Patterns of use and effectiveness of early invasive strategy in non–ST-segment elevation acute coronary syndromes: An assessment by propensity score

Article Outline

• Abstract
• Methods
  • Study design and population
  • Follow-up
  • Statistical analysis
  • Missing data management
• Results
  • Baseline characteristics and inhospital management
  • Predictors of the EIS
  • Outcomes
    • Crude inhospital and follow-up outcomes
    • Adjusted inhospital mortality and follow-up mortality or readmission for ACS
• Discussion
  • Limitations
• Appendix A. Missing data management
• Appendix B. MASCARA study researchers
• References
• Copyright

Background

The patterns of use and the benefit of an early invasive strategy (EIS) in patients with non–ST-segment elevation acute coronary syndrome in a real-life population are not well established.

Methods

All consecutive patients hospitalized because of non–ST-segment elevation acute coronary syndrome between November 2004 and June 2005 in 32 randomly selected hospitals were prospectively included. Patients were stratified by their baseline risk profile using the Global Registry of Acute Coronary Events (GRACE) risk score in 2 groups. Inhospital mortality and 1- and 6-month mortality or rehospitalization for acute coronary syndromes were analyzed. To ensure optimal adjustment propensity score, conventional logistic regression and Cox regression were used.

Results

Of 2,856 patients analyzed, 1,616 (56%) had low/intermediate risk (GRACE ≤140) and 1,240 had high risk (GRACE >140). Patients who underwent EIS had lower risk than those who did not (GRACE score 128.2 ± 41 vs 138.5 ± 43, P < .001). Coronary angiography facility emerged as the strongest predictor of EIS (odds ratio [OR] 13.7 [95% CI 7.1-25]). Patients who underwent EIS had lower rate of the 6-month outcome in both the whole population (9% [95% CI 6.6-11.9] vs 14% [95% CI 12.5-15.6], P = .003) and in high-risk patients (16.5% [95% CI 11-23] vs 23.6% [95% CI 20.8-26.5], P = .04). However, this benefit of EIS was not apparent after statistical adjustment in the whole population (OR 0.8, CI 0.55-1.1, P = .17) or in high-risk patients (OR 0.7, CI 0.46-1.1, P = .16).

Conclusions

In a real-life population, EIS was mainly performed in patients of low/intermediate risk. An obvious benefit of this strategy could not be found.

In 2002, European, American, and Spanish guidelines recommended early invasive strategy (EIS) for high-risk patients with non–ST-segment elevation acute coronary syndrome (NSTEACS).¹ Since their publication, caveats about the overall appropriateness of EIS have been put forth.² The recent European Society of Cardiology Task Force on NSTEACS recommends EIS (ie, coronary angiography during the first 72 hours) followed by revascularization for patients with intermediate- to high-risk features (I-A).³ The Global Registry of Acute Coronary Events (GRACE) risk score is the preferred classification to establish the risk of patients with NSTEACS at admission and thus the indication of EIS.³

It could be expected that during the 2004-2005 period, the 2002 recommendations would enhance the use of EIS for patients with high-risk NSTEACS. The MASCARA (Manejo del Sindrome Coronario Agudo Registro Actualizado) study, a prospective registry of acute coronary syndrome (ACS) in Spanish hospitals, was designed to assess the impact that guidelines had had on patterns of practice and clinical outcomes in 2004-2005.

Using data from the MASCARA registry, we sought to (a) define the profiles of patients with NSTEACS selected for EIS (defined as coronary angiography in the first 72 hours) and (b) assess its impact on inhospital mortality and on the end point “mortality or new ACS” at 1 month and at 6 months from the index episode, using accurate adjustment techniques, in the whole population and in the high-risk baseline profile stratum as assessed by GRACE score.

Back to Article Outline

Methods 

Study design and population 

The MASCARA study design has been previously reported.⁴, ⁵ Thirty-two Spanish hospitals of the public health care system fulfilled the quality requirements of the registry,⁵ and thus participated in the present study (22 had angiography capability).

From October 2004 to June 2005, all consecutive patients ≥18 years old within 24 hours of the onset of angina at rest and who were hospitalized in any study center were eligible. Patients were included if ACS was finally confirmed in case a diagnosis of ACS was made and if the patient had any of the following criteria: (1) cardiac biomarkers above the higher normal limit of each laboratory, (2) ST-segment deviation on electrocardiogram (ECG), (3) stress testing during hospitalization showing ischemia, or (4) known history of coronary vessel disease. The only exclusion criteria were (1) noncardiac illness with expected survival <1 year, (2) ischemia due to noncardiac causes, or (3) impossibility of follow-up. Participating hospitals were encouraged to enroll consecutive patients.

