Written by: Chetan P HudedJohn A SpertusPhilip G JonesSean M O'BrienDaniel B MarkSripal BangaloreGregg W StoneDavid O WilliamsHarvey D WhiteWilliam E BodenHarmony R ReynoldsJudith S HochmanDavid J MaronISCHEMIA Research Group
Find these authors on Roon
DM
Dr. Daniel Mark
Cardiac Critical Care · Duke
GS
Dr. Gregg Stone
Interventional Cardiology · ISMMS
METHODS
The design of the ISCHEMIA trial and its clinical and health status outcomes between INV and CON strategy arms have been previously reported.5, 10, 11 In brief, ISCHEMIA was a multi-center international randomized controlled trial of 5179 patients with CCD and at least moderate ischemia on stress testing comparing an initial INV management strategy (coronary angiography and revascularization, if possible, plus guideline-directed medical therapy) and an initial CON strategy (guideline-directed medical therapy alone). Stress testing was performed at individual study sites, with subsequent review by a core laboratory after randomization. Key trial exclusion criteria included coronary computed tomography angiography (CCTA) with left main stenosis >50% or no significant coronary disease, left ventricular ejection fraction <35%, New York Heart Association functional class III or IV heart failure, and unacceptable angina despite maximally tolerated anti-anginal medical therapy. For this analysis, 481 participants from 5 sites with invalid health status form completion were excluded, 81 participants without baseline or any follow-up Seattle Angina Questionnaire (SAQ) data were excluded, as were 196 with prior CABG. The population thus comprised 4,421 patients randomized in the ISCHEMIA trial with available health status assessments. The trial was approved by the institutional review board at each study site, and written informed consent was obtained from all participants. To minimize the possibility of unintentionally sharing information that can re-identify private information, a subset of the data generated for this study is available at BioLINCC (https://biolincc.nhlbi.nih.gov/studies/ischemia/).Intensive guideline-directed medical therapy for CCD was recommended for all participants, regardless of assignment to INV or CON treatment arms. By protocol, participants randomized to the INV strategy additionally underwent invasive coronary angiography within 30 days and subsequent revascularization to all feasible ischemic territories. The method of revascularization (PCI, CABG, none) for each individual participant was determined by the local heart team after diagnostic angiography. The study provided an algorithm for using fractional flow reserve (FFR) testing to supplement the non-invasive ischemia evaluation to guide PCI.11The present analysis stratified participants based on the mode of revascularization, if any, undertaken within 90 days (Figure 1). Invasively managed participants treated with PCI and CABG were restricted to those who were revascularized within 90 days. For 16 participants in whom both PCI and CABG were performed within 90 days, treatment assignment was based on the intended management strategy documented by the local heart team, most commonly the first procedure performed. Participants in the INV arm who did not receive revascularization within 90 days (i.e., those who did not undergo an invasive angiogram, those with no significant obstructions, and those with such extensive disease that the risks of revascularization were deemed by the local heart team to exceed its benefits) were described separately. The CON group included all participants randomized to a CON strategy, regardless of whether revascularization was ultimately performed. The rate of revascularization within 6 months was <9.0% in this group, with most occurring after suspected clinical events, 1.6% for unacceptable symptoms, and 2.6% for protocol non-adherence.12
Participants’ health status was assessed prior to randomization, at 1.5 months, 3 months, 6 months, and every 6 months thereafter until study completion with the SAQ,2, 8 Rose Dyspnea Scale,13 and the European Quality of Life 5 Dimensions (EQ-5D) visual analogue scale.14
The SAQ is a disease-specific patient reported outcome tool to assess angina-related health status in patients with CCD.2 The shortened 7-item version was the primary health status instrument used in the trial.15 Validated translations were used in each country (www.cvoutcomes.org). The SAQ provides separate scores for angina frequency (AF), physical limitation (PL), and quality of life (QoL). The average of these individual scores generates the SAQ summary score (SS). Scores range from 0–100 with higher scores indicating better health status. SAQ AF scores of 0–30, 31–60, 61–99, and 100 correspond to daily, weekly, monthly, and no angina respectively. For the SAQ SS, PL, and QoL scores, values of 0–24, 25–49, 50–74, and 75–100 can be interpreted as very poor to poor, poor to fair, fair to good, and good to excellent respectively. A 5-point change in the SAQ SS has been suggested as a minimum difference that would be considered clinically important to most patients.16
