European Heart Journal (2023) 44, 2833–2842 CLINICAL RESEARCH

https://doi.org/10.1093/eurheartj/ehad369 Ischaemic heart disease

PCI or CABG for left main coronary artery

disease: the SWEDEHEART registry
Jonas Persson 1*, Jacinth Yan 2, Oskar Angerås 3, Dimitrios Venetsanos 4
,
Anders Jeppsson 5,6, Iwar Sjögren 7, Rikard Linder 1, David Erlinge 8,

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

Torbjörn Ivert 9, and Elmir Omerovic 3
1
Division of Cardiovascular Medicine, Department of Clinical Sciences, Karolinska Institutet, Danderyd University Hospital, Entrevägen 2, 182 88 Stockholm, Sweden; 2Division of
Biostatistics, Institute of Environmental Medicine, Karolinska Institutet, Nobels väg 13, 17177 Stockholm, Sweden; 3Department of Cardiology, Sahlgrenska University Hospital, Blå stråket 5,
413 45 Gothenburg, Sweden; 4Division of Cardiology, Department of Medicine, Karolinska Institutet Solna and Karolinska University Hospital, Eugeniavägen 3, 171 76 Stockholm, Sweden;
5
Department of Cardiothoracic Surgery, Sahlgrenska University Hospital, Blå stråket 5, 413 46 Gothenburg, Sweden; 6Department of Molecular and Clinical Medicine, Institute of Medicine,
Sahlgrenska Academy, University of Gothenburg, Blå stråket 5B, 413 45 Gothenburg, Sweden; 7Department of Cardiology, Falu Hospital, Lasarettsvägen 10, 791 82 Falun, Sweden; 8Clinical
Sciences, Lund University, Sölvegatan 19, BMC I12, 221 84 Lund, Sweden; and 9Department of Cardiothoracic Surgery, Karolinska University Hospital and Department of Molecular
Medicine and Surgery, Karolinska Institutet, Eugeniavägen 3, 171 76 Stockholm, Sweden

Received 20 September 2022; revised 21 March 2023; accepted 25 April 2023; online publish-ahead-of-print 8 June 2023

See the editorial comment for this article ‘How to treat left main coronary artery disease: the complementary lessons from trials and regis­
tries’, by M. McEntegart and T. Gori, https://doi.org10.1093/eurheartj/ehad287.

Abstract

Aims An observational nationwide all-comers prospective register study to analyse outcomes after coronary artery bypass graft­
ing (CABG) or percutaneous coronary intervention (PCI) in unprotected left main coronary artery (LMCA) disease.
.............................................................................................................................................................................................
Methods All patients undergoing coronary angiography in Sweden are registered in the Swedish Web-system for Enhancement and
and results Development of Evidence-based care in Heart disease Evaluated According to Recommended Therapies registry. Between
01/01/2005 and 12/31/2015, 11 137 patients with LMCA disease underwent CABG (n = 9364) or PCI (n = 1773). Patients
with previous CABG, ST-elevation myocardial infarction (MI) or cardiac shock were excluded. Death, MI, stroke, and new
revascularization during follow-up until 12/31/2015 were identified using national registries. Cox regression with inverse
probability weighting (IPW) and an instrumental variable (IV), administrative region, were used. Patients undergoing PCI
were older, had higher prevalence of comorbidity but lower prevalence of three-vessel disease. PCI patients had higher mor­
tality than CABG patients after adjustments for known cofounders with IPW analysis (hazard ratio [HR] 2.0 [95% confi­
dence interval (CI) 1.5–2.7]) and known/unknown confounders with IV analysis (HR 1.5 [95% CI 1.1–2.0]). PCI was
associated with higher incidence of major adverse cardiovascular and cerebrovascular events (MACCE; death, MI, stroke,
or new revascularization) than CABG, with IV analysis (HR 2.8 [95% CI 1.8–4.5]). There was a quantitative interaction
for diabetic status regarding mortality (P = 0.014) translating into 3.6 years (95% CI 3.3–4.0) longer median survival time
favouring CABG in patients with diabetes.
.............................................................................................................................................................................................
Conclusion In this non-randomized study, CABG in patients with LMCA disease was associated with lower mortality and fewer MACCE
compared to PCI after multivariable adjustment for known and unknown confounders.

* Corresponding author. Tel: +46 700 891412, Email: jonas.persson@regionstockholm.se

© The Author(s) 2023. Published by Oxford University Press on behalf of the European Society of Cardiology.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/), which permits
non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com
2834 Persson et al.

Structured Graphical Abstract

Key Question
How do the outcomes of coronary artery bypass grafting (CABG) compare to percutaneous coronary intervention (PCI) in an
all-comer population with left main coronary artery disease? How can this guide the selection of revascularization strategy to improve
patient outcomes?

Key Finding
In this observational study, CABG was associated with lower mortality and fewer major adverse cardiovascular and cerebrovascular
events in patients with left main coronary artery disease compared to PCI after adjustment for confounders during a median follow-up
of 4.7 years.

Take Home Message

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

This observational study may inform the choice between CABG and PCI for left main coronary artery disease, in particular for patients
who are not represented in randomized clinical trials.

patients with left main CAD in Sweden 2005–2015

Hazard ratio
0 1 2 3 4 5 6 7
Mortality
Crude
IPW adj.
IV adj.
MACCE
Crude
IPW adj.
IV adj.

