1 INTRODUCTION

Significant left main (LM) coronary artery disease is a Class 1 indication for revascularisation, given its prognostic importance and the large myocardial territory it supplies.1 While there have been recent debates about the optimal revascularisation strategy for LM disease, both percutaneous coronary intervention (PCI) and coronary artery bypass graft (CABG) surgery are viewed as acceptable revascularisation strategies, based on patient comorbidities and overall anatomical complexity.2-8

The coronavirus pandemic has led to a significant strain on healthcare services, with previous reports demonstrating a substantial reduction in cardiac procedural activity since the start of the pandemic.9-14 This is particularly relevant for procedures requiring admission to an intensive care unit (ICU) and prolonged hospitalization (such as CABG surgery), at a time when such resources were prioritized for critically unwell patients with COVID-19. Moreover, patients with cardiac conditions are at a higher risk of COVID-19 related mortality than the general population and, in the United Kingdom, were advised to shield meaning that significant numbers of elective cardiac procedures were canceled.15, 16

Little is known about the impact these national policy level changes and healthcare service restructures may have had on the revascularisation of significant LM disease. Furthermore, it is unclear whether there has been in change in the choice of revascularisation strategy in patients with significant LM disease during the pandemic, and if such changes in strategy were associated with higher post-revascularisation mortality in this group of patients.

The United Kingdom is unique in that administrative and clinical data are systematically and routinely collected for all National Health Service (NHS) admissions, including for CABG surgery, PCI procedures, and mortality. The present study sought to examine impact of the COVID-19 pandemic on LM procedural activity, choice of revascularisation strategy and associated postprocedural mortality for patients with significant LM disease, in an unselected and contemporary nationwide cohort in England between January 2017 and August 2020.

2 METHODS 2.1 Data source, study design and population

Data for all adults (aged ≥18 years) with significant LM disease (≥50% stenosis) who underwent coronary revascularisation through PCI or CABG surgery between January 1, 2017 and August 19, 2020 in England were drawn from the British Cardiovascular Intervention Society (BCIS) national audit of PCI (NAPCI) and the British Society of Cardiothoracic surgery (SCTS) national adult cardiac surgery (NACSA) registries, both of which are managed by the National Institute for Cardiovascular Outcomes Research (NICOR), and commissioned by the Healthcare Quality Improvement Partnership (HQIP).17-19 Each contain data on clinical and procedural characteristics, as well as in-hospital death for all procedures undertaken in England. Cases were deterministically linked with data of death as recorded in the Civil Registrations of Death dataset using each individual's unique NHS identifier (final follow up October 1, 2020).20

Significant LM disease was defined on the basis of angiographic data available in both the NAPCI and NACSA datasets. We excluded patients with missing data on death (PCI: 633 cases; CABG: 61 cases) and excluded CABG patients who received PCI in the preceding 30 days (n = 222) (flow diagram: Figure S1). Patients were stratified according to revascularisation modality (CABG surgery and PCI) as well as the period during which the procedure was undertaken (pre-COVID: January 1, 2017-February 29, 2020; COVID: March 1, 2020-August 19, 2020). Procedural risk was assessed using the Logistic EuroSCORE and British Cardiovascular Intervention Society (BCIS) 30-day mortality score for CABG and PCI cases, respectively, using coefficients previously described for both scoring systems.21, 22

2.2 Outcomes

The primary outcomes were (a) the receipt of CABG or PCI for significant LM disease and (b) in-hospital and 30-day mortality from the date of the procedure.

