Results

Over the period studied, 4775 PTCAs and 5691 CABGs were undertaken; 576 of the CABGs involved additional procedures such as valve replacement during the same operation and were excluded from further analysis. Of the 4775 PTCAs, 2417 (51%) included insertion of one or more stents. Of the 5115 CABGs undertaken as isolated procedures, 265 (5%) involved a single graft, 1284 (25%) two grafts, and 3566 (70%) three or more grafts. Three hundred and fifteen PTCAs (7%) were undertaken as primary or rescue procedures or for postinfarction angina, 1688 (35%) were performed for unstable angina, and 2089 (43%) for stable angina. By contrast, 3418 CABGs (67%) were undertaken for stable angina (p < 0.0001). Only 149 patients (3%) required intravenous nitrates before CABG, compared with 794 (17%) before PTCA (p < 0.0001). Seven hundred and two PTCA patients (15%) were on angiotensin converting enzyme (ACE) inhibitors, compared with 1142 CABG patients (22%) (p < 0.0001). PTCA was more likely to be undertaken as an emergency or urgent procedure (table1).

The median age for PTCA was 61 years (interquartile range (IQR) 53 to 67) compared with 63 years for CABG (IQR 57 to 68) (Mann–Whitney U test, p < 0.0001). Significantly more PTCA patients were women (table 1). CABG patients had evidence of worse cardiovascular disease in terms left ventricular function and the presence of left main stem stenoses or triple vessel disease (table 1). Comorbidity and other risk factors, such as obesity, cigarette smoking, hypertension, severe respiratory disease, and cerebrovascular disease, were also more prevalent among CABG patients (table 1).

Previous attempts at coronary revascularisation were more common before PTCA procedures. Of the 4775 PTCAs, 513 (11%) had been preceded by CABG and 695 (15%) by a previous attempt at PTCA. The corresponding figures for CABG were 154 (3%) and 111 (2%), respectively. Overall, 1021 PTCAs (21%) and 253 CABGs (5%) followed at least one previous attempt at some form of revascularisation (p < 0.001).

Periprocedural myocardial infarction was reported for 47 PTCAs (1%) and 88 CABGs (2%) (p < 0.01). However, this difference was not significant after adjusting for case mix. One hundred and ninety one CABG patients (4%) had delayed closure or reopening of their sternal wounds. Thirty PTCA patients (0.6%) required urgent cardiac surgery. Periprocedural death occurred in 15 PTCAs (0.3%) and 166 CABGs (3.2%) (p < 0.0001).

On univariate analysis, patients who underwent PTCA were less likely to die within two years but more likely to be readmitted for ischaemic heart disease or undergo further angiography or revascularisation (table 2). The overall likelihood of suffering any further event, whether death, readmission, coronary angiography or revascularisation, was higher following PTCA (table 3). Restriction of the analyses to elective procedures undertaken on patients who did not have a significant left main stem stenosis and had not previously undergone revascularisation reduced the frequency of end points in both groups but not the direction of the differences.

After adjustment for age, sex, left ventricular function, the number of arteries with significant stenoses, operative urgency, and comorbidity, CABG was associated with higher overall case fatality, a reduced likelihood of readmission, repeat angiography, and further revascularisation (table 4), and a lower overall risk of suffering any event (table 4).

The presence of diabetes mellitus was associated with poorer survival. On subgrouping, the risk of dying up to two years was significantly higher following CABG than PTCA in non-diabetic patients. By contrast, it was lower in diabetic patients but this did not reach significance (hazard ratio 0.59, 95% confidence interval 0.32 to 1.09).

Discussion

CABG and PTCA have been used to treat coronary artery disease for more than 30 years and 20 years, respectively. Various randomised controlled trials in the early 1990s1-6 and subsequent meta-analyses9 ,10 concluded that there was no significant difference in overall survival, but PTCA was more likely to be followed by symptom recurrence, restenosis, and repeat revascularisation. Our ability to generalise the results of trials to everyday practice is limited by their use of restrictive and varying selection criteria.

Observational data have been published from two registries in the USA.10 ,11 The case mix of patients in the Scottish register differs greatly from these series. In the New York register, 40% of all revascularisation procedures were performed for single vessel disease and 32% for triple vessel disease.11Exclusion of left main stem stenoses from our own data produces corresponding figures of 23% and 45%, respectively. The differences in case mix are even greater for PTCA. In Scotland, 43% of PTCA patients had single vessel disease and 19% triple vessel disease. This compares with 71% and 4%, respectively, in New York.