CABG, n=9364 PCI, n=1773 PCI better CABG better

Median survival CABG vs PCI Difference in years (95%CI; P-value)

CABG -2.58 (-2.81 to -2.35; <0.001)

PCI
0 2 4 6 8 10 12 Years

CABG = coronary artery bypass grafting, CAD = coronary artery disease, CI = confidence interval, PCI = percutaneous coronary intervention,
IPW adj. = inverse probability weighting adjustment, IV adj. = instrumental variable adjusted; MACCE = major adverse cardiovascular and cerebro­
vascular events.
.............................................................................................................................................................................................
Keywords Left main coronary artery disease • Percutaneous coronary intervention • Coronary artery bypass grafting • Mortality •
Instrumental variable analysis • Cox regression

Introduction follow-up when comparing CABG with PCI.4 There is a significant and
beneficial interaction between time and PCI with regard to MI and peri-
Patients with untreated significant left main coronary artery (LMCA) interventional stroke, however this is offset by risk of spontaneous MI
disease have a poor prognosis and revascularization with coronary ar­ and new revascularization during follow-up.4
tery bypass grafting (CABG) or percutaneous coronary intervention The recommendations for choosing mode of revascularization of un­
(PCI) is indicated.1–3 The 2018 ESC/EACTS Guidelines on myocardial protected LMCA disease is based on coronary anatomical complexity
revascularization, which relies on evidence from randomized clinical assessed by the Synergy between PCI with Taxus and Cardiac
trials (RCTs) and meta-analyses, suggest equivalent results for the com­ Surgery (SYNTAX) score,5 individual cardiac and extracardiac charac­
posite of death, myocardial infarction (MI), and stroke up to 5 years of teristics, and patient preference. For patients with a low SYNTAX
PCI or CABG for left main coronary artery disease 2835

score (≤22), PCI and CABG have the same class and level of recom­ generation drug-eluting stent (DES) (Figure 1). The study was reviewed
mendation (I A), but for patients with intermediate SYNTAX score and approved by the local ethics committee in Stockholm, Sweden
(23–32) the recommendation for CABG is I A and for PCI (2015/1258–31).
IIa A. Only CABG is recommended (I A) for patients with very complex
coronary anatomy (SYNTAX score ≥33) and PCI is not recommended Outcomes
(III B) for such patients.4
All subjects were followed through the Swedish Population Register for
The recommendations are, to a large extent, based on RCTs that death, the Swedish National Patient Registry (NPR) for MI and stroke,
have been conducted in highly selected cohorts. Consequently, the and the SWEDEHEART registry for MI and repeat revascularization after
guidelines and the RCTs may not be fully applicable to the diversity the index angiography until the onset of outcome or until 31 December
of real-life patients and circumstances associated with revascularization 2015. New revascularization was defined as new PCI or new CABG during
of LMCA disease in clinical practice. Thus, we analysed the outcome of follow-up. Major adverse cardiovascular and cerebrovascular events
all subjects with LMCA disease undergoing revascularization with PCI (MACCE) were defined as death, MI, stroke, or new revascularization
or CABG in Sweden between 1 January 2005 and 31 December 2015. (whichever occurred first) within the follow-up period.

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

Co-variates
Methods Variables associated with mortality available in the SWEDEHEART and the
NPR registers, as well as administrative region and year of treatment were
Study population used for adjustments. Multivariable analysis included the year of treatment,
All patients undergoing coronary angiography, PCI and CABG in Sweden administrative region, body mass index, diabetes mellitus (diabetes mellitus
are registered in the Swedish Web-system for Enhancement and diagnosis known to the patient, independently of treatment), insulin treat­
Development of Evidence-based care in Heart disease Evaluated ment (yes/no), estimated creatinine clearance, indication for revasculariza­
According to Recommended Therapies (SWEDEHEART) registry which in­ tion [non-ST-elevation MI (NSTEMI)/unstable angina, stable angina],
cludes the Swedish Coronary Angiography and Angioplasty Registry and the number of diseased coronary arteries (left main, left main + one-vessel,
Swedish Cardiac Surgery Registry. All subjects that underwent coronary left main + two-vessel, left main + three vessel disease), smoking status
angiography in 28 PCI centres in Sweden during the 11 years from 1 (no, former smoker, current smoker), age at diagnosis, gender, previous
January 2005, to 31 December 2015 were screened. We included 11 137 MI, previous PCI, chronic obstructive pulmonary disease (COPD), periph­
patients who had undergone revascularization with either PCI or CABG eral artery disease, history of cancer, and dialysis. Hypertension was defined
due to LMCA disease within 3 months after the index angiography. as treatment with anti-hypertensive drug and hyperlipidaemia as treatment
Exclusion criteria were previous CABG (excluding protected LMCA dis­ with statins. COPD was defined as medication with bronchodilators or
ease), ST-elevation MI and/or cardiogenic shock, subjects deemed and regis­ steroids for lung disease. Administrative region was categorized
tered not being eligible for CABG, patients with data error or age <18 (Stockholm, Uppsala-Örebro, South-east, South, West, North). The risk
years, no revascularization within 90 days after coronary angiography, of surgery was assessed using the European System for Cardiac
and PCI in left main with another device than first-, second-, or third- Operative Risk Evaluation Score (EuroSCORE) II6 in CABG patients only.

Figure 1 Flowchart of patient selection. *Bare metal stent, bioresorbable vascular scaffold or self-expandable stent.
2836 Persson et al.