2.3 Statistical analysis

We examined rates and patient and procedural characteristics of patients undergoing PCI and CABG, as well as in-hospital and 30-day mortality before and during the COVID pandemic (pre-COVID: January 1, 2017-February 29, 2020; COVID: March 1, 2020-August 19, 2020). Similar comparisons were performed for each calendar year. The number of PCI and CABG surgery procedures for each of the months January–July were compared between the 2017–2019 average and 2020 to estimate the percentage change (Δ) in procedural activity and, in turn, the projected deficit in cases in 2020 as a result of the pandemic. Data between August 1, 2020 and August 19, 2020 was not used to calculate procedural activity as there may have been a lag between data submission by participating hospitals and availability in the NHS Digital database. Continuous variables were summarized using median and interquartile range (IQR) and compared using the Kruskal-Wallis test. Categorical variables were summarized as percentages and analyzed using the Chi squared (X2) test. Multiple imputation with chained equations was performed for variables with missing data prior to model fitting, with a total of 10 imputations and model estimates combined using Rubin's rules.23 The frequency of missing data prior to imputation is provided in Table S1. Multivariable logistic regression models were fitted to examine (a) the likelihood of receipt of CABG compared with PCI and (b) associated in-hospital and 30-day mortality in the COVID-19 period (with pre-COVID as reference), adjusting for the variables summarized in Appendix A. We report the association of COVID-19 period with the outcomes using odds ratios (OR) with corresponding 95% confidence intervals (CI). In order to estimate the adjusted probability of death in the CABG and PCI groups in both the pre-COVID and COVID time periods, multivariable logistic regression models were performed with an interaction term between revascularisation strategy (PCI vs. CABG) and time period (pre-COVID vs. COVID), adjusting for the variables in Appendix A, followed by the margins command to generate adjusted probabilities of mortality in each group. Statistical analyses were performed using Stata 16 MP (College Station, TX).

2.4 Ethics statement

The UK Secretary of State for Health and Social Care has issued a time limited Notice under Regulation 34 of the NHS (Control of Patient Information Regulations) 2002 (COPI) to share confidential patient information. The study complies with the Declaration of Helsinki. This work was part of a work stream endorsed by the Scientific Advisory Group for Emergencies (SAGE), the body responsible for ensuring timely and coordinated scientific advice is made available to UK government decision makers. SAGE supports UK cross-government decisions in the Cabinet Office Briefing Room (COBR)) and by NHS England, which oversees commissioning decisions in the NHS, and NHS Improvement, which is responsible for overseeing quality of care in NHS hospitals.

3 RESULTS

A total of 22,235 cases of revascularisation involving significant LM disease were recorded between January 1, 2017 and August 19, 2020, of which 62.9% (n = 13,994) were PCI and 37.1% (n = 8,241) were CABG.

3.1 Procedural activity

Overall, there was a decline in procedural activity for LM revascularisation during the pandemic, with an estimated total deficit of 1,354 cases between March 1, 2020 and July 31, 2020 compared with previous years' (2017–2019) averages (Δ −48.8%), with the greatest decline observed in July 2020 (Δ-60.6%). (Table S2) This was evident in both PCI and CABG groups, with a significant decline in the 7-day rolling average number of procedures beginning at the start of UK-wide lockdown (March 23, 2020). (Figure 1).

Moving average of procedural activity for PCI and CABG procedures over the study period. UK-wide lockdown was introduced on March 23, 2020; restrictions were lifted on July 4, 2020. CABG, coronary artery bypass graft surgery; PCI, percutaneous coronary intervention [Color figure can be viewed at wileyonlinelibrary.com] 3.2 Choice of revascularisation strategy

Pre-COVID, PCI accounted for 61.9% of all left main revascularisation procedures, whereas during the COVID period, this has risen to 77.6%. (Table 1) Compared with 2017, the odds of receipt of CABG over PCI increased in 2018 and 2019 (OR 1.85 95% CI 1.70, 2.03 and 1.46 95% CI 1.33, 1.61, respectively) but were similar in the start of 2020 before the COVID pandemic (January–February 2020 OR: 1.13 95% CI 0.95, 1.34, p = .176). However, there was a decline in odds of receipt of CABG (vs. PCI) in the COVID period compared with the pre-COVID period (OR 0.46 95% CI 0.39, 0.53). (Figure 2, Table S3).

