Change in Left Ventricular Ejection Fraction with Coronary Artery Revascularization and Subsequent Risk for Adverse Cardiovascular Outcomes

Abstract

Background

Coronary revascularization is recommended to treat ischemic cardiomyopathy. However, the relations of revascularization-associated ejection fraction (EF) change to subsequent outcomes have not been elucidated.

Methods

In 10,071 veterans (mean age 67 years; 1% women; 15% non-Caucasian) who underwent a first percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) between 01/01/1995 and 12/31/2010 and had pre- and post-revascularization EF measured, we calculated delta-EF (post-procedure EF–pre-procedure EF). We related delta-EF as a continuous measure and as categories (≤−5, −5<delta-EF<0, delta-EF=0, 0<delta-EF<5, delta-EF≥5) to death (using Cox regression) and heart failure (HF) hospitalization days (using negative binomial regression) in multivariable-adjusted models, for total sample, and PCI and CABG strata.

Results

Over follow-up (mean/maximum 5/14 years) 56% died. Each 5% improvement in delta-EF was associated with statistically significant reductions in death and HF hospitalization days of 5% [95% CI: 3-7%] and 10% [95% CI: 5-15%] respectively in the total sample and 6% [95% CI: 4-8%] and 10% [95% CI: 5-16%] respectively in the PCI subgroup. Patients in the highest delta-EF category had 27% [95% CI: 19-34%] lower mortality (30% [95% CI: 21-37%] lower in PCI stratum) and ~40% lower HF hospitalization days in total sample and PCI stratum, compared with those in the lowest category. Relations of delta-EF and outcomes in CABG subgroup did not reach statistical significance.

Conclusions

Revascularization-associated EF improvement was associated with significant reductions in mortality and HF hospitalization burden, particularly in the PCI subgroup.

Graphical Abstract

Introduction

Revascularization, particularly coronary artery bypass grafting [CABG], is recommended for patients with coronary artery disease (CAD) and heart failure (HF) with reduced ejection fraction (HFrEF)¹, based on the concept that restoration of blood flow to under-perfused but viable myocardium reverses left ventricular systolic dysfunction. ^2.3 An association between treatment-associated EF improvement and outcomes has been demonstrated with medical management in HFrEF patients.⁴ However, it is unclear whether the observed benefit of multivessel revascularization in patients with CAD and reduced EF is mediated by an improvement in EF leading to a reduction in death and HF hospitalization, or by a reduction in CAD events and cardiac death similar to that seen in subjects without HF, or both mechanisms. Understanding the magnitude and mediating effect of revascularization-associated EF improvement on subsequent outcomes will provide a firmer evidence base in support of revascularization in patients with CAD and HFrEF. In addition, whether percutaneous coronary intervention (PCI) is associated with an EF improvement that correlates with subsequent clinical outcomes has been less well studied.⁵

We examined the relationship between revascularization-associated EF change and subsequent outcomes in a real-world setting. We (a) quantify the magnitude of EF improvement associated with coronary revascularization, either with PCI or CABG; and (b) elucidate the relations of revascularization-associated EF change with subsequent clinical outcomes.

Methods

Study Sample

Owing to organizational restrictions, the authors will not make the data and study materials related to this investigation available to other researchers. The study was approved by the Veterans Affairs (VA) Boston Healthcare System Institutional Review Board. As the analyses were based on anonymized data from a large pre-configured dataset, informed consent was waived. We included patients from the national VA patient care databases who fulfilled the following criteria in addition to an HF diagnosis code:

1. Underwent PCI or CABG between 01/01/1995 and 12/31/2010

2. At least one EF measurement in the year preceding revascularization (EF₁)

3. EF₁<50%

4. At least one EF measurement within the year succeeding revascularization (EF₂).

Figure 1 displays the cohort compilation process. Of all revascularization patients considered for inclusion in this study, approximately 90% had an EF measurement available within one year from procedure date. The PCI and CABG procedures were identified using International Classification of Disease (ninth revision) procedure codes and Current Procedural Terminology codes (listed in Supplementary Table S1). The utility and accuracy of these codes in identifying cardiovascular procedures within the VA dataset have been reported.⁶

Figure 1: Schematic for study sample compilation.