Creatinine clearance was calculated using the Cockcroft and Gault NSTEMI and/or unstable angina, but chronic coronary syndrome
equation.7 (CCS) was significantly more common in the CABG group (Table 1).
The CABG patients had a median EuroSCORE II risk of 4% (inter­
quartile range 3%–7%), and 73% of the patients had normal left ven­
Statistics tricular function. The left internal thoracic artery (ITA) was used in
Baseline characteristics of patients in the PCI and CABG groups were de­ more than 90% of the cases, whereas both ITA in 2% of the cases.
scribed as frequencies for categorical variables and the mean value with
Three or more distal anastomoses were performed in 77% of the pa­
standard deviations for continuous variables. The chi-square test and
tients (Table 1).
t-test were applied separately to compare the distribution of characteristics
in the two groups at baseline. Due to an imbalance between the groups, we Two-hundred and eighty-five patients (16%) had first generation,
applied the inverse probability weighting (IPW) method to create a 1324 (75%) had second generation and 164 (9%) patients had third-
weighted population after which the distribution of characteristics was simi­ generation DES in the LMCA (see Supplementary data online,
lar in the two groups.8 Propensity scores and standard mean differences Figure S3). One thousand five-hundred fifty-seven patients (88%) had
prior and following IPW adjustments are presented in Supplementary one stent in the left main, 207 patients (11.5%) had two stents, and

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

data online, Figures S1 and S2. nine patients (0.5%) had three stents positioned in the left main.
A standard Cox regression model reporting hazard ratios (HRs) together
with Laplace regression to compare the median survival time9 was used as
the primary analysis to compare the PCI and CABG groups. The models Outcome
were built on weighted population to calculate HR with 95% confidence
The median follow-up for the whole cohort was 4.7 years (interquartile
interval (CI) and median of event free time between PCI and CABG patients
as two comparable groups.
range 2.1–7.6). The crude incidence rates for death, MI, revasculariza­
We performed instrumental variable (IV) analyses10,11 to adjust for un­ tion, and MACCE were higher in the PCI group compared to the
known confounders. For an IV analysis, one must identify a naturally varying CABG group (Figure 2). IPW-adjusted HR (aHR) comparing PCI with
variable in the observed data which predicts the treatment that will be as­ CABG were significant for mortality (aHR 2.0, 95% CI 1.5–2.7), MI
signed to the individual patient. The IV variable must fulfil the following cri­ (aHR 4.0, 95% CI 2.9–5.5), new revascularization (aHR 5.1, 95% CI
teria: (i) it must be associated with the received treatment and (ii) it must 3.8–7.0), and MACCE (aHR 2.5, 95% CI 1.9–3.2). After adjustment
not be associated directly or indirectly with the outcome, except through for unknown confounders in the IV analysis, the differences between
the effect of the treatment itself. We identified the variable administrative groups were still evident for mortality (IV-HR 1.5, 95% CI 1.1–2.0),
region (Stockholm, Uppsala-Örebro, South-east, South, West, North) MI (IV-HR 6.1, 95% CI 1.4–26.3), new revascularization (IV-HR 14.0,
to be associated with the received treatment, PCI or CABG
95% CI 5.8–33.6), and MACCE (IV-HR 2.8, 95% CI 1.8–4.5). There
(see Supplementary data online, Table S1). The variable administrative re­
were no significant differences in HRs for stroke before (crude HR
gion is not expected to be associated with the outcome, except through
the effect of the treatment itself. To test for the strength of the instrument 1.2, 95% CI 0.95–1.2) or after adjustments (aHR 1.2, 95% CI 0.77–
variable, we examined the partial F-test, which predicts treatment as a func­ 1.9 and IV-HR 5.1, 95% CI 0.72–35.8) (Figure 2).
tion of the instrument and covariates. The partial F-test has the null hypoth­ Crude and IPW-adjusted one-minus survival curves for mortality, MI
esis that the coefficient for the effect of the instrument in the first-stage and stroke are presented in Supplementary data online, Figures S4–S9.
regression model is zero.12 An F-statistic >10 indicates that the instrument Thirty-day event rates are presented in Supplementary data online,
is not weak. The F-value for the association of the IV, administrative region, Table S2.
with treatment (PCI or CABG) was 153 and the IV was considered valid. Of the 580 new revascularizations in the CABG group, 26 (4.5%)
The IV might be imperfect, thus the following variables were entered in were done by CABG and 128 (22.0%) included revascularizations of
the IV regression to also adjust for known confounders; age, gender, num­ one or more grafts. There were 235 new revascularizations in the
ber of diseased coronary arteries, hypertension, hyperlipidaemia, year of
PCI group of which 41(17%) included left main revascularization and
treatment, body mass index, diabetes mellitus, insulin treatment, estimated
creatinine clearance, indication for revascularization, smoking status, previ­
18 (7.7%) were done by CABG.
ous MI, previous PCI, COPD, peripheral artery disease, history of cancer, The incidence of hospitalizations for serious non-cardiac events
and dialysis. All the analyses were performed using Stata MP 17.1. The sig­ (Alzheimer/Parkinson, bleeding, infection, chronic pulmonary obstruct­
nificance level of the statistic tests was set to 0.05. ive disease, malignant tumor, peripheral artery disease, renal insuffi­
ciency, any trauma) was higher in the PCI group than in the CABG
group (Supplementary data online, Figure S10).
Results
A total of 11 137 patients were included of which 9364 (84%) patients Outcome in subgroups
had undergone CABG and 1773 (16%) PCI. The proportion of patients IPW- aHRs for mortality (Figure 3) and MACCE (Figure 4) for PCI vs.
treated with PCI increased from 7% in 2005 to 34% in 2015 (see CABG are presented for (i) left main and left main plus one vessel
Supplementary data online, Table S1). There was a large regional vari­ and left main plus two or three vessels, (ii) diabetes and no diabetes,
ation in the choice of revascularization method ranging from 8% PCI (iii) age >70 years and age ≤70 years, (iv) women and men, and (v)
in the South and West regions to 34% in the Uppsala-Örebro regions for patients undergoing angiography at hospital with and without thor­
(see Supplementary data online, Table S1). Subjects undergoing PCI acic surgery. There was a significant quantitative interaction for diabetes
were three years older than CABG patients. Hyperlipidaemia, previous status with a larger benefit for CABG vs. PCI in patients with diabetes
MI, previous PCI, reduced renal function and isolated left main stenosis with regard to mortality compared to non-diabetic patients (Figure 3).
were more common in the PCI than in the CABG group (Table 1). A There was also an interaction for MACCE with regard to number of
history of smoking, diabetes mellitus, peripheral artery disease and diseased vessels showing a larger benefit for CABG vs. PCI in patients
LMCA disease combined with three vessel disease were more common with left main plus two or three vessels disease compared to left main
in the CABG group. Most of the patients were revascularized due to only or left main plus one vessel disease (Figure 4). Patients ≤70 years
PCI or CABG for left main coronary artery disease 2837