At each site, the designated physician or study coordinator identified those patients with inclusion criteria and no exclusion criteria; requested the informed consent; and classified the patients into STEACS, NSTEACS, and unclassified ACS according to the ECG findings at admission. Thereafter, specifically trained external researchers recorded demographic and clinical data, inhospital treatment, and outcome on standardized case report forms, which were then forwarded to a coordinating center and scanned into an electronic database (Teleform; Cardiff Software Inc, San Diego, CA). In the present study, only the patients with confirmed diagnosis of NSTEACS are considered.

Follow-up 

Patients were followed up by telephone call at 1 and 6 months after discharge to assess vital status and readmissions presumably due to ACS. All calls were centralized and made by trained interviewers who specifically inquired patients or their relatives about the diagnosis reported in the discharge form in case of readmission.

Statistical analysis 

Continuous variables are summarized as means (±SD) or medians (and interquartile range) where appropriate; and categorical data, as percentages.

Baseline GRACE risk score (range 1-372) was computed in all patients with valid data in the 8 variables required. The population was categorized into the 3 GRACE score strata. Thereafter, the first 2 strata were collapsed into 1 because they did not show important differences in most characteristics including prognosis. Thus, the baseline population differences between the 2 strategies (EIS vs non-EIS) were compared in the whole population and in the low/intermediate-risk (range 1-140) vs high-risk strata (>140). Group differences in continuous and categorical variables were compared by Student t and χ² tests, respectively.

The adjusted effect of EIS was assessed by conventional logistic regression and by propensity score analysis⁶ in the whole population and in the high-risk subgroup. Potential predictive variables were selected on the basis of clinical plausibility and their association with outcome in bivariate analysis (selecting all variables with P < .2). The final model was obtained using a backward stepwise method with a threshold for exit set at P > .10. First-order covariate interactions were assessed. We forced the variable EIS into the best model to obtain the adjusted effect. The propensity score represents the probability that a patient received EIS and was computed using extensive, nonparsimonious, logistic regression modeling with the following covariates: age, gender, hypertension, diabetes, peripheral vascular disease, hypercholesterolemia, prior myocardial infarction, prior coronary intervention, prior cardiac surgery, ST deviation on first ECG, elevated cardiac biomarkers, heart rate, systolic blood pressure, Killip class, serum creatinine ≥1.4 mg/dL, concomitant treatment during first 24 hours, and catheterization facility in the first center where the patient was admitted. The selection of the variables was made so as to get the best discriminating model as assessed by the C-statistic.

The resulting propensity score was then used for adjustment of the effect of the EIS on the outcome events including both variables into a logistic regression as the independent variables.⁶ In addition, the propensity score was used to define 5 risk strata for receiving EIS, corresponding to the quintiles of propensity score or quintiles of the computed probability to receive EIS. Afterward, a common odds ratio (OR) across the 5 quintiles was computed using the Maentel-Haenszel method.⁷ Because we observed that a certain degree of residual confounding remained in some quintiles, we performed a propensity score matched-paired analysis as a sensitivity analysis. We matched each treated subject to the closest available nontreated subject based on the estimated propensity score using a greedy-matching algorithm.⁸ The logistic regression model to assess the impact of EIS on outcomes was then estimated using generalized estimating equation methods with robust estimation of variance to incorporate the matched-pairs design.⁹ This strategy was applied to the whole and to the high-risk stratum population separately.

Cumulative survivals for each strategy (EIS vs non-EIS) were illustrated by the Kaplan-Meier method. In addition, we used Cox proportional hazard models to calculate the hazard ratio of the EIS on the outcome “death or ACS” at 6 months. As with logistic regression, the final model was obtained using a backward stepwise method with a threshold for exit set at P > .10.

Missing data management 

Between October 2004 and June 2005, 3,929 consecutive patients with confirmed diagnosis of NSTEACS were eligible. One hundred eighty-seven of 3,929 (4.7%) were excluded, most of them (80%) because of denied informed consent. From 3,742 included patients, GRACE score could not be calculated in 886 (23.6%) mainly because of the lack of ECG information. These patients were excluded from the main analyses, which are based on the 2,856 patients with complete data. To assess the impact of the potential bias, we did a sensitivity analysis with the overall population imputing the missing data (Appendix A available online).

In all cases, P < .05 was considered significant. Corrections were not made for multiple comparisons. All statistical analysis was performed with SPSS 13.0 (SPSS Inc, Chicago, IL).

Back to Article Outline

Results 

Baseline characteristics and inhospital management 

Early invasive strategy was performed in 565 (19.8%) of 2,856 patients included in the analysis (Table I). Early invasive strategy was more frequent in low/intermediate-risk (n = 369, 22.8%) than in high-risk stratum (n = 196, 15.8%). Most high-risk features were more prevalent in the non-EIS. This pattern was very similar in both risk strata. Thus, in all strata, GRACE score tended to be higher in the non-EIS group.