HF = heart failure; LVEF = left ventricular ejection fraction.

Exposure

EF was extracted from documents in the electronic medical record using a natural language processing algorithm. The methods used to create the tool⁷ and its validity in curating EF-based patient groups⁸ have been published. During initial development, the algorithm’s accuracy was validated by three independent reviewers performing chart reviews.⁸ The algorithm achieved precision of 0.986 to 1.000 for extracting EF measurements.⁷ Details regarding how EF values were selected (e.g. when multiple values were available, or ranges were reported etc.) are described in Supplementary Figures S1 and S2. We then calculated delta-EF, defined as

$$\text{Delta-EF}={\text{EF}}_2-{\text{EF}}_1$$

Covariates

Age at the time of EF₂ (the beginning of the follow-up period for outcome assessment) was modeled as a continuous variable. Race was categorized as Caucasian vs. non-Caucasian. Information regarding body mass index (BMI), systolic blood pressure, diastolic blood pressure, low density lipoprotein cholesterol (LDL-C), diabetes status, smoking status (ever vs. never) and serum creatinine were obtained from clinical records. For medications, systolic blood pressure, diastolic blood pressure, serum creatinine and BMI, values recorded closest to within 3 months on or before revascularization procedure date were used. For LDL-C, measurements closest to within the six-month period prior to the procedure date were used. For diabetes status, a single inpatient or two outpatient diagnosis codes ever recorded on or before procedure date was categorized as “yes” otherwise “no”. For “hypertension therapy” and “statin therapy”, if patients were receiving treatment within the 3-month period prior to the procedure, they were coded “yes”; otherwise “no”.

Outcomes

Our endpoints of interest were

1. All-cause mortality

2. Cumulative hospitalization days for HF per year

As in our previous real-world studies, we chose cumulative hospitalization days instead of number of hospitalizations as the former has greater discriminative capacity for capturing disease burden and effects of risk factors and interventions.^9,10 We excluded patients with any inpatient stay >180 days from the hospitalization outcome analyses. Patients were followed from the date of EF₂ measurement until death, last VA visit, or the end of the follow-up period (12/31/2013).

Statistical Analysis

We calculated event proportions and multivariable-adjusted event rates per 100 person-years of follow-up and event rates by category of delta-EF, for the total sample and separately for PCI and CABG strata. To assess the linearity of the relationship between delta-EF and mortality, we fit Cox proportional hazards regression cubic spline plots with delta-EF as a continuous predictor of all-cause mortality (Figure 2).

Figure 2: Cubic spline relating delta-EF to mortality.

EF = ejection fraction

The spline was created using a SAS macro, specifying 9 knots placed based on Harrell’s approach, with 3 degrees of freedom, and a reference value of no change in EF.

Relations of delta-EF to death were evaluated using multivariable Cox proportional hazards regression models. To assess the associations of delta-EF to HF hospitalization days, we calculated incidence density ratios using negative binomial regression models. All multivariable models adjusted for age, sex, race, BMI, systolic blood pressure, diastolic blood pressure, hypertension treatment, LDL-C, statin therapy, diabetes mellitus, smoking status and serum creatinine. We then performed the same analyses as described above, separately by type of revascularization: PCI and CABG.

To make results practical and clinically relevant, we performed the following secondary analyses. We grouped patients into delta-EF categories (≤−5, −5<delta-EF<0, delta-EF=0, 0<delta-EF<5, delta-EF≥5) and repeated the aforementioned multivariable regression analyses with delta-EF categories as the exposure and the lowest delta-EF category as the referent group. A two-sided p-value threshold of 0.05 was used to ascribe statistical significance. Analyses were performed using SAS Enterprise Guide 7.1 (SAS Institute Inc., Cary, NC).

Results

Study Cohort

Baseline characteristics of the study sample are shown in Table 1. There were some statistically significant (but clinically modest in magnitude) differences in baseline characteristics between the PCI and CABG strata. Patients who underwent PCI were older, were more likely to be smokers, had higher blood pressure, higher creatinine and less likely to receive statin therapy; the CABG group had a slightly higher prevalence of diabetes.

Table 1.

Baseline characteristics of study participants.

