Competing risk analysis of events 10 years after revascularization

Abstract

Objectives. To evaluate the influence of competing risk (CR) non-cardiac death during long-term follow-up of revascularized patients on the interpretation of the cardiac outcomes. Methods. Retrospectively, we compared outcomes estimated with the Kaplan-Meier and the cumulative incidence function (CIF) methods after a median 10.8 years follow-up in 1 234 consecutive patients (594 CABG, 640 PCI) undergoing first time non-emergent revascularization in a community cohort. Results. Overall 301 (24.4%) patients died (27.3% in the CABG vs. 21.7% in the PCI group, p=0.02). The causes of death were cardiac (10.3%) and non-cardiac (14.1%). CR analysis showed a similar probability of cardiac death (CIF 0.10 (95% CI 0.092, 0.18) vs. 0.093 (0.07, 0.12)) in the CABG and PCI treated patients, respectively. The probability for acute myocardial infarction (CIF 0.12 vs. 0.16 p<0.001), congestive heart failure (CIF 0.15 vs. 0.09 p=0.007) in the CABG and PCI group respectively, differed. The differences were also statistically significant after multivariate adjustment for the competing risks of death. For all outcomes the Kaplan-Meier method overestimated risk estimates. Conclusions. The competing risk adjusted probability for cardiac death, but not other cardiac endpoints are comparable in patients treated with either CABG or PCI after very long-term follow-up. The risk for all-cause death was mainly predicted by the occurrence of non-cardiac diseases.

Key words::

• CABG
• PCI
• long-term survival
• competing risk

The relative effectiveness of revascularization in preventing death and adverse ischemic events with either coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI) in patients with symptomatic coronary artery disease (CAD) has been evaluated in a number of randomized clinical trials (RCT) and observational studies (1). While the longevity of up to 5–7 years follow-up has been shown to be comparable, it has been speculated that a possible benefit of surgical revascularization over PCI with regard to major cardiac events (cardiac death, acute myocardial infarction (AMI), development of ischemic congestive heart failure (CHF) and repeat revascularizations) would become apparent after extended follow-up. Few studies however, support this hypothesis (2,3), and studies with very long follow-up are therefore warranted.

Death in these patients is often assumed to be of cardiac cause as CAD is progressive and life-long (4,5). The improvements in cardiac therapy together with aging in the last decades have however (6) increased the probability of developing and dying of non-cardiac diseases. Evaluation of only total mortality and not cause-specific deaths in studies comparing long-term outcome after either CABG or PCI may therefore conceal the unexpected occurrence of non-cardiac deaths, obscuring the interpretation of the results (7–9). Statistically, a particular situation arises when interest is focused on a specific cause of failure in the presence of other different causes, which alter the probability of experiencing the event of interest. Death due to non-cardiac competing causes during study follow-up before the occurrence of the cardiac event of interest poses special analytic challenges.

How the competing risk of non-cardiac death influence the evaluation of the very long-term outcome in revascularized patients in the present era has not, to our knowledge, been previously studied. The primary aim of the present study was therefore to evaluate the competing risk adjusted risk for fatal or non-fatal cardiac events in a well defined cohort of patients with stable or unstable CAD treated initially with either CABG or PCI. Secondly, we also explored the predictors for all cause mortality in this cohort.

Methods

Patient selection

Haukeland University Hospital is the sole tertiary cardiac center in western Norway and provides CABG and PCI procedures for the all residents in Hordaland county (450 000 inhabitants). Clinical characteristics and detailed findings at coronary angiography and PCI for all patients who are treated at the hospital are entered prospectively into a database by the cardiologist who performs the procedure.

Included in the present study are all patients referred to angiography with either stable angina pectoris or recent acute coronary syndromes (ACS: Unstable angina pectoris or non-ST elevation AMI) from January 1, 1995 to December 31, 1998 who thereafter underwent isolated CABG (the CABG group) or PCI (the PCI group) (Figure 1). Patients with prior revascularization, significant left main coronary artery stenosis, acute or recent ST elevation myocardial infarction or who required emergency procedures were excluded. The treatment choice for each patient was the joint decision of the cardiologists and cardiac surgeons.

