Author + information
- Received March 3, 2005
- Revision received October 23, 2005
- Accepted November 21, 2005
- Published online September 5, 2006.
- Alexandre C. Pereira, MD⁎ (, )
- Neuza H.M. Lopes, MD, PhD,
- Paulo R. Soares, MD, PhD,
- Jose Eduardo Krieger, MD, PhD,
- Sergio A. de Oliveira, MD, PhD,
- Luiz A.M. Cesar, MD, PhD,
- Jose A.F. Ramires, MD, PhD and
- Whady Hueb, MD, PhD
- ↵⁎Reprint requests and correspondence:
Dr. Alexandre C. Pereira, Heart Institute, Av. Dr. Enéas de carvalho Aguiar, 54, 10 Andar, Bloco 2, São Paulo, Brazil.
Objectives This study examined the predictive power of clinical judgment in the incidence of cardiovascular end points in a group of individuals with multivessel coronary artery disease (CAD) followed up in the MASS II (Medicine, Angioplasty, or Surgery Study II).
Background There is still no consensus on the best treatment for patients with stable multivessel CAD and preserved left ventricular function.
Methods Preferred treatment allocation was recorded for each of the 611 randomized patients in the MASS II trial before randomization. We have divided our sample according to physician-guided decision and randomization result into two categories: concordant or discordant. The incidence of the points of cardiac death, myocardial infarction, and refractory angina was compared between concordant and discordant patients.
Results The number of concordant individuals was 292 (48.2%), and this number was not different between the three studied treatments (p = 0.11). A significant difference (p = 0.02) was disclosed because of an increased incidence of combined end point events in discordant patients. In the multivariate Cox hazard model, clinical judgment was a powerful predictor of outcome (p = 0.01) even after adjustment for other covariates. The main subgroup explaining this difference was a significant shift toward a worse outcome in the subgroup of discordant patients who underwent percutaneous coronary intervention (PCI) (p = 0.003).
Conclusions Angiographic variables were more often used in making clinical decisions regarding PCI than clinical variables, and the only independent predictor of concordance status in the PCI group was the number of diseased vessels (p = 0.01). Our data are a reminder that physician judgment remains an important predictor of outcomes.
Clinical judgment in chronic illness involves knowledge of the natural history of the disease, the ability to assess the validity of therapeutic claims, and a means of applying what is known about the individual patient. In this scenario, even when different treatments result in equal outcomes for a particular overall population, differential effects may be present in subsets of patients, such that physician judgment could result in an optimal allocation of treatments.
Despite the progress in coronary artery disease (CAD) risk stratification, no consensus exists about the best treatment for patients with stable multivessel CAD and preserved left-ventricular function. Recently, our group published the 1-year follow-up of the MASS II (Medicine, Angioplasty, or Surgery Study II) (1). In this study, compared with angioplasty, medical treatment for multivessel CAD was associated with a lower incidence of short-term events and a reduced need for additional revascularization procedures. Surgical revascularization was superior to medical treatment for eliminating anginal symptoms. All three therapeutic regimens yielded relatively low rates of cardiac-related deaths. Nevertheless, a clear answer to the best treatment modality in these patients remains unknown.
The evaluation of outcomes of physician-guided versus random assignment of therapeutic modalities for chronic CAD has been the subject of several studies (2,3). However, previous work has relied on retrospective or registry data. Here, we examine the predictive power of clinical judgment in the incidence of mortality and cardiovascular end points in this group of individuals with multivessel CAD prospectively followed up in the MASS II trial. In addition, we examine what biologic variables are most commonly used by physicians in deciding what therapy should be offered to patients with stable multivessel CAD and no left ventricular dysfunction.
The participants of this study were selected from the prospective, randomized, controlled clinical trial MASS II. The MASS II trial was designed to compare medical treatment (MT), angioplasty/stent placement (PCI), and surgical myocardial revascularization (CABG) in patients with stable multivessel CAD with preserved left ventricular function (1). Briefly, 2,076 candidates who were indicated for myocardial revascularization were evaluated between May 1995 and May 2000. Of these, 611 patients were eligible and met all entry criteria to be randomly assigned to one of the three therapeutic groups: MT (n = 203), PCI (205), or CABG (203).
The inclusion criteria were symptomatic multivessel coronary disease, preserved left ventricular function, and the presence of coronary lesions (>70% of stenosis) amenable to angioplasty. Treatment was defined according to the intention-to-treat principle. Randomization date was regarded as the time treatment began. All subjects gave informed consent, and the ethics committee of the Heart Institute of the University of São Paulo-Brazil approved the study. All procedures were performed in accordance with the Helsinki Declaration of 1975 (revised 1983).