The Rose Dyspnea Scale is a 4-item patient reported outcome tool assessing the presence or absence of dyspnea under 4 circumstances, yielding a score of 0–4.13 A score of 0 indicates no dyspnea while a score of 4 indicates major limitation due to dyspnea. Although the Rose Dyspnea Scale is not a disease-specific tool for CCD, higher scores on the Rose Dyspnea Scale have been associated with lower quality of life in patients with CCD, independent of angina frequency.17
The EQ-5D visual analogue scale is patient reported outcome tool for general health status assessment. Patients rate their health on a scale from 0 (worst imaginable health status) to 100 (best imaginable health status).3
Categorical variables are presented as number (%), and continuous variables are presented as mean ± standard deviation. The primary aim of this analysis was to describe health status outcomes, stratified by revascularization method, among participants in the ISCHEMIA trial. To achieve this aim, the primary outcome was the unadjusted SAQ summary score. Secondary outcomes were the unadjusted SAQ AF, SAQ PL, SAQ QoL, Rose Dyspnea Scale, and EQ-5D Visual Analogue Scale. Unadjusted health status scores to 48 months are presented.
A secondary aim of the study was to estimate adjusted differences in 1-year and 3-year health status outcomes between CON, PCI and CABG groups. There were 651 participants in the INV arm that did not undergo PCI or CABG within 3 months or had a primary endpoint prior to revascularization, and these were excluded from risk-adjusted comparisons, leaving a population of 3,770 for the risk-adjusted analyses. Given the inherent differences between participants selected for PCI and CABG, and the high likelihood of selection bias, risk-adjustment with a combination of propensity scoring and regression was used. All comparisons included adjustment for age, sex, geographic region, hypertension, diabetes, smoking status, prior MI, heart failure, prior stroke, cerebrovascular disease, peripheral arterial disease, prior PCI, left ventricular ejection fraction, body mass index, estimated glomerular filtration rate, New York Heart Association heart failure classification, core lab-assessed degree of ischemia, and baseline SAQ, Rose Dyspnea Scale, and EQ-5D scores. For the comparison of PCI and CABG, we additionally included variables obtained from the invasive coronary angiogram: Duke Jeopardy Score, SYNTAX score, left main disease, left anterior descending (LAD) disease, proximal LAD disease, left circumflex disease, right coronary artery disease, number of chronic total occlusion lesions, number of calcified lesions and number of tortuous lesions. Finally, for comparisons of PCI to CON and CABG to CON, we repeated the analyses among the 74% of participants for whom CCTA was performed, including additional adjustment for CCTA-derived SYNTAX score and the number of diseased vessels, in addition to the sociodemographic and clinical variables described above.
For each pairwise comparison of treatment groups, the following approach was used. First, propensity scores for treatment were generated using logistic regression on the relevant set of covariates. The overlap of propensity score distributions between the two groups (plotted visually using nonparametric kernel density estimates; Figure 2 and Figure S1) was then examined and participants whose scores exceeded the 1st or 99th percentile of scores in the alternate treatment group were excluded. For example, when comparing CON and PCI, CON participants whose probability of PCI was lower than the 1st percentile of scores in the PCI group were excluded, as were PCI participants whose probability of PCI was higher than the 99th percentile of scores in the CON group. Within the restricted cohort, proportional odds regression models of health status outcomes on treatment group were fit, adjusting for the same variables included in the propensity score as covariates. While the propensity scoring step served to limit inferences to participants who have at least some likelihood of receiving either treatment, covariate adjustment (i.e. rather than propensity matching or weighting) produces conditional, or patient-specific, estimates of the effect of treatment, which we believe to be generally more relevant to clinicians and patients for health status outcomes than population-average estimates. To examine potential heterogeneity of treatment effects, we re-fit the outcome models including only treatment group, the (logit) propensity score as a linear effect, and a treatment-by-propensity score interaction. This evaluates whether the benefit of treatment increases or decreases with the likelihood of receiving the treatment. For both propensity and outcome models, nonlinear effects of continuous covariates were explored by using restricted cubic splines. Treatment effects are reported as estimated odds ratios and 95% confidence intervals for better health status with PCI vs. CON, CABG vs. CON and CABG vs. PCI. Treatment effects were estimated across all participants and stratified by baseline angina frequency defined by the SAQ Angina Frequency score (daily/weekly angina: 0–60; monthly angina: 70–90; no angina: 100).
