Clinical Predictors of In-Hospital Prognosis in Unstable Angina : ECLA 3

O. Bazzino, MD, FACC, R. Daz, MD, C. Tajer, MD, C. Paviotti, MD, E. Mele, MD, M. Trivi, MD, A. Piombo, MD, A. Hirschson Prado, MD, E. Paolasso, MD, Servicio de Cardiologa, Hospital Italiano, Buenos Aires, Argentina; for the ECLA Collaborative Group.

Am Heart J13721999

1372

Abstract

Objectives : Because of recent changes in the treatment of unstable angina , we wanted to reassess the short-term prognostic value of clinical and echocardiographic variables .
Methods : This was an observational , prospective study that included 1038 nonselected consecutive patients admitted to coronary care units for unstable angina .
Results : Baseline characteristics were age 60.18 +-16 years , history of prior myocardial infarction in 336 patients ( 32 % ) , and a history of previous angina in 817 patients ( 78.7 % ) .
Angina during the 48 hours before admission was observed in 1004 patients ( 96.7 % ) and ST-segment changes on admission electrocardiogram occurred in 385 patients ( 37 % ) .
In-hospital treatment consisted of nitrates in 81.4 % of patients , aspirin in 88.6 % , b-blockers in 71 % , intravenous heparin in 34.5 % , subcutaneous heparin in 23 % , and angioplasty or coronary artery bypass grafting in 25.1 % .
After admission , angina occurred in 443 patients ( 40.8 % ) , refractory angina in 223 patients ( 21.5 % ) , and death or myocardial infarction in 84 patients ( 8.1 % ) .
At admission , the independent predictors of myocardial infarction or death identified by multivariate logistic regression analysis were ST-segment depression ( odds ratio [ OR ] 2.13 , 95 % confidence interval [ CI ] 1.23 to 3.68 , P = .006 ) , prior angina ( OR 2.23 , 95 % CI 0.98 to 5.05 , P = .05 ) , number of episodes of angina within the previous 48 hours ( OR 1.63 , 95 % CI 0.98 to 2.70 , P = .05 ) , and history of smoking ( OR 0.69 , 95 % CI 0.56 to 0.85 , P = .
004 ) .
Age greater than 65 years ( OR 1.49 , 95 % CI 1.09 to 2.03 , P = 0.03 ) was significantly related to in-hospital death .
The area under the receiver operating characteristic curve for application of this model was 0.59 .
Sensitivity was 80 % with a specificity of only 33 % .
Refractory angina after admission showed a strong relation with an adverse short-term outcome .
Conclusions : With current therapy , clinical and electrocardiographic variables provide useful information about the short-term outcome of unstable angina .
However , this model has low specificity to identify high-risk patients .
Future studies about the incremental value of the new serum markers such as troponin T and C-reactive protein to assist in identification of high-risk patients are necessary .

Introduction

Management of unstable angina has become more aggressive to include full medical therapy with anti-thrombotic and anti-ischemic drugs , often followed by coronary angiography and revascularization procedures .

Much of the published data about the natural history of unstable angina corresponds to selected patients included in randomized clinical trials .
Therefore these patients may not represent the general population admitted to community hospitals .

Unstable angina is also quite heterogeneous , with a wide spectrum of clinical manifestations and prognostic categories .
in addition , most of the data on the evolution and prognosis of unstable angina were obtained before the development of new drugs and interventional procedures for management of this disease .

This multicenter study enrolled a large cohort of patients admitted with unstable angina to identify the clinical and electrocardiographic markers associated with in-hospital death and myocardial infarction , in the context of a nonselected population admitted consecutively to community hospitals and treated according to current therapeutic strategies .