Table 1 Baseline characteristics in 11 137 patients with left main coronary artery disease who had percutaneous
coronary intervention (PCI) or coronary artery bypass grafting (CABG) from January 1st 2005 to December 31st 2015

PCI (n = 1773, 16%) CABG (n = 9364, 84%) P-value

......................................................................................................................................................................................
Age (years), mean (SD) 72.8 (10.7) 69.6 (8.8) <0.001
2
BMI (kg/m ), median (IQR) 26.2 (24.0–28.9) 26.6 (24.4–29.3) 0.001
2
Creatinine clearance (mL/min/1.73 m ), median (IQR) 69.9 (52.1–92.3) 79.2 (62.0–99.2) <0.001
Female sex, n (%) 500 (28.2) 1963 (21.0) <0.001
Smoking, n (%) <0.003
Former 639 (37.6) 3649 (42.4)

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

Present 226 (13.3) 1340 (15.6)
Hypertension, n (%) 1124 (64.3) 5648 (63.8) 0.70
Hyperlipidaemia, n (%) 922 (52.9) 5403 (61.2) <0.001
Diabetes, n (%) 0.027
Medical treated 187 (10.6) 1124 (12.6)
Insulin treated 151 (8.6) 833 (9.4)
Previous myocardial infarction, n (%) 332 (19.1) 1428 (16.3) 0.004
Previous PCI, n (%) 133 (7.5) 491 (5.2) <0.001
COPD, n (%) 91 (5.1) 301 (3.2) <0.001
Peripheral vascular disease, n (%) 7 (0.4) 62 (0.7) 0.19
History of cancer, n (%) 147 (8.3) 413 (4.4) <0.001
Indication, n (%) <0.001
CCS 533 (30.1) 3574 (38.2)
NSTEMI/Unstable angina 1240 (69.9) 5790 (61.8)
Dialysis, n (%) 17 (1.0) 65 (0.7) 0.22
Number of diseased vessels, n (%) <0.001
Isolated left main stenosis 183 (10.3) 414 (4.4)
Left main + one vessel 500 (28.2) 1043 (11.1)
Left main + two vessel 589 (33.2) 2498 (26.7)
Left main + three vessel 501 (28.3) 5409 (57.8)
Number of stents, n (%) NA
1 565 (31.9) NA
2 538 (30.3) NA
3 300 (16.9) NA
4 205 (11.6) NA
5 83 (4.7) NA
≥6 82 (4.6) NA
Distal anastomoses, n (%) NA
1 NA 83 (1.0)
2 NA 1439 (18.0)
3 NA 3216 (40.3)
4 NA 2219 (27.8)
≥5 NA 702 (8.8)
Unknown NA 315 (4.0)
Continued
2838 Persson et al.

Table 1 Continued

PCI (n = 1773, 16%) CABG (n = 9364, 84%) P-value

......................................................................................................................................................................................
Left internal mammary artery, n (%) NA 7201 (90.3) NA
Right internal mammary artery, n (%) NA 207 (2.6) NA
Bilateral internal mammary artery, n (%) NA 194 (2.1) NA

BMI: body mass index; COPD: chronic obstructive pulmonary disease; CCS: chronic coronary syndrome; NSTEMI: non-ST-elevation myocardial infarction.

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

Figure 2 Outcome for PCI and CABG in left main coronary artery disease. The number of deaths, MI, stroke, revascularization, MACCE, and cor­
responding incidence rates for CABG and PCI, and crude and adjusted hazard ratios for PCI compared to CABG. IPW = inverse probability weighting,
IV = instrumental variable.