Table I. Baseline characteristics and inhospital management in the whole population and each risk stratum

	Whole population (n = 2856)			Low/intermediate risk (n = 1616)			High risk (n = 1240)
	EIS (n = 565)	Non-EIS (n = 2291)	P	EIS (n = 369)	Non-EIS (n = 1247)	P	EIS (n = 196)	Non-EIS (n = 1044)	P
Baseline characteristics
Age (mean, SD)	64.5 ± 11.7	70 ± 11.7	<.001	60.1 ± 11.2	63.9 ± 11	<.001	72.9 ± 7.3	77.2 ± 7.6	<.001
Gender (% female)	25.7	31.8	.005	21.4	26.6	.04	33.7	37.9	.2
Hypertension (%)	62.3	67.6	.02	59.3	61.8	.4	67.9	74.4	.04
Diabetes (%)	29.7	37.9	<.001	24.7	31.1	.02	39.3	46.1	.08
Hypercholesterolemia (%)	51.7	51.6	.8	52	55.8	.2	51	46.6	.25
Smokers (%)	30.3	16.9	<.001	37.7	23.4	<.001	16.3	9.2	.003
History of MI (%)	23.7	30.3	.002	20.3	23.6	.01	30.1	34.8	.2
Prior PCI (%)	13.8	16.2	.2	14.6	19	.05	12.2	12.8	.8
Angina (FC III/IV) (%)	25	24.2	.7	23.6	23	.8	27.6	25.7	.6
Elevated biomarkers (%)	75.2	65.3	<.001	68	52.2	<.001	88.8	80.8	.008
ST deviation (%)	52	50.7	.6	36.3	29.7	.02	81.6	75.9	.08
Killip class 1st 24 h (%)			.002			.7			.3
K-I	84.4	77.7		96.7	95.7		61.2	56.2
K-II	11.2	15		3	4		26.5	28.1
K-III/IV	4.4	7.3		0.2	0.3		12.2	15.7
Creatinine >1.4 g/dL (%)	10.9	16.8	<.001	5.7	8	.15	20.4	27.5	.04
Peripheral arterial disease (%)	13.1	14	.6	11.9	10.3	.4	15.3	18.4	.3
SBP (mean, SD)	147.9 ± 29	148.2 ± 29	.8	153 ± 29	153.6 ± 28	.7	138.3 ± 28	141.7 ± 29	.2
Heart rate (mean, SD)	79.7 ± 20	80 ± 20	.7	76.7 ± 17	74.2 ± 15	.008	85.1 ± 23	87 ± 23	.3
GRACE score (mean, SD; %)	128.2 ± 40.7	138.5 ± 42.8	<.001	105.1 ± 24	107 ± 22	.2	171.5 ± 27	176 ± 28	.028
Inhospital management
Catheterization facility (%)	97.9	79.3	<.001	97.8	80	<.001	98	78.4	<.001
Aspirin (%)⁎	91.3	86	.001	93.2	88	.005	87.8	83.5	.1
Clopidogrel (%)⁎	46.4	42	.06	48	42.3	.07	43.4	41.2	.6
β-Blockers (%)⁎	69.2	57.6	<.001	75.3	63.4	<.001	57.7	50.6	.07
IIb/IIIa inhibitors⁎	48.3	16.5	<.001	48.5	17.7	<.001	48	15	<.001
Low–molecular weight heparin (%)⁎	83.7	81.2	.2	84.8	81.2	.1	81.6	81.3	.9
Coronary angiography (%)	100	54.9	<.001	100	64.9	<.001	100	43	<.001
PCI (%)	61.2	29.3	<.001	64.5	36.6	<.001	55.1	20.9	<.001
Days of hospitalization (median; IQR)	6 (4-9)	8 (6-13)		5 (3-8)	8 (6-11)		8 (5-14)	9 (6-14)

MI, Myocardial infarction; FC, functional class; SBP, systolic blood pressure; IQR, interquartile range.

Most patients with EIS were first admitted to centers with catheterization facility and received more often therapies recommended in guidelines. Percutaneous coronary intervention (PCI) was performed much more frequently in the EIS group. Hospital stay was longer in the non-EIS group.

Predictors of the EIS 

The propensity score model reached C-statistics of 0.78 (95% CI 0.76-0.80), 0.77 (95% CI 0.75-0.80), and 0.79 (95% CI 0.76-0.83) in the whole population, low/intermediate-risk stratum, and high-risk stratum, respectively.

The strongest predictor of undergoing EIS was, by far, hospitalization in a center with catheterization facility (Table II). Remarkably, the younger the patient was, the higher was the probability of receiving EIS in all groups.

