OUP user menu

Patients enrolled in coronary intervention trials are not representative of patients in clinical practice: results from the Euro Heart Survey on Coronary Revascularization

Marjo Hordijk-Trion, Mattie Lenzen, William Wijns, Peter de Jaegere, Maarten L. Simoons, Wilma J.M. Scholte op Reimer, Michel E. Bertrand, Nestor Mercado, Eric Boersma
DOI: http://dx.doi.org/10.1093/eurheartj/ehi731 671-678 First published online: 19 January 2006

Abstract

Aims Revascularization in patients with coronary artery disease changed over the last two decades, favouring the number of patients treated by means of percutaneous coronary interventions (PCI) when compared with coronary artery bypass grafting (CABG). Many randomized controlled trials (RCTs) have been performed to compare these two competing revascularization techniques. Because of the strict enrolment criteria of RCTs in which highly selected patients are recruited, the applicability of the results may be limited in clinical practice. The current study evaluates to what extent patients in clinical practice were similar to those who participated in RCTs comparing PCI with CABG.

Methods and results Clinical characteristics and 1-year outcome of 4713 patients enrolled in the Euro Heart Survey on Coronary Revascularization were compared with 8647 patients who participated in 14 major RCTs, comparing PCI with CABG. In addition, we analysed which proportion of survey patients would have disqualified for trial participation (n=3033, 64%), aiming at identifying differences between trial-eligible and trial-ineligible survey patients. In general, important differences were observed between trial participants and survey patients. Patients in clinical practice were older, more often had comorbid conditions, single-vessel disease, and left main stem stenosis when compared with trial participants. Almost identical differences were observed between trial-eligible and trial-ineligible survey patients. In clinical practice, PCI was the treatment of choice, even in patients who were trial-ineligible (46% PCI, 26% CABG, 28% medical). PCI remained the preferred treatment option in patients with multi-vessel disease (57% in trial-eligible and 40% in trial-ineligible patients, respectively, P<0.001); yet, the risk profile of patients treated by PCI was better than that for patients treated either by CABG or by medical therapy. In the RCTs, there was no mortality difference between PCI and CABG. In clinical practice, however, we observed 1-year unadjusted survival benefit for PCI vs. CABG (2.9 vs. 5.4%, P<0.001). Survival benefit was only observed in trial-ineligible patients (3.3 vs. 6.2%, P<0.001).

Conclusion Many patients in clinical practice were not represented in RCTs. Moreover, only 36% of these patients were considered eligible for participating in a trial comparing PCI with CABG. We demonstrated that RCTs included younger patients with a better cardiovascular risk profile when compared with patients in everyday clinical practice. This study highlights the disparity between patients in clinical practice and patients in whom the studies that provide the evidence for treatment guidelines are performed.

  • Euro Heart Survey
  • Randomized controlled trials
  • PCI
  • CABG

Introduction

Cardiovascular diseases are the major cause of mortality in the Western world, and it is expected that this will remain so during the foreseeable future.13 Treatment of patients with coronary artery disease (CAD) includes risk factor management, drug therapy, and revascularization techniques. During the last decades, changes in revascularization techniques were observed, favouring the number of patients treated by means of percutaneous coronary interventions (PCI) when compared with the more invasive coronary artery bypass grafting (CABG). As these competitive techniques are both feasible in many patients, randomized controlled clinical trials (RCTs) have been performed in order to compare the two procedures. On the basis of these RCTs, the results of registries, and consensus of experts, international societies developed guidelines in order to support physicians in clinical decision-making.47 In these guidelines, RCTs are valued as the highest level of evidence in the ranking order. It should be noted, however, that RCTs have strict enrolment criteria. Consequently, patients who participate in trials may represent a selected group of patients that is poorly representative of the majority of patients treated in routine clinical practice.8 This may complicate the applicability of the results of clinical trials in everyday practice and limit the generalizability of recommendations.9

In the current study, we aimed to evaluate to what extent patients in clinical practice were similar to those who participated in RCTs comparing CABG with PCI. Patient outcome was compared as well between the selected treatment options.