TABLE 1. Characteristics of patients undergoing PCI and CABG in the pre-COVID and COVID periods Pre-COVID (n = 20,774) COVID (n = 1,461) PCI (n = 12,860) CABG (n = 7,914) p-value PCI (n = 1,134) CABG (n = 327) p-value Patient characteristics Age, median (IQR) 72 (63,79) 69 (62,75) <.001 72 (62,79) 69 (61,75) <.001 Age groups (years) Column % Row % Column % Row % <.001 Column % Row % Column % Row % <.001 <50 5.1 66.0 4.3 34.0 5.7 85.5 3.4 14.5 51–60 14.5 57.9 17.1 42.1 15.3 73.1 19.6 26.9 61–70 25.5 55.2 33.7 44.8 24.3 71.2 33.9 28.8 71–80 33.4 59.2 37.5 40.8 35.0 75.9 38.5 24.1 >80 21.5 82.3 7.5 17.7 19.7 93.7 4.6 6.3 Males, % 74.3 83.1 <.001 76.1 87.5 <.001 Elective, % 39.3 37.6 <.001 33.1 30.3 .342 ACS, % 61.2 34.1 <.001 66.0 38.2 <.001 Previous/current smoker, % 60.4 63.5 <.001 55.9 64.8 .004 Diabetes, % 30.4 31.7 .041 29.7 28.1 .580 Previous MI, % 37.9 48.8 <.001 33.4 54.1 <.001 Previous PCI, % 34.1 14.1 <.001 31.5 14.1 <.001 Previous CABG, % 22.3 — — 15.9 — — Previous cardiac surgery (including CABG), % — 1.8 — — 2.1 — Previous CVA, % 7.0 8.0 .013 5.6 7.3 .256 PVD, % 9.0 11.8 <.001 8.1 11.3 .072 Hypertension, % 65.7 74.6 <.001 64.4 76.5 <.001 Left ventricular ejection fraction, % <.001 .300 Good (>50%) 56.0 68.2 58.9 63.7 Moderate (30–50%) 35.0 26.1 33.8 29.8 Poor (<30%) 9.0 5.7 7.3 6.5 Hypercholesterolemia, % 58.2 — — 50.6 — — Cardiac transplant, % 0.1 — — 0.1 — — Creatinine clearance (ml/min), median (IQR)a 63 (38,89) 81 (63,103) <.001 64 (34,90) 83 (65,105) <.001 PCI risk scoreb/logistic EuroSCORE, median % (IQR) 1.0 (0.5, 2.6) 3.3 (1.9, 6.4) <.001 1.1 (0.5, 2.6) 3.4 (1.9, 6.5) <.001 Procedural characteristics

Cardiogenic shock

(pre-procedure), %

8.5 1.0 <.001 7.1 1.8 <.001 Circulatory support (IABP or inotropes), % 4.4 8.0 <.001 2.8 8.6 <.001 Radial access, % 74.5 — — 81.2 — — Femoral access, % 30.3 — — 21.5 — — Intravascular ultrasound, % 42.4 — — 52.6 — — Drug eluting stents, % 90.1 — — 92.0 — — Chronic total occlusion, % 6.4 — — 5.5 — — On-pump CABG, % — 85.5 — — 93.3 — Concomitant valve surgery, % — 11.5 — — 7.0 — No of vessels attempted, % <.001 <.001 1 24.0 3.1 19.4 1.2 2 46.9 24.7 43.6 20.2 3 26.5 49.9 34.5 55.4 4+ 2.6 22.3 2.6 23.2 Abbreviations: ACS: acute coronary syndrome; CABG: coronary artery bypass surgery; CTO: chronic total occlusion; CVA: cerebrovascular accident including stroke and transient ischaemic attack; IABP: Intra-aortic balloon pump; IQR: interquartile range; MI: myocardial infarction; PCI: percutaneous coronary intervention.

Adjusted odds of receipt of CABG (reference is PCI) at different time points over the study period. Reference is year 2017; ptrend < .001

3.3 Patient and procedural characteristics

Overall, patients undergoing PCI were older, more frequently male, more likely to be admitted electively and to receive PCI for an ACS indication compared with those undergoing CABG. (Table 1) Patients undergoing PCI also had a higher prevalence of left ventricular (LV) impairment (moderate or severely impaired) compared with CABG. However, all these differences were similar in the pre-COVID and COVID time periods (Table 1), as well as in individual year subgroups of the pre-COVID period (Table S4). Patients undergoing CABG had a higher prevalence of smoking history, diabetes (only in the pre-COVID group), previous history of myocardial infarction (MI) and cerebrovascular accidents (CVA, including transient ischaemic attack and stroke), hypertension and peripheral vascular disease (PVD) compared with those undergoing PCI. Furthermore, patients undergoing CABG were at a higher risk of all-cause post procedure mortality compared with PCI, as evidenced by the Logistic EuroSCORE and PCI risk score. This pattern of differences, however, was consistent in the pre-COVID and COVID periods.

