Long-term Survival Benefit From Revascularization Compared With Medical Therapy in Patients With or Without Diabetes Undergoing Myocardial Perfusion Single Photon Emission Computed Tomography

Diabetes is one of the major risk factors for coronary artery disease (CAD) and is increasingly prevalent among contemporary populations (1). While optimal medical treatment is the foundation of CAD management in patients with diabetes (2,3), coronary artery revascularization is an additional treatment option, depending on the patient’s clinical characteristics, risk level by noninvasive testing, and CAD burden (4). Current guidelines emphasize individualized consideration for the need and optimal selection of revascularization strategy for patients with diabetes (3,5).

A major goal of CAD management, including medical treatment and revascularization, is to improve survival. Previous observational studies suggested the survival benefit of revascularization might vary depending on patients’ ischemic burden (6–9). Compared with patients without diabetes, patients with diabetes have a higher prevalence of abnormal results on single photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) (10) and a higher risk of adverse outcomes even at a similar degree of myocardial ischemia (11,12). However, whether there is a difference in potential benefit of revascularization according to ischemic burden in patients with diabetes compared with patients without diabetes has not been evaluated previously. The major objective of the current study was to explore the potential differences in the benefit of early revascularization between patients with and without diabetes using a large observational cohort of patients who underwent SPECT-MPI with long-term follow-up for all-cause mortality (ACM).

The current study included 52,853 consecutive patients who underwent stress/rest SPECT-MPI at Cedars-Sinai Medical Center from 1 January 1998 to 31 December 2017. After excluding patients with known valvular heart disease, cardiomyopathy, or history of cardiac transplantation (n = 1,586), missing essential data (diabetes status, early revascularization status, or key prognostic risk factors; n = 8,016), and those lost to follow-up (n = 1,269), 41,982 patients were included. For patients who underwent more than one test during the study period, only the first test was evaluated. All patients were prospectively enrolled in a research database at the time of testing and were followed for the occurrence of mortality. The study was approved by the Cedars-Sinai Medical Center Institutional Review Board and complied with the Declaration of Helsinki.

Data regarding early revascularization were prospectively collected and included patients who underwent revascularization with percutaneous coronary intervention (PCI) or coronary artery bypass graft surgery (CABG) within the first 90 days after the SPECT-MPI. Hereafter, these patients are referred as the “early revascularization” group, and the patients who did not undergo early revascularization are referred to as the “medical treatment” group.

Demographic information was obtained at the time of the SPECT-MPI, including age, sex, cardiac symptoms, CAD risk factors, BMI, medication use, and prior CAD status. Patient CAD risk factors included hypertension, hyperlipidemia, diabetes, current smoking, family history of premature CAD, peripheral vascular disease, chronic kidney disease, and stroke. Patients with diabetes were categorized as insulin treated diabetes versus non-insulin treated diabetes according to insulin use status at the time of scanning.

All patients underwent symptom-limited exercise or pharmacological stress testing using adenosine, regadenoson, or dobutamine. Stress/rest SPECT-MPI was performed according to standard protocols. Patients were imaged using a dual-isotope rest thallium-201/stress technetium-99m (^99mTc)-sestamibi protocol until 2007, a rest/stress ^99mTc-sestamibi protocol until 2012, and then a stress/rest ^99mTc-sestamibi protocol thereafter. Semiquantitative visual interpretation of SPECT-MPI images was performed by experienced observers according to a 5-point score (0 = normal to 4 = absence of tracer uptake) for each myocardial segment, with division of images into 20 myocardial segments before February 2005 and 17 myocardial segments thereafter (13). Summed stress score (SSS), summed rest score (SRS), and summed difference score (SDS) were generated and converted to percent myocardium by dividing summed scores by 80 for studies with 20-segment analysis or by 68 for studies with 17-segment analysis, and then multiplying by 100 (13). We defined SDS <5% as the presence of no ischemia, and SDS 5–9.9% as mild, 9.9–14.9% as moderate, and ≥15% as severe myocardial ischemia (14). Left ventricular ejection fraction (LVEF) was measured with 8- or 16-frame gating using Quantitative Gated SPECT software (Cedars-Sinai Medical Center, Los Angeles, CA).

The primary end point was ACM. Follow-up for ACM was performed by using internal hospital medical records as well as the Social Security Death Index, California Noncomprehensive Death File, and National Death Index. The last date of access for the Social Security Death Index was 9 April 2012, the last date of access for California Noncomprehensive Death File was 22 July 2020, and the last date of access for the National Death Index was 12 February 2018. The median follow-up duration was 11.3 years (interquartile range [IQR], 7.7–14.1).