Methods

Study Design

This was a multicenter prospective study of consecutive patients admitted to coronary care units with unstable angina .
To select a representative sample and safeguard against selection bias , all consecutive patients admitted to the coronary care unit with a suspicion of an acute ischemic syndrome were included in the database .
If the admission diagnosis turned out to be Q-wave or non-Q-wave myocardial infarction , or if a nonischemic cause of chest pain became clearly evident , the process of data collection was interrupted and the patient censored from subsequent analysis .
The diagnosis of Q-wave or non-Q-wave acute myocardial infarction present at admission was made on the basis of the finding of an elevated total creatine phosphokinase ( CPK ) plasma level ( greater than 2-fold increase above normal value ) at entry to the study or at any time up to 12 hours thereafter .
for patients prescribed thrombolytic treatment or primary coronary angioplasty at entry , the enzymatic criteria were not required and the diagnosis was made on the basis of the finding of chest pain lasting more than 30 minutes plus > = 1 mm ST elevation in at least 3 precordial or 2 standard electrocardiogram ( ECG ) leads .

Between March 1991 and May l 992 , 1172 consecutive patients with suspected acute ischemic syndrome were prospectively enrolled .
This group represented 23 % of the total population admitted to the coronary care units during that period of time .
One hundred and thirty-four patients ( 11.4 % ) were excluded : 46 patients because of an evolving myocardial infarction unequivocally present at hospital admission and 88 patients because of noncardiac chest pain .
The remaining 1038 patients with unstable angina formed the study population .

Data collected include demographics and baseline clinical variables .
Electrocardiograms were routinely obtained at admission , during subsequent episodes of chest pain , and at hospital discharge .
Serum total CPK levels were determined at admission , every 12 hours during the first 24 hours , and after a new chest pain episode during hospital stay .
Changes in the ECG were considered positive if there was > = 1 mm of ST-segment elevation or depression in at least 2 limb leads or 3 precordial leads or symmetric T-wave inversion in at least 3 limbs or precordial leads .

The treatment received , any coronary intervention , and all the in-hospital events ( angina , acute myocardial infarction , and death ) were also recorded .
The investigator who was responsible for the patient was instructed to carry a daily prospective and systematic register of all new chest pain episodes in addition to all the information related to the associated electrocardiographic and enzymatic alterations .
All the ECG tracings were examined by an ECG core lab reader who was blinded to the patient clinical condition to validate ST-segment changes > = 1 mm , T-wave changes , left bundle branch block , and left ventricular hypertrophy .
Serial electrocardiograms were compared to evaluate if the changes were transient or fixed .
The medical treatment was entirely left to the discretion of the treating physician except for a strong recommendation to use aspirin in all patients .

End Points and Definition of Events

New episodes of ischemic chest pain after admission were classified as ( 1 ) recurrent angina , defined as angina occurring in the absence of full medical therapy or ( 2 ) refractory angina , angina occurring while the patient was receiving full medical therapy .

Full medical therapy was defined as the combination of aspirin and / or intravenous heparin , nitrates , and b-blockers or calcium antagonists at clinically effective doses .
The adequacy for full medical therapy was left to the decision of the treating physician with suggested criteria of heart rate < 60 beats / min and systolic blood pressure < 120 mm Hg .

Major in-hospital events were new myocardial infarction and death from any cause .
Hospital mortality was defined as death occurring in the hospital at any moment from arrival through discharge .
The criteria for new in-hospital acute myocardial infarction included all of the following : ( 1 ) new prolonged episode of chest pain after admission ; ( 2 ) elevation of total CPK level greater than twice the limit of normal , and ( 3 ) the development of new Q-wave or ST-T changes on the ECG .

Events that occurred during or within 24 hours after an invasive procedure ( coronary angiography , coronary angioplasty , or coronary bypass surgery ) were considered procedure related .
The remaining events were prospectively classified as spontaneous .

An ad-hoc committee blinded to the clinical variables and the administered treatment examined and validated all major events ( acute myocardial infarction or death during hospitalization ) with a systematic review of serial electrocardiograms , enzymatic curves , and clinical charts of all patients who had major events reported or adjudicated by investigators .