The collected demographic and laboratory data included age, gender, ethnicity, angiographic findings, and traditional risk factors such as history of previous coronary events, hypertension, diabetes, body mass index, severity of angina, smoking status, and cholesterol and triglyceride profile. A current smoker was considered one who had smoked more than 10 cigarettes per day for at least 6 months, and ex-smokers were those who had quit smoking for 6 months or more. Because of the study design, nonsmokers were composed of never-smokers and ex-smokers.
Before randomization, the treatment decision based on clinical judgment was collected after the consensus of two cardiologists with known experience in the care of patients with chronic CAD. Cardiologists participating in this phase of the study (N.H.M.L. and W.H.) were clinical cardiologists (i.e., not interventionalists or angiographers) participating in the inclusion phase of the protocol. These individuals are coauthors of the present work. This information was kept confidential and was not used in any other way after randomization.
Data are presented as mean ± SD for continuous variable and as frequencies for categorical variables. We have arbitrarily divided our sample according to the physician-guided decision before randomization, and the randomization result in two dichotomous categories: concordant (patients randomized to the same therapeutic option as chosen by physician-guided decision) or discordant (patients randomized to a different treatment option than chosen by physician-guided decision). Baseline characteristics of patients regarding concordance between physician and randomization were compared. Chi-square tests, ttests, and analysis of variance were used for baseline comparisons. Long-term survival comparisons were conducted for the entire study group and for the subgroups of treatment modalities. Logistic regression was used to estimate the cross-sectional association of this variable with each of the MASS II trial end points (death, myocardial infarction, and recurrent ischemia requiring revascularization), as well as the combined end point after the 1-year follow-up period. Survival curves were calculated with the Kaplan-Meier method, and differences between the curves were evaluated with the log-rank statistic. We assessed the relationship between baseline variables and composite end points by using a Cox proportional hazards survival model. Hazard ratios (relative risks [RR] with 95% confidence intervals [CI]) show the risk of combined events. A value of p < 0.05 was considered significant for comparisons.
Statistical analyses were performed with the statistical package StatView for Windows, version 5.0 (SAS Institute Inc., Cary, North Carolina).
Among the 611 individuals participating in the MASS II trial, 40 (6.5%) had refractory angina requiring revascularization, 45 (7.4%) had myocardial infarction, and 28 (4.6%) deaths occurred during the follow-up period of 1 year. The number of at least one of the combined end points in this population was 90 events (14.7%). We were not able to collect data on physician preference regarding their therapeutic decision on five patients for logistical reasons. There was a tendency of clinicians to select less PCI and surgery and more medical therapy. Physician-guided therapeutic decisions among the three possible groups were as follows: 234 (38.6%) individuals for MT, 192 (31.7%) individuals for CABG, and 180 (29.7%) individuals for PCI. The number of concordant individuals (physician-guided × randomization) was 292 (48.2%), and this percentage was not statistically different among the three study treatments (MT, PCI, and CABG) (p = 0.11).
Sample characteristics regarding concordance status are shown in Table 1.As expected based on randomization, none of the risk factors studied significantly differed between concordant and discordant individuals.
Survival curves were calculated with the Kaplan-Meier method, and differences between curves were evaluated with the log-rank test and are shown in Figure 1.A statistically significant difference (p = 0.02) was disclosed showing an increased incidence of combined end point events in the group of discordant patients, as compared with patients with a concordant option between clinical judgment and randomization (Fig. 1A). The main end point explaining this difference was a significant shift toward a worse outcome regarding refractory angina requiring revascularization (p = 0.007) (Fig. 1B). No difference could be observed regarding mortality (p = 0.85) or incidence of myocardial infarction (p = 0.42). In a multivariate Cox proportional hazards model, clinical judgment was still a powerful predictor of outcome (p = 0.01, RR 0.57, 95% CI 0.36 to 0.89), even after adjustment for age, gender, treatment allocation, hypertension, diabetes, smoking status, and total cholesterol.
Interestingly, stratification of our analysis into different treatment options shows that no differences can be observed for either the medical (p = 0.58) or the surgical (p = 0.64) groups (Fig. 2).A clear difference was found, however, in the subgroup of patients who underwent PCI (p = 0.003). One-year probabilities of event-free survival are shown in Figure 3.In this figure, a direct comparison between concordant (white boxes) and discordant (blue boxes) patients can be made easily. It is interesting to note the reduced survival of individuals randomized to PCI whose clinician’s first choice would be toward CABG or MT.