Categorical variables are shown as n (%). Continuous variables are shown as mean ± SD or median values (IQR). Categorical variables were compared by the χ² test, and continuous variables were compared by the Student t test or Mann-Whitney U test, as appropriate.

To compare the survival benefit from early revascularization compared with medical treatment between patients with and without diabetes, we adjusted for confounding factors using propensity score matching for reducing the impact of differences in baseline characteristics among patients with and without diabetes (15). Propensity scores were calculated from the predicted probabilities of a multiple logistic regression model predicting diabetes by using the following variables: early revascularization, ischemic category (no [SDS, <5%], mild [SDS, 5–9.9%], moderate [SDS, 10–14.9%], or severe ischemia [SDS, ≥15%]), age, sex, BMI, hypertension, hyperlipidemia, family history of CAD, smoking, prior history of CAD, chronic kidney disease, peripheral vascular disease, a history of stroke, stress test type (pharmacological or exercise stress), LVEF, %SRS, and presence of chest pain or shortness of breath (16). We used a caliper width equal to 0.2 of the SD of the propensity score for 1:1 nearest neighbor propensity score matching. Adequacy of matching was assessed with absolute standardized differences, with residual differences >0.1 considered significant (17).

Kaplan-Meier survival curves, stratified by patients with or without diabetes and early revascularization or medical treatment in each ischemic category, no ischemia (SDS <5%), mild ischemia (SDS, 5–9.9%), or moderate to severe ischemia (SDS ≥10%), were used to assess the primary outcome of ACM and compared using the log-rank test, followed by the Holm post hoc test. A Cox regression model was used to assess associations between early revascularization versus medical treatment and ACM among the propensity-matched cohort with and without diabetes overall and in each ischemic category (no, mild, moderate, and severe ischemia). The following variables were included in all multivariable models for adjustment: age, sex, BMI, diabetes (no diabetes, non–insulin-treated diabetes, and insulin-treated diabetes), hypertension, hyperlipidemia, family history of CAD, smoking, prior history of CAD, chronic kidney disease, peripheral vascular disease, a history of stroke, stress test type (pharmacological or exercise stress), LVEF, %SDS, %SRS, and cardiac symptoms (8). We used a multivariable model to account for nonrandomization early revascularization. Our model included the same components as the analysis performed by Patel et al. (8), where ischemia was only accounted for in the multivariable model. The hazard ratios (HRs) of early revascularization versus medical treatment for ACM at levels of each percent ischemic myocardium (i.e., %SDS) were calculated and plotted to identify the threshold of ischemic myocardium at which the patient may have survival benefit from early revascularization compared with medical treatment, where the upper 95% CI crosses an adjusted HR of 1.00. The interactions between early revascularization versus medical treatment and diabetes and %SDS were assessed in predicting ACM. Among patients with diabetes, we evaluated potential interactions between insulin use (insulin-treated or non–insulin-treated) and type of revascularization (CABG or PCI) with mortality outcomes using the Cox model. We conducted a subanalysis after excluding patients with impaired LVEF (<35%). The interactions between insulin use, type of revascularization, and prognosis were also assessed. To assess whether the association between revascularization and ACM changed over time in relationship to the changing proportion of myocardial ischemia and improvement of medical treatment, we assessed for interaction between revascularization and two time periods (1998–2007 and 2008–2017). A two-sided P < 0.05 was considered statistically significant. All statistical analyses were performed with R 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria) or Stata 16 (StataCorp LLC, College Station, TX) software.

The study included 41,982 patients, and 8,328 (19.8%) patients had diabetes (Table 1). Patients with diabetes were older (63.3 vs. 61.6 years; P < 0.001), more commonly had lower LVEF (60.5% vs. 63.9%; P < 0.001), moderate to severe ischemia (8.5% vs. 4.8%), higher mean percent ischemic myocardium (2.5% vs. 1.4%; P < 0.001), and a higher rate of early revascularization (7.9% vs. 4.6%; P < 0.001) (Table 1). Patients with diabetes more frequently underwent a pharmacological stress test compared with patients without diabetes. After propensity score matching, the number of patients with diabetes decreased to 8,046 matched to 8,046 patients without diabetes. The diabetes and nondiabetes groups were well matched, with no significant differences in baseline characteristics. Standardized differences of all matching covariates between groups were <0.05, indicating excellent covariate balance (Supplementary Table 1) (15,17).

