Outcomes after coronary artery bypass grafting in patients with myocardial infarction, cardiogenic shock and unresponsive neurological state: analysis of the Society of Thoracic Surgeons Database

Abstract

OBJECTIVES

Previous studies have demonstrated a 20% mortality rate among patients undergoing isolated coronary artery bypass grafting (CABG) for cardiogenic shock. However, outcomes following CABG for cardiogenic shock in patients who are neurologically unresponsive preoperatively are unknown.

METHODS

Utilizing the Society of Thoracic Surgeons Adult Cardiac Surgery Database between July 2011 and December 2013, patients undergoing urgent or emergent CABG within 7 days of an acute myocardial infarction complicated by cardiogenic shock were identified. Patients were stratified on the basis of whether they had a non-medically induced unresponsive state within 24 h of surgery.

RESULTS

Of the 5259 patients with acute myocardial infarction complicated by cardiogenic shock who underwent CABG during the study period, 243 (4.62%) patients had an unresponsive preoperative neurological state. The unresponsive cohort had a higher 30-day operative mortality than the responsive cohort (33.74% vs 16.91%, P < 0.001). Unresponsive neurological state was associated with increased odds for mortality (adjusted odds ratio 1.81, 95% confidence interval 1.37–2.4; P < 0.001), postoperative stroke (adjusted odds ratio 2.17, 95% confidence interval 1.27–3.73; P = 0.0048) and encephalopathy (adjusted odds ratio 2.08, 95% confidence interval 1.44–3.01; P < 0.001). Among survivors in the unresponsive cohort, 78 (46.15%) were discharged home and 62 (36.69%) were discharged to extended care facilities.

CONCLUSIONS

Although cardiac surgery in unresponsive patients in the setting of acute myocardial infarction complicated by cardiogenic shock is associated with considerable neurological disability and mortality, the majority survive to discharge. These findings may help guide patient and family discussions regarding goals of care.

INTRODUCTION

Patients who develop cardiogenic shock after myocardial infarction (MI) have an associated mortality ranging from 15% to 40% with revascularization with either percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) [1–3]. According to the Should We Emergently Revascularize Occluded Coronaries for Cardiogenic Shock (SHOCK) trial, European Society of Cardiology and European Association for Cardio-Thoracic Surgery Guidelines on myocardial revascularization, emergent revascularization in the setting of acute myocardial infarction complicated by cardiogenic shock (AMI-CS) has a significant survival advantage after 6 months compared to the initial medical stabilization [4–6]. Survival was similar between the PCI and CABG groups in the SHOCK trial. Therefore, emergent CABG is a reasonable option in those patients with AMI-CS and extensive coronary disease [7].

However, there are limited data regarding the influence of preoperative neurological status on the potential for postoperative recovery following CABG in the setting of AMI-CS. A single-centre study by Hosmane et al. [8] investigated predictors of neurological recovery after undergoing resuscitation for cardiac arrest due to ST-elevation MI. Of the 98 total patients, only 7 patients with preoperative unresponsive neurological state underwent revascularization via CABG, and of those patients, 3 had full recovery, 2 had residual deficits and 2 died. The aim of this study was to determine the contemporary outcomes associated with the use of CABG in patients with cardiogenic shock and an unresponsive neurological state by using a nationwide, multi-institutional database. Specific emphasis was placed on postoperative mortality and discharge location, as these clinical outcomes are important to patients, caregivers and providers.

MATERIALS AND METHODS

Data source

The Society of Thoracic Surgeons (STS) Adult Cardiac Surgery Database (ACSD) contains more than 5 million records for adult cardiac surgery. This captures 95% of all cardiac surgical procedures performed in the USA. The rationale and methodology of data collection for the STS ACSD have been previously described [9, 10].

Patients

Registry records of patients in the STS ACSD who had MI complicated by cardiogenic shock within 7 days prior to undergoing non-elective CABG were included. Only records from operations between July 2011 and December 2013 coinciding with version 2.73 of the STS data collection form were included. Patients with a preoperative ventricular assist device or undergoing concomitant ventricular assist device implantation with CABG were excluded (Fig. 1).

