Reduced post-cardiopulmonary bypass echocardiographic inotropy index is associated with complications following cardiac surgery: an institutional outcomes study

Abstract

Objective:

Despite advances in echocardiography and hemodynamic monitoring, limited progress has been made to effectively quantify left ventricular function during cardiac surgery. Traditional measures including left ventricular ejection fraction (LVEF) and cardiac index remain dependent on loading conditions; more complex measures remain impractical in a dynamic surgical setting. However, the Smith-Madigan Inotropy Index (SMII) and potential-to-kinetic energy ratio (PKR) offer promise as measures calculable during cardiac surgery and potentially predictive of outcomes. Using echocardiographic and hemodynamic monitoring data, we aimed to calculate SMII and PKR values after cardiopulmonary bypass, and understand associations with postoperative outcomes, adjusting for previously identified risk factors.

Design:

Observational cohort study.

Setting:

Tertiary care academic hospital.

Patients:

189 elective adult cardiac surgical procedures from 2015 to 2016.

Intervention:

None.

Measurements and Main Results:

The primary outcome was postoperative mortality or organ system complication (stroke, prolonged ventilation, reintubation, cardiac arrest, acute kidney injury, new-onset atrial fibrillation). After adjustment, SMII <0.83 W/m² independently predicted the primary outcome (adjusted odds ratio 2.19, 95% confidence interval 1.08–4.42) whereas PKR, LVEF, and cardiac index demonstrated no associations. When SMII and PKR were incorporated into a EuroSCORE II risk model, predictive performance improved (net reclassification index improvement 0.457, p = 0.001) whereas a model incorporating LVEF and cardiac index demonstrated no improvement (0.130, p = 0.318).

Conclusion:

We demonstrate that SMII, but not PKR, as a measure of cardiac function is associated with major complications. Our data may guide investigations of more suitable perioperative goal-directed therapies to reduce complications following cardiac surgery.

Within a cardiac surgical context, left ventricular ejection fraction (LVEF) and cardiac index provide simple, rapid means for approximating ventricular function; however, such methods are heavily dependent on loading conditions^1–3 and routinely fail to characterize inotropic state.⁴ Conversely, alternative echocardiographic measures of LV function, which have gained traction in cardiology literature, have failed to penetrate the perioperative setting due to practicality: calculations either require concomitant findings (e.g. mitral regurgitation for dP/dt⁵), repeated measurements under varying preload conditions potentially harmful to the cardiac surgery patient,^6,7 or precise images challenging to obtain in the acute perioperative setting^8,9. Furthermore, echocardiography has inherent limitations to precisely modelling indices of myocardial contractility (e.g. end-systolic elastance, preload recruitable stroke work) derived invasively from LV pressure-volume relations.¹⁰

Beyond limitations inherent to widely used echocardiographic measures, relationships between LV function, inotropy, and postoperative outcomes remain unresolved. Potentially addressing these shortcomings is the Smith-Madigan inotropy index (SMII), a pragmatic echocardiographic and hemodynamic measure describing the total kinetic and potential cardiac energy expenditure during systole. The SMII has been previously described¹¹ and is summarized in Figures 1 and 2. Additionally, a related measure derived from similar parameters is the potential to kinetic energy ratio (PKR), used to describe myocardial efficiency for yielding perfusion pressure versus blood flow.¹¹ SMII and PKR represent underutilized but potentially useful measurements for describing hemodynamic derangements in a clinically actionable way; however their uptake into clinical practice has been limited, potentially due to a lack of studies demonstrating the utility of such measures for predicting adverse events, including postoperative complications following cardiac surgery. In this study of a cardiac surgical population, we aimed to calculate SMII and PKR as measures of inotropy and myocardial efficiency, respectively, and explore their independent associations with postoperative complications. We hypothesized that cases with favorable states of SMII and PKR are associated with improved outcomes compared to those with unfavorable states, when adjusted for previously identified risk factors. We secondarily explored the utility of SMII and PKR compared to traditional measures of cardiac function - LVEF and cardiac index - for improving existing cardiac surgery risk prediction models.