In the propensity-matched cohort, 2,983 patients (37.1%) died in the diabetes group and 2,566 patients (31.9%) died in the nondiabetes group during a median follow-up time of 10.3 years (IQR 6.6–13.1). The annualized mortality rate was higher in the diabetes group compared with the nondiabetes group (4.7 vs. 3.6%; P < 0.001). The higher annualized mortality rate was observed in the diabetes group versus nondiabetes group in patients with ischemia (6.8 vs. 5.2%) or without ischemia (4.3 vs. 3.2%; all P < 0.001). Kaplan-Meier curves for mortality according to diabetes and revascularization status in each ischemic category are shown in Fig. 1. In patients with no ischemia (SDS <5%), early revascularization was associated with increased mortality risk in both diabetes and nondiabetes groups. In patients with mild ischemia (SDS, 5–9.9%), early revascularization was associated with lower mortality in the diabetes group. There was no significant difference between early revascularization and medical treatment in the nondiabetes group with mild ischemia. In patients with moderate or severe ischemia (SDS ≥10%), early revascularization was associated with lower mortality in both diabetes and nondiabetes groups.

Table 2 reports the results of the univariable and multivariable analyses for the survival benefit from early revascularization among the propensity-matched cohort. In multivariable analysis, early revascularization was associated with a survival benefit for the patients with severe ischemia (SDS ≥15%) in both diabetes and nondiabetes groups. In the patients with moderate ischemia (SDS, 10–14.9%), early revascularization was associated with survival benefit in the diabetes group but not in the nondiabetes group. In patients with no or mild ischemia (SDS <10%), there was no survival benefit from early revascularization for both diabetes and nondiabetes groups. Figure 2 shows the association between early revascularization and ACM at each level of percent ischemic myocardium. There was a significant association with survival benefit from early revascularization in patients with >8.6% ischemic myocardium for the diabetes group (Fig. 2A) and those with >12.1% ischemic myocardium for the nondiabetes group (Fig. 2B).

Significant interactions were observed in the prediction of mortality between early revascularization and myocardial ischemia (interaction-adjusted HR 0.97, 95% CI 0.96–0.98, P < 0.001) and early revascularization and diabetes (interaction-adjusted HR 0.91, 95% CI 0.82–0.998, P = 0.046) in the overall population before propensity matching (n = 41,982). The three-way interaction of early revascularization, percent myocardial ischemia, and diabetes was not significant (P = 0.252) in the overall population. In patients without diabetes (n = 33,654), there was significant interaction in early revascularization and percent myocardial ischemia (interaction adjusted HR 0.97, 95% CI 0.95–0.98, P < 0.001). In patients with diabetes (n = 8,328), significant interaction was also observed between early revascularization and percent myocardial ischemia (interaction-adjusted HR 0.98, 95% CI 0.97–0.997, P = 0.020) (Supplementary Table 2).

In a total of 8,328 patients with diabetes before propensity matching, 2,832 patients (35.2%) were treated with insulin at baseline. Supplementary Fig. 2 shows Kaplan-Meier curves for mortality according to insulin use in patients with diabetes. During a median follow-up time of 9.8 years (IQR 6.2–12.6), patients with insulin treatment had a higher annualized mortality rate than those without insulin treatment (4.1% for non–insulin-treated diabetes vs. 6.2% for insulin-treated diabetes, P < 0.001). There was no interaction in the prediction of mortality between insulin use and percent myocardial ischemia (P = 0.335) and early revascularization (P = 0.405) (Supplementary Table 2).

A total of 658 patients (7.9%) underwent early revascularization among 8,328 patients with diabetes. Of those, 484 patients (73.6%) underwent PCI, and 174 patients (26.4%) underwent CABG. Patients who underwent CABG had greater myocardial ischemia and lower LVEF (Supplementary Table 3). By multivariable Cox regression analysis, CABG strategy was not associated with an increased risk of ACM compared with PCI strategy (adjusted HR 0.95, 95% CI 0.73–1.25, P = 0.733). Supplementary Fig. 2 shows Kaplan-Meier curves for mortality according to the type of revascularization in patients with diabetes. During a median follow-up time of 10.7 years (IQR 7.4–13.3), there was no significant difference in the annualized mortality rate between revascularization strategies (5.6% for PCI vs. 5.8% for CABG, P = 0.728). In addition, there was no interaction in the prediction of mortality between revascularization strategies and percent myocardial ischemia (P = 0.781) and insulin use (P = 0.346) (Supplementary Table 2). There was no interaction between early revascularization and the two time periods, 1998–2007 and 2008–2017, in the overall population (interaction HR 1.02, 95% CI 0.82–1.27, interaction P value = 0.878).