Data definitions

This study utilizes the American College of Cardiology Foundation/American Heart Association definition of MI [11]. A patient is considered to be in cardiogenic shock if there was a sustained (>30 min) clinical state of hypoperfusion due to cardiac failure prior to the procedure. Cardiac failure is defined as persistent hypotension (systolic blood pressure <80 mmHg or mean arterial pressure 30 mmHg lower than baseline) with a severe reduction in cardiac index (<1.8 l/min/m² without mechanical or inotropic support or <2.0 l/min/m² with mechanical or inotropic support), in the setting of adequate or elevated filling pressures. Urgent operative status is defined as the procedure required during same hospitalization to minimize the chance of further clinical deterioration. Emergent operative patients are defined as those having ongoing, refractory, unrelenting cardiac compromise, with or without haemodynamic instability, and unresponsive to any form of therapy except cardiac surgery. Emergent salvage patients are defined as those undergoing cardiopulmonary resuscitation en route to the operating room, before anaesthesia induction or having ongoing ECMO to maintain life. Resuscitation is defined as cardiopulmonary resuscitation within 1 h of the start of the operative procedure. STS defines unresponsive neurological state as a non-medically induced unresponsive state prior to surgery and encephalopathy as any patient who developed coma and/or encephalopathy postoperatively [12]. STS predicted risk of mortality (PROM) was stratified to 4 categories including <4%, 4–12%, 12–20% and >20% as previously performed using the STS ACSD [13]. Operative mortality includes all deaths occurring during the index hospitalization or within 30 days of the procedure. Other definitions are located on the STS website [12].

Statistical analysis

Standard summary statistics were used to analyse baseline characteristics, operative variables and postoperative outcomes of the overall cohort stratified by responsive and unresponsive neurological state preoperatively. Categorical variables were calculated as frequency and percentage, while continuous variables were calculated as median and interquartile range. The χ² tests were used to compare selected categorical data, and Wilcoxon tests compared continuous data between the responsive and non-responsive patients. Kaplan–Meier curves assessed survival up to 30 days. For patients whose mortality status at 30 days was deceased but date of death was missing, time of death was assigned at 30 days. Patients with time to death beyond 30 days were censored as alive at 30 days. The stroke variable had missing values for 12 (0.23%) patients who were excluded from the model. Log-rank test compared survival between the responsive and unresponsive cohorts. Multivariable logistic regression models were used to evaluate predictors of mortality, stroke and encephalopathy in the overall cohort and of mortality in the unresponsive cohort. Initial variables were those included in the STS CABG mortality risk model with the addition of left main disease, centre census region and centre teaching status [10]. Stepwise logistic regression was utilized with an entry and stay of alpha = 0.05. Continuous outcomes, including intensive care unit hours and length of stay, were analysed with adjusted Poisson regression models.

All analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC, USA). Exemption from institutional review board approval was obtained prior to data analysis.

RESULTS

Baseline characteristics

Between July 2011 and December 2013, a total of 5259 patients underwent a non-elective CABG within 7 days of AMI-CS. These patients represented 1.45% of all patients who underwent CABG during the study period. Of the AMI-CS cohort, 4.62% (n = 243) were identified as having a non-responsive state preoperatively.

Baseline characteristics such as age, gender, race, prior cerebrovascular accident and prior CABG were similar in responsive and non-responsive patients (Table 1). Of the responsive patients, 20.41% (n = 1024) required resuscitation compared to 59.26% (n = 144) of unresponsive patients (P < 0.001). The unresponsive cohort had a larger proportion of patients with higher STS PROM compared to the responsive cohort, such that 46.91% of unresponsive patients were in the >20% PROM group compared with only 29.31% of responsive patients (P < 0.001).

Outcomes

The unresponsive cohort had a significantly higher 30-day operative mortality than the responsive cohort (33.77% vs 16.91%, P < 0.001) (Fig. 2). The unresponsive group also had higher incidence of permanent stroke (7.41% vs 3.25%, P < 0.001) and postoperative encephalopathy (16.87% vs 8.67%, P < 0.001) compared to the responsive group. The unresponsive cohort spent a median of 126.8 h (IQR 61.0–262.8) in the intensive care unit, whereas the responsive cohort spent a median of 106.7 h (IQR 58.8–207.0, P = 0.089). Overall length of stay was similar (median of 8.0 days) for both cohorts (P = 0.27) (Table 2).