	Total sample (n = 10,071)	PCI stratum (n = 7,397)	CABG stratum (n = 2,674)	p-value*
Age, years	67 (10)	67 (10)	66 (9)	<0.0001
Race, % non-Caucasian	1,369 (15)	981 (15)	388 (16)	0.33
Sex, % women	98 (1.0)	78 (1.1)	20 (0.8)	0.17
Body mass index, kg/m2	29.4 (5.9)	29.5 (6.0)	29.3 (5.9)	0.18
Systolic blood pressure, mmHg	129 (22)	130 (22)	127 (21)	<0.0001
Diastolic blood pressure, mmHg	71 (12)	72 (13)	70 (12)	<0.0001
Hypertension treatment, %	9,794 (97)	7,201 (97)	2,593 (97)	0.30
LDL cholesterol, mg/dl	94 (35)	94 (35)	95 (35)	0.20
Statin therapy, %	8,387 (83)	6, 072 (82)	2,315 (87)	<0.0001
Diabetes mellitus, %	5,821 (58)	4,225 (57)	1,596 (60)	0.02
Smoking (current/former), %	8,811 (88)	6,502 (88)	2,309 (86)	0.04
Serum creatinine, mg/dl	1.3 (1.0)	1.4 (1.0)	1.2 (0.8)	<0.0001

PCI = percutaneous coronary intervention; CABG = coronary artery bypass surgery; LDL = low density lipoprotein.

Cells present mean (standard deviation) for age, body mass index, systolic and diastolic blood pressure, LDL cholesterol and serum creatinine; for others they show number (%).

^*For continuous variables we used ANOVA tests and for categorical variables we used Chi-square tests. The tests determined if there was a significant difference between the PCI and CABG groups.

Mean EF₁ (standard deviation) was 35 (10) and the mean follow-up period was 5 years (maximum 14 years). The time (median [interquartile range]) between EF₁ and revascularization procedure was 1 (5) days for CABG subgroup and 2 (9) days for PCI subgroup. The time between revascularization procedure and EF₂ was 242 (236) days for CABG stratum and 246 (231) days for the PCI stratum. The distribution of patients across the five delta-EF categories was 16%, 12%, 23%, 14% and 34% for the CABG stratum and 16%, 12%, 29%, 14% and 29% for the PCI stratum.

Associations of Improvement in EF to Outcomes

Results of models relating delta-EF as a continuous variable to subsequent outcomes are displayed in Table 2. Overall, a 5% improvement in EF was associated with a statistically significant 5% [95% CI: 3 to 7%] and 10% [95% CI: 5 to 15%] reduction in death and HF hospitalization days respectively. Similarly, in the PCI subgroup, each 5% improvement in EF was associated with statistically significant 6% [95% CI: 4 to 8%] and 10% [95% CI: 5 to 16%] reductions in death and HF hospitalization days respectively. In the CABG stratum, reductions in mortality and HF hospitalization days associated with delta-EF were not statistically significant.

Table 2.

Crude events rates and multivariable-adjusted hazards ratios and incidence density ratios relating delta-EF to clinical outcomes of interest.

	Mortality	HF hospitalization days
Total sample	(n=10,071)	(n=9,906)
Crude event rate*	11.48	95.78
Hazards/Density ratio (CI)	0.95 (0.93, 0.97)	0.90 (0.85, 0.95)
PCI stratum	(n=7,397)	(n=7,264)
Crude event rate*	12.42	105.07
Hazards/Density ratio (CI)	0.94 (0.92, 0.96)	0.90 (0.84, 0.95)
CABG stratum	(n=2,674)	(n=2,642)
Crude event rate*	8.80	69.37
Hazards/Density ratio (CI)	0.97 (0.94, 1.01)	0.95 (0.85, 1.07)

HF = heart failure; PCI = percutaneous coronary intervention; CABG = coronary artery bypass surgery; CI = confidence interval

^*For mortality, this represents deaths/100 person-years; for HF hospitalization days, this reflects number of hospitalized days/100 person years.

Hazards ratios/Density ratios (confidence intervals) are from a multivariable Cox regression/negative binomial regression models that included age, sex, race, body mass index, systolic and diastolic blood pressures, hypertension treatment, LDL cholesterol, statin therapy, diabetes, smoking and serum creatinine.