had a larger benefit, with regard to MACCE for CABG vs. PCI, than pa­ There are several factors that make the comparison between
tients >70 years (Figure 4). CABG and PCI challenging: (i) the unequal number of patients in
To illustrate the difference in expected survival between CABG the groups; (ii) substantial differences in baseline characteristics; (iii)
and PCI, IPW-adjusted differences in median survival time for mortal­ unknown selection mechanisms for choice of the revascularization
ity in the whole cohort and subgroups are presented in Figure 5. method; and (iv) the PCI group, consisting of more frail patients,
There was a benefit in median survival time for CABG vs. PCI in had more hospitalisations due to non-cardiac events. Despite these
the whole cohort as well as in the subgroups. Furthermore, there challenges, it is important to analyse revascularization mode in rela­
was a notable difference in IPW-adjusted median survival time of tion to outcomes in observational registry studies to complement
3.6 years (95% CI 3.3–4.0) in favour of CABG compared to PCI in pa­ RCTs. In RCTs, the external validity is restrained by the selection of
tients with diabetes. patients, omitting older, more frail patients with a large number of co­
morbidities that are treated in the cardiology department in everyday
clinical practice.13–15
Discussion
The main finding in this observational non-randomized study of all- Adjustments for baseline characteristics
comers with LMCA disease was that CABG was associated with lower The treatment decision, CABG or PCI, for patients with left main cor­
mortality and fewer cardiovascular or cerebrovascular events com­ onary stenosis is based on a patient’s pre-treatment characteristics and
pared to PCI before and after multivariable adjustment for known personal preference.4,16,17 If these are imbalanced and associated with
and unknown confounders (Structured Graphical Abstract). the study outcome, the assessment of the treatment effect from an
PCI or CABG for left main coronary artery disease 2839

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

Figure 3 Inverse probability weighting adjusted hazard ratios for mortality for PCI vs. CABG in subgroups.

Figure 4 Inverse probability weighting adjusted hazard ratios for MACCE for PCI vs. CABG in subgroups.

observational study suffers from bias caused by confounding by indica­ the PCI group), site of LMCA lesion (shaft vs. bifurcation), proportion
tion. We have thus adjusted for known confounders with the IPW of calcified lesions or chronic total occlusions and heart team assess­
method.8 ment (yes vs. no) are missing. The selection of revascularization
mode is to a large extent associated with administrative region (see
Supplementary data online, Table S1). We used statistical modelling
Adjustments for unknown confounders based on the IV to reduce bias due to unmeasured confounders. To
In our study, important confounders such as SYNTAX score, use IV analysis, one must identify a naturally varying phenomenon in
EuroSCORE (in the PCI group), left ventricular ejection fraction (in the observed data, which, like randomization in an RCT, predicts the
2840 Persson et al.

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

Figure 5 Inverse probability weighting adjusted median survival for mortality for PCI vs. CABG in the whole cohort and subgroups. Thin bar at point
estimate = 95% confidence interval.

treatment assigned to the individual patient. A valid instrument must In the EXCEL trial, PCI and CABG showed comparable results with
fulfil some necessary criteria. First, the variable has to be strongly asso­ regard to the primary endpoint, a composite of death, stroke, or MI at 3
ciated (F-test >10) with the treatment received. Second, it must not be and 5 years of follow-up. Of note, total mortality was significantly higher
directly or indirectly associated with the outcome except through the in the PCI group than in the CABG group and CABG proved to be
effect of the treatment itself. The variable with these statistical qualities beneficial over PCI when adding revascularization to outcome at 5
is called an IV. We used the administrative region as the treatment- years.19 The European Association for Cardio-Thoracic Surgery has
preference instrument. The administrative region is frequently withdrawn its support from the current recommendation on treat­
employed as an instrument because this variable type usually fulfils ment of left main disease in the 2018 ESC/EACTS myocardial revascu­
the theoretical criteria for a valid instrument.10,11 Because administra­ larization guidelines. They state that there is a significant survival
tive region may be an imperfect instrument, the following variables advantage of CABG over PCI in the EXCEL trial and that the EXCEL
were entered into IV regression: age, gender, number of diseased cor­ investigators adopted a new definition for MI leading to results appear­
onary arteries, hypertension, hyperlipidemia, year of treatment, body ing to favour the PCI option.22
mass index, diabetes mellitus, insulin treatment, estimated creatinine Patients in our observational study differ from the patients included
clearance, indication for revascularization, smoking status, previous in RCTs. The patients in our cohort were approximately 4 years older
MI, previous PCI, COPD, peripheral artery disease, history of cancer, compared to subjects in the NOBLE and EXCEL trials. CCS was the
and dialysis.10,11 predominant indication for index revascularization in the NOBLE and
EXCEL trials (82% and 53%, respectively) whereas the rate of CCS
was 30.9% in our cohort. In the NOBLE trial, complex lesions (chronic
total occlusions, non-left main bifurcation lesions requiring two stent
Comparison to other studies on
techniques or lesions with calcified or tortuous vessel morphology)
revascularisation strategies for were excluded and there was a high proportion of isolated left main dis­
LMCA disease ease.18 In the EXCEL trial, patients with high anatomical complexity of
The two large landmark trials, the Nordic-Baltic-British Left Main coronary artery disease as defined by a SYNTAX score of ≥33, were
Revascularization Study (NOBLE) and Evaluation of XIENCE excluded. In our analysis, we have not excluded subjects based on
Everolimus Eluting Stent Versus Coronary Artery Bypass Surgery the complexity of the coronary artery disease. In fact, almost half of
for Effectiveness of Left Main Revascularization (EXCEL) trial, evaluat­ the PCI patients had four or more stents. Furthermore, we have in­
ing PCI vs. CABG for LMCA disease have reported conflicting re­ cluded patients that were stented with first-, second- and third-
sults.18–21 The NOBLE trial reported that CABG was superior to generation DES.
PCI in 1201 subjects up to 5 years of follow-up. The primary endpoint, In a meta-analysis of individual patient data from four clinical rando­
a combination of all-cause mortality, non-procedural MI, any repeat mised trials comparing 5-year outcomes for PCI with DES (n = 2197)
coronary revascularization, and stroke, was more frequent in subjects with CABG (n = 2197) for LMCA disease, there was no significant dif­
assigned to PCI. ference in 5-year mortality between groups.18,19,23–26 However,
PCI or CABG for left main coronary artery disease 2841