Of the 3,770 participants included in these analyses, 5.4% were missing 1-year health status scores (3.1% missed assessments, 1.2% lost to follow-up, 0.9% died, 0.2% study withdrawal) and 45.4% were missing 3-year scores (37.2% due to study termination, 3.3% died, 2.8% missed assessments, 1.7% lost to follow-up, 0.4% study withdrawal). There were no missing covariates, except for angiographic factors and CCTA factors (which we addressed by conducting secondary analyses). For the analyses estimating treatment effects, we imputed missing SAQ domain scores using multiple imputation methods (R package ‘mice’).18 The imputation model included treatment group, all covariates described above, and all health status scores between baseline and 1 year. Forty imputed data sets were generated and the results were pooled using Rubin’s method to obtain final treatment effect estimates.19 All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC) and R version 4.1.3.
RESULTS
The study population comprised 2,232 participants in the CON group, 1,198 in the PCI group, and 340 in the CABG group. There were 146 participants randomized to the INV group who did not undergo an invasive angiogram within 90 days, 230 who had no significant disease identified by invasive angiography, and 249 who had significant disease not amenable to revascularization.
Characteristics of the study population stratified by revascularization method are summarized in Table 1. Mean age was between 63 to 67 years, and most participants in each group were men of White race. Cardiovascular risk factors and comorbidities were common in each group. Moderate or severe ischemia on stress testing, confirmed by the core laboratory, was present in >85% of each group, but was most common in the CABG group (92.4%). Coronary anatomy was also more complex in the CABG group. SYNTAX score <23 was observed in 81.6% in the PCI group vs. 33.1% in the CABG group and SYNTAX score >32 was observed in 3.7% in the PCI group vs 30.2% in the CABG group. Left main disease was rare but was observed in a higher proportion of CABG-treated participants. At least 1 chronic total occlusion was found in 38.0% of the PCI group vs. 64.0% of the CABG group.
Baseline health status stratified by revascularization groups is summarized in Table 2. Mean baseline SAQ SS ranged from 72–79 points across groups, with the lowest baseline score in the PCI group (72.1). Daily/weekly angina frequency was observed in 15–24% of patients, while monthly angina (38–46% of patients) and no angina (30–43% of patients) were more commonly observed in all groups.
Unadjusted health status scores through 48 months are provided in Figures 3–4 and Table S1 (Data Supplement). For the SAQ SS, SAQ quality of life (QoL), and SAQ physical limitation (PL) scores, similar patterns were observed in which the PCI group demonstrated the largest early changes from baseline to 3-months. Changes in the CABG group were somewhat delayed, but the highest scores were consistently observed in the CABG groups from 6 through 48 months. SAQ AF score and freedom from angina rates were highest in the CABG group at all time points from 3 through 48 months. Table 3 provides the proportion of participants with no angina from 3 through 48 months, as assessed by the SAQ AF category. Rose Dyspnea Scores were similar between PCI and CABG at 3 months and lowest (more favorable) in the CABG group from 6 through 48 months. EQ-5D Visual Analog score was highest in the PCI group at 3 months and highest (more favorable) in the CABG group from 6 through 48 months.
Among randomized participants in the INV arm who did not undergo revascularization, there was considerable heterogeneity, but their scores during follow-up were more like CON-managed participants than those managed with PCI or CABG.