Table II. Predictors of undergoing EIS in the whole population and in each risk stratum

	OR	95% CI	P
Whole population
Coronary angiography facility	13.2	7.1-25	<.001
GP IIb/IIIa inhibitors (1st 24 h)	3.9	3.2-4.9	<.001
Elevated biomarkers	1.6	1.3-2	<.001
β-Blockers (1st 24 h)	1.52	1.2-1.9	<.001
Clopidogrel (1st 24 h)	1.27	1.03-1.6	.025
Age	0.97	0.96-0.98	<.001
Prior coronary disease	0.7	0.6-0.9	.002
Diabetes	0.8	0.6-1	.05
Low/intermediate risk
Coronary angiography facility	13.6	6.2-29	<.001
GP IIb/IIIa inhibitors (1st 24 h)	3.7	2.8-4.9	<.001
Elevated biomarkers	1.6	1.2-2	<.001
β-Blockers (1st 24 hours)	1.7	1.2-2.2	.001
Clopidogrel (1st 24 hours)	1.3	1-1.7	.05
Age	0.98	0.97-0.99	.004
Prior coronary stenosis >50%	0.7	0.5-0.9	.04
High risk
Coronary angiography facility	11.7	4.2-32.5	<.001
GP IIb/IIIa inhibitors (1st 24 h)	4.3	3-6	<.001
Elevated biomarkers	1.7	1-2.8	.05
Diabetes	0.7	0.5-1.02	.06
Age	0.94	0.92-0.96	<.001

GP, Glycoprotein.

Outcomes 

All outcome events were much more frequent in the high-risk than in the low/intermediate-risk stratum (Table III).

Table III. Unadjusted outcomes in the whole population and in each risk strata

There were no statistically significant differences in hospital outcomes between the EIS versus the non-EIS group in the whole population or in each stratum. One- and 6-month follow-up could be completed in 88% and 86.2% of patients, respectively. Although EIS did not show benefit on inhospital mortality, the longer the follow-up was, the greater was the benefit of EIS, which reached statistical significance on the outcome mortality or readmission due to ACS at 6 months, especially in the high-risk stratum. Cumulative survival free from death or new ACS was higher for the EIS group in the whole population (Figure 1) and in the high-risk stratum (Figure 2) but not in the low/intermediate-risk stratum (P = .40) (Figure 3).

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1. 

  Cumulative survival free from death or new ACS for the EIS and the non-EIS groups in the whole population.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 2. 

  Cumulative survival free from death or new ACS for the EIS and the non-EIS groups in the high-risk population.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 3. 

  Cumulative survival free from death or new ACS for the EIS and the non-EIS groups in the low/intermediate-risk population.

Table IV shows the strongest independent predictors of inhospital mortality and late events in both the whole population and the high-risk stratum. Table V shows the effect of EIS adjusted by conventional logistical regression models and propensity score multivariable analysis, using only the patients with all valid data (n =2,856) and having imputed the missing data (n =3,742). Adjusted effect using only patients with all valid data overestimated the effect of EIS in the whole population, but there were no important differences in the high-risk stratum. Initially, there was a trend to higher inhospital mortality. However, this progressively inverted along time in the high-risk subgroup, which showed a nonsignificant trend for a better survival free of new episodes of ACS at 6 months among patients who underwent an EIS. The results were similar using propensity score stratified analysis and propensity score matched-paired analysis (data not shown).

Table IV. Predictors of inhospital mortality and follow-up events

	Whole population			Grace >140
	OR	95% CI	P	OR	95% CI	P
Inhospital mortality
Age	1.04	1.02-1.07	<.001	1.03	0.99-1.06	.07
Elevated biomarkers	3.3	1.6-7	.001	2.7	1.1-6.4	.02
Killip class II	2.6	1.5-4.3	<.001	2.3	1.3-4	.002
Killip class III/IV	5.5	3.2-9.5	<.001	4.4	2.5-7.7	<.001
Creatinine ≥1.4 g/dL	2.8	1.8-4.5	<.001	2.6	1.6-4.2	<.001
ST deviation	1.6	0.99-2.5	.06	1.01	0.99-1.03	.3
Peripheral arterial disease	1.86	1.2-3	.01	1.7	1.01-2.9	.04
SBP	0.991	0.98-0.998	.009	0.990	0.98-0.997	.006
β-Blockers	0.61	0.4-0.95	.03	0.6	0.4-0.9	.04
Clopidogrel (1st 24 h)	0.76	0.5-1.2	.3	0.6	0.4-1.04	.07
Mortality or readmission for ACS at 1 m
Age	1.04	1.02-1.06	<.001	1.045	1.015-1.08	.004
Elevated biomarkers	1.8	1.08-2.9	.02	1.9	1.02-3.8	.04
Killip class II	2.3	1.5-3.7	<.001	2.4	1.5-4	.001
Killip class III/IV	4.4	2.6-7.2	<.001	4.4	2.6-7.5	<.001
Creatinine ≥1.4 g/dL	2.5	1.7-3.7	<.001	2.1	1.4-3.3	.001
ST deviation	1.5	1.01-2.2	.048	1.44	0.9-2.3	.1
Peripheral arterial disease	1.6	1.01-2.5	.048	1.7	1.03-2.8	.04
SBP	0.993	0.987-0.998	.014	0.992	0.985-0.99	.025
Mortality or readmission for ACS at 6 m
Age	1.03	1.02-1.04	<.001	1.022	1.01-1.044	.04
Biomarkers elevated	1.6	1.2-2.2	.002	1.7	1.1-2.6	.02
Killip class II	2	1.4-2.6	<.001	1.8	1.2-2.5	.002
Killip class III/IV	2.7	1.9-4	<.001	2.6	1.7-4	<.001
Creatinine ≥1.4 g/dL	2.1	1.6-2.8	<.001	1.88	1.3-2.6	<.001
ST deviation	1.6	1.2-2.1	<.001	1.4	1-2.1	.05
Peripheral arterial disease	1.7	1.2-2.3	.002	1.7	1.1-2.4	.008
SBP	0.995	0.991-0.999	.015	0.995	0.99-1	.08
β-Blocker therapy	0.7	0.5-0.9	.003	0.8	0.6-1.1	.3