Figure 1. Flow diagram for the selection of study patients.

Event classification during follow-up

Almost all of the study population (98.7%) remained residents of western Norway throughout the study period. All out-patient consultations, hospitalizations and surgical procedures in the region were tracked and verified by reviewing all related patient records and querying the hospital information system (HIS). The following study events was categorized: any subsequent revascularizations, hospitalizations for ACS, cerebral stroke, cancer, primary pulmonary hypertension, chronic renal failure (CRF) or diabetes mellitus (DM). Vital status was continuously updated by The National Statistics of Norway, and available for all patients in the HIS up to November 15, 2007. The underlying cause of death (the disease or injury that initiated the train of morbid events resulting in death) was defined as the cause of death (10). For patients dying in-hospital (n=148 (49.0% of all deaths)) the cause of death was classified on the basis of discharge notes and diagnosis. The classification of out of hospital deaths was based on the death certificates, but adjusted according to recent hospital discharge notes if the diagnosis was equivocal or non-informative. Sudden deaths and deaths not clearly attributable to non-cardiac disease were classified as cardiac deaths (n=33, 11.0% of all deaths) (11). The competing risk is defined as an event whose occurrence either precludes the occurrence of another event under examination, or fundamentally alters the probability of the occurrence of this other event (12). If for example cardiac death (AMI, CHF, and malignant tachy-and bradyarrhytmias, end stage valvular disease) is the outcome of interest, cause-specific death due to cancer or other non-cardiac diseases is a competing risk of death.

Statistics/data analysis

Baseline continuous data are presented as either mean±SD or median (interquartile range), and categorical data as frequencies (%). Baseline differences between the CABG and PCI groups were compared with the t-test, Wilcoxon rank sum test or χ² test where appropriate, and should be regarded as exploratory data analyses. The distribution of age and ejection fraction were skewed, and these variables were entered dichotomizes in the analysis. An extended Cox proportional hazard model was used to isolate independent predictors for all-cause mortality at long-term follow-up. New serious diseases during the follow-up period were entered as time-dependant covariates in the analysis. The determination of the functional forms of the continuous variables was guided by analysis of the Martingale residuals, and the proportional hazard assumption of the regressional models was verified by analysis of the scaled Schoenfeld residuals.

In the analysis of the study events of interest (cardiac death, AMI, CHF or repeat revascularizations) neither the Kaplan-Meier (KM) method nor the Cox proportional hazard model are appropriate when competing causes of death are present, or the study events are not independent (12,13). Therefore, the cumulative incidence function (CIF) was used to estimate the probability for the events of interest during follow-up. Gray's or Pepe-Mori test (13) was used to test for statistical significant differences between the CABG and PCI group respectively. Multivariate competing risk analyses of predictors for cardiac, vs. non-cardiac death, and for AMI and CHF (with deaths not due to AMI or CHF as the competing risks, respectively), was performed with the Fine and Gray proportional subdistribution hazard regression model (13). The R statistical software version 2.6.2 (The R Foundation for Statistical Computing, Vienna, Austria; available at http://www.r-project.org) with the cmprsk package was used for the competing risk analyses. In order to minimize the impact of confounding when comparing the occurrence of cardiac events in the CABG vs. PCI group, the analyses were also repeated after balancing the differences in baseline variables by propensity match scoring (using the R package twang which calculates the scores with a boosted logistic regression method). The propensity score is defined as a subject's probability of receiving a specific treatment conditional on the observed covariates. The use of propensity scores was effective in reducing baseline differences between the two study groups (all p-values >0.25). All other analyses were performed with SPSS version 15 (SPSS Inc. Chicago, Ill). Missing data are assumed to occur by random. A p-value less than 0.05 was considered statistically significant. The study was approved by the regional ethics committee.