Figure 1:

Calculation of Smith-Madigan Inotropy Index and Potential:Kinetic Energy Ratio

Figure 2:

Measurement Methods for Parameters Required to Calculate Smith-Madigan Inotropy Index and Potential-to-Kinetic Energy Ratio

METHODS

Institutional Review Board approval was obtained for this retrospective study within our academic tertiary care medical center (Ann Arbor, MI; HUM00052066) and patient consent was waived. The study plan was registered within an internal research forum on November 2, 2016 prior to analysis and is available upon request.¹² The Strengthening the Reporting of Observational Studies guidelines were followed throughout the conduct of this study.¹³

Patient Population

Adult patients (≥18 years) undergoing elective cardiac surgical procedures involving echocardiographic and pulmonary arterial catheter monitoring at a tertiary care medical center were eligible for inclusion. A sample of 189 cases was analyzed over a 14-month study period with intensive data collection from January 2015 to February 2016. Cases were included based on the availability of an experienced cardiac anesthesiology fellow as a supplementary clinical provider, dedicated to acquiring echocardiographic measurements simultaneously with cardiac index measurements. Cardiac surgical procedures included coronary artery bypass, valve replacement, and ascending aortic (excluding arch) procedures. For patients undergoing repeated cardiac surgical procedures during the study period, only the first (index) case was included. Exclusion criteria were American Society of Anesthesiologists physical status classification 5 or 6, emergent procedures, use of postoperative mechanical circulatory support, or cardiac surgery within 365 days (Supplementary Material 1). For all cases, complete post-cardiopulmonary bypass echocardiography examinations were performed and documented using templated electronic procedure notes.

Data Collection

We collected data from three sources: our institution’s local perioperative electronic clinical database as available within the Multicenter Perioperative Outcomes Group, our institution’s electronic health record (Epic Systems Corporation, Verona, WI, USA), and the Society for Thoracic Surgeons Adult Cardiac Surgery Database. Methods for data collection, validation, and extraction of data from each source are described elsewhere^14,15 and utilized in multiple prior studies.^16–19 Data quality were ensured using pre-specified definitions, validated by nurses with training in data definitions, and via manual review by the study team.

Clinical Care Processes

Anesthetic management was performed at the discretion of the attending anesthesiologist. Intraoperative hemodynamic management was guided by standard monitoring as well as invasive arterial catheter, central venous pressure, and pulmonary artery catheter monitors (as per routine institutional practice); cardiac function was assessed via echocardiography. Hemodynamic targets included a cardiac index >2.2 L/min/m², mean arterial pressure >65 mmHg, mixed venous oxygen saturation >65%, and an echocardiographic assessment of post-bypass cardiac function compared to pre-bypass function. Echocardiographic assessments were performed using Philips iE-33 probes (Philips Healthcare ©, Andover, MA, USA); cardiac indices were calculated via pulmonary artery catheters (Arrow®, Morrisville, NC, USA) using the Fick thermodilution method. Postoperative management was at the discretion of the intensive care unit team, as based on local protocols and targeting goals discussed during handoff.

Exposure Variables

In addition to collecting data on traditional measures of cardiac function - including LVEF by visual estimation and cardiac index by pulmonary artery catheter measurement - we collected data on novel SMII and PKR values for each case. All values were calculated post-bypass during sternal closure as a reliable index time point indicative of completion of the critical portions of the surgical procedure. To facilitate real-time intraoperative calculation of SMII and PKR as described in Figures 1 and 2, we provide a calculation tool in Supplementary Material 2, with key components of the indices including: i) systolic flow time and mean velocity of aortic valve blood flow (measured via an aortic valve velocity-time integral), ii) cardiac index (measured via a pulmonary artery catheter), iii) heart rate, and iv) mean arterial pressure.