Methods

We performed a comparison between participants in RCTs and patients enrolled in the Euro Heart Survey on Coronary Revascularization (EHS-CR). Details of the survey were published previously.10 Briefly, between September 2001 and March 2002, a total of 5619 patients from 130 hospitals throughout 31 countries belonging to the ESC were included in this survey. All consecutive patients entering the catheterization laboratory were screened, and patients with >50% diameter stenosis in at least one major epicardial vessel were asked to participate. Data were collected on medical history, demography, clinical, haemodynamic, and angiographic status and sent by Internet connection to a central database located at the European Heart House. A follow-up was performed at 1-year (median 11–13 months). Follow-up information, including vital status, was available in 83%, as 14 hospitals (11%) were not able to provide follow-up information. The survey was approved by the relevant national authorities.

Selection of trials

We intended to identify all major randomized phase III clinical trials of CABG vs. PCI, which were published in the English language during 1980–2005, using the Medical Subject Heading terms ‘angioplasty, transluminal, percutaneous coronary,’ ‘coronary artery bypass,’ ‘RCT,’ and ‘comparative study,’ as was used in a recent meta-analysis by Hoffman et al.11 We identified 15 major trials comparing initial strategies of PCI and CABG. As the AWESOME trial12 included patients who would have been excluded in the other trials (i.e. patients with severe heart failure or very recent myocardial infarction), this trial was excluded from the analysis. Table 1 shows the characteristics of 14 trials with a total of 8647 patients who were selected for this analysis. In 10 of these trials, only patients with multi-vessel coronary disease were included, whereas in one trial patients with single- or multi-vessel disease could be enrolled.8,1322 In the remaining three trials, only patients with single-vessel disease were included.2325 Medical therapy alone was added to the two invasive treatment options in one trial.22,24 The tabulated patient characteristics, as presented in the main publication article of the separate clinical trials, were compiled in an electronic database and the data were pooled. In case of a discrepancy between the text of the manuscript and the table, we used the values as shown in the tables.

View this table:
Table 1

Characteristics of the major coronary intervention trials comparing CABG with PCI

StudyNo. of patientsPublication yearVessel diseaseTrial exclusion criteriaOutcome (CABG vs. PCI, respectively)
Follow-up (years)Mortality (%)Non-fatal MI (%)Repeat revascularizations (%)P-value(death/MI)
ERACI131271993multiLeft main disease, EF≤35%, valvular disease10–32–33–320.17
RITA1410111993multi/singlePrior PCI, prior CABG, left main disease, valvular disease2.54–35–77–310.50
Lausanne231341994singlePrior PCI, prior CABG, EF≤50%2.52–03–123–25nsa
EAST153921994multiPrior PCI or CABG, left main disease, EF≤25%, PCI or CABG unsafe36–720–1513–540.81
GABI163591994multiPrior PCI, prior CABG, age ≥75 years, left main disease16–37–43–39<0.05b
CABRI810541995multiPrior PCI or CABG, age ≥76 years, left main disease, EF≤35%, valvular disease13–413–660.30c
MASS242141995singlePrior PCI, prior CABG, left main disease, left ventricular dysfunction, valvular disease31–11–30–29nsa
BARI1718291996multiPrior PCI or CABG, age ≥80 years, left main disease, valvular disease, unsuitable for CABG or PCI5.411–1420–218–540.84
Toulouse181521997multiPrior PCI, prior CABG, left main disease, valvular disease, contraindication to CABG or PCI511–131–59–290.69c
SIMA251212000singlePrior PCI, prior CABG, EF≤45%2.43–27–70–240.90
ERACI II194502001multiPrior stenting, PCI in last year, prior CABG, older age, EF≤35%1.58–36–25–17<0.05b
ARTS2012052001multiPrior PCI, prior CABG, concomitant valve surgery13–34–54–170.71
SoS219882002multiPrior thoracotomy, prior coronary revascularization, valvular disease, EF≤30%22–58–56–210.80
MASS II226112004multiPrior PCI, prior CABG, valvular disease, left main, contraindication to CABG or PCI, EF<40%14–52–81–13<0.05b

EF, ejection fraction.

aActual P-value not given, but difference stated as non-significant.

bActual P-value not given, but difference stated as significant.

cOnly mortality.

All cardiac events include coronary angiography, revascularization, angina.