Table V. Adjusted effect of EIS in inhospital mortality and follow-up events

	Whole population			High risk
	OR	95% CI	P	OR	95% CI	P
Inhospital mortality
Valid data (n = 2856)
Conventional LR	1.7	1-3.03	.05	1.7	0.9-3.2	.07
PS	1.4	0.8-2.4	.26	1.5	0.8-2.8	.2
All data (n = 3742)
Conventional LR	2	1.3-3.1	.001	1.5	0.8-2.8	.19
PS	1.9	1.2-2.8	.004	1.4	0.8-2.6	.24
Mortality or readmission for ACS at 1 m
Valid data (n = 2856)
Conventional LR	1	0.6-1.7	1	1.1	0.6-2	.75
PS	0.97	0.6-1.6	.9	1.1	0.6-2	.7
All data (n = 3742)
Conventional LR	1.3	0.9-1.9	.2	0.9	0.5.1.7	.8
PS	1.39	0.9-2	.1	1	0.6-1.9	.8
Mortality or readmission for ACS at 6 m
Valid data (n = 2856)
Conventional LR	0.8	0.55-1.1	.17	0.72	0.46-1.1	.16
PS	0.84	0.6-1.2	.6	0.82	0.5-1.3	.4
All data (n = 3742)
Conventional LR	1.1	0.8-1.46	.47	0.7	0.4-1.06	.09
PS	1.2	0.9-1.6	.16	0.8	0.5-1.2	.3

LR, Logistic regression; PS, propensity score.

Cox regression analysis provided similar results, with a hazard ratio for mortality or new ACS at 6 months of 0.82 (95% CI 0.62-1.14, P = .2) and 0.79 (95% CI 0.53-1.12, P = .2) in the whole population and in the high-risk stratum, respectively.

Back to Article Outline

Discussion 

In this cohort of patients with NSTEACS in randomly selected hospitals across Spain, a definite tendency toward selecting patients for EIS with lower rather than higher risk was detected. The major determinant of EIS was the availability of catheterization facility, the risk profile having a much smaller influence. No statistical association between the strategy and the inhospital and 1-month outcomes could be detected even after accurate risk adjustment. Nevertheless, there was a tendency favoring EIS at 6 months, suggesting the possibility of a long-term benefit. These results of EIS were consistently similar when the analysis only included the patients with all valid data (n = 2,856) or when imputing the missing data (n = 3,742).

As expected, in Spain, there was an increase in the use of invasive procedures after the publication of the guidelines.¹⁰ However, contrary to what might have been expected, this increase was not concentrated in those patients in whom EIS is indicated. When patients were categorized into 2 risk strata using the GRACE score,¹¹ we found that EIS was preferentially used for patients in the low/intermediate-risk stratum. Similar findings were also detected in the CRUSADE (Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early Implementation of the ACC/AHA Guidelines) registry in the United States¹² and in the GRACE registry across several countries. Together with CRUSADE and GRACE, our results again suggest widespread inappropriate compliance with recommendations. The reasons for this tendency should be specifically investigated across countries, maybe using qualitative research methods as has been done with primary PCI.¹³

We did not find any statistical association between the EIS, which was performed in only a limited proportion of candidates, and outcomes during short-term follow-up. In this sense, the benefit of EIS is still controversial in experimental designs. Whereas some randomized clinical trials showed potential benefits for the EIS,¹⁴, ¹⁵ others showed no substantial advantage²; and no study has shown a mortality benefit at 6 to 12 months.¹⁶

Assessing the effectiveness of the therapy in real life, as in the present study, is particularly challenging in ACS because patients' risk, clinical practice patterns, compliance with guidelines, quality of care, and availability of procedures¹⁷ may vary widely between centers. This may explain differences among registries in hard variables such as inhospital mortality. For instance, whereas inhospital mortality for NSTEACS in some registries was as low as 1.5%,¹⁴ in the CRUSADE registry, it reached 4.3%.¹² In our study, inhospital mortality (3.7%, 95% CI 3.1-4.5) is not surprising because the baseline population risk was quite high, with >40% of patients having a GRACE score >140. In this sense, although our 6-month results should be considered with caution because follow-up could only be completed in 86% of patients, they point to a poor 6-month prognosis of patients in the high-risk subgroup.