Results

The clinical characteristics and findings at preoperative coronary angiography are presented in Table I. A total of 1 234 patients were identified in the database of which 640 and 594 underwent isolated PCI or CABG respectively (Figure 1). Patients undergoing CABG were on average older, a larger proportion had prior AMI and reduced EF compared to the patients treated with PCI. On the other hand, more patients in the PCI group had angiographic features suggesting recent ACS. When only patients with multivessel CAD were compared, the differences between the two groups were less obvious. Of the patients treated with CABG, 93.1% either had triple vessel disease or two vessel disease involving the LAD. The corresponding percentage in the PCI treated patients was 58.3%, implying that 40.3% of these patients with extensive CAD in the total cohort were treated with PCI. There was no trend for change in the relative proportion of patients undergoing either CABG or PCI during the study period.

Table I. Baseline characteristics according to treatment group for 1 234 non-emergent CAD patients undergoing first-time revascularization.

Characteristic	% of CABG group (n=594)	% of PCI group (n=640)	P-value
Age (yrs) median (IQR)	66 (59.0–71.3)	61 (53.0–69.0)	<0.001
Female gender	19.9	25.9	0.01
Family history of CAD	43.1	46.6	0.2
Hypertension	23.4	20.8	0.3
Treated hyperlipidaemia	54.8	57.2	0.7
BMI (median, IQR)	25.6 (23.8–28.1)	25.3 (23.7–27.5)	0.3
Smoker	20.4	12.8	<0.001
Ex-smoker	34.5	33.1	0.6
Diabetes Mellitus	9.4	6.4	0.05
CRF	0.8	0.0	0.03
Clinical presentation			<0.001
Stable angina	79.1	70.3
Acute coronary syndrome*	20.9	29.7
Angina functional class (CCS)			0.007
I–II	54.9	57.7
III–IV	45.1	42.3
Parsonnet score	7.0±7.6	4.9±6.5	<0.001
Euroscore value	2.7±2.0	2.4±2.1	0.02
Findings at angiography
Prior myocardial infarction (%)
None	55.9	65.3	0.001
Inferior wall	20.5	15.8	0.03
Anterior wall	13.0	15.0	0.3
Anterior + inferior wall	10.6	3.9	<0.001
Ejection fraction (%) (median (IQR))	65.0 (55–70)	70 (60–70)	<0.001
Lesion characteristics in pts. with multivessel disease
No. of patients (% of group)	574 (96.6)	491 (76.7)	<0.001
No. of lesions (mean±SD)	3.9±1.4	2.3±1.2	0.03
Long lesions > 10 mm	5.0	5.8	0.4
Total occlusions	22.1	21.0	0.5
Eccentric lesions	2.3	4.3	0.003
Bifurcation lesions	2.5	3.8	0.04
Thrombotic lesions	0.5	1.5	0.1
Visible collaterals	16.9	16.3	0.6

All values are percentages unless otherwise stated.

CRF=Chronic renal failure; CAD=Coronary artery disease; CCS=Canadian Cardiovascular Society angina classification; IQR=Interquartile range.

*Either recent unstable angina or non-ST elevation AMI.

Procedural outcomes

In PCI treated patients the average number of lesions dilated (1.4 lesions per patient) and the success rate (93%) remained unchanged from 1995 to 1998. Of the lesions treated 20% were chronic total occlusions, of which 75% were successfully dilated. From 1995 to 1998, the yearly percentage of PCI patients receiving a stent increased from 31 to 87% (overall 77% stent procedures). Stents were implanted in 71% of all successfully dilated proximal LAD stenosis, increasing from 39% in 1995 to 93% in 1998. There were three same-day rescue CABG after PCI (0.5%), nine patients experienced non fatal Q-wave AMI (1.4%), and subacute stent-thrombosis occurred in three patients (0.5%).