In secondary analyses, when delta-EF was modeled as categories (Table 3), we noted that there was a statistically significant trend towards decreased mortality across groups, with patients in the higher delta-EF categories having lower mortality compared to those in the referent group. These results were similar in the PCI stratum, with a 30% [95% CI: 21 to 37%] lower mortality hazards in patients within the category with the largest delta-EF, compared with those in the referent group. However, in the CABG stratum, trend in mortality across categories of delta-EF did not reach statistical significance.

Table 3.

Secondary analyses. Mortality rates and multivariable-adjusted hazards ratios across categories of delta-EF

	Category 0 EF ≤ −5	Category 1 −5 < EF < 0	Category 2 EF = 0	Category 3 0 < EF < 5	Category 4 EF ≥ 5
Total sample (n=10,071)	n=1,591	n=1,221	n=2,781	n=1,421	n=3,057	p-value for trend
No. of events (%)	1,009 (63.42)	667 (54.63)	1,605 (57.71)	763 (53.69)	1,555 (50.87)	<0.0001
Crude event rate *	14.61	11.32	11.65	11.08	10.16
Adjusted event rate *	12.74	10.46	10.46	10.24	9.39
Hazards ratio (CI)	Referent	0.80 (0.71, 0.91)	0.77 (0.70, 0.85)	0.78 (0.70, 0.88)	0.73 (0.66, 0.81)
PCI Stratum (n=7,397)	n=1,154	n=903	n=2,158	n=1,040	n=2,142	p-value for trend
No. of events (%)	809 (70.10)	534 (59.14)	1,339 (62.05)	613 (58.94)	1,193 (55.70)	<0.0001
Crude event rate *	16.13	12.31	12.45	12.05	10.90
Adjusted event rate *	14.37	11.58	11.49	11.20	10.26
Hazards ratio (CI)	Referent	0.79 (0.69, 0.91)	0.75 (0.64, 0.84)	0.75 (0.66, 0.86)	0.70 (0.63, 0.79)
CABG Stratum (n=2,674)	n=437	n=318	n=623	n=381	n=915	p-value for trend
No. of events (%)	200 (45.77)	133 (41.82)	266 (42.70)	150 (39.37)	362 (39.56)	0.13
Crude event rate *	10.57	8.56	8.78	8.33	8.31
Adjusted event rate *	9.19	7.83	7.83	7.93	7.51
Hazards ratio (CI)	Referent	0.83 (0.64, 1.07)	0.80 (0.64, 0.99)	0.85 (0.66, 1.09)	0.81 (0.66, 1.00)

PCI = percutaneous coronary intervention; CABG = coronary artery bypass surgery; CI = confidence interval; EF = ejection fraction.

^*For mortality, this represents deaths/100 person-years; for HF hospitalization, this reflects number of hospitalized days/100 person years.

Hazard ratios (confidence intervals) are from a multivariable Cox regression model that included age, sex, race, body mass index, systolic and diastolic blood pressures, hypertension treatment, LDL cholesterol, statin therapy, diabetes, smoking and serum creatinine.

In analyses relating delta-EF categories to subsequent HF hospitalization days (Table 4), we observed a graded, statistically significant trend for reduction in HF hospitalization days across the groups; those in the higher delta-EF categories had lower HF hospitalization days compared with the referent group. Similar results were observed in the PCI stratum; patients with the largest delta-EF had an approximately 42% [95% CI: 15 to 60%] lower HF hospitalization days compared with the referent group. We did not observe a significant trend for hospitalization days across delta-EF categories in the CABG stratum.

Table 4.

Secondary analyses. HF Hospitalization rates and multivariable-adjusted incidence density ratios across categories of delta-EF