These differences may, in part, reflect a higher threshold for revascularisation in Scotland owing to a higher incidence of coronary artery disease combined with a lower intervention rate. The difference may also reflect trends over time. The Duke University data relate to 1984–1990,10 the New York data to 1993–1995,11 and our own data to 1997–1999. Worsening case mix over time has been reported for both PTCA12 and CABG.13

Despite our worse case mix, our in-hospital complication rates following PTCA compared favourably with previously published results. Only 0.6% of patients required urgent referral for cardiac surgery, compared with 4.3% in the meta-analysis by Pocock and colleagues.9 Similarly, our perioperative mortality rate of 0.3% is lower than the rate of 1.0% reported by Pocock and associates,9 and comparable with the figure of 0.4% in the New York register,11 despite their better case mix. Pocock and colleagues reported a one year survival rate following PTCA of 2.9%.9 Our figures of 2.8% in all procedures and 1.5% in the subgroup suggest that medium term survival is at least as good and may be superior. Previous studies suggested that around one third of PTCA patients required further revascularisation within one year,9-11 compared with 14.1% in our study.

Our favourable results following PTCA despite a poorer case mix may in part reflect increased use of stenting. Stents were used in 12% of PTCAs performed in New York between 1993 and 1995.11 Lee and colleagues reported an increase from 4% to 46% between 1993/1994 and 1995/199612 and in our own study they were used in 51% of cases between 1997 and 1999. Elective stenting can reduce the risk of restenosis and avoid the need for repeat revascularisation.7 ,8 Bail out stenting may reduce the need for urgent surgery by avoiding dissection or acute vessel closure following PTCA. Two ongoing trials will determine whether stenting significantly improves survival following PTCA.14 ,15

On subgroup analysis, 4.4% of CABG patients were dead at one year. The meta-analysis by Pocock and colleagues reported an overall figure of 2.3%.9 However, this included trials restricted to patients with single vessel disease who account for only 6% of our CABGs. Two of the component trials reported figures of 4.7%6 and 5.1% in patients with multivessel disease.3 Therefore, our results appear to be in line with previously published figures. In our study 0.6% of CABG patients required further revascularisation within one year compared with 3% in previously published studies.9-11

As with previous studies, our results showed significantly higher repeat revascularisation rates after PTCA than after CABG. The number of arteries with significant stenoses does not necessarily equate to the number of arteries which are tackled or successfully revascularised during PTCA or CABG, and more complete revascularisation may be easier to achieve by surgery. Subsequent revascularisation procedures are not necessarily directed at a lesion operated on during the index procedure. Repeat revascularisation may be undertaken because of restenosis or disease progression or to tackle multiple lesions as a staged procedure. Technical improvements cannot be expected to affect those procedures required because of progression of the underlying atherosclerotic disease elsewhere in the coronary arterial tree. Conversely, not all patients who experience restenoses undergo further revascularisation procedures and there may be a bias between the procedures in terms of threshold for reintervention. The higher risk of perioperative death associated with redo surgery may result in a greater reluctance to undertake further revascularisation in those who have previously undergone CABG than in those who have undergone PTCA.

Our finding of better overall survival following PTCA reflected better survival in the 88% of patients who did not have diabetes mellitus. Among those with diabetes mellitus, CABG was associated with non-significantly lower odds of dying. Previous studies have shown that diabetic patients have better medium term survival following CABG than following PTCA.16-18

The principal use of our study is to provide contemporary information on outcomes, as previously published results have been rapidly outdated by technical and therapeutic advances. Although we have compared overall outcomes between PTCA and CABG, the inherent limitations of using observational data for this purpose must be acknowledged. Overall results depend on patient selection which, as demonstrated, varies considerably between countries. Although some patients are eligible for both procedures, some are not. In addition to analysing all procedures, our analyses were repeated on a subgroup of patients which excluded patients who had had previous revascularisation procedures, those who had left main stem stenoses, and those who underwent urgent or emergency procedures. This was in part to provide a more meaningful comparison with previous studies, and in part to exclude patients who were not eligible for both procedures. Even restriction to a subgroup does not ensure that the two treatment groups are truly comparable. We have adjusted for case mix in so far as this was possible, but statistical adjustment within an observational study is never as robust as a randomised trial. Also, as with all registries, data completeness can always be improved.

Several factors influence the choice of PTCA or CABG in individual patients who are eligible for both procedures. Patient preference, availability, age, and waiting time are inevitably considered. However, the principal determinant is the relative outcomes of these procedures in terms of survival and the need for further interventions. The balance of risks and benefits is not the same in all patients. Previous studies suggested that CABG is associated with better survival in those with triple vessel disease and two vessel disease, which includes stenosis of the proximal left anterior descending artery, but poorer survival in those with single vessel disease.10 ,11Therefore it would be inappropriate to interpret our finding that PTCA is associated with better overall survival as supporting the choice of PTCA in all patients. The relative merits of these procedures must continue to be judged on an individual basis.