In the group which initially was treated with CABG, 98% had internal mammary artery (IMA) graft to the LAD. All patients in the PCI group who were treated with CABG during follow-up (n=49, 7.7%) also received an IMA.

Competing risk analysis of cardiac events

The overall cohort mortality was 24.4% (n=301) of which 42.5% were classified cardiac deaths after a median 10.8 years follow-up for surviving patients (IQR 9.7–11.9 years) (Table II). An extended Cox model showed that long-term risk for death of any cause was similar in the CABG vs. PCI group (Table III). Traditional CAD risk-factors, the extent of CAD and medication recorded at baseline did not influence long-term prognosis, while the development of CHF and the advent of other serious diseases predicted all cause mortality in this study cohort. Figure 2 shows the 1-KM plots for survival. The crossing of the curves indicates that different levels of competing risk of death occurred during the long-term follow-up of the CABG and PCI patients, respectively. The probability for any death is the sum of probabilities of cardiac and non-cardiac death, and equals the KM estimate (Table IV). After decomposing the survival curves with the CIF method, the risk for cardiac death was comparable after long-time follow-up, while the CABG treated patients had higher risk for non-cardiac death compared to PCI treated patient. In the multivariate competing risk analysis, high age and reduced EF predicted cardiac death, while initial revascularization mode did not (Table V).

Figure 2. The 1-Kaplan-Meier plot of all-cause mortality in each study group. 95% CI and number of patients at risk at yearly timepoints are indicated (upper panel). Cumulative incidence function curves for cardiac (event of interest) and non-cardiac death (the competing risk event) in relation to study group (the lower panels).

Table II. Causes of death according to initial treatment for 1 234 non-emergent CAD patients undergoing first-time revascularization.

	All patients (n=1234)	CABG group (n=594)	PCI group (n=640)	P-value
Total mortality	301 (24.4)	162 (27.3)	139 (21.7)	0.02
Cause specific deaths
Cardiac	128 (10.3)	66 (11.1)	62 (9.7)	0.4
Stroke	28 (2.3)	20 (3.4)	7 (1.1)	0.06
Infection	24 (1.9)	16 (2.7)	8 (1.3)	0.07
Cancer	83 (6.7)	42 (7.1)	41 (6.4)	0.6
CRF	5 (0.4)	4 (0.7)	1 (0.2)	0.2
CPD	5 (0.4)	1 (0.2)	4 (0.6)	0.2
Gastro-intestinal diseases	14 (1.1)	5 (0.8)	9 (1.5)	0.4
Neurologic diseases	4 (0.3)	3 (0.5)	1 (0.2)	0.3
Trauma	6 (0.5)	3 (0.5)	3 (0.3)	0.9
Other	5 (0.5)	2 (0.4)	3 (0.5)	0.7

The data are the number of deaths (percentage of the group).

CPD = Chronic pulmonary disease; CRF=Chronic renal failure.

Table III. Multivariate Cox hazard ratios (HR) showing predictors for all-cause mortality for 1 234 non-emergent CAD patients undergoing first-time revascularization.

Variable	HR	95% CI	P-value
Age ≥65 years	1.9	1.2, 2.5	<0.001
PCI vs. CABG	1.2	0.9, 1.5	0.2
CHF*	4.0	3.0, 5.4	<0.001
Stroke*	3.2	2.2, 4.8	<0.001
CPD*	2.4	1.6, 3.4	<0.001
CRF*	3.3	2.2, 4.8	<0.001
Cancer*	2.8	1.9, 4.1	0.001

All variables were included in the same model. The marked *variables were entered time-dependant into the model counting from the time of first diagnosed.

CHF=Congestive heart failure; CPD=Chronic pulmonary disease; CRF=Chronic renal failure; CI=Confidence interval.

Table IV. Comparison of probability estimates for death and cardiac events calculated with the 1- KM and CIF methods for 1 234 non-emergent CAD patients undergoing first-time revascularization.