Selection of patients

Patients enrolled in the EHS-CR were considered eligible for the comparison between trial participants and clinical practice unless the primary diagnosis was ST-elevation myocardial infarction (n=4713), as these patients were excluded from all trials comparing CABG with PCI. In addition, we analysed which proportion of survey patients would have been disqualified for participation in a coronary intervention trial, aiming at separating trial-eligible from trial-ineligible patients in our clinical practice population. Because all patients in this survey had at least one >50% diameter stenosis in a major epicardial vessel, we considered patients as trial-eligible if no major exclusion criteria were observed. As Table 2 clearly reveals, we only selected the most important exclusion criteria. If we would use the enrolment criteria of the individual trials, which we were able to identify in our database, ∼11–25% of the survey patients would be eligible for participation in the individual trials. However, by using only major exclusion criteria, we aimed at comparing an average of patients as seen in clinical practice with RCT participants, decreasing the risk of identifying a highly selective group of survey patients. It should be noted, however, that this selection is, by necessity, crude. We would also like to state that defining patients from clinical practice as trial-eligible or trial-ineligible was done in retrospect.

View this table:
Table 2

Major exclusion criteria, used in RCTs8,13–25 applied to patients included in this study

Exclusion criteriaRCTsEHS-CR
Number of trialsNumber of patients (%)
Age ≥80 years (A)2195 (4)
Prior PCI (B)131054 (22)
Prior CABG (C)13566 (12)
Ejection fraction ≤35% (D)5294 (6)
Left main disease (E)8427 (9)
Valvular heart disease (F)8319 (7)
CABG or PCI not suitable (G)141757 (37)
Three of the above (B,C,G)142585 (55)
Five of the above (B,C,D,E,G)142841 (60)
Any of the above (A–G)143033 (64)

Statistical analyses

Differences in baseline characteristics between survey patients and pooled intervention trials were analysed by χ2 tests, using EpiInfo (version 5.0). To analyse continuous variables, we used the highest standard deviation presented in the selected coronary intervention trials. Comparison between trial-eligible and trial-ineligible survey patients was analysed by χ2 tests, Fisher's exact tests or t-tests as appropriate, using SPSS for Windows (version 12.0). Data are presented as percentage and 95% confidence interval (95% CI), unless indicated otherwise. A P-value of less than 0.001 was considered significant (two-sided).

As we acknowledged the fact that patients who were treated medically differ considerable from those who were intended to undergo revascularization, we repeated all analyses excluding medically treated patients. As the results of these analyses were highly consistent, we report our original choice on the basis of all three treatment groups (i.e. PCI, CABG, and medical).

Results

In total, 8647 patients enrolled the identified RCTs that compared CABG with PCI. The duration of follow-up varied between 1 and 5.4 years. As Table 1 visualizes, the competitive procedures differed hardly in survival rates and non-fatal myocardial infarction. The need for repeat revascularization, however, differed largely between the two revascularization techniques, favouring CABG.

On the basis of the selection of major exclusion criteria, we estimated that almost two-thirds (n=3033, 64%) of the 4713 EHS-CR patients would have disqualified for trial participation (Table 2). Most frequently observed obstacles for trial participation were prior revascularization and non-suitability of the diseased vessels for CABG and/or PCI. The remaining 36% (n=1680) was considered trial-eligible.

There were important differences in clinical and angiographic characteristics between participants of the identified RCTs and patients as seen in clinical practice (Table 3). In general, patients in clinical practice were older and more often had comorbid conditions when compared with trial participants. Interestingly, unstable angina was more frequent in trial populations as well as in trial-eligible patients. As most RCTs included only patients with multi-vessel disease, we identified survey patients with multi-vessel disease and compared this major subgroup (65%) with those who participated in RCTs. On top of the observed differences between survey patients and trial participants, patients enrolled in the EHS-CR with multi-vessel disease were more likely to have three-vessel disease when compared with trial participants.