Our study shows only a trend to a short-term benefit from EIS, which mainly concentrates in the high-risk stratum. This is consistent with other studies¹⁸ and reinforces previous views about EIS effect. The absence of benefit of EIS on inhospital mortality in crude and adjusted results is in contrast with other studies. Thus, in the CRUSADE study, the unadjusted inhospital mortality was 2% versus 6.2% for the early and nonearly invasive care, respectively (adjusted 2.5% vs 3.7%).¹² However, CRUSADE, in contrast with our study, only included patients with either elevated cardiac biomarkers or ST-segment changes; and most importantly, all patients referred from other hospitals or transferred from the institution, as well as those admitted to hospitals without catheterization or PCI or coronary artery bypass graft facility, were excluded. Therefore, in addition to differences in the care process, the comparability between both studies is arguable.

Associated therapies with demonstrated benefit such as glycoprotein IIb/IIIa inhibitors, aspirin, clopidogrel, and β-blockers could be claimed to influence the absence of benefit observed of the EIS in MASCARA registry. However, the rate of use of these treatments in MASCARA registry was similar to CRUSADE study, except for β-blocker therapy, which was quite higher in the latter (77.7% vs 69.2%).¹²

The results of our study may be important for developing implementation strategies, as they suggest that, in the real world, recommendation of a complex intervention in guidelines, even accepting its efficacy, is not enough to ensure its correct use. Our study was not intended to clarify the indications of current guidelines, but it shows the complexity of properly selecting patients in real life for EIS¹⁷ and the lack of effect of its indiscriminate use. Our findings are also a caveat regarding the widening of the indications of invasive procedures without considering the demands of change on the care process and on clinical practice.

Limitations 

As any other observational study, our results should be interpreted with caution. The GRACE score could not be calculated in 23.9% of patients, and 6-month follow-up could not be achieved in 13.8%. Although we made an important effort to account for this in the analysis, some degree of selection bias cannot be ruled out. The unadjusted follow-up outcome event rates, especially at 6 months, may reflect the selection of an especially high-risk population. However, it is unlikely that the adjusted results related with the main study objectives are significantly biased because (a) the difference in the baseline risk between the population with valid and missing data was negligible; (b) the difference in hospital mortality suggested that the bias, if any, would favor the EIS; (c) after imputing the missing data, the adjusted results did not show clinically relevant differences with the valid data analysis; (d) the percentage of patients without complete follow-up was similar between the EIS (12.3%) and the non-EIS group (14.1%) (P = .2), thus making a differential selection bias during follow-up unlikely; and (e) we obtained similar results with multivariate Cox models, which limit the potential bias from missing data during follow-up. Finally, even if patients available at 6-month follow-up analysis were those at higher risk, the potential for benefit of the EIS in these patients should have been apparent if such potential were substantial.

Back to Article Outline

Appendix A. Missing data management 

To assess selection bias, we compared the baseline characteristics of these 886 patients with those of the main cohort in the whole population and in each risk stratum. Although there were no important baseline differences between both cohorts (Table I), we detected a difference in hospital mortality between the “not-missing” cohort (3.7%, 95% CI 3-4.4) and the “missing” cohort (4.6%, 95% CI 3.3-6.2) (P = .2) (Table II). This difference was negligible in patients who did not undergo EIS, but it was significant for the patients who underwent EIS: 3.5% (95% CI 2.2-5.2) for the not-missing cohort and 9.1% (95% CI 5.2-14) for the missing cohort (P = .003). These results suggested that the exclusion of the 886 patients from the analysis could lead to “overestimate” of the possible benefit of the EIS. Therefore, to assess the impact of the potential bias, we did a sensitivity analysis. We first imputed the missing values of the GRACE score using the expectation-maximization algorithm to predict the missing value. The Little test was not significant (P = .7), thus indicating that the missing values were completely at random. Once the GRACE score for the full cohort (3,742 patients) was obtained, we included this variable as covariate in a logistic regression model along with the EIS variable and the variables that were independent predictors of the outcome events in the analysis carried out in the no-missing cohort. This strategy was done for the whole population and for the high-risk stratum.