	CABG group			PCI group
		CIF			CIF
	1-KM	Event of interest	Competing risk death	1-KM	Event of interest	Competing risk death
Overall death	0.24 (0.19, 0.27)	–	–	0.21 (0.17, 0.24)	–	–
Cardiac death	–	0.10 (0.09, 0.18)	0.15 (0.12, 0.18)	–	0.09 (0.07, 0.12)	0.12 (0.09, 0.14)^†
AMI	0.13 (0.10, 0.16)	0.12 (0.09, 0.14)	0.20 (0.16, 0.23)	0.17 (0.15, 0.21)	0.16 (0.13, 0.19)^†	0.15 (0.12, 0.18)^‡
CHF	0.16 (0.12, 0.18)	0.15 (0.13, 0.19)	0.17 (0.14, 0.21)	0.10 (0.07, 0.12)	0.09 (0.05, 0.13)^‡	0.16 (0.07, 0.12)
PCI during follow-up	0.04 (0.03, 0.06)	0.02 (0.01, 0.04)	0.23 (0.19, 0.26)	0.24 (0.20, 0.27)	0.10 (0.07, 0.12)^‡	0.19 (0.16,0.22)^†
Cardiac surgery during follow-up	0.04 (0.03, 0.06)	0.03 (0.02, 0.05)	0.23 (0.20, 0.26)	0.6 (0.04, 0.08)	0.05 (0.03, 0.07)^†	0.19 (0.16, 0.22)^†

The numbers are the cumulative probabilities (95% confidence interval) for experiencing the events of interest at 10 years follow-up.

KM=Kaplan Meier method; CIF=Cumulative incidence function; AMI=acute myocardial infarction; CHF=congestive heart failure.

Significant differences for the CIF comparisons between the CABG and PCI groups are indicated: ^†p<0.05; ^‡p<0.01.

Table V. Multivariate competing risk regression analysis with and without propensity score adjustment for predictors of cardiac events in 1 234 non-emergent CAD patients undergoing first-time revascularization.

	CIF Hazard ratios (95% CI)
Multivariate models of CIF	Cardiac death	AMI	CHF
Age ≥65 years	2.9 (1.9, 4.3)^‡	–	2.1 (1.5, 3.0)^‡
Prior AMI	2.7 (1.9, 4.0)^‡	1.2 (1.0, 1.4)^†	2.2 (1.6, 3.1)^‡
Stable angina vs. ACS	–	0.6 (0.4, 0.8)^‡	–
CABG vs. PCI	1.3 (0.9, 1.8)	0.7 (0.5, 0.9)^‡	1.4 (1.0, 1.9)5
AMI during follow-up	–	–	2.4 (1.7, 3.5)^§
Propensity score weighted multivariate CIF models*
Age ≥65 years	3.1 (2.1, 4.3)^‡	–	–
Prior AMI	3.0 (2.0, 4.3)^‡	–	–
Stable angina vs ACS	–	0.6 (0.41, 0. 8)^‡	–
CABG vs PCI	1.2 (0.8, 1.7)	0.71 (0.5, 1.0)^†	1.7 (1.2, 2.3)^‡
AMI during follow-up	–	–	2.6 (1.9, 3.7)^‡
Propensity weight score	1.2 (0.9, 1.6)^†	0.84 (0.6, 1.1)	1.2 (1.0, 1.4)

The regressional CIF hazard ratios (HR) for cardiac events during follow-up were adjusted for other competing risk of death using proportional subdistribution hazards regression models by the Gray crr function (R cmprsk package).

*The propensity of undergoing either CABG or PCI based on baseline features was added as covariate in the model.

CIF=Cumulative incidence function; AMI=myocardial infarction; ACS=either recent unstable angina or non-ST elevation AMI.

Only significant predictors in addition to treatment group and propensity weight scores were retained in the models.