Patients with severely impaired LVEF (<35%, 3.8% of the total population) had higher a prevalence of moderate to severe ischemia (325 of 1,580 [20.6%]) and a higher burden of CAD risk factors compared with those without impaired LVEF (Supplementary Table 4). After excluding these patients, significant interactions remained between revascularization and myocardial ischemia in patients both with and without diabetes (interaction P value = 0.017 for patients with diabetes and interaction P value <0.001 for patients without diabetes). In patients with severe ischemia without impaired LVEF, revascularization was significantly associated with a reduced risk of ACM in patients both with and without diabetes (Supplementary Table 5). The threshold for receiving survival benefit was 10.2% for patients with diabetes and 16.2% for patients without diabetes (Supplementary Fig. 3).

In the current study, we explored the differences in long-term survival benefits from early revascularization between patients with and without diabetes. The principal findings of the current study are as follows: 1) significant interactions existed between early revascularization and percent myocardial ischemia regarding mortality benefit in patients both with and without diabetes; 2) the threshold of ischemic burden to receive survival benefit from early revascularization was lower in patients with diabetes compared with patients without diabetes; 3) while insulin-treated diabetes had higher mortality risk compared with non–insulin-treated diabetes, there was no interaction between early revascularization and insulin use for mortality outcome; 4) in patients with diabetes who underwent early revascularization, there was no significant mortality difference between PCI versus CABG revascularization strategies.

Several prior studies have investigated whether early revascularization has a prognostic benefit for the management of CAD across the range of ischemia by SPECT-MPI. Hachamovitch et al. (6,7) showed that patients with >10–12.5% myocardial ischemia might benefit from early revascularization to decrease the risk of death. A recent study reported a similar mortality benefit between PCI and CABG in patients with severe ischemia (18). Similar findings regarding the relationship of revascularization and survival benefit have been found in previous studies, including a large multicenter, multinational registry (8,9).

In the current study, we identified a significant interaction for ACM between diabetes and early revascularization (interaction HR 0.91, interaction P value = 0.046). Consistent with this, the threshold of percent myocardial ischemia related to survival benefit from early revascularization was lower in patients with diabetes compared with patients without diabetes (>8.6% vs. >12.1%, respectively). Multiple studies using MPI have shown that the risk of adverse cardiac events is higher in patients with diabetes than in those without (11,19). A recent study showed that the rate of ACM, myocardial infarction, unstable angina, and late revascularization (>3 months after image acquisition) in patients with >10% ischemic myocardium was more than doubled in patients with diabetes compared with those without diabetes (12). Our findings suggest that the ischemic threshold for a survival benefit from revascularization is lower in patients with diabetes. In terms of categorical ischemic burden, patients with moderate or greater myocardial ischemia (≥10%) and diabetes and those with severe myocardial ischemia (≥15%) and without diabetes may benefit from revascularization.

The findings of this study are discordant with those of the randomized ISCHEMIA (International Study of Comparative Health Effectiveness with Medical and Invasive Approaches) trial, in which there was no benefit from a strategy of early catheterization with planned revascularization in either the overall population or in patients with diabetes who had moderate to severe ischemia (20). The reasons for the discordance between the current study and ISCHEMIA are not clear, but may be partly related to the differences that are inherent in randomized clinical trials versus observational registries (21,22). In ISCHEMIA, a referral bias may have been operative in which the patients with the highest perceived clinical risk may have been less likely to be recruited or may not have participated in the trial. Further, ISCHEMIA excluded patients who had low LVEF (<35%), left main coronary artery stenosis (>50%), or unacceptable angina (23), but who were included in this study. Taking into account the exclusion of patients with low LVEF (<35%) in the ISCHEMIA trial, we repeated the analysis in patients with LVEF ≥35%. Within this cohort, significant interaction between revascularization and myocardial ischemia was observed in patients with and without diabetes who had severe ischemia, similar to our recent findings in patients with normal LVEF (≥45%) (24). In ISCHEMIA, optimal medical management was part of the trial design, was monitored monthly, and was largely achieved by objective measurements (e.g., 100% use of antithrombotic agents and 95% use of statin) (23). It is unlikely that the high level of optimal medical therapy observed in ISCHEMIA was achieved in this observational study. Further, treatments have improved over time, such as sodium–glucose cotransporter 2 inhibitors or glucagon-like peptide 1 receptor agonists, which were not available during the time frame of this study but were available for use in the later parts of the ISCHEMIA trial (3). Additionally, since the current study was based on an observational cohort, all patients in the early revascularization group underwent revascularization. On the other hand, a high cross-over rate was observed in ISCHEMIA with an intention-to-treat analysis (revascularization was performed in 79% of patients in the invasive-strategy group and in 21% of patients in the conservative-strategy group) (14).