Outcomes

The primary outcome was a composite index of mortality or major organ system-based complication, selected a priori on the basis of availability within the electronic health record or Society of Thoracic Surgeons database, and having a biologically plausible relationship with the study exposure variables, SMII and PKR. Major organ system-based complication components included stroke, cardiac arrest, new-onset atrial fibrillation, prolonged postoperative ventilation >24 hours, reintubation, and acute kidney injury (Appendix 1). Secondary outcomes were postoperative ventilator duration, postoperative inotrope/vasopressor duration, and intensive care unit length of stay.

Covariate Data

For descriptive purposes as well as for risk adjustment, we collected perioperative characteristics as study covariates (Table 1). Using covariate data, we calculated EuroSCORE II values for all patients, as has been previously used for cardiac surgical risk adjustment.^20,21 Additional a priori selected covariate data used for risk adjustment but not incorporated into a preoperative EuroSCORE II calculation included body mass index and cardiopulmonary bypass duration. Finally, beyond the preoperative measures of cardiac function including in the EuroSCORE-II calculation (preoperative LVEF and pulmonary artery systolic pressure), intraoperative post-bypass measures of cardiac function (LVEF and cardiac index) were also used for risk adjustment.

Table 1:

Study Cohort Characteristics and Bivariate Analyses

		Entire cohort N = 189		Mortality or Major Organ System Dysfunction, n / Mean (% / SD)
Category	Descriptor			Yes (n = 81)		No (n = 108)		p-value
		n / mean	% / SD	n / mean	% / SD	n / mean	% / SD
Anthropometric/Demographic Data	Age (yr)	61	13	66	10	57	14	<0.001
	Height (cm)	172	10	171	10	173	10	0.210
	Weight (kg)	90	21	89	22	91	20	0.598
	Body-Mass Index (kg/m2)	30.3	6.3	30.2	6.3	30.3	6.4	0.941
	Female sex	63	33	25	31	38	35	0.640
	Race							0.377
	Caucasian	169	89	95	88	74	91
	Black	13	6.9	7	6.5	6	7.4
	Multi-racial/Other	7	3.7	6	5.6	1	1.2
Patient Medical History	ASA Physical Status Classification							0.188
	3	55	29	19	23	36	33
	4	134	71	62	77	72	67
	Renal Impairment							<0.001
	Normal (CC >85 mL/min)	113	60	37	46	76	70
	Moderate (CC ≥50 and ≤85 mL/min)	60	32	32	40	28	26
	Severe (CC <50 mL/min)	12	6.3	10	12	2	1.9
	Dialysis (Regardless of CC)	4	2.1	2	2.5	2	1.9
	Extracardiac arteriopathy	119	63	55	68	64	59	0.287
	Poor mobility	109	58	51	63	58	54	0.260
	Previous cardiac surgery	31	16	14	17	17	16	0.932
	Chronic lung disease	38	20	21	26	17	16	0.122
	Active endocarditis	5	2.6	2	2.5	3	2.8	0.999
	Critical preoperative state	2	1.1	1	1.2	1	0.9	0.999
	Diabetes on insulin	23	12	14	17	9	8.3	0.102
	NYHA Class							0.011
	I	170	90	70	86	100	93
	II	3	1.6	0	0.0	3	2.8
	III	7	3.7	3	3.7	4	3.7
	IV	9	4.8	8	9.9	1	0.9
	CCS class 4 angina	25	13	9	11	16	15	0.598
	Preoperative LV Function							0.927
	Good (LVEF >50%)	158	84	67	83	91	84
	Moderate (LVEF 31–50%)	25	13	11	14	14	13
	Poor (LVEF 21–30%)	6	3.2	3	3.7	3	2.8
	Very Poor (LVEF 20% or less)	0	0.0	0	0.0	0	0
	Recent MI within 90 days	9	4.8	2	2.5	7	6.5	0.305
	Preoperative pulmonary hypertension							0.317
	Moderate (PA systolic 31–55 mmHg)	117	62	49	61	68	63
	Severe (PA systolic >55 mmHg)	27	14	15	19	12	11
Intraoperative Data	Type/Weight of Intervention							0.084
	Isolated CABG	23	12	5	6.2	18	17
	Single Procedure non-CABG	72	38	29	36	43	40
	2 Procedures	78	41	38	47	40	37
	3 Procedures	16	8.5	9	11	7	6.5
	Surgery on thoracic aorta	60	32	25	31	35	32	0.946
	Urgency Elective Urgent	184 5	97 2.6	78 3	96 3.7	106 2	98 1.9	0.653
	CPB Duration	161	71	171	73	153	69	0.092
Perioperative Risk Summary	EuroSCORE II - Mortality Risk	5.5	6.6	7.7	9.0	3.9	3.0	<0.001
Exposure Variables	Post-CPB SMII (W/m2)	0.86	0.30	0.82	0.31	0.90	0.28	0.059
	Post-CPB PKR	16.3	14.4	16.4	15.3	16.3	13.8	0.941
	Post-CPB LVEF (%)	55.2	8.7	53	11	57	7	0.019
	Post-CPB Cardiac Index (L/min/m2)	2.4	0.6	2.3	0.6	2.5	0.6	0.027