Cause of death stratified by preoperative neurological status is presented in Table 2. The cause of death was cardiac in 78.05% of unresponsive patients and 75.47% of responsive ones (P = 0.23). Cause of death secondary to a neurological event was 4.95% (n = 42) in the responsive cohort and 10.98% (n = 9) in the unresponsive cohort (P = 0.23) (Table 2).

Regarding discharge location, a substantial proportion of both unresponsive and responsive patients were discharged home; however, a larger proportion of responsive (vs unresponsive) patients were discharged to home (58.84% vs 46.15%, P < 0.001). A higher proportion of the unresponsive cohort was discharged to extended care or nursing facilities (38.5% vs 31.5%, P < 0.001).

Multivariable analysis: overall cohort

For the overall cohort, multivariable predictors of operative mortality included unresponsive neurological state, age, atrial fibrillation, New York Heart Association (NYHA) Class IV congestive heart failure, creatinine, diabetes, ejection fraction, dialysis, female gender, PCI within 6 h of surgery, peripheral vascular disease, operative status and reoperation. Predictors of postoperative stroke were unresponsive neurological state, NYHA IV congestive heart failure, creatinine, cardiovascular disease and prior cerebrovascular accident and urgent status. Predictors of postoperative encephalopathy included unresponsive neurological state, creatinine, dialysis, peripheral vascular disease and immunosuppression (Table 3).

Having an unresponsive preoperative neurological state was associated with an adjusted odds ratio (OR) of 1.81 for mortality, 2.17 for postoperative stroke and 2.08 for postoperative encephalopathy (Table 3). Unresponsive neurological state was not an independent predictor of ICU hours or length of hospital stay in survivors.

Multivariable analysis: unresponsive group

For the unresponsive cohort, multivariable predictors of mortality were CHF (non-NYHA Class IV) [OR 0.26, 95% confidence interval (CI) 0.08–0.84; P = 0.024], creatinine (OR 5.32, 95% CI 2.51–11.28; P < 0.001) PCI ≤6 h (OR 3.43, 95% CI 1.51–7.79; P = 0.003), peripheral vascular disease (OR 4.03, 95% CI 1.80–9.00; P < 0.001) and emergent/emergent salvage status (OR 8.28, 95% CI 2.20–31.22; P = 0.002) (Table 4).

DISCUSSION

The leading cause of death after MI is cardiogenic shock [1]. Multiple studies, including the SHOCK trials, have demonstrated a survival advantage with early revascularization with either PCI or CABG [2, 4–6]. Survival was similar between PCI and CABG, suggesting emergency CABG as an important treatment modality for appropriately selected patients [7]. However, there is hesitation in operating on these very high-risk patients, especially when neurological status is in question.

Several questions merit considerations in the complex decision on whether to operate on patients with neurological dysfunction or unresponsive neurological state. First, what is the likelihood of reversibility of neurological dysfunction in resuscitated, but comatose patients with MI? An analysis of 28 studies with over 4500 such patients with ST-elevation myocardial infarction managed with therapeutic hypothermia and early coronary angiography (and PCI in those with revascularizable disease) showed 60% survival to discharge with favourable neurological outcome in 86% of survivors [14]. Therefore, an unresponsive neurological state on presentation does not necessarily indicate irreversibility or a hopeless prognosis. Second, what is the anticipated benefit of revascularization? A young patient with few comorbidities, witnessed arrest, short resuscitation time and clear culprit lesions on angiogram is likely to recover from surgery, despite an unresponsive neurological state at presentation. However, if there are features of poor prognosis including unwitnessed arrest, no bystander resuscitation, prolonged resuscitation, persistent unresponsiveness several days after arrest, end-organ dysfunction and very advanced age, limited benefit may be gained by CABG, and careful assessment of the risks and benefits of surgery is warranted. Third, what is the cause of the unresponsive neurological state, particularly in patients without cardiac arrest? If it is stroke, drugs or respiratory failure, coronary revascularization with CABG will not improve prognosis. Finally, is CABG likely to reverse the underlying process? In patients with profound shock or multisystem organ failure, or those with late presentation and potentially irreversible myocardial injury, revascularization alone may not improve outcomes, and haemodynamic support with mechanical circulatory support may also need to be considered.