Statistical Analysis

The initially targeted sample of 1000 patients was based on a projected 24 % rate of events ( 16 % refractory angina , 6 % acute myocardial infarction , and 2 % death ) .
An a level of significance < .05 was considered for all purposes .
Unadjusted odds ratios and 95 % confidence intervals were obtained for selected clinical factors judged to be associated with the outcome of interest .
Univariate and forward stepwise multiple logistic regression models were used to evaluate the independent association of variables present at admission to in-hospital acute myocardial infarction or death .
From 28 clinical and electrocardiographic variables , those identified of potential significance from the univariate analysis were chosen for forward stepwise selection into the model .
The variables included in the model were age ( with a cutoff point of 65 years ) , prior myocardial infarction , prior angina ( angina preceding the 48-hour interval before admission ) , smoking ( prior or current ) , the number of angina episodes during the last 48 hours before admission ( with a cutoff rate of 3 hours ) , the presence of precipitating factors of angina ( anemia , tachyarrythmias , fever , hypertension ) , and the finding of ST-segment depression on the admission ECG .
Progressive angina was not included because of its significant association with prior angina ( Pearson 's r = 0.70 ) .

The model was validated with the split-sample technique .
A simple random sample of approximately 70 % of the 1038 cases ( the training set , n = 728 ) was used to derive the predictive model ; the remaining 30 % sample ( the test set , n = 310 ) was used to test it .
The rule obtained from the training set was applied to each member of the test set .
The efficiency of the model was assessed by comparison of the area under the receiver operating characteristic ( ROC ) curves of the training and test samples .

The model was also tested on all 1038 cases to determine its overall predictive ability ( by use of the area under the ROC curve ) , sensitivity , and specificity in predicting in-hospital death or myocardial infarction .
The goodness of fit of the model was assessed by the Hosmer-Lemeshow test .

Kaplan-Meier survival curves were generated to calculate the cumulative probability of having a major event develop during hospitalization in the overall population and in the subgroups with and without refractory angina .
Statistical analysis was performed with the SPSS system 7.5 package ( Statistical Package for the Social Sciences , 1997 ) and the ROC Analyzer Program version 6.0 .

Results

Population Demographics and Characteristics

The patients in this study were mostly elderly ( age 60.18 +-16 years ) , 65 % were male , 78 % reported prior angina ( angina preceding the 48 hours interval before admission ) , and 32 % had experienced a prior myocardial infarction ( Table I ) .
The study population was clearly unstable as manifested by the high frequency of resting chest pain ( 78 % ) , prolonged ( > 30 minutes ) chest pain ( 29 % ) and the last episode of pain within the 48 hours preceding admission ( 96.7 % ) .
Angina within 4 weeks and 2 weeks after infarction ( postinfarction angina ) was observed in 9.7 % and 3.3 % of cases , respectively .
Precipitating factors of unstable angina were not identified in 76 % of the population .
There were relatively few smokers ( 21.3 % ) , and the most prevalent risk factors were hypertension ( 60.3 % ) and hyperlipidemia ( 43.3 % ) .
The presence of heart failure at admission was infrequently reported ( 1.5 % ) .

Left ventricular hypertrophy or complete left bundle branch block precluded the analysis of ST-T changes in 12.7 % of patients .
ST segment elevation was frequently observed ( 16 % ) and , in contrast , ST-segment depression was relatively infrequent ( 21 % ) .
Only 16 % of the patients had a normal ECG at entry .

Before hospitalization , treatment consisted of aspirin in half of the patients ( 50.6 % ) , b-blockers in 31 % , nitrates in 44 % , and calcium antagonists in 42 % .
Combined treatment with aspirin , nitrates , and b-blockers or calcium antagonists was administered to 27 % of the study population .

Table II depicted the baseline characteristics of patients with and without major complications during hospitalization .
Patients with complications were older and had greater frequency of prior angina , ST-segment depression on admission electrocardiogram , and repetitive angina ( > 3 episodes ) during the last 48 hours preceding admission .

Conversely , the history of previous infarction was less frequent among patients without complications during hospitalization .