In risk-adjusted analyses, higher odds of favorable SAQ SS and SAQ AF at 1 year were observed for both PCI (SAQ SS OR 1.98 [95% CI 1.73, 2.26] and SAQ AF OR 2.10 [95% CI 1.78, 2.48] vs. CON) and CABG (SAQ SS OR 2.56 [95% CI 2.04, 3.22] and SAQ AF OR 3.46 [95% CI 2.52, 4.76] vs. CON) compared with the CON group (Figure 5). The findings of higher SAQ scores at 1 year with PCI vs. CON and CABG vs. CON were confirmed in a sensitivity analysis restricted to participants with baseline CCTA that allowed for further risk-adjustment by including CCTA assessments of coronary disease severity (Figure S2). At 1 year, CABG was associated with a higher likelihood of favorable SAQ AF than PCI (OR 1.54 [95% CI 1.03, 2.31]) but no significant difference in SAQ SS was observed between CABG and PCI at 1 year.
At 3 years, a higher odds of favorable SAQ SS and SAQ AF scores were observed with both PCI (SAQ SS OR 1.58 [95% CI 1.35, 1.84] and SAQ AF OR 1.48 [95% CI 1.22, 1.79] vs. CON) and CABG (SAQ SS OR 1.77 [95% CI 1.36, 2.31] and SAQ AF OR 1.59 [95% CI 1.12, 2.25] vs. CON; Figure 6). The sensitivity analysis of patients with inclusion of baseline CCTA for risk-adjustment showed similar findings (Figure S3). By 3 years, there was no difference in the odds of a higher SAQ SS or SAQ AF score between CABG and PCI.
In a secondary analysis of health status outcomes stratified by baseline angina frequency, PCI and CABG both showed higher risk-adjusted odds of favorable 1-year SAQ SS and SAQ AF compared with CON, regardless of baseline angina frequency (Figure S4). Among participants with daily/weekly baseline angina, CABG was associated with higher odds of a more favorable SAQ SS than PCI at 1 year (OR 2.09 [95% CI 1.05, 4.15]).
At 3 years, the odds of a more favorable SAQ SS were generally higher in both CABG and PCI compared with CON, regardless of baseline angina severity. However, benefits in terms of SAQ AF score among CABG and PCI treated participants compared with CON were generally restricted to those with daily/weekly angina at baseline. Findings of the 1 and 3-year models of PCI vs. CON and CABG vs. CON among patients with baseline CCTA for risk-adjustment, stratified by baseline angina frequency, are provided in Figures S6 and S7.
DISCUSSION
Following the initial recovery period from CABG, some, but not all, prior studies have demonstrated greater late health status benefits with CABG as compared with PCI. This secondary analysis of the ISCHEMIA trial thus explored whether the sustained health status benefits of an INV strategy were primarily driven by CABG. Significant health status benefits among participants managed with both PCI and CABG compared with CON (Central Illustration) were found. Durable improvements in health status were observed in both groups, and these improvements with both PCI and CABG were greater than those observed in the CON group based on risk-adjusted comparisons through 3 years of follow-up. When directly comparing CABG and PCI, CABG was observed to have better angina control at 1 year, but not at 3 years.