complementary Bayesian analysis suggested a probable mortality differ­ ≤70 years. The relatively larger procedural risks with CABG compared
ence (more likely than not <0.2% per year) in favour of CABG.26 This to PCI might lessen the beneficial effects compared to PCI in older
low, yet probable, excess risk for death within five years with PCI com­ patients.
pared to CABG in patients eligible for RCTs could result in larger dif­
ferences between groups in our study of consecutive all-comers
including older, frailer patients with more comorbidities than patients IPW-adjusted median survival time
in RCTs. Since HRs are measurements of relative effects and not absolute treat­
In a recent study from Canada, revascularization strategy in LMCA ment effects, they may be difficult to communicate to patients and phy­
disease has been compared using clinical and administrative databases sicians when discussing revascularization options for LMCA disease. In a
in Ontario.27 Like our study, Tam et al. reported that CABG was the study of patients’ preferences, only 38% of the subjects with LMCA dis­
preferred method for revascularization and propensity score matched ease or three-vessel disease consented to go through CABG, although
analysis showed that PCI was associated with higher mortality (HR 1.63 CABG was the recommended option over PCI (n = 763).17 Thus, we
[95% CI 1.42–1.87]) and higher MACCE rates (HR 1.77 [95% CI 1.57– analysed the IPW-adjusted median survival time for PCI and CABG

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

2.00]). Additionally, differences in patient characteristics between the for the whole cohort and subgroups. The median survival time might
PCI and CABG groups were similar to our study. Patients who under­ be helpful when discussing revascularization modes with the patients.
went PCI were older, more often women, more likely presented with The benefits of CABG with a longer median survival time must be scru­
an acute coronary syndrome event, but had fewer diseased coronary tinized against the disadvantages with open-heart surgery. Longer hos­
vessels.27 pital stays and recovery time, and higher risk of perioperative
complications can all be discussed between the patient and the heart
team. The differences in median survival time favour CABG in all sub­
Possible mechanisms for differences
groups and there is a large difference in survival for CABG vs. PCI in dia­
in mortality betic patients.
Invasive treatment with PCI or CABG restores flow in obstructive le­
sions in fundamentally different ways. With PCI, a DES is positioned
in flow-limiting lesions after balloon dilatation, leaving non-flow-limiting Limitations
lesions untreated. With CABG, a blood conduit (arterial or venous The data in the SWEDEHEART registry are prospectively collected and
graft) is sewn distal to the flow-limiting lesion, creating a collateral the analysis is a post hoc non-randomized comparison of patients under­
flow. Through distal surgical collateralization, the graft prevents going PCI or CABG for LMCA disease. The data are observational, and
non-flow-limiting lesions from causing MIs. Establishing distal collateral we have tried to adjust for known imbalances in patient characteristics
blood flow is an important feature of CABG since new MIs frequently with IPW adjustment. We do not have data in all patients about the
develop in non-flow-limiting coronary segments.28,29 Most deaths in SYNTAX score, EuroSCORE, LVEF, site of LMCA lesion, heart team
high-risk patients with proximal and extensive coronary artery disease assessment, or completeness of revascularization. However, IV adjust­
are caused by MIs.30 RCTs in patients with left main and three-vessel ment allows estimation of treatment effects in the presence of unmeas­
disease have shown that CABG decreased the occurrence of new ured confounding. A disadvantage with the IV analyses is a somewhat
MIs more than PCI.21,31,32 Our study showed that CABG was asso­ higher statistical uncertainty than with IPW adjustments, which mani­
ciated with fewer MIs and supports evidence from RCTs, thus adding fests itself in larger CIs with less precise estimates and enhanced prob­
weight to the hypothesis that surgical collateralization and a larger pro­ ability of type 2 error. Although IV adjustment models are designed to
portion of complete revascularization33 are possible mechanisms re­ adjust for unknown confounders, residual confounding cannot be ruled
sponsible for the mortality benefit for CABG vs. PCI. out. The cohort of studied patients were entered into the registry be­
tween January 1st 2005 and December 31st 2015 and the selection of
Subgroup analyses revascularization mode for patients with LMCA disease might differ in
There was a significant quantitative interaction for diabetes with a large our study from today’s clinical practice.
difference in aHR for mortality favouring CABG over PCI. The inter­
action translates into a median survival benefit of 3.6 years for CABG
compared to PCI. In the EXCEL20 and NOBLE18 trials, there were Conclusion
no interactions for diabetes for the primary outcomes. However, we
In this non-randomized study, CABG in patients with LMCA disease
know from the randomized Future Revascularization Evaluation in
was associated with lower mortality and fewer MACCE compared to
Patients with Diabetes Mellitus: Optimal Management of Multivessel
PCI after multivariable adjustment for known and unknown
Disease trial31 that patients with diabetes and multivessel disease (ex­
confounders.
cluding left main coronary stenosis) have a better outcome with
CABG in comparison with PCI, including death from any cause. Our
data reinforce diabetes as an important clinical factor when deciding Acknowledgements
on revascularization mode in LMCA disease. Effect modification for We would like to acknowledge the work collecting data from the pro­
MACCE between concomitant coronary artery disease and type of re­ cedures performed by the staff at each PCI- and Thoracic surgery cen­
vascularization has shown higher benefit for CABG in patients with tre in Sweden.
more severe disease. This finding is line with the results from RCTs
showing less repeat revascularization for CABG.18,20 This interaction
could be mediated by more complete revascularization achieved with
CABG.33 There was also a quantitative interaction for age group with
Supplementary data
a larger benefit in MACCE for CABG compared to PCI in patients Supplementary data is available at European Heart Journal online.
2842 Persson et al.