In - Hospital Management

Antithrombotic treatment was administered in the hospital to 94 % of patients , with heparin in 57 % ( intravenous in 34 % , subcutaneous heparin in 23 % ) , aspirin in 88 % , and ticlopidine in 5 % of the patients .
Heparin was used in 34.7 % and 63.9 % of patients with or without precipitating factors , respectively .
Anti-ischemic therapy consisted of the administration of nitrates in 81 % , b-blockers in 71 % , and calcium antagonists in 56 % .

A coronary angiography was performed in 50.2 % of the population .
The most relevant angiographic findings were normal or nonsignificant stenosis in 5.4 % , 1-vessel disease in 32 % , multivessel disease in 61 % , and left main involvement in 10 % .

A quarter of the population was submitted to revascularization procedures ( coronary artery bypass grafting was performed in 14 % and coronary angioplasty in 10.8 % ) .
There was a positive relation between the presence and severity of angina after admission and the need for invasive procedures , either diagnostic or therapeutic : 29.5 % and 44.2 % in patients with recurrent and refractory angina , respectively , and 11.1 % in the subgroup without angina after admission ( P < .
001 ) .
The indications for revascularization procedures were angina after admission in 71 % , inducible ischemia in 6 % , angiographic findings in 21 % , and preference of the patient or his / her physician in the remainder .

Clinical Events

During hospital stay , 40.8 % of the patients had at least 1 new episode of cardiac pain ( Table III ) .
Recurrent angina was diagnosed in 200 patients ( 19.3 % ) and refractory angina in 223 ( 21.5 % ) patients .

New acute myocardial infarction was diagnosed in 54 patients ( 5.2 % ) , and there were 42 in-hospital deaths ( 4.1 % ) .
The combined end point of death and myocardial infarction occurred in 83 patients ( 8.1 % ) .

The incidence of new acute myocardial infarction was higher during the first day and declined thereafter .
On the other hand , cases of death were evenly distributed during hospitalization .
The Kaplan-Meier event-free survival curve for in-hospital infarction or death is shown in Fig 1 .
Fig 2 shows the in-hospital survival curve of patients free of infarction , death , or refractory angina .

The rate of " spontaneous " ( non-procedure-related ) acute myocardial infarction and death was 3.1 % and 1.4 % , respectively .
Acute myocardial infarction and death related to an interventional procedure rates were 1.0 % and 3.8 % .

Relation Between Clinical Characteristics and Outcome

Univariate analysis

The clinical variables present at admission and associated with an adverse outcome through the hospitalization period are summarized in Table IV .
New acute myocardial infarction or death was associated with ST-segment depression ( odds ratio [ OR ] 2.00 , 95 % confidence interval [ CI ] 1.20 to 3.40 ; P = .008 ) , the number of chest pain episodes within the last 48 hours ( OR 1.84 , 95 % CI 1.12 to 3.07 ; P = .01 ) , prior angina ( OR 2.70 , 95 % CI 1.34 to 5.57 ; P = .001 ) , and age > 65 years ( OR 1.64 , 95 % CI 1.00 to 2.70 ; P = .
01 ) .
A negative trend was noted for smoking history ( OR 0.51 , 95 % CI 0.31 to 0.83 ; P = .005 ) , and prior acute myocardial infarction ( OR 0.58 , 95 % CI 0.32 to 1.03 ; P = .
039 ) .
Age ( OR 2.26 , 95 % CI 1.06 to 4.69 ; P = .01 ) , and prior angina ( OR 4.9 , 95 % CI 1.16 to 12.5 , P = .009 ) , were the only univariate predictors of in-hospital mortality .

Multivariate analysis

The model was constructed with all the variables present at admission and found significant at the univariate analyses at a level of P < .
10 .
As shown in Table V , the multivariate analysis retained the following independent admission predictors of myocardial infarction or death : ischemic ST-segment depression in the ECG upon admission ( OR 2.13 , 95 % CI 1.23 to 3.68 , P = .006 ) , prior angina ( OR 2.23 , 95 % CI 0.98 to 5.05 , P = .05 ) , the number of angina episodes within the last 48 hours before admission ( OR 1.63 , 95 % CI 0.98 to 2.70 , P = .05 ) , and smoking history ( OR 0.69 , 95 % CI 0.56 to 0.85 , P = .
004 ) .
Age > 65 years ( OR 1.49 , 95 % CI 1.09 to 2.03 , P = .03 ) was significantly related to in-hospital death .