Next, we proceeded to exploratory analysis, trying to identify the clinical variables most likely used by clinicians in their decisions to choose for or against PCI. We show in Table 2clinical variables regarding concordance status in the group of patients who underwent PCI. None of the clinical variables studied could reliably predict whether clinical judgment would be for or against PCI. In Table 3,we show angiographic variables regarding concordance status in the same group of patients randomized to PCI. Interestingly, different than with clinical variables, angiographic variables apparently were more discriminative of the clinical decision regarding PCI. A significant difference in the number of diseased vessels could be disclosed (p = 0.006), and also a trend toward a higher frequency of type C lesions in the discordant group.
Indeed, in a multiple logistic regression model adjusted for age, gender, hypertension, diabetes, total cholesterol, and smoking status, the only independent predictor of concordance status was the number of diseased vessels (p = 0.01, odds ratio 2.1, 95% CI 1.2 to 3.8) (having three-vessel disease increased the odds of a clinical decision against PCI by 2.1 times).
Another interesting view of the same data is acquired by comparing treatment results stratifying our analysis in 2 groups: individuals with a clinical decision toward PCI and individuals with a clinical decision toward CABG or MT. In Figure 4A,it is interesting that PCI is as good an option for treating multivessel CAD as is medical or surgical treatment in the subgroup of patients in whom PCI would be clinically chosen. A rather different picture (Fig. 4B), and very similar to the results of the 1-year follow-up study of the MASS II trial, is seen in individuals for whom clinical judgment is in favor of medical or surgical therapy. In this subgroup of patients, clearly PCI showed an increased incidence of cardiovascular events (p < 0.0001).
The fact that the clinical decision was collected before randomization in the MASS II trial allows a unique overview of the selection process and outcome in patients with stable multivessel CAD without left ventricular dysfunction. Fundamental to the design of the MASS II trial was the requirement that all patients entered into the randomization would be anatomically suitable for both CABG and PCI. Within this study, no significant difference was noted in the 1-year mortality between MT, PCI, or CABG before or after the adjustment for baseline differences in either the overall population or any of the subgroups analyzed (1). However, a significant difference was observed that showed an increased incidence of combined end points in the subgroups of patients randomized to, but not necessarily clinically chosen for, PCI.
Unlike previous reports (2), subsequent revascularization procedures were significantly more common in the discordant PCI group than in the concordant PCI group. It seemed that physicians could identify those patients who would do well with PCI (concordant PCI patients). The opposite also seemed to hold true: physicians can identify patients not suited for PCI (discordant PCI patients) (Fig. 4).
The single variable most strongly associated with a discordance regarding PCI was the presence of triple-vessel disease. This fact is in agreement with findings of previous studies of clinical decision-making in patients with multivessel CAD. In the BARI (Bypass Angioplasty Revascularization Investigation) registry (3), CABG patients also had more severe coronary angiographic profiles than PCI patients did, with more significant lesions, a higher prevalence of triple-vessel disease, significant proximal left anterior descending coronary artery (LAD) lesions, and diffuse disease. Unlike this study, however, significant proximal LAD lesions and diffuse disease did not seem to influence the clinical decision in a significant manner (3). The same trend was also observed in the EAST (Emory Angioplasty Versus Surgery Trial) registry (2). Nevertheless, this does not mean that individuals with three-vessel disease were not present in our group of individuals whose first clinical decision was in favor of PCI. In fact, even in this group, 48.4% of individuals had three-vessel disease. A similar scenario was also observed regarding plaque morphology. Almost all patients in the MASS II trial had at least one lesion classified as type B2 or C. Although we could not identify any confounding variable that could explain the described association between three-vessel disease and concordance status in three-vessel disease individuals randomized to PCI (data not shown), it is not clear that the improved outcome in concordant patients is etiologically related to three-vessel disease. Alternatively, three-vessel disease may just be an epiphenomena, confounded by more diffuse disease in patients who also have three-vessel disease. The use of severity scores for atherosclerosis burden in future studies may indeed shed light on this issue.
Different from the previously published BARI registry (3), PCI was not the initial revascularization strategy favored by patients and their physicians. Medical treatment was the preferred option for 38.6% of the study individuals. It is interesting to note that physician-guided treatment options were more equitably distributed among the 3 possible therapeutic options for this particular population of patients. This figure is quite different from the one reported by the BARI registry, in which 59% of patients were allocated to receive percutaneous transluminal coronary angioplasty, 31% CABG, and only 10% MT. This fact may be explained by the more uniform subgroup of CAD patients in the MASS II trial (stable, multivessel CAD, with no left ventricular dysfunction), representing a subset of CAD patients in whom treatment allocation is not consensual. Rather than what could be erroneously interpreted as a lack of consensus, these data can be interpreted as an indication of the judicious use of clinical variables in the allocation of patients to a particular treatment modality.