Table I. Baseline differences between the population with missing and nonmissing data

	Whole population (n = 3742)			Early invasive (n = 729)			Non–early invasive (n = 3013)
	Not missing (n = 2856)	Missing (n = 886)	P	Not missing (n = 565) (19.8%)	Missing (n = 164) (19.2%)	P	Not missing (n = 2311) (80.2%)	Missing (n = 702) (80.8%)	P
Age (mean, SD)	69 ± 11.9	68.6 ± 11.4	.5	64.6 ± 11.7	66.1 ± 12	.13	70 ± 11.7	69.1 ± 11.3	.1
Gender (% female)	30.6	32	.4	25.7	26.2	.9	31.8	33.5	.4
Hypertension (%)	66.5	67.4	.6	62.3	59.1	.5	67.6	69.2	.4
Diabetes (%)	36.3	33.3	.1	29.7	25.6	.3	37.9	35.1	.3
History of MI	29	29.2	.9	23.7	21.3	.5	30.3	30.9	.8
Prior PCI	15.7	23	<.001	13.8	17.7	.2	16.2	23.9	<.001
Prior coronary surgery	7.6	10.1	.02	6.4	6.7	.9	7.9	11	.01
Coronary stenosis >50%	36.7	41.1	.2	29	29.3	.9	38.5	43.4	.01
Angina (FC III/IV)	24.4	22.3	.2	25	24.4	.9	24.2	21.9	.2
Elevated biomarkers	67.2	64.5	.16	75.2	80.1	.2	65.3	61	.06
ST deviation	51	47.6	.4	52	47.6	.7	50.7	49.6	.8
Killip class 1st 24 h			.3			.03			.3
K-I	79.1	80.8		84.4	81.2		77.7	80.4
K-II	14.2	12.1		11.2	9.1		15	12.8
K-III/IV	6.7	6.1		4.4	9.7		7.3	6.8
First creatinine			.5			.4			.4
0-0.39	0.2	0		0	0		0.3	0
0.4-0.79	14.3	15		15.7	14.1		13.9	15.4
0.8-1.19	53.3	53.5		57.8	52.8		52.2	53.5
1.2-1.59	19.4	20.2		18.2	23.9		19.8	19.6
1.6-1.99	5.8	5.2		4.5	6.1		6.2	4.9
2-3.99	5.2	5.1		2	2.5		6	5.4
≥4	1.8	1		2	0.6		1.7	1.2
Peripheral arterial disease	13.8	15.6	.2	13.1	14.6	.6	14	16.3	.1
University hospital	82.9	88.1	<.001	97.9	98.2	.8	79.3	85.5	<.001
GRACE score	88.1 ± 32	88 ± 32	1	79.3 ± 31	85.2 ± 35	.06	90.2 ± 32	88.8 ± 31	.4

FC, Functional class.

Table II. Differences in management and in inhospital death between the population with missing and not missing data

Back to Article Outline

Appendix B. MASCARA study researchers 

Dr Radován and Dr Maulén (Hospital de Campdevanol; Girona). Dr Ortiz de Murua, Dr Marcos, and Dr Arribas (Hospital Virgen de la Concha; Zamora). Dr Laperal and Dr Casado (Hospital de Calatayud; Zaragoza). Dr Bisbe (Hospital Sant Jaume de Olot; Girona). Dr Bartomeu, Dr Carrillo, and Asunción Mateu (Hospital Univerisitario Sant Joan d’Alacant). Dr Gutierrez and Dr Benítez (Hospital Virgen del Puerto; Plasencia). Dr De Miguel, Dr Martínez, and Dr Soriano (Hospital de Terrasa). Dr Arias e Isabel Gómez (Hospital de Montecelo; Pontevedra). Dr Ortega and Dr Molina (Hospital Sta María del Rossell; Cartagena). Dr Herreros and Dr Azcárate (Clínica Universitaria de Navara). Dr Worner, Dr Piqué, and Purificación Cascant (Hospital Arnau de Vilanova; Lérida). Dr Salvador and Dr Aguar (Clínica Dr Pesset; Valencia). Dr Arós and Dr Sanz (Hospital de Txagorritxu; Vitoria). Dr Velasco and Dr Belchi (Hospital Gral Universitario de Valencia). Dr Pagola and Ma. Amparo Pérez (Hospital Ciudad de Jaén). Dr Sogorb and Dr Oliver (Hospital Gral Universitario de Alicante). Teresa Martorell, Dr Borquez, and Dr Verbal (Hospital Clìnic i Provincial; Barcelona). Dr Esplugues, Dr Ribas, and Cristina Carvajal (Ciudad Sanitaria de Bellvitge; Barcelona). Dr Martín and Dr Pabón (Hospital Universitario de Salamanca). Dr Froufe, Dr Leon, and Dr Montes (Hospital de Cruces; Bilbao). Dr Poveda, Dr Ruíz, and Marta Calvo (Hospital Universitario Marqués de Valdecilla; Santander). Dr Alcalde, Dr Alguersuari, Dr Otaegui, and Purificación Cascant (Hospital Vall d’Hebron; Barcelona). Dr Juan, Dr Barrio, and Dr Estévez (Hospital Universitario Gregorio Marañón; Madrid). Dr Moreno and Dr Martín (Hospital San Cecilio; Granada). Dr Fernández Avilés and Dr Sánchez (Hospital Clínico Universitario de Valladolid). Dr Bruguera, Dr Soriano, and Dr Recasens (Hospital del Mar; Barcelona). Dr Abizanda and Dr Micó (Hospital Gral de Castellón). Dr Huelmos (Fundación Hospital de Alcorcón). Dr Ortigosa and Dr Silva (Clínica Puerta de Hierro; Madrid). Dr Bardají, Dr Serrano, and Purificación Cascant (Hospital Joan XXIII; Tarragona). Dr Sala, Isabel Ramió, and Ruth Martì (Hospital Josep Trueta; Girona). Dr Montón (Hospital Gral Yagüe; Burgos). All hospital locations are in Spain.

