS insertion or removal status with change in VIS, the proportion of patients with a VIS decrease ≥10 was numerically greater among those undergoing MCS insertion between these timepoints compared to those with MCS removal (17.5% vs. 6.1%, p=0.06), which may represent a possibly meaningful trend in this group.
Our findings have potential implications for clinical practice and for research applications. While conventional risk scores including both SOFA and SCAI include simplified capture of vasoactive agents, our findings suggest that the more comprehensive quantification of vasoactive burden offers important prognostic value.4 Moreover, it is well appreciated that CS presents with a range of severity, and that CS trajectory, in particular the response to therapies or more importantly, the lack of response or deterioration despite initial attempts, informs prognosis and should guide escalation or even palliation if consistent with patient and family wishes.1,7 To this end, the 2021 Society of Cardiovascular Angiography and Interventions (SCAI) Shock Stages update showcases potential recovery and deterioration pathways in CS, emphasizing the need for iterative assessment of shock severity based on trajectory given the potential implications for clinical decision-making.6 In prior work, we demonstrated how changes in the intensity of vasoactive support could be used to reclassify shock severity staging within the SCAI staging framework.4 While the VIS is not intended to be calculated by the clinician at the bedside, integration of iterative or even continuous VIS assessment leveraging dosing information within contemporary EHR systems could be helpful to seamlessly alert busy clinicians to the need to reconsider treatment approaches based on changes in trajectory. Moreover, such information could be integrated with the wealth of other data routinely available within EHRs including clinical variables, biomarkers of end-organ injury, and invasive hemodynamic data, possibly in the form of electronic clinical decision support tools that more comprehensively consider a patient’s trajectory and response to therapies. Such data may even be helpful to guide optimal allocation of advanced therapies including cardiac transplantation in patients presenting with CS.
Additionally, such assessments of trajectory may also be helpful to guide optimal selection of patients for entry into clinical trials of adjunctive CS therapies. Notably, the majority of randomized clinical trials to date have not demonstrated a clinical benefit with the use of MCS in patients with CS.14–16 However, MCS may be unlikely to benefit patients who are otherwise able to be stabilized with only transient low-dose pharmacologic support while addressing other contributing factors (e.g., volume status or active ischemia), particularly given the tradeoffs in MCS-associated complications. Conversely, in those with persistent instability despite rapid escalation of pharmacologic support, equipoise may no longer exist to allow for randomization or considerations of futility may become relevant to clinical decision-making. Accordingly, identification of patients with CS between these two extremes, when there is an early signal of worsening but before irreversible organ-injury and there remains both equipoise and plausibility of benefit for adjunctive therapies, may be useful to support future well designed clinical trials in this space. Such precise automated quantification of pharmacologic support could be leveraged as a trial enrollment criterion in this context.
Several limitations of our study should be acknowledged. We used VIS to quantify aggregate vasoactive burden due to its ability to include a wide array of both vasopressor and inotropic agents; prognostic associations across different vasoactive scores may vary. Additionally, dose equivalents across vasoactive agents remains poorly defined, with agent-specific weights in the VIS determined arbitrarily to allow for dose comparison on a similar scale of magnitude as opposed to being determined by associations with mortality; future iterations of vasoactive scores might develop alternative weighting guided by the prognostic relationships. Serial VIS assessments in our study were performed only at 4h and 24h from CICU admission. Accordingly, the distribution of change in VIS observed in our study should be interpreted in the context of these timepoints of assessment, which reflect the initial adjustment of vasoactive medications on CICU arrival. Other time points or different durations may also carry relevant prognostic information. For analyses examining changes in VIS, patients with missing 24h VIS were excluded by necessity. While we explored the consistency of the prognostic associations across several subgroups, information on concomitant sedation was not captured, limiting our ability to explore the effects of sedation-related vasoplegia in this dataset. The CCCTN registry predominantly includes academic tertiary care CICUs and thus, the absolute risk estimates from this cohort may not be generalizable to other settings. In addition, the clinical course and testing at the external referral hospital may offer incremental information for the subset of patients presenting as outside hospital transfers; however, we observed consistent prognostic associations irrespective of transfer status.
Conclusion
Among patients presenting with CS, early changes in the magnitude of vasoactive support were associated with a gradient of risk for mortality. These data suggest that early serial assessment of aggregate vasoactive support can provide important incremental prognostic information, which has the potential to be leveraged for clinical decision making and research applications.
SC
Dr. Sunit-Preet Chaudhry
Cardiology · Advanced Heart Failure & TransplantationSt Vincent Hospitalsunitpreet