ASA = American Society of Anesthesiologists, CABG = coronary artery bypass graft, CC = creatinine clearance, CCS = Canadian Cardiovascular Society, CPB = cardiopulmonary bypass, LVEF = left ventricular ejection fraction, MI = myocardial infarction, NYHA = New York Heart Association, PA = pulmonary artery, PKR = potential:kinetic energy ratio, SMII = Smith-Madigan Inotropy Index

Statistical Analysis

We performed all analyses in R version 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria) and calculated descriptive statistics for all variables. We graphically assessed normality of continuous variables via histograms and Q-Q plots. We next examined variables via univariate analysis, with categorical variable p-values calculated using Fisher’s exact or Pearson chi-square tests, and continuous variable p-values calculated using Student t--tests or Mann-Whitney U tests depending on normality. Using logistic regression and the ‘OptimalCutpoints’ package in R,²² we assessed best thresholds for SMII and PKR by selecting the values that simultaneously maximized the sensitivity and specificity of each covariate for predicting the primary outcome. We performed univariate linear regressions, with post-bypass SMII, PKR, cardiac index, LVEF as independent variables, to assess associations with the continuous outcomes.

We next analyzed independent associations between SMII, PKR (assessed as binary variables based on best thresholds to enable ease of interpretation and target levels for future goal-directed therapy investigations) and outcomes, risk-adjusted for EuroSCORE II components via multivariable logistic regression (primary outcome) and linear regression (continuous secondary outcomes). A threshold of p <0.05 denoted statistical significance. We additionally assessed for improvement in EuroSCORE II multivariable model discrimination with post-bypass SMII and PKR added as binary variables, using net reclassification indices and c-statistics. We compared this to multivariable model discrimination using EuroSCORE II and commonly used thresholds for traditional measures of cardiac function: cardiac index <2.0 L/min/m² and LVEF <40%.^23,24 Finally, we assessed the performance of a multivariable model including EuroSCORE II, exposure variables, and all additional covariates.

Sensitivity Analyses

Given the use of a composite primary outcome, we performed two sensitivity analyses with modified primary outcome definitions and analyzed via similar multivariable regressions. These sensitivity analyses were (i) the acute kidney injury component of the composite outcome restricted to severe stage only (Kidney Disease - Improving Global Outcomes Stage 3 acute kidney injury only) and (ii) new-onset postoperative atrial fibrillation removed from the composite outcome.

Power Analysis

Based upon primary outcome rates of 10% versus 30% in patients with favorable versus unfavorable post-bypass indices respectively, and a sampling ratio of 1, a study with a power of 0.80 to detect a statistically significant difference p <0.05 would require a sample size of 118 patients for analysis. Differences in outcome rates for power analysis were selected a priori based upon prior literature^25–27 and clinical judgment of the study team.