The overall predictive efficiency was similar in the training and test samples ( ROC curves areas 0.57 +-0.05 and 0.53 +-0.04 respectively , P = .
29 ) .

On all 1038 patients , the area under the ROC curve for the model was 0.59 +-0.03 .
The sensitivity and specificity of the model were computed .
To ensure at least 80 % sensitivity , a predicted probability cutpoint of 0.06 was required ( ie , if the predicted probability was > = 0.06 , the patient was classified as positive ) .
This cutpoint resulted in a sensitivity of 80 % and a specificity of 33 % .
The Hosmer-Lemeshow goodness-of-fit statistic was 3.51 ( degrees of frequency = 7 , P = .
83 ) .

Postadmission Angina and Prognosis

The occurrence of angina after admission ( recurrent or refractory ) showed a strong univariate relation with the incidence of in-hospital acute myocardial infarction or death ( 14.4 % vs 3.2 % , or 5.04 , 95 % CI 2.86 to 8.87 ; P < .0001 ) and death alone ( 6.9 % vs 1.1 % , or 6.45 , 95 % CI 2.61 to 5.9 ; P < .0001 ) ( Table III ) .
Fig 3 shows the Kaplan-Meier infarction-free survival curves in patients with or without refractory angina .
Absence of refractory angina from admission to discharge was associated to a better hospital outcome .
This association persisted even when only non-procedure-related ( " spontaneous " ) events were taken into consideration 10.3 % vs 2.3 % of acute myocardial infarction or death in patients with or without refractory angina , respectively ( OR 4.40 , 95 % CI 2.35 to 8.22 ; Table IV ) .

Discussion

Despite improved treatment of unstable angina during the last years , there remains a considerable risk of events upon admission to the hospital .
Unstable angina includes a wide variety of patients with different prognoses ; accordingly , a reliable risk stratification for everyday practice is necessary .
This is a major challenge because most of the short-term events are concentrated in an small number of patients .
In addition , to select appropriate treatment the risk assessment should be performed as early as possible , ideally during the first hours after admission .

Prognostic Assessment at Hospital Presentation

This study indicates that with current treatment the clinical variables continue to be useful short-term predictors at the time of hospital admission .

After performing multivariate logistic regression analysis , prior angina , ST-segment depression at the admission ECG , the number of pain episodes in the last 48 hours preceding admission , smoking history , and increased age were all independently associated to adverse in-hospital outcome .
However , this model showed low specificity ( 33 % ) .

In this study it was also confirmed that at the time of hospital presentation , the severity of preadmission symptoms is an important determinant of in-hospital prognosis .
Prior angina and the number of episodes of angina during the 48 hours preceding admission were independent predictors of an adverse outcome by multivariate logistic regression analysis .
This finding is a contribution to the elucidation of the controversy regarding the importance of these clinical features as indicators of high risk .
In the past , both a long history of ischemic disease and the lack of preceding symptoms have been suggested to identify high-risk subgroups .

Our study confirms that prognosis is worse in patients with ST-segment depression on admission ECG .
As shown in previous studies , T-wave changes , although common , have no such prognostic value .
This analysis did not take into account the extension and severity of ST-segment or T-wave changes and therefore we cannot rule out the possibility of a worse prognostic significance associated with more severe ECG abnormalities .

In agreement with other reports , patients who smoke cigarettes showed a better in-hospital course .
In another study by Barbash et al , more advanced coronary artery disease explained the worse prognosis of nonsmokers .
One explanation might be that smokers have less-advanced atherosclerotic disease at the time of their first clinical episode of ischemic disease .

Some common clinical variables , such as diabetes and heart failure , did not appear as predictors in this analysis .
Also , these variables did not turn out to be independent predictors in an analysis involving 3171 patients with non-ST-segment elevation ischemic syndrome in the ESSENCE trial .