View this table:
Table 3

Characteristics of patients enrolled in randomized trials when compared with EHS-CR patientsa

CharacteristicsPooled trialsNumber of trialsEHS-CRP-valuebEHS-CR MVDP-valuec
Patients (n)86471447133081
Mean age, years (SD)60±121263±11<0.0164±10<0.01
Male gender (%, 95%CI)77 (76–78)1476 (75–77)0.2977 (76–79)0.40
History (%, 95%CI)
 Hyperlipidaemia56 (54–57)867 (66–69)<0.00168 (67–71)<0.001
 Hypertension47 (46–48)1262 (60–63)<0.00163 (62–66)<0.001
 Diabetes20 (19–20)1324 (22–26)<0.00127 (25–28)<0.001
 Peripheral vascular disease8 (7–9)513 (12–14)<0.00114 (13–16)<0.001
 Renal failuren.a.144 (4–5)6 (5–7)
 Prior myocardial infarction44 (43–45)1341 (39–42)<0.00144 (42–46)0.91
 Congestive heart failuren.a.1417 (16–19)20 (19–21)
 Cerebrovascular disease4 (4–5)48 (7–9)<0.0019 (8–10)<0.001
Presentation (%, 95%CI)
 Unstable angina42 (41–43)835 (34–37)<0.00135 (33–37)<0.001
 Stable anginan.a.1562 (61–64)63 (61–65)
   CCS class 3–4 (%)42 (41–43)735 (34–37)<0.00139 (37–41)0.006
Chronic medication (%, 95%CI)
 Antiplatelet agents83 (81–84)485 (84–86)0.0286 (84–87)0.006
 ACE-I/angiotensin-II blockers23 (20–26)250 (49–51)<0.00152 (50–53)<0.001
 Bèta-blocker68 (66–69)669 (67–70)0.3570 (68–71)0.12
 Calcium-antagonist65 (63–66)628 (26–29)<0.00128 (27–30)<0.001
 Nitrates61 (59–63)656 (54–57)<0.00159 (57–61)0.30
 Statinesn.a.52 (51–54)52 (50–54)
Coronary angiogram (%, 95%CI)
 Diseased left anterior descending73 (72–75)1174 (72–75)0.7386 (84–87)<0.001
 Diseased right coronary artery68 (66–69)462 (60–63)<0.00178 (78–81)<0.001
 Diseased left circumflex69 (67–71)459 (57–60)<0.00179 (77–80)<0.001
 Diseased left main stem2 (1–3)29 (8–10)<0.00114 (13–15)<0.001
Number of diseased vessels<0.001
 111 (10–12)1435 (33–36)0
 253 (52–55)1431 (29–32)47 (45–49)
 335 (34–36)1434 (33–36)53 (51–54)
Treatment (preference/option) (%)
 PCI495446 (44–47)
 CABG482333 (31–35)
 Medical only32321 (20–23)

MVD, multi-vessel disease.

aPatients, admitted with ST-elevation infarction were excluded.

bPooled trials vs. EHS-CR population.

cPooled trials vs. EHS-CR population with MVD.

PCI was clearly the preferred treatment in clinical practice, whereas CABG and medical treatment were evenly distributed in RCTs, the choice of treatment being dictated by randomization. In contrast to the observed differences, it is important to note that the under-representation of women, as seen in RCTs, was also observed in clinical practice.

Within the study population, a comparison between trial-eligible and trial-ineligible patients was made, and it revealed that patients who would be excluded from trial participation had a worse clinical profile when compared with trial-eligible patients (Table 4). Trial-ineligible patients were older, more likely to have a comorbid condition, and had a higher proportion of diseased coronary arteries including left main stenosis. By means of the EuroSCORE,26 trial-ineligible patients would have a higher estimated peri-procedural risk. These trial-ineligible patients were treated more often surgically or medically when compared with trial-eligible patients, though PCI remained the treatment of choice. When excluding patients with single-vessel disease from the analyses, PCI remained the preferred treatment option in 57% of trial-eligible and 40% of trial-ineligible patients (P<0.001). CABG did not differ between patients with multi-vessel disease in the two subgroups (31 vs. 34%, respectively). The majority of patients with left main disease (n=427) was treated by means of CABG (n=239), whereas 96 patients were treated percutaneously and 92 did not undergo revascularization. Of the 96 patients undergoing PCI, the left main was not protected by means of a prior CABG in 45 patients (47%).