RESULTS

Patient Population - Baseline Characteristics

Characteristics of the study cohort are detailed in Table 1. Complete data were available on all patients. The study population had a mean (standard deviation) age of 61 (13) years, 67% were men, and the mean (standard deviation) EuroSCORE II predicted risk of mortality was 5.5 (6.6)%. Common EuroSCORE II risk factors included pulmonary hypertension (72%), extracardiac arteriopathy (63%), and poor mobility (58%). Cardiac surgical procedures included isolated coronary artery bypass grafting (12%), single non-coronary artery bypass grafting (38%), and multiple procedures (50%). Among single non-coronary artery bypass grafting procedures, 50% were aortic valve repairs or replacements, 36% were aortic root and ascending aorta replacements, and 14% were mitral valve repairs or replacements. Among the patients undergoing multiple procedures, 32% were procedures on the aortic valve and ascending aorta, 27% were procedures on multiple valves, 26% were coronary artery bypass and valve procedures, 9% were valve and Maze procedures, 5% were coronary artery bypass and ascending aortic procedures, and 1% were coronary artery bypass, valve, and ascending aorta procedures.

Among traditional measures of cardiac function, the median LVEF observed on post-bypass intraoperative echocardiography was 55% (interquartile range 55–60%; full range 20–80%). The median post-bypass cardiac index was 2.3 L/min/m² (interquartile range 2.0–2.8; range 1.2–5.0). For the novel measures of cardiac function, median post-bypass SMII and PKR were 0.84 W/m² (interquartile range 0.63–1.02 W/m²; range 0.30–2.04 W/m²) and 12.3 (interquartile range 8.3–19.3; range 1.1–111) respectively. Each measure demonstrated rightward skew except LVEF (Figure 3).

Figure 3:

Post-Cardiopulmonary Bypass Measures of Cardiac Function: Left Ventricular Ejection Fraction, Cardiac Output, Smith-Madigan Inotropy Index, Potential:Kinetic Energy Ratio

Postoperative Outcomes

Among the 189 cardiac surgical patients studied, 81 (43%) experienced the primary outcome, including 3 (1.6%) mortalities. Individual components of the primary outcome are described in Table 2. Among secondary outcomes studied, 56% of patients required postoperative inotropic support and 49% of patients required postoperative vasopressor support. All except five patients received postoperative mechanical ventilation; the median postoperative ventilator duration was 6 hours (interquartile range 4–10 hours). The median intensive care unit length of stay was 49 hours (interquartile range 28–78 hours).

Table 2:

Study Outcomes

		n / median	% / IQR
Primary Outcome & Components	Composite index, mortality or major organ system dysfunction*	81	43
	In-hospital all-cause mortality	5	2.6
	Neurologic - Stroke	3	1.6
	Pulmonary - Prolonged Ventilation >24 hours	16	8.5
	Pulmonary – Reintubation	2	1.1
	Cardiovascular - Cardiac Arrest	2	1.1
	Cardiovascular - New-onset Atrial Fibrillation	60	32
	Renal – Acute Kidney Injury (KDIGO definition)**	32	17
Secondary Outcomes	Postoperative inotrope support duration (hours)	6	[0, 22]
	Postoperative vasopressor support duration (hours)	0	[0, 13]
	Postoperative ventilator duration (hours)	6	[4, 10]
	ICU length of stay (hours)	49	[28, 78]

KDIGO = Kidney Disease - Improving Global Outcomes

^*Comprised of all individual organ system dysfunction outcome components

^**Any stage, defined as a serum creatinine ≥1.5 times baseline with 7 days postoperative, or ≥0.3 mg/dL increase within 48 hours postoperative, among patients without pre-existing chronic renal failure (Kidney Disease - Improving Global Outcomes Stage 5). Among patients with acute kidney injury of any stage, 4 patients (2.1% of study cohort) developed KDIGO Stage 3 acute kidney injury, defined as serum creatinine ≥3.0 times baseline or need for renal replacement therapy within 7 days postoperative.