As mentioned, very little data exist on outcomes after CABG in unresponsive patients with AMI-CS. Hosmane et al. [8] investigated survival and neurological recovery in patients with ST-elevation myocardial infarction requiring resuscitation from cardiac arrest. Within their cohort of 98 patients, 59 (60.2%) patients were unresponsive after resuscitation, and 40 of the 59 patients underwent revascularization. Of the unresponsive patients, those who underwent revascularization had significantly lower in-hospital mortality compared to no revascularization (42% vs 84%). Only 7 (17.5%) patients had intervention with CABG. Within the CABG group, 5 (71%) patients survived and 3 (42.8%) patients had full neurological recovery. Predictors of neurological recovery included shorter time to return of spontaneous circulation, neurological status post-resuscitation and younger age.

This analysis of a large nationwide registry includes by far the largest number of unresponsive patients undergoing CABG for AMI-CS in the current literature. It shows that patients with unresponsive preoperative neurological status have worse outcomes compared to responsive patients but can achieve 66.3% survival, with half of the survivors discharged directly to home. Lengths of ICU stay and hospitalization were not significantly prolonged compared to responsive patients. Given the observational nature of this study, there is almost certainly some degree of selection bias in that patients who were offered and underwent CABG in the registry may not be reflective of all patients who present with AMI-CS and an unresponsive neurological state, and the results of this study should not be taken to advocate surgery for all such patients. Nevertheless, it is encouraging that in this somewhat desperate situation, while morbidity and mortality are high, the absolute benefit of intervention may also be substantial, particularly in previously healthy patients who have a potentially reversible cause of neurological dysfunction. Multivariable analyses showed that measures of clinical acuity such as resuscitation and operative status, degree of heart failure, measures of end-organ function (particularly renal function) and unresponsive neurological state were important predictors of outcomes. An individualized approach incorporating these clinical variables, clinical history, comorbidities and angiographic and neurological parameters should be used to make decisions regarding appropriateness of surgery.

The primary cause of mortality was cardiac in both responsive and unresponsive patients, which is consistent with prior studies [2]. This finding suggests that patients whose neurological status improves or normalizes after CABG but who continue to have haemodynamic instability and poor end-organ perfusion should be evaluated for short-term mechanical support until myocardial stunning subsides or sometimes for durable left ventricular assist device support. We also found that the STS PROM was higher in the unresponsive cohort as compared to patients with a responsive preoperative neurological status, reflecting the clinical acuity. However, despite 59.3% of unresponsive patients having had cardiopulmonary resuscitation within an hour prior to surgery, only 46.9% had an expected mortality >20%, suggesting some underestimation of risk with current models.

This study quantifies perioperative outcomes of unresponsive patients undergoing CABG for AMI-CS and provides a basis for discussion with families of patients regarding therapeutic options and prognosis. These data may also be useful to payers as well as public reporting and quality improvement organizations. In high-acuity situations where current metrics may not adequately capture risks, reluctance to operate on very high-risk patients for fear of worse public reporting performance measures may also be associated with higher mortality from under-treatment [15].

Further investigation is necessary to identify patients who would benefit most from surgery and those in whom intervention is futile.

Limitations

There were several limitations to this study. Patients with an unresponsive neurological state who were operated on were likely a selected population who were felt to be more likely to do well following surgery than the general unresponsive population. Given that the STS database includes only patients who underwent surgery, there is no comparator medically or percutaneously managed group to directly assess the absolute benefit of surgical therapy. Furthermore, the STS database does not have adequately detailed information to distinguish the cause of neurological unresponsiveness, for example from metabolic, vascular, anoxic injury versus patients who were unconscious due to persistently poor haemodynamic status. In addition, although discharge location can be used as a surrogate for functional status, it is unclear exactly how well patients were functioning and whether their neurological status was back to baseline following surgery. Finally, many variables which are important in risk stratification, such as duration of arrest before the onset of resuscitation, duration of resuscitation, vasopressor requirements, indices of perfusion such as pH and lactate and results of neurological investigations to determine other causes of unresponsive state were not available, limiting our ability to identify patients who derive the most benefit from surgery.

CONCLUSION

In conclusion, despite higher risks of mortality and neurological dysfunction, data support continued selective use of CABG in patients with AMI-CS and an unresponsive preoperative neurological state. Additionally, future risk models should incorporate preoperative neurological status.

Funding

This work was supported by the Society of Thoracic Surgeons Access & Publications Task Force.

Conflict of interest: none declared.

REFERENCES

Author notes