We surveyed very evident heart failure with overt respiratory distress only .
Thus the low frequency of heart failure at the time of presentation may be a reflection of these diagnostic criteria .
It was not possible to assess either the predictive value of exertional unstable angina or that of subacute angina at rest ( patients without angina within the preceding 48 hours ) because 96.7 % of our patients belonged to Braunwald 's type III classification .

In - Hospital Prognostic Assessment

Recently reported comparisons of early conservative versus early invasive therapeutic strategy indicates that there is no benefit associated with the routine use of any of these strategies .
Therefore it is still necessary to adjust the therapeutic recommendation to the individual needs of each patient .
Because of the lack of other adequate markers of instability , clinical risk stratification performed at admission and during hospitalization is still essential to identify patients at high risk in which an early interventional strategy may be recommended .
The findings of this study are consistent with this appreciation : the appearance of refractory angina , the clinical hallmark of instability , was followed by a rise in the rate of interventions and of major events of 44.2 % and 22.7 % , respectively , whereas persistent relief of symptoms after admission was associated with a favorable in-hospital outcome and a lower procedural rate ( 11.1 % ) .

Although thrombotic complications related to procedures driven by symptoms can be considered to belong to the unstable state , interventions can confound the predictive analyses because recurrence of angina prompts the physician to intervene .
The existence of an intrinsic adverse significance associated with refractory angina is supported by our finding of a significant relation between refractory angina and major nonprocedural ( " spontaneous " ) events .

Limitations

Because no valid universal definition of refractory angina exists , the criteria used in this study may be different from those used by others .
Because of the pragmatic character of the study , neither the presence of ECG evidence of ischemia nor the use of heparin during the episodes of chest pain were required for the diagnosis of refractory angina .

In this study the use of intravenous heparin was less than what might be anticipated ( 57 % ) because in recent years there has been a greater trend toward increased use .
However , the use of heparin was higher in patients without precipitating factors ( 69.7 % ) .
This finding is not unexpected because in primary unstable angina ( Braunwald class B ) the presence of a thrombus can be presumed and the use of heparin treatment is well substantiated .
The studies by Theroux and Wallentin showed the effectiveness of intravenous heparin and low-molecular-weight heparin in preventing major complications during the acute phase of unstable angina .
Consequently , in our study , a lower incidence of refractory angina and a better outcome could have been observed if heparin had been used more frequently or if a more stringent definition of refractory angina had been used .
Nonetheless , mortality rate in this series was similar to that reported in recent trials .
At 30 days the death rate was 3.6 % and 3.9 % in the heparin arms of the studies ESSENCE and GUSTO II B , respectively .
In the PURSUIT trial , 30-day mortality rate was 3.7 % in the control arm .
In the TIMI-IIIB study , the death rate of the early conservative arm was 4.7 % at 42 days .
In the PRISM PLUS study , the 30-day mortality rate was 4.0 % in patients treated with heparin .

Clinical Implications

The relative benefits of early invasive and early conservative treatment strategies have been debated for many years .
Either of these approaches is considered clinically appropriate for many patients .
Accordingly , the identification of candidates for a more aggressive strategy has become of utmost importance .
Because of its simplicity and wide availability , clinical and ECG markers are widely used for this purpose .
Unfortunately , the specificity of the clinical predictors is insufficient to fulfill the requirements of risk stratification under the current therapeutic modalities .

In recent years , serum markers such as troponin T and C-reactive protein have become the focus of interest for risk stratification of unstable angina , with promising results .
It is very likely that routine measurements of these markers will contribute to the identification of those patients at highest risk for cardiac events .
A major question to be explored is how much information these new biochemical markers will add to what is already provided by the classic clinical predictors .

Acknowledgements

We thank Gianni Tognoni , MD , for his assistance in the design of the study ; Dr Pierre Theroux for helpful comments on early drafts of this paper ; and Maria Elena Aizpurua for secretarial assistance .

REFERENCES