^†p<0.05, ^‡p<0.001, ^§p=0.07.

AMI and CHF

Significantly more patients in the PCI group experienced fatal or non-fatal AMI during follow-up as compared to the CABG group (16.7% vs. 11.8% in the PCI and CABG group respectively, p=0.013). This was due to more spontaneous and not procedural-related AMI in the PCI group. The risk difference was highly significant in the multivariate competing risk analyses (Table V).

Significantly more patients in the CABG group had a diagnosis of CHF during follow-up (15.8% vs. 10.2% in the CABG and PCI group respectively, p=0.003). The higher incidence of CHF in the CABG group appeared in the second half of the study period (Figure 3). In the multivariate analyses the difference in risk for CHF persisted after adjustment for prior medical history including age, extent of CAD, need for repeat revascularization and the occurrence of AMI during follow-up.

Figure 3. Major cardiac events in the CABG vs. PCI group. For congestive heart failure (CHF) and acute myocardial infarction (AMI) the figures show the cumulative incidence probability (CIF) of developing the disease, conditional on not dying of other competing diseases. The CIF for the first repeat PCI and cardiac surgery are conditional on survival. P-values are according to the Pepe-Mori test.

Need for repeat procedures

The need for repeat revascularizations was higher in the PCI group compared to the CABG group (Figure 3). Overall 7.4% in the CABG group and 27.4% of the patients in the PCI group subsequently underwent any revascularization during follow-up. The yearly rate of early (<6 months) repeat procedures due to restenosis and subacute vessel occlusion decreased from 26% in 1995 to 13% in 1998 (p=0.025) parallel to increased use of stents. Only 28 primary PCI procedures were performed in connection with AMI during follow-up. Most repeat procedures after the first follow-up year were due to progression of CAD in native vessels or due to attrition of venous grafts. The KM method which censors patients dying during follow-up, overestimated the risk for repeat revascularizations markedly compared the CIF method (Table IV).

Discussion

We have evaluated the very long-term competing risk adjusted outcomes in first-time revascularized CAD patients in a real life setting at a single hospital serving a geographically well-defined region. After a median of 10 years follow-up, a quarter of the patients were dead, and more than half of the patients died from non-cardiac causes. After adjusting for competing risks of death, the probability for cardiac events, except cardiac death, differed in the CABG and PCI treated patients.

The competing risk analysis of revascularization studies

Analysis of the underlying cause of death is seldom reported in neither RCT's nor observational studies comparing CABG vs. PCI (1,8). Observational studies are prone to treatment selection bias, and outcome evaluations can be influenced by competing endpoints which are often not available or adjusted for in the analysis (2,7). Differences in baseline characteristics can be adjusted for by propensity score matching, and standard multivariable outcome analysis can be used to identify predictors for overall death. However, both the KM method with use of the log rank test, and the Cox proportional hazard model treats competing risks of the event of interest as censored observations, and thus wrongly inflates the probability of experiencing the event of interest (12,13). The error increases with the length of follow-up and the number of competing endpoints. Competing risk analysis demands detailed knowledge of the events in the follow-up period. When competing risk is present, the CIF is recognized in both the medical (14) and statistical literature (15,16) as the right tool to use, but is seldom used in clinical studies. This may in part be due to a lack of awareness that the KM method produces misleading results, and that the procedures needed to compute CIF only recently have become available in mainstream statistical software packages.

Adjusted long-term risks for cardiac events

In our study we were able to track nearly all events in a community cohort of revascularized patients due to detailed registries covering data from all hospitals in a single health care system. We found that after adjusting for the competing risks of death, a different pattern of cardiac disease progression emerged in patients treated initially with CABG as compared to PCI during 10 years follow-up.