Back to Article Outline

References 

1. Lopez BL, Aros BF, Lidon RM, et al. 2002 Update of the guidelines of the Spanish Society of Cardiology for unstable angina/without ST-segment elevation myocardial infarction. Rev Esp Cardiol. 2002;55:631–642
   • View In Article
   • MEDLINE
2. Hirsch A, Windhausen F, Tijssen JG, et al. Long-term outcome after an early invasive versus selective invasive treatment strategy in patients with non–ST-elevation acute coronary syndrome and elevated cardiac troponin T (the ICTUS trial): a follow-up study. Lancet. 2007;369:827–835
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (216 KB)
   • CrossRef
3. Bassand JP, Hamm CW, Ardissino D, et al. Guidelines for the diagnosis and treatment of non–ST-segment elevation acute coronary syndromes. The Task Force for the Diagnosis and Treatment of Non–ST-Segment Elevation Acute Coronary Syndromes of the European Society of Cardiology. Eur Heart J. 2007;28:1598–1660
   • View In Article
   • MEDLINE
   • CrossRef
4. Permanyer-Miralda G, Ferreira-Gonzalez I, Marrugat J, et al. Rationale and conceptual design of MASCARA study: a challenge in the evaluation of the effectiveness. Med Clin (Barc). 2005;125:580–584
   • View In Article
   • MEDLINE
   • CrossRef
5. Ferreira-Gonzalez I, Permanyer-Miralda G, Marrugat J, et al. MASCARA (Manejo del Síndrome Coronario Agudo. Registro Actualizado) study. Overall results. Rev Esp Cardiol. 2008;(in press)
   • View In Article
6. D'Agostino RB. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17:2265–2281
   • View In Article
   • MEDLINE
   • CrossRef
7. Lachin JM. Biostatistical methods: the assessment of relative risks. New York: John Wiley and Sons Ltd; 2000;
   • View In Article
8. Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med. 1997;127(8 Pt 2):757–763
   • View In Article
   • MEDLINE
9. Austin PC, Mamdani MM. A comparison of propensity score methods: a case-study estimating the effectiveness of post-AMI statin use. Stat Med. 2006;25:2084–2106
   • View In Article
   • MEDLINE
10. Bueno H, Bardaji A, Fernandez-Ortiz A, et al. Descripción del Estado de los Síndromes Coronarios Agudos en un Registro Temporal ESpañol. Estudio DESCARTES. Rev Esp Cardiol. 2005;58:242–250
    • View In Article
11. de Araujo GP, Ferreira J, Aguiar C, et al. TIMI, PURSUIT, and GRACE risk scores: sustained prognostic value and interaction with revascularization in NSTE-ACS. Eur Heart J. 2005;26:865–872
    • View In Article
    • MEDLINE
    • CrossRef
12. Bhatt DL, Roe MT, Peterson ED, et al. Utilization of early invasive management strategies for high-risk patients with non–ST-segment elevation acute coronary syndromes: results from the CRUSADE Quality Improvement Initiative. JAMA. 2004;292:2096–2104
    • View In Article
    • CrossRef
13. Bradley EH, Curry LA, Webster TR, et al. Achieving rapid door-to-balloon times: how top hospitals improve complex clinical systems. Circulation. 2006;113:1079–1085
    • View In Article
    • CrossRef
14. Collinson J, Flather MD, Fox KA, et al. Clinical outcomes, risk stratification and practice patterns of unstable angina and myocardial infarction without ST elevation: Prospective Registry of Acute Ischaemic Syndromes in the UK (PRAIS-UK). Eur Heart J. 2000;21:1450–1457
    • View In Article
    • MEDLINE
    • CrossRef
15. Wallentin L, Lagerqvist B, Husted S, et al. Outcome at 1 year after an invasive compared with a non-invasive strategy in unstable coronary-artery disease: the FRISC II invasive randomised trial. FRISC II Investigators. Fast Revascularisation during Instability