View this table:
Table 4

Comparison between trial-eligible and trial-ineligible EHS-CR patients

Patient characteristicsTrial-eligible EHS-CR patientsTrial-ineligible EHS-CR patientsP-value
Patients (n)168030330
Mean age, years (±SD)62±1064±11<0.001
Male gender (%, 95%CI)74 (72–76)77 (75–78)0.052
History (%, 95%CI)
 Hyperlipidaemia66 (64–69)68 (67–70)0.23
 Hypertension59 (57–61)63 (61–65)0.006
 Diabetes21 (19–23)26 (25–28)<0.001
 Peripheral vascular disease10 (8–11)14 (13–15)<0.001
 Renal failure3 (2–4)5 (4–6)<0.001
 Prior myocardial infarction32 (30–35)46 (43–47)<0.001
 Congestive heart failure12 (11–14)20 (19–22)<0.001
 Cerebrovascular disease7 (6–9)8 (7–9)0.31
Presentation (%, 95%CI)
 Unstable angina41 (39–44)32 (30–34)<0.001
 Stable angina57 (55–59)65 (64–67)<0.001
  CCS class 3–4 (%)35 (31–37)36 (34–38)0.21
Chronic medication (%, 95%CI)
 Antiplatelet agents85 (83–86)85 (84–86)0.76
 ACE-inhibitors/angiotensin-II blockers45 (42–47)53 (51–55)<0.001
 Bèta-blocker68 (66–71)69 (67–70)0.69
 Calcium-antagonist25 (23–27)29 (28–31)0.003
 Nitrates54 (52–56)56 (55–58)0.11
 Statines46 (44–49)55 (44–49)<0.001
Coronary angiogram(%, 95%CI)
 Diseased left anterior descending73 (71–76)74 (72–75)0.78
 Diseased right coronary artery58 (56–61)64 (62–65)<0.001
 Diseased left circumflex54 (52–57)61 (59–63)<0.001
 Diseased left main stem014 (13–15)<0.001
Number of diseased vessels<0.001
 140 (37–42)32 (30–33)
 234 (31–36)29 (27–31)
 326 (24–28)39 (37–41)
RISK SCORE (mean, ±SD)
 EuroSCOREa3.6 ±2.35.0 ±3.0<0.001
Treatment (preference/option) (%)<0.001
 PCI6946
 CABG1926
 Medical only1328

aEuroSCORE was calculated from the available variables.26

Identical to the outcomes in RCTs, the most important difference between PCI and CABG was observed in the lower need for repeat revascularizations in CABG patients (Figure 1). In contrast to the trials, the overall unadjusted 1-year survival differed between PCI and CABG (2.9 vs. 5.4%, P<0.001). As Figure 1 clearly reveals, this survival benefit in PCI patients was only observed in those who would be excluded from trial participation (3.3 vs. 6.2%, P<0.001). In trial-eligible patients, neither treatment had a clear advantage over the other for preventing death.

Figure 1 Outcome at 1 year.

Discussion

This study revealed that participants of RCTs are not representative of patients treated in daily clinical practice. Moreover, only a minority of patients in clinical practice (36%) was potentially eligible for participation in one of the RCTs. These trial-eligible patients had a different clinical profile when compared with RCT participants.

The importance of this finding is related to the fact that RCTs are valued as highest in the hierarchy of evidence that is used in the guidelines and for the formulation of recommendations. That patients enrolled in these trials may not be representative for the general clinical practice27 obviously depends on the inclusion and exclusion criteria. As a result of this and other factors, such as physician preference, 58–96% of the screened patients and 84–98% of the eligible patients are eventually not enrolled.8,13,1517,23,28 As a result, many RCTs are known to have a limited generalizability.9,29

These observations highlight the difficulties experts have in writing guidelines, as well as for clinicians who have to choose the most appropriate treatment for individual patients. In addition to this, observational studies can be useful adjuncts to RCTs, as they are more likely to reflect clinical practice, and consequently can provide information on subpopulations that were disregarded in trials, as well as on the effectiveness of evidence-based treatments in routine practice.