Optimal SMII and PKR Thresholds

Across unadjusted comparisons, patients who developed the primary outcome more commonly were older and had a history of heart failure or renal impairment (Table 1). Such patients more commonly exhibited reduced measures of post-bypass cardiac function, including LVEF, cardiac index, and SMII; however no association between PKR and postoperative mortality or complications was observed. Best thresholds maximizing the sum of univariate sensitivity and specificity were <0.83 W/m² for SMII (sensitivity 62%, specificity 61%) and <12.6 for PKR (sensitivity 49%, specificity 51%).

Associations with Postoperative Mortality or Major Organ System Complication

Given a total of 81 patients with primary outcomes, all covariates were used in the primary multivariable analyses. Following multivariable adjustment, we observed that a post-bypass SMII <0.83 W/m², but not PKR, LVEF, or cardiac index, was independently associated with complications (adjusted odds ratio 2.19, 95% confidence interval 1.08–4.42) (Table 3).

Table 3:

Risk-Adjusted Associations of Exposure Variables with Primary Outcome and Overall Model Performance

Multivariable Model	Variables Included	Adjusted Odds Ratio	95% CI	Net reclassification index versus EuroSCORE II	95% CI	C-statistic	95% CI
A	EuroSCORE II	1.14	(105, 1.22)	N/A	N/A	0.641	(0.56, 0.72)
B	EuroSCORE II	1.13	(104, 1.22)	0.457	(0.18, 0.74)	0.673	(0.59, 0.75)
	SMII <0.83 W/m2	2.22	(120, 4.11)
	PKR < 12.6	1.26	(0.68, 2.35)
C	EuroSCORE II	1.12	(104, 1.21)	0.130	(−0.13, 0.38)	0.645	(0.56, 0.73)
	LVEF < 40%	3.21	(0.61, 16.81)
	Cardiac Index < 2.0 L/min/m2	1.38	(0.68, 2.82)
D	EuroSCORE II	1.11	(103, 1.20)	0.228	(−0.06, 0.51)	0.699	(0.62, 0.78)
	SMII <0.83 W/m2	2.19	(108, 4.42)
	PKR < 12.6	1.31	(0.69, 2.48)
	LVEF < 40%	3.62	(0.68, 19.22)
	Cardiac Index < 2.0 L/min/m2	0.92	(0.41, 2.10)
	Cardiopulmonary Bypass Duration	1.00	(0.96, 1.06)
	Body Mass Index	1.01	(100, 1.01)

LVEF = left ventricular ejection fraction, SMII = Smith-Madigan Inotropy Index, PKR = Potential:Kinetic Energy Ratio

Exploring Risk Model Performance

In an exploratory multivariate analysis, when novel measures of cardiac function (SMII and PKR) were added to a risk model with EuroSCORE II components, more patients were correctly identified as having complications (net reclassification index improvement of 0.457, p = 0.001) but the overall discrimination of the model yielded a non-significant improvement (c-statistic 0.673 versus 0.641; p = 0.347). Conversely, when traditional measures of cardiac function (intraoperative LVEF and cardiac index) were added to a EuroSCORE II risk model, there was neither an improvement in identifying patients with complications (net reclassification index improvement of 0.130, p = 0.318) nor an improvement in overall model discrimination (c-statistic 0.645 versus 0.641; p = 0.806).

Sensitivity Analyses

Using a primary outcome definition restricting acute kidney injury complications to Stage 3 only, 69 patients developed the composite primary outcome, and SMII <0.83 W/m² remained independently associated with complications (adjusted odds ratio 2.22, 95% confidence interval 1.20–4.11), whereas PKR, LVEF, and cardiac index continued to lack independent associations (Supplementary Material 3). A multivariable model incorporating SMII and PKR continued to improve model net reclassification index, whereas a model incorporating LVEF and cardiac index continued to lack improvement. When we excluded new-onset postoperative atrial fibrillation from the primary outcome definition, only 43 patients suffered an adverse outcome, and neither SMII nor any other measure of cardiac function was independently associated with complications; however a model including SMII and PKR continued to significantly improve net reclassification index whereas a model including LVEF and cardiac index continued to lack improvement (Supplementary Material 4).