In terms of procedural characteristics, patients in the PCI group were more likely to be in cardiogenic shock at presentation in the pre-COVID and COVID time periods, whereas the CABG group were more likely to require circulatory support in the form of intra-aortic balloon pump or inotropes at any point during the admission.

Compared with the pre-COVID time period, patients undergoing PCI in the COVID time period were more likely to undergo a procedure via radial access (81.2% vs. 74.5%), using intravascular ultrasound (52.6% vs. 42.4%) and with drug eluting stents (DES; 92.0 vs. 90.1%).

Patients undergoing CABG in the COVID period were less likely to undergo concomitant valve surgery (7.0% vs. 11.5%) and more likely to have an on-pump CABG (85.5% vs. 93.3%) than in the pre-COVID period.

3.4 In-hospital and 30-day mortality

Overall, the crude rate of 30-day mortality after LM revascularisation declined over the study period (2017:6.7% to March–August 2020:5.7%). However, after adjustment, there was no difference in the odds of mortality after LM revascularisation over different time points when compared with 2017. (Table 2, Figure 3)

TABLE 2. Crude and adjusted odds of 30-day mortality after left main disease revascularization over the study period Time period % OR [95% CI] p-value 2017 6.7 Reference 2018 6.0 1.05 [0.76. 1.45]a .757 2019 5.9 1.04 [0.76, 1.44]a .792 Jan-Feb 2020 5.1 0.88 [0.46, 1.69]a .711 Mar-Aug 2020 (COVID period) 5.7 1.26 [0.83, 1.91]b .281 Abbreviations: OR: odds ratio; CI: confidence interval; Adjusted for the following: age (years), sex, smoking status, diabetes, indication for intervention (ACS vs. CCS), previous MI, PCI and CVA, LV function category (good, moderate or poor); cardiogenic shock pre-procedure, intra-aortic balloon pump; hypertension; peripheral vascular disease; creatinine clearance (ml/min- Cockcroft and Gault formula).

Adjusted odds of 30-day mortality after left main disease revascularization over the study period. Reference is year 2017

The unadjusted rates of in-hospital and 30-day mortality were higher in PCI than CABG in the pre-COVID period (6.5% vs. 2.2% and 8.4% vs. 2.5%, respectively) (Table 3), a pattern that was consistent in individual years of the pre-COVID period (Table S5). However, only 30-day mortality was higher in PCI than CABG in the COVID period (6.5% vs. 3.0%) with no difference in in-hospital mortality between PCI and CABG (4.4% vs. 2.1%, p = .062). (Table 3).

TABLE 3. Unadjusted and adjusted mortality rates in PCI and CABG groups during the pre-COVID and COVID time periods Pre-COVID COVID PCI (n = 12,860) CABG (n = 7,914) p-value PCI (n = 1,134) CABG (n = 327) p-value Unadjusted In-hospital, % 6.5 2.2 <.001 4.4 2.1 .062 Post discharge to 30 days, % 1.9 0.3 <.001 2.1 0.9 .167 30-day, % 8.4 2.5 <.001 6.5 3.0 .020 Adjusted In-hospital, % 5.1 (4.7,5.4) 4.4 (3.8. 5.0) .095 3.9 (3.0, 4.8) 3.7 (1.5, 6.0) .900 Post discharge to 30 days, % 1.5 (1.3, 1.8) 0.5 (0.3, 0.7) <.001 1.8 (1.1, 2.5) 1.4 (−1.4, 2.9) .635 30-day, % 6.6 (6.2, 7.0) 4.9 (4.2, 5.5) <.001 5.7 (4.6, 6.8) 5.1 (2.6, 7.6) .663 Abbreviations: Adjusted for the following: age (years), sex, smoking status, diabetes, indication for intervention (ACS vs. CCS), previous MI, PCI and CVA, LV function category (good, moderate or poor); cardiogenic shock pre-procedur