The value of physician-guided treatment allocation is even more impressive if one analyzes these data against the conclusion of the 1-year follow-up of the MASS II trial recently published (1). Different from the whole study population, if only concordant patients are analyzed, no difference between randomized treatments could be disclosed after a 1-year follow-up (Fig. 4). A possible interpretation of these data is that, by using all available variables, physicians are able to assign a particular treatment (including MT or PCI) better than a randomization procedure, and that results from randomized trials involving this population of patients should be interpreted in light of this fact. In addition, our data suggest that physicians are able to define a subgroup of patients with multivessel CAD and preserved left ventricular function in which PCI treatment is as good a treatment option as is CABG or MT.
In previous work, cardiac mortality, as well as total mortality, was usually lower in the registry patients (2,3). Apart from physicians’ and patients’ decisions, this could also have been observed because of unmeasured baseline differences between registry and randomized patients. The design of the current work provides a rather different appraisal of a quite similar situation because all patients were randomized to a particular treatment, and clinical end points are being compared within this population.
In addition, an important strength of the MASS II trial design is the presence of a medically treated group. This subgroup is not present in other important studies dealing with this population of CAD patients (4). It is important to note that MT is not underrepresented in the physician-guided treatment allocation group, a fact that could lead to bias. On the contrary, MT would be the treatment of choice in 38.6% of patients with two-vessel disease, and in 38.7% of patients with three-vessel disease. In addition, concordant patients for MT did not have an increased incidence of cardiovascular end points compared with patients treated with PCI or CABG.
Our study has potential limitations. First, in the study design we define as clinical preference the consensus of two clinical cardiologists with known experience in the care of patients with CAD. This definition does not exclude the possibility of potential selection bias based on their individual backgrounds. This fact may be contemplated when one analyses the distribution of treatment modalities according to clinical preference: 234 for MT, 192 for CABG, and 180 for PCI. The differences regarding concordant percentages across treatment modalities were, however, not statistically significant and do not invalidate the conclusions reached, especially when stratification for treatment modalities was applied. It should also be noted that a direct comparison between randomized and randomizable (i.e., registry) patients might reflect the clinical decisions of a wider number of physicians as they practice in the real world, and this may in many aspects differ from the described approach. Second, one of the possible main reasons for the reported results may reside in the restenosis rate of patients who underwent PCI. This scenario is likely to change with the use of drug-eluting stents, and one could envision a reduced impact of clinical decision in the situation of a patient with a high prior probability of restenosis. Our data, however, are reassuring in that PCI, even without the use of drug-eluting stents, is a good treatment alternative for multivessel CAD patients whenever this treatment is found to be clinically appropriate. Finally, there was an overrepresentation of concordant individuals in our sample. This increased frequency of concordant individuals was present in all treatment groups and could not be explained by any systematic bias we could identify in our study design. It is important to note, however, that this bias toward concordance in our sample is uniformly distributed in the three different treatment groups and does not explain the particularities of our findings regarding PCI-treated patients.
In conclusion, our data suggest that the best treatment option for individuals with multivessel CAD and preserved left ventricular function defined by the physician is a predictor of the incidence of cardiovascular events, mainly because of the need for additional revascularization procedures. Angiographic variables seem to be more commonly used in the decision-making process regarding PCI. Clinical decision-making seems to be able to define a particular subgroup of patients in whom PCI treatment is as good as CABG or MT.
Our data, together with that of others, are a reminder that physician judgment remains an important predictor of outcomes, even in a time when evidence-based medicine is considered the gold standard of medical practice.
The authors thank Dr. Katherine Detre for encouraging the writing of the manuscript and for always enlightening manuscript review and discussion.
- Abbreviations and Acronyms
- coronary artery bypass graft surgery
- coronary artery disease
- high-density lipoprotein
- left anterior descending coronary artery
- low-density lipoprotein
- MASS II
- Medicine, Angioplasty, or Surgery Study II
- medical treatment
- percutaneous coronary intervention
- Received March 3, 2005.
- Revision received October 23, 2005.
- Accepted November 21, 2005.
- American College of Cardiology Foundation
- Hueb W.A.,
- Soares P.R.,
- Gersh B.J.,
- et al.
- King S.B. 3rd.,
- Barnhart H.X.,
- Kosinski A.S.,
- et al.,
- Emory Angioplasty Versus Surgery Trial Investigators
- Feit F.,
- Brooks M.M.,
- Sopko G.,
- et al.,
- BARI Investigators
- Detre K.M.,
- Guo P.,
- Holubkov R.,
- et al.