Data availability revascularization studies: trials vs real-world registries. JACC Cardiovasc Interv 2022;15:
1441–1449. https://doi.org/10.1016/j.jcin.2022.05.023
16. Gripenberg T, Jokhaji F, Östlund-Papadogeorgos N, Ekenbäck C, Linder R, Samad B,
The data underlying this article will be shared on reasonable request to the
et al. Outcome and selection of revascularization strategy in left main coronary artery
corresponding author.
stenosis. Scand Cardiovasc J 2018;52:100–107. https://doi.org/10.1080/14017431.2018.
1429648
17. Kim C, Hong SJ, Ahn CM, Kim JS, Kim BK, Ko YG, et al. Patient-centered decision-
Conflict of interest making of revascularization strategy for left main or multivessel coronary artery disease.
Am J Cardiol 2018;122:2005–2013. https://doi.org/10.1016/j.amjcard.2018.08.064
J.P. has received unrestricted grants from Abbott Inc., unrelated to the pre­ 18. Mäkikallio T, Holm NR, Lindsay M, Spence MS, Erglis A, Menown IBA, et al.
sent work. J.Y. has no conflicts of interest. O.A. has received research grant Percutaneous coronary angioplasty versus coronary artery bypass grafting in treatment
and lecture fees from Abbott Inc. D.V. has no conflicts of interest. A.J. has of unprotected left main stenosis (NOBLE): a prospective, randomised, open-label,
received fees for consultancy or lectures from AstraZeneca, Werfen, non-inferiority trial. Lancet 2016;388:2743–2752. https://doi.org/10.1016/S0140-
Portola, Baxter and LFB Biotechnologies, all unrelated to the present 6736(16)32052-9
work. I.S. has no conflicts of interest. R.L. has no conflicts of interest. D.E. 19. Stone GW, Kappetein AP, Sabik JF, Pocock SJ, Morice MC, Puskas J, et al. Five-year out­
comes after PCI or CABG for left main coronary disease. N Engl J Med 2019;381:
has no conflicts of interest. T.I. has no conflicts of interest. E.O. has no con­

Downloaded from https://academic.oup.com/eurheartj/article/44/30/2833/7191997 by guest on 17 August 2023