Secondary Outcomes

Among univariate analyses of secondary outcomes, elevated values for both novel measures of cardiac function (SMII and PKR) and traditional measures (LVEF and cardiac index) were associated with shorter durations for all secondary outcomes (Supplementary Material 5).

Following multivariable adjustment, elevated intraoperative LVEF was associated with shorter postoperative inotrope duration as well as shorter intensive care unit length of stay, but not shorter postoperative vasopressor duration or ventilator duration (Supplementary Material 6). Within the same multivariable models, cardiac index was not independently associated with any secondary outcome. Conversely, among multivariable models incorporating SMII and PKR, we observed no independent associations between them and secondary outcomes (Supplementary Material 7).

DISCUSSION

In this observational study, we found that decreased intraoperative post-bypass SMII - but not PKR, LVEF, or cardiac index - was independently associated with postoperative complications. These data reveal potential shortcomings of traditional post-bypass measures of cardiac function, LVEF and cardiac index. These shortcomings may be in part addressed by the consideration of SMII, but not PKR, as a novel measure potentially calculable during cardiac surgery and informative for postoperative risk prediction.

Although the observed mortality rate of 1.6% was substantially less than that predicted by the EuroSCORE-II risk model (5.5%), the mortality rate was comparable to previous studies among similar cardiac surgical populations,^28,29 as was the composite complication rate of 43%, primarily driven by new-onset postoperative atrial fibrillation (32%) and acute kidney injury (17%).^29,30 In contrast to studies selecting more severe component complications to define a composite primary outcome, we selected component complications for which mechanistic relationships between poor cardiac function and outcomes were biologically plausible - including atrial fibrillation and acute kidney injury of any severity. For a composite outcome excluding new-onset postoperative atrial fibrillation and restricting acute kidney injury to severe stages only - as are commonly excluded from Society of Thoracic Surgeons database-defined major postoperative complications³¹ - we observed a complication rate of 11%, also comparable to previously reported rates of severe complications.^28,29 The performance of prediction models based upon sensitivity analyses excluding these outcome components yielded similar findings to the primary analysis.

In the cardiac surgical cohort studied, the distribution of SMII values observed were similar to those previously reported in a SMII validation study which reliably differentiated healthy patients from patients with acute left ventricular failure presenting to the emergency department or coronary care intensive care unit (SMII mean 0.73 W/m²; range 0.43–0.97 W/m²).¹¹ Conversely, the distribution of PKR values observed were in contrast to the higher values previously reported (PKR mean 124; range 96–174), suggestive of either differences in a cardiac surgical cohort exhibiting greater degrees of vasodilation potentially due to a systemic inflammatory response from the cardiac surgical physiologic insult, or bias in echocardiographic- and pulmonary artery catheter-derived measurements. Other recent studies have demonstrated the value of SMII and PKR for hemodynamic monitoring of neonates,^32,33 patients with hypertensive disorders of pregnancy,³⁴ and potentially intensive care unit patients.³⁵ Within the context of cardiac surgery, determinants of SMII may include the use of inotropic and vasoactive medications, cardiovascular depressant anesthetic agents, myocardial stunning and the inflammatory response from exposure to cardiopulmonary bypass, and structural heart changes from the surgical intervention. Further studies are needed to understand the independent contributions of each intraoperative factor influencing SMII, and to validate these findings across similar cardiac surgery and acute care populations.