	Category 0 EF ≤ −5	Category 1 −5 < EF < 0	Category 2 EF = 0	Category 3 0 < EF < 5	Category 4 EF ≥ 5
Total sample (n=9,906)	n=1,565	n=1,201	n=2,742	n=1,391	n=3,007	p-value for trend
Crude event rate *	112.10	122.05	93.36	89.90	83.09	0.0003
Adjusted event rate *	122.71	127.68	96.43	79.86	87.22
Negative binomial density ratio (CI)	Referent	0.99 (0.65, 1.52)	0.70 (0.49, 1.00)	0.65 (0.43, 0.97)	0.59 (0.42, 0.83)
PCI Stratum (n=7,264)	n=1,131	n=888	n=2,126	n=1,014	n=2,105	p-value for trend
Crude event rate *	124.27	136.53	105.85	98.00	86.32	0.0005
Adjusted event rate *	144.69	150.93	111.65	85.48	89.67
Negative binomial density ratio (CI)	Referent	1.07 (0.67, 1.71)	0.74 (0.50, 1.10)	0.66 (0.42, 1.03)	0.58 (0.40, 0.85)
CABG Stratum (n=2,642)	n=434	n=313	n=616	n=377	n=902	p-value for trend
Crude event rate *	80.24	81.89	49.22	67.44	74.99	0.63
Adjusted event rate *	72.33	68.56	50.98	66.26	79.73
Negative binomial density ratio (CI)	Referent	0.81 (0.34, 1.93)	0.62 (0.29,1.31)	0.82 (0.36, 1.87)	0.78 (0.39, 1.55)

HF = heart failure; PCI = percutaneous coronary intervention; CABG = coronary artery bypass surgery; CI = confidence interval

^*For mortality, this represents deaths/100 person-years; for HF hospitalization, this reflects number of hospitalized days/100 person years.

Incidence density ratios (confidence intervals) are from a negative binomial regression model that included age, sex, race, body mass index, systolic and diastolic blood pressures, hypertension treatment, LDL cholesterol, statin therapy, diabetes, smoking and serum creatinine.

Discussion

In this large, hospital-based cohort of older patients with substantial comorbidity burden and HFrEF who underwent coronary revascularization, revascularization, especially PCI, was associated with a modest improvement in EF and a significantly lower burden of subsequent mortality and HF hospitalization days. There was a continuous, graded association between improvement in EF and reduced risk for subsequent death and HF hospitalization days. Our results demonstrate the clinical benefit of coronary revascularization in HFrEF as well as the role of EF improvement in mediating the beneficial effects of revascularization.

Many prior observational studies,^11–15 metanalyses,^3,16 one randomized controlled trial (Surgical Treatment for Ischemic Heart Failure¹⁷ [STICH]), and its extended follow-up study (Surgical Treatment for Ischemic Heart Failure Extension Study¹⁸ [STICHES]) evaluated the benefit of revascularization in patients with CAD and HFrEF. These investigations focused on clinical outcomes alone (survival and/or other non-fatal cardiovascular events), without evaluating the mediating effect of EF improvement. They mostly report that revascularization leads to better outcomes compared with medical therapy alone. Several studies have reported similar survival with CABG and PCI,^11,14–16 especially if complete revascularization is achieved.^11,14 whereas others showed better survival with CABG.¹⁹ A recent sub-analysis of the ISCHEMIA (International Study of Comparative Health Effectiveness With Medical and Invasive Approaches) trial also confirmed the benefit of revascularization in those with CAD and HF with mild-moderately reduced EF²⁰. Other studies focused only on quantifying magnitude of EF improvement associated with completeness of revascularization²¹ and evaluating the determinants of improvement.²² Small prior studies also indicate that method of revascularization does not seem to influence likelihood of EF improvement²³. Our investigation is unique in that it is a large, “real world” cohort without exclusion of any clinical subgroup and that we simultaneously evaluated the magnitude of EF change associated with revascularization and the relations of such EF change with risk for subsequent clinical outcomes.