1820–1830. https://doi.org/10.1056/NEJMoa1909406
flicts of interest. 20. Stone GW, Sabik JF, Serruys PW, Simonton CA, Généreux P, Puskas J, et al.
Everolimus-eluting stents or bypass surgery for left main coronary artery disease. N
Engl J Med 2016;375:2223–2235. https://doi.org/10.1056/NEJMoa1610227
Funding 21. Holm NR, Mäkikallio T, Lindsay MM, Spence MS, Erglis A, Menown IBA, et al.
Percutaneous coronary angioplasty versus coronary artery bypass grafting in the treat­
This work has been funded by the regional cooperation for medical re­ ment of unprotected left main stenosis: updated 5-year outcomes from the rando­
search and healthcare development in Stockholm (ALF; 20140076; mised, non-inferiority NOBLE trial. Lancet 2020;395:191–199. https://doi.org/10.
20150422; 20130339). 1016/s0140-6736(19)32972-1
22. European Association for Cardio-Thoracic Surgery. Changing Evidence, Changing
Practice. https://www.eacts.org/changing-evidence-changing-practice/. 19 December
References 2019.
1. Conley MJ, Ely RL, Kisslo J, Lee KL, McNeer JF, Rosati RA. The prognostic spectrum of 23. Serruys PW, Morice MC, Kappetein AP, Colombo A, Holmes DR, Mack MJ, et al.
left main stenosis. Circulation 1978;57:947–952. https://doi.org/10.1161/01.cir.57.5.947 Percutaneous coronary intervention versus coronary-artery bypass grafting for severe
2. Bittl JA, He Y, Jacobs AK, Yancy CW, Normand SL. Bayesian methods affirm the use of coronary artery disease. N Engl J Med 2009;360:961–972. https://doi.org/10.1056/
percutaneous coronary intervention to improve survival in patients with unprotected NEJMoa0804626
left main coronary artery disease. Circulation 2013;127:2177–2185. https://doi.org/10. 24. Morice M-C, Serruys PW, Kappetein AP, Feldman TE, Ståhle E, Colombo A, et al.
1161/circulationaha.112.000646 Outcomes in patients with de novo left main disease treated with either percutaneous
3. Yusuf S, Zucker D, Peduzzi P, Fisher LD, Takaro T, Kennedy JW, et al. Effect of coronary
coronary intervention using paclitaxel-eluting stents or coronary artery bypass graft
artery bypass graft surgery on survival: overview of 10-year results from randomised
treatment in the synergy between percutaneous coronary intervention with TAXUS
trials by the coronary artery bypass graft surgery trialists collaboration. Lancet 1994;
and cardiac surgery (SYNTAX) trial. Circulation 2010;121:2645–2653. https://doi.org/
344:563–570. https://doi.org/10.1016/s0140-6736(94)91963-1
10.1161/CIRCULATIONAHA.109.899211
4. Neumann FJ, Sousa-Uva M, Ahlsson A, Alfonso F, Banning AP, Benedetto U, et al. 2018
25. Park S-J, Kim Y-H, Park D-W, Yun S-C, Ahn J-M, Song HG, et al. Randomized trial of
ESC/EACTS guidelines on myocardial revascularization. Eur Heart J 2019;40:87–165.
stents versus bypass surgery for left main coronary artery disease. N Engl J Med 2011;
https://doi.org/10.1093/eurheartj/ehy394
364:1718–1727. https://doi.org/10.1056/NEJMoa1100452
5. Sianos G, Morel MA, Kappetein AP, Morice MC, Colombo A, Dawkins K, et al. The
26. Sabatine MS, Bergmark BA, Murphy SA, O’Gara PT, Smith PK, Serruys PW, et al.
SYNTAX score: an angiographic tool grading the complexity of coronary artery disease.
Percutaneous coronary intervention with drug-eluting stents versus coronary artery by­
EuroIntervention 2005;1:219–227.
pass grafting in left main coronary artery disease: an individual patient data meta-analysis.
6. Nashef SA, Sharples LD, Roques F, Lockowandt U. EuroSCORE II and the art and sci­
Lancet 2021;398:2247–2257. https://doi.org/10.1016/S0140-6736(21)02334-5
ence of risk modelling. Eur J Cardiothorac Surg 2013;43:695–696. https://doi.org/10.1093/
27. Tam DY, Fang J, Rocha RV, Rao SV, Dzavik V, Lawton J, et al. Real-World examination of
ejcts/ezs468
7. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. revascularization strategies for left main coronary disease in Ontario, Canada. JACC
Nephron 1976;16:31–41. https://doi.org/10.1159/000180580 Cardiovasc Interv 2023;16:277–288. https://doi.org/10.1016/j.jcin.2022.10.016
8. Heinze G, Jüni P. An overview of the objectives of and the approaches to propensity 28. Ambrose JA, Tannenbaum MA, Alexopoulos D, Hjemdahl-Monsen CE, Leavy J, Weiss
score analyses. Eur Heart J 2011;32:1704–1708. https://doi.org/10.1093/eurheartj/ M, et al. Angiographic progression of coronary artery disease and the development of
ehr031 myocardial infarction. J Am Coll Cardiol 1988;12:56–62. https://doi.org/10.1016/0735-
9. Bottai M, Zhang J. Laplace regression with censored data. Biom J 2010;52:487–503. 1097(88)90356-7
https://doi.org/10.1002/bimj.200900310 29. Falk E, Shah PK, Fuster V. Coronary plaque disruption. Circulation 1995;92:657–671.
10. Rassen JA, Schneeweiss S, Glynn RJ, Mittleman MA, Brookhart MA. Instrumental vari­ https://doi.org/10.1161/01.cir.92.3.657
able analysis for estimation of treatment effects with dichotomous outcomes. Am J 30. Milojevic M, Head SJ, Parasca CA, Serruys PW, Mohr FW, Morice MC, et al. Causes of
Epidemiol 2009;169:273–284. https://doi.org/10.1093/aje/kwn299 death following PCI versus CABG in Complex CAD: 5-year follow-up of SYNTAX. J Am
11. Li J, Fine J, Brookhart A. Instrumental variable additive hazards models. Biometrics 2015; Coll Cardiol 2016;67:42–55. https://doi.org/10.1016/j.jacc.2015.10.043
71:122–130. https://doi.org/10.1111/biom.12244 31. Farkouh ME, Domanski M, Sleeper LA, Siami FS, Dangas G, Mack M, et al. Strategies for
12. Bound J, Jaeger DA, Baker RM. Problems with instrumental variables estimation when multivessel revascularization in patients with diabetes. N Engl J Med 2012;367:
the correlation between the instruments and the endogeneous explanatory Variable 2375–2384. https://doi.org/10.1056/NEJMoa1211585
is weak. J Am Stat Assoc 1995;90:443–450. https://doi.org/10.2307/2291055 32. Head SJ, Davierwala PM, Serruys PW, Redwood SR, Colombo A, Mack MJ, et al.
13. Frieden TR. Evidence for health decision making—beyond randomized, controlled trials. Coronary artery bypass grafting vs. percutaneous coronary intervention for patients
N Engl J Med 2017;377:465–475. https://doi.org/10.1056/NEJMra1614394 with three-vessel disease: final five-year follow-up of the SYNTAX trial. Eur Heart J
14. Rothwell PM. External validity of randomised controlled trials: “to whom do the results 2014;35:2821–2830. https://doi.org/10.1093/eurheartj/ehu213
of this trial apply?”. Lancet 2005;365:82–93. https://doi.org/10.1016/s0140-6736(04) 33. Takahashi K, Serruys PW, Gao C, Ono M, Wang R, Thuijs D, et al. Ten-year all-cause
17670-8 death according to completeness of revascularization in patients with three-vessel dis­
15. Megaly M, Buda K, Mashayekhi K, Werner GS, Grantham JA, Rinfret S, ease or left main coronary artery disease: insights from the SYNTAX extended survival
et al. Comparative analysis of patient characteristics in chronic total occlusion study. Circulation 2021;144:96–109. https://doi.org/10.1161/circulationaha.120.046289