Among recent high-quality clinical trials investigating goal-directed therapies for hemodynamic management of cardiac surgical patients, parameters used have included central venous pressure, pulmonary artery pressure, and cardiac output,^36–39 invasive arterial blood pressure,^36–38,40 non-invasive cardiac output monitoring,^36–39, central venous oxygen saturation,^36–38,41 esophageal doppler,^40,42 arterial lactate,^38,41 hemoglobin/hematocrit,^36,37,39 and urine output^36,38. In such trials, inotropes were commonly used, and were specifically initiated and titrated based upon measures of cardiac index,^36,37,39,41 central venous oxygen saturation,³⁷ mean arterial pressure,⁴⁰ stroke volume index,^39,40 systemic vascular resistance,³⁶ and pulmonary capillary wedge pressure⁴¹. Although such trials commonly used hemodynamic monitors permitting SMII to be calculated, the SMII has not yet been studied as a hemodynamic goal for goal-directed therapies. Goal directed therapies commonly demonstrated benefits during the post-bypass period in such cardiac surgical patient populations^39–41 however it remains unclear as to whether the incorporation of SMII into therapeutic goals may lead to further improved outcomes.

Limitations

Our study has multiple limitations which must be considered in the context of generalizing findings to other populations or developing research protocols testing the potential efficacy of SMII as a parameter for goal-directed therapy. First, although the effects of preload and afterload are theoretically independent of SMII,¹¹ this independence between loading conditions and SMII has not been verified using LV pressure-volume loops; further studies are needed to determine how SMII relates to contractility in the setting of manipulations to preload and afterload. Second, although the cohort studied represented a broad range of cardiac surgical procedures, this study is limited by performance at a single center across a limited sample size. Patient populations and practice patterns, such as management of inotropes, vasopressors, and fluids, may not reflect those encountered at other institutions. Third, this study remains observational and was analyzed within the context of clinical care rather than a controlled experimental setting. Intraoperative LVEF, cardiac index, SMII, PKR and were analyzed only during a single time point at sternal closure, and subject to a level of accuracy and precision available from echocardiographic and hemodynamic monitoring used in a clinical setting. Conversely however, our findings represent real-world health care data and reflect what information may be available and clinically actionable for providers at the point of care. Such an approach may be advantageous for informing pragmatic prospective trial design, including investigations of goal-directed therapies using SMII.

Additionally, components of the composite primary outcome may have varying clinical significance, strength of relationship, or effect size, in relation to the modifiable potential risk factors (traditional versus novel measures of cardiac function) studied. Furthermore, patient characteristics such as mitral valve disease and left atrial dilation²⁷ influencing new-onset postoperative atrial fibrillation, the most frequently occurring component complication of the primary outcome, were unaccounted for due to lack of availability or limited sample size precluding a more comprehensive prediction model. Although primary outcome components were selected on the basis of biologically plausible mechanistic relationships with measures of cardiac function and examined via sensitivity analyses using modified component definitions, all potential long-term health consequences were not investigated. Future studies may wish to emphasize the risk of persistent organ dysfunction.⁴³ However, even milder complications such as stage 1 acute kidney injury have been shown to be associated with increased risk of late mortality.⁴⁴

Finally, whereas LVEF and cardiac index values were available to clinicians at the point of care and were potentially used to inform clinical-decision making, SMII and PKR values were calculated retrospectively. Associations with postoperative inotrope and vasopressor support were limited by LVEF and cardiac indices potentially serving as indications for such therapies rather than markers of true disease severity, and thus such associations must be interpreted with caution. Conversely, this bias did not exist for SMII and PKR values, which were not used to guide clinical care and thus not considered as an indicator for postoperative inotropic or vasopressor support. Across all cases, the indications for inotropic or vasopressor support upon which clinicians acted upon were not documented, and thus existed as a limitation for the secondary outcomes describing postoperative inotropic and vasopressor support.

Conclusions

In summary, we describe the SMII as an innovative measure of cardiac function, independently associated with major complications following cardiac surgery. SMII may offer greater utility for postoperative risk prediction than traditional intraoperative measures of cardiac function including LVEF and cardiac index, and remains practical to obtain intraoperatively by the multi-tasking cardiac anesthesiologist. Given the lack of measures accurately characterizing the inotropic state of patients undergoing cardiac surgery, our results may inform the design of future interventional studies further investigating the potential efficacy of SMII-based goal-directed inotropic therapies for cardiac surgical patients.