Insulin treatment was suggested to be associated with platelet dysfunction (25) and has a risk of hypoglycemia that induces secretion of adrenaline and increases the risk of adverse cardiovascular events (26,27). In previous studies, insulin-treated diabetes had a higher mortality rate and adverse cardiac events compared with non–insulin-treated diabetes (20,28). In line with the earlier observations, insulin-treated diabetes in the current study was associated with higher mortality risk compared with those without insulin use. However, we found that no significant interaction existed between early revascularization and insulin use (interaction P value = 0.405) for mortality risk.

A pooled analysis of patients with diabetes from three major trials found a significant reduction of the composite rate of death, myocardial infarction, or stroke in patients who underwent CABG compared with those with PCI or medical treatment alone during a median follow-up of 4.5 years (29). The reduction was mainly driven by myocardial infarction, and the mortality rate was comparable between CABG and PCI (29). In another pooled analysis from 11 randomized trials including 11,518 patients, 976 patients died in a mean follow-up of 3.8 years, and CABG had a lower mortality rate than PCI, especially for patients with multivessel disease or diabetes (30). In the current study, there was no significant difference in mortality risk between the PCI versus CABG revascularization strategies in patients with diabetes during a median follow-up time of 10.7 years. In addition, there were no significant interactions between the revascularization strategy and percent myocardial ischemia or insulin use. Although the revascularization strategy was not randomized in the current study, our findings may suggest no significant difference in long-term mortality if the revascularization strategy is appropriately determined.

The current study has several limitations. Since ACM was the end point, it is possible that noncardiovascular causes of death were more prevalent in the medically managed group. However, we are not able to ascertain cardiovascular mortality in this large, retrospective study. In the current study, <10% of patients in this study had moderate or severe ischemia; however, this low prevalence of myocardial ischemia is consistent with recent findings showing a decrease in ischemia over time (31,32). Although all patients in the early revascularization group underwent revascularization, it was unknown how many patients underwent revascularization >90 days of the SPECT study.

The intensity of medical management and lifestyle changes after testing was unknown. Recommended practices for optimal medical therapy for patients with diabetes did not include blood pressure and statin targets as strong as those more recently adopted. Pharmacological therapies for patients with diabetes and cardiovascular disease at the time of testing were not as effective as those more recently used as noted and were not available in the current study (3).

Finally, this is an observational study, and there are unmeasured confounders that could not be taken into account. However, these confounding factors are likely to be distributed evenly between patients with and without diabetes; therefore, our finding of different thresholds for survival benefit from revascularization between patients with and without diabetes would not be explained by the potential unmeasured confounders.

In conclusion, the benefit of early revascularization was more pronounced with an increased burden of myocardial ischemia in patients both with and without diabetes in this large, long-term observational study. The threshold of ischemic burden to receive survival benefit from early revascularization was lower in patients with diabetes compared with patients without diabetes. In patients with diabetes, the survival benefit from early revascularization was independent of insulin use and revascularization strategy.

Funding. This work was supported by a grant from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. K.K. receives funding support from the Society of Nuclear Medicine and Molecular Imaging Wagner-Torizuka Fellowship Grant and the Nihon University School of Medicine Alumni Association Research Grant.

Duality of Interest. D.B. and P.S. participate in software royalties for QPS software at Cedars-Sinai Medical Center. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.K., D.H., and R.J.H.M. drafted the manuscript. K.K., D.H., R.J.H.M., A.R., H.G., D.D., S.W.H., J.D.F., L.T., P.J.S., and D.S.B. contributed to data collection. K.K., D.H., R.J.H.M., A.R., H.G., D.D., S.W.H., J.D.F., L.T., P.J.S., and D.S.B. approved the final article. K.K., D.H., and H.G. analyzed and interpreted the data. R.J.H.M., A.R., H.G., S.W.H., J.D.F., and D.S.B. contributed to critically revising the manuscript. D.S.B. is the guarantor of this work and, as such, has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.