Nearly all prior trials have demonstrated better early symptom relief after PCI than CABG during the post-operative recovery period of 1–6 months. After this period, some prior studies have suggested that CABG is associated with better intermediate-term health status compared with PCI, but long-term differences in health status after these procedures appear to be minimal.6, 7, 20, 21 In the FREEDOM trial of patients with CCD, diabetes, and multivessel coronary disease randomized to PCI or CABG, the CABG arm showed better health status through 2 years, but not later.7 Similarly, data from a large Canadian registry of diabetic patients with multivessel CCD treated with PCI and CABG demonstrated higher (better) SAQ scores across multiple domains with CABG at 1 year, with attenuated differences by 3 years.21 The 5-year outcomes of the SYNTAX trial, which compared PCI with drug-eluting stents vs. CABG for the treatment of left main or 3-vessel CCD, reported a small but significant improvement in SAQ AF with CABG over PCI that persisted out to 5 years.6 In the EXCEL trial of left main disease revascularization, however, health status outcomes were better with PCI compared with CABG at 1 month but then similar at 1 and 3 years.20 In a recently published systematic meta-analysis of all 5 randomized trials of PCI vs CABG in which QoL was formally assessed in 10,760 patients, the mean difference in SAQ-AF scores were similar between procedures at 1 month, favored CABG at 1 year, and were similar between procedures between 36 and 60 months.9
The health status outcomes of PCI vs. medical therapy have also been evaluated in multiple prior trials with mixed findings. The COURAGE trial, which randomized patients with CCD by coronary angiography to PCI or medical therapy, identified a small but significant health status benefit early (SAQ QoL 73+/−22 with PCI vs. 68+/−23 with medical therapy at 3 months) which dissipated by 3 years of follow-up.4 Patients with more severe angina had greater health status benefit with PCI in that analysis, similar to the findings in ISCHEMIA.5. Of note, however, most PCI patients in COURAGE were treated with bare metal stents rather than drug-eluting stents. The substantially high restenosis rates with BMS, coupled with performance of PCI in the OMT arm during follow-up in nearly one-third of patients, may explain the modest QOL benefits at 1 year and loss of a durable effect of PCI over time The Bypass Angioplasty Revascularization Investigation 2 Diabetes study was a randomized trial of revascularization (with stratum for CABG and PCI) compared with medical therapy in patients with CCD. In that study, improved health status was observed in the revascularization group, but this effect was primarily driven by patients treated with CABG while those treated with PCI had similar health status to medically treated patients.8 However, the PCI and CABG arms were not randomized in BARI-2D, and a substantially greater amount of myocardium was at risk and was revascularized in the CABG strata. More recently, the ORBITA study of medically optimized patients with CCD demonstrated no improvement in 6-week exercise time with PCI compared with a sham PCI procedure,22 whereas the ORBITA-2 study of CCD patients on no anti-anginal medical therapy demonstrated reduced angina with PCI compared with sham PCI.23
Given the better health status observed with CABG over PCI in FREEDOM and SYNTAX, and the questionable long-term effects of PCI on health status observed in COURAGE, BARI-2D, and ORBITA, one might have hypothesized that the health status benefits achieved with an INV strategy in ISCHEMIA was driven primarily by CABG. However, the present analysis demonstrates that both PCI and CABG, when selected in the context of a multidisciplinary heart team guided by evidence-based study recommendations, were associated with improved angina-related health status through 3-year follow-up. These findings are congruent with the 3-year outcomes of the Fractional Flow Reserve Versus Angiography for Multivessel Evaluation (FAME) 2 trial, which randomized patients with CCD and documented ischemia by FFR assessment to PCI vs. medical therapy and demonstrated a significant reduction in angina severity with PCI out to 3 years.24 More recently, the FAME 3 trial, which randomized patients to FFR-guided PCI vs. CABG for multivessel CCD, demonstrated no difference in EQ-5D scores at 1 year between PCI and CABG.25 In this regard only the EXCEL and FAME trials used second generation contemporary drug-eluting stents, which may have contributed to similar QOL improvements with PCI compared with CABG, and more durable effects of PCI demonstrated in this analysis of ISCHEMIA.
Other potential explanations may have contributed to the durable health status benefits of PCI vs. CON and the similar long-term health status of CABG vs. PCI observed in this study. First, revascularization methods in the ISCHEMIA trial were non-randomized and guided by evidence-based protocol recommendations and local heart team decisions. This may have improved selection of patients treated with PCI based upon their providers’ beliefs in the success and sustainability of percutaneous revascularization. Second, in the ISCHEMIA trial, revascularization was performed with the goal of relieving all significant areas of ischemia. Although most participants in the study did not have complete anatomic revascularization, the rates of both anatomic and functional complete revascularization were approximately 13% higher in the INV-PCI group compared with the INV-CABG group, supporting providers’ judgements in recommending PCI. We have previously demonstrated from ISCHEMIA that complete ischemic revascularization may optimize clinical and health status benefits.26, 27 These results, when taken in the context of the prior results of FAME 2 and FAME 3, suggest PCI targeting complete relief of objective ischemia may achieve better long-term health status improvements than PCI based only on anatomic assessments of coronary stenoses.