In our observational study, the risk for cardiac death was comparable in patients treated with either CABG or PCI at median 10 years follow-up after competing risk adjustment for non-cardiac deaths. More AMI deaths in the PCI group were balanced by more deaths due to CHF and sudden deaths in the CABG group. The risk for non-cardiac death became higher than for cardiac death after six years follow-up, and the risk differed for CABG and PCI treated patients. In the survival analysis of studies with longer than five years follow-up, the influence of competing risk of death should therefore be accounted for. Moreover, during the very long-term follow-up in these patients non-cardiac diseases become increasingly important. More focus on preventing and detecting these conditions may therefore be warranted.

The increased risk for AMI in the PCI group compared to the CABG group persisted after adjusting for differences in baseline features and competing risks during follow-up. A possible explanation for this advantage in CABG treated patients may be that more patients achieve complete revascularization after surgery compared to PCI, and that the midvessel graft insertion in CABG should give both proximal and distal protecting in case of ACS (17,18).

Development of CHF during the very long term follow-up of revascularized patients has not previously been reported in detail. Over time the incidence of hospitalizations for CHF was significantly higher in the CABG group as compared to the PCI group. The difference was also present after adjustments for baseline characteristics and other cardiac event during follow-up. Angiographic studies have shown up to 50% vein graft disease after five years or longer follow-up (19,20), and at least in a subset of these patients this was associated with adverse outcome. Perioperative myocardial injury and early and late veingraft failure might in part explain the increased rate of CHF in the CABG group in our study (21,22). Chronic ischemia may also contribute to the progression of left ventricular dysfunction without an obvious clinical ischemic event being recorded (23,24).

In our study the need for repeat revascularizations in the short-term was comparable to other studies in the stent/IMA era (25), with a higher rate in patients initially treated with PCI than CABG (1) due to restenosis and incomplete revascularization. The risk estimate was, however, markedly lower by the CIF than the KM method.

Limits and strengths

We have no follow-up data on the adherence to lifestyle adjustments or medical treatment which might also influence outcome. The proportion taking statins and aspirin in the follow-up period was, as in other patients treated in the current era in our region (26), likely very high in both study groups. The study population was well defined, but too small to perform subgroup analysis. In particular, the proportion of diabetic patients was low among interventional patients in Norway during this era. In our observational study unrecorded risk factors or angiographic findings prohibiting one of the revascularization procedures may have been present at baseline. Differences in baseline features were however adjusted for, and the outcome comparison between the CABG and PCI groups is relevant as the study population to some extent reflects current patient selection and treatment practice in a real-life setting (2,7,27). Innovations in PCI technology during the last decade have not reduced mortality in non-acute patients (28), and veingraft disease after CABG is still an unresolved issue (19,20). RCT's comparing CABG and PCI typically include only 5–10% of the eligible patients, and tend to exclude high-risk patients. Our comparison of CABG versus PCI in unselected patients is further justified by the fact that the differences in baseline characteristics in the two groups were modest, and a large proportion of the patients with extensive CAD were treated with PCI.

The recorded study events are based on a near complete registration of hospitalizations and surgical procedures for a well-defined study population. The community based selection limits referral bias, and ascertained that only first-time revascularized patients were included. The verification of the cause of death may be difficult to ascertain especially for those occurring out of hospital. Causes listed in the death certificates may be non-conclusive, and less diagnostic tests may be performed in terminally ill or aged patients. In our experience, however, the vast majority of death could be attributed to either cardiac or non-cardiac causes.

Conclusion

This study demonstrates that the 10 year risk for cardiac death after PCI or CABG in unselected patients is comparable after adjustments of competing risks. In the very long-term, the manifestation of progression of CAD might differ between patients initially revascularized with either CABG or PCI.

Acknowledgements

We thank Janne Dyngeland RN, Department of Heart Disease, Haukeland University Hospital and Alf Aksland, The IT department of Western Norway Health Authority who contributed to the study data extraction and verification process. The study was supported by grants from the Western Norway Health Authority and the Western Norway Cardiology Society. There are no conflicts of interest to be declared.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.