Regarding the observed differences between RCT participants and patients enrolled in the EHS-CR, we would like to address the under-representation of women (25%) in both groups (i.e. pooled RCTs and the EHS-CR). This observation erroneously suggests that men are more likely to have CAD. In fact, ∼50% of all patients with CAD are known to be women, although women are approximately 10 years older.30 Apparently, and in line with previous findings, women are less likely to undergo invasive investigations and consequently revascularization.31,32

As most trials included only patients with multi-vessel disease, it was not surprising to observe that patients with two-vessel disease dominated in the RCTs. In clinical practice, the number of diseased coronary vessels was more evenly distributed. Although this suggests a more extensive CAD in RCT participants, it should be noted that the left main disease was seen more often in clinical practice. Regarding demographics, risk factors, and comorbid conditions, patients in clinical practice were somewhat older and had a worse clinical profile when compared with the trial patients. In addition to this, a high proportion of patients had a history of heart failure. Although no numbers were given regarding heart failure patients in the selected trials, it is unlikely that they approximate to the observed proportion as seen in the EHS-CR. This study therefore provides a valuable perspective on the disparity between patients in clinical practice and patients in whom the studies that provide the evidence for treatment guidelines are performed. It should be noted, however, that the AWESOME trial, which was excluded from the selection of trials, has specifically addressed the impact of PCI or CABG on patients with high-risk clinical characteristics and thus trying to overcome the gap between clinical practice and RCTs. In the AWESOME trial, as in the other selected RCTs, mortality rates were similar between the two treatment groups.12

In clinical practice, treatment preference was unmistakable in favour of PCI (54%), whereas surgically and medically treated patients were comparable. Although PCI seemed to be the treatment of choice, we observed a shift towards fewer percutaneous interventions in those who did not qualify for trial participation. In contrast to RCTs, the choice of treatment in clinical practice is not dictated by randomization, but influenced by the weight clinicians and patients gave to a variety of factors. In this respect, the major advantages of PCI when compared with CABG (e.g. relative ease of use, no need for general anaesthesia, thoracotomy, and extracorporeal circulation) seemed superior to the disadvantages (e.g. higher risk of early restenosis, and lower ability to achieve complete revascularization).

Not surprisingly, as observed in the selected trials,8,1221,33 the need for repeat revascularization in clinical practice was considerably higher in the PCI group when compared with CABG. However, with respect to irreversible adverse events such as death, myocardial infarction, and cerebrovascular events, this study only partly supports the conclusion of most RCTs that PCI and CABG patients have similar outcome. Indeed, trial-ineligible patients undergoing CABG had a worse survival when compared with trial-ineligible patients undergoing PCI. This difference is only partly explained by the fact that patients with unprotected left main stenosis were treated by CABG rather than by PCI10 in line with the guidelines, advocating CABG in high-risk patients.5,7 Previous real-life studies likewise showed that patients selected for CABG have more extensive disease, more comorbidities, higher procedural risk, and therefore unadjusted event rates tend to be worse with CABG than with PCI. However, propensity analysis showed that under those circumstances CABG actually improves outcome.34,35 The implications are that trial results are indeed confirmed in real life, but only in trial-eligible patients. Clearly, this does not appear to be the case in trial-ineligible patients.

Limitations

The limitations of this study are those inherent to observational surveys involving voluntarily participating hospitals. Although we have attempted to include a wide spectrum of hospitals in different countries, almost certainly the results are biased towards better than average practices. The sample size only represents a small fraction of all patients admitted in catheterization laboratories throughout Europe during the study period. Patient inclusion was consecutive in all participating sites, therefore reflecting the ongoing clinical practice. Consequently, even though the results of this survey reflect the real world better than the RCTs, it should be noted that this might still be too far away from daily practice. Regarding the selection of trial-ineligible patients, we focused on the major exclusion criteria of the pooled trials. Obviously, in retrospect, we could not trace the complete decision-making process leading to trial-(in)eligibility of patients enrolled in this survey.

Conclusions

The present study revealed that only a minority of patients enrolled in the EHS-CR would have qualified for participation in an RCT comparing PCI and CABG. Furthermore, we demonstrated that trial participants had a better clinical profile when compared with patients in everyday clinical practice. Nevertheless, the less invasive approach with percutaneous intervention was the preferred treatment over surgical and medical treatment.

Conflict of interest: none declared.

References

View Abstract