Medical therapy is the cornerstone of cardiovascular risk reduction in CCD patients,28–30 and revascularization does not further reduce the risk of mortality or myocardial infarction in the absence of high-risk coronary anatomy.31–33 Therefore, the decision to revascularize a patient with CCD should be based largely on the health status benefits of treatment and the incorporation of patients’ preferences. In this study, it was observed that 61.4% of participants in the CON group were angina-free at 1 year and 70.4% were angina-free at 3 years, highlighting the potential benefits of medical therapy in managing patients with CCD. However, while the ISCHEMIA trial demonstrated that CCD participants managed invasively had better health status through 4 years than participants managed conservatively, this analysis expands upon those results, demonstrating that both PCI and CABG were associated with durable improvements in health status. This approach is consistent with the American College of Cardiology/American Heart Association guidelines on coronary revascularization, which give a class 1A indication for revascularization in patients with CCD and refractory symptoms despite medical therapy.32 To aid in the selection of INV and CON options for CCD management, a shared decision-making tool for patients and physicians was developed using data from the ISCHEMIA trial to estimate an individual patient’s 1-year health status and 4-year survival with INV and CON approaches to manage their CCD (https://myhealthoutcomes.org/tools/shared-decision-making-tool/). 34, 35
This was an observational study of health status changes over time among patients treated with PCI and CABG in a non-randomized fashion, compared with patients who were managed conservatively. The treatment groups in this study were clearly different in terms of baseline characteristics and coronary anatomy, with the CABG group showing higher anatomic complexity than the PCI group. This highlights that the decision to perform PCI vs. CABG was greatly influenced by coronary anatomy. The findings of this analysis support that this approach is effective in achieving long-term health status improvements regardless of the revascularization method ultimately selected, when guided by a multidisciplinary heart team, with the goal of complete revascularization. Our group has previously reported that complete revascularization in the ISCHEMIA trial was associated with better health status than incomplete revascularization in symptomatic patients.26 The use of a multidisciplinary heart team for patients with coronary disease is now supported with a class IB recommendation in the 2021 American College of Cardiology/American Heart Association Coronary Revascularization Guidelines.32
The study results should be interpreted within the context of the following limitations. First, as highlighted above, CABG and PCI revascularization strategies were not randomly assigned and there were clear differences in baseline characteristics and coronary anatomy, which cannot be fully accounted for despite extensive risk adjustment. For example, factors such as anxiety, depression, chronic pain, frailty, and functional limitations were not included in the risk-adjustment, and these may have influenced treatment allocation and response to intervention. Therefore, this study is a descriptive summary of the health status outcomes of these procedures when selected within the framework of a multidisciplinary heart team and not representative a randomized comparison of these strategies. An advantage of this approach, however, is that the study findings may more closely represent real-world clinical practice in which patients are selected for revascularization strategies in a non-randomized fashion. Second, the ISCHEMIA trial did not use sham control procedures when comparing INV and CON arms. Furthermore, the comparison of PCI and CABG was adjusted for both clinical and angiographic variables, but the comparison of each revascularization strategy with CON could not be adjusted for angiographic variables, as conservatively managed participants did not undergo a diagnostic angiogram. To account for this limitation, a sensitivity analysis was conducted among participants with baseline CCTA studies performed. CCTA findings were used to adjust comparisons of PCI vs. CON and CABG vs. CON for anatomic severity of baseline coronary disease. However, baseline CCTA data were not available in all participants and are not equivalent to invasive angiography adjudicated by a core lab, although there was a high correlation between these techniques in the ISCHEMIA trial.36 Finally, the cohort that underwent CABG was small (n=340) compared with the cohort that underwent PCI (n=1,198).
Conclusion.
Among participants with CCD in the ISCHEMIA trial, both PCI and CABG were associated with better angina-related health status than CON through 3 years when these procedures were selected within the context of a multidisciplinary heart team. Angina relief was better with CABG than PCI at 1 year, but no difference was observed between CABG and PCI at 3 years.
DM
Dr. Daniel Mark
Cardiology · Cardiac Critical CareDuke University Health Systemdanielmark