While our study was not intended to directly compare the effects of PCI vs. CABG, the different results observed in the PCI and CABG groups in our study warrant discussion. Mortality after CABG is higher in patients with CAD and HFrEF (compared to those with normal EF), since preoperative reduced EF is a strong risk factor for early mortality after CABG.²⁴ For example, in the STICH trial, 30-day mortality in the CABG group was ~4-fold higher compared to the medical management group. However, in both STICH and other CABG clinical trials, a “catch up” effect has been noted wherein long-term mortality benefit in those who survive the initial postoperative phase makes the overall results in the CABG arm of the clinical trial better. Since this effect may take several years to manifest, as evident when comparing the results of STICH and STICHES, it is possible that the mean follow-up period of 5 years in our study may not have been sufficient to uncover the long-term mortality benefit after CABG. Alternately, since clinical trial patients are younger compared to real-world patients and age strongly modifies the effect of CABG on subsequent outcomes,^25,26 it is possible that the older age of our cohort may have contributed to the lesser benefit of CABG. PCI, which does not have an early mortality penalty similar to CABG, may conceivably be a better option in terms of harmonized risk vs. benefit in older patients with CAD and reduced EF needing revascularization; in younger patients and those with a lower comorbidity burden, CABG is likely still the preferred revascularization strategy. And lastly, it is possible that the mechanism mediating the benefit of CABG and PCI is different. CABG, which reduces rates of subsequent non-fatal cardiovascular events and coronary death in patients with multivessel CAD may be exerting its beneficial effects on mortality independent of EF improvement (e.g. by reduction of fatal myocardial infarctions), whereas PCI’s effect might be largely mediated via EF improvement.

Another important observation in our investigation is the wide interindividual variation in delta-EF. Prior studies have also noted that post-revascularization EF improvement is generally modest and variable, and a significant number of patients have lower post-procedure EF. This could be for various reasons. Although controversy exists regarding whether revascularization decisions should be made based on myocardial viability, prior studies do indicate that the presence of viability predicts likelihood of EF recovery. Another factor could be the extent of revascularization (complete vs. incomplete). The use of invasive physiological assessment to demonstrate flow limitation and guide revascularization may also influence likelihood of EF improvement. And lastly, variations in post-revascularization medical management may explain some of the variability of delta-EF.

The clinical implications of our findings are two-fold. If EF improvement is a mediating mechanism for improvement in outcomes, but is highly variable, this implies we have a tremendous opportunity for patient selection, such that revascularization (which is expensive and carries procedural risks) can be reserved for those with higher likelihood of EF improvement and the rest treated with guideline derived optimal medical therapy. Second, since PCI is associated with EF improvement that seems to translate to improved clinical outcomes and does not require survival long enough to see a catch-up benefit like CABG, older patients and those with higher surgical risk may be better served with PCI.²⁷

Limitations

We could not ascertain the specific indication for revascularization. In addition, we were unable to systematically ascertain the receipt of various guideline derived medical therapy or if there was variable intensification of such therapy pot-revascularization. Both of these factors could have an impact on delta-EF; however, this may not be systematically different between PCI and CABG cohorts. We also do not have data on the severity of CAD (e.g. SYNTAX score for each patient) or extent of revascularization (complete vs. incomplete), myocardial viability, or specific adjunctive procedural medications or techniques used. Advances in both medical management and procedural and technological innovations might mean that contemporary patients may be deriving a larger, or more consistent improvement in EF with revascularization compared with what we observed, and advanced imaging modalities might provide more precise estimates.²⁸ The observational nature of the study cannot completely rule out confounding despite statistical adjustment.

Conclusions

Patients with HFrEF are at high risk for mortality and recurrent hospitalizations. Our results demonstrate that EF improvement after revascularization was variable but was associated with a significant reduction in mortality and hospitalization burden, especially after PCI. Further research is needed to examine any differences between PCI and CABG on EF improvement and outcomes and to refine patient selection, such that those with the most expected benefit in terms of EF improvement can be offered the most appropriate method of revascularization in addition to optimal medical therapy.

What Is Known

• Surgical revascularization is usually performed to treat coronary artery disease in patients with heart failure and reduced ejection fraction and leads to improved long-term survival compared with medical management.

• However, whether ejection fraction improvement is the mechanism mediating this benefit is still unclear.

• In addition, whether percutaneous revascularization achieves comparable improvements in systolic function and clinical outcomes is not established.

What the Study Adds

• In patients with heart failure with reduced ejection fraction and coronary artery disease, revascularization-associated change in ejection fraction was associated with significantly lower rates of mortality and heart failure hospitalization burden.

• The improvements in clinical outcomes were particularly notable in those undergoing percutaneous coronary intervention.

List of Abbreviations

CABG

coronary artery bypass grafting

CAD

coronary artery disease

EF

ejection fraction

HF

heart failure

HFrEF

heart failure with reduced ejection fraction

PCI

percutaneous coronary intervention

VA

veterans affairs

BMI

body mass index

LDL-C

low density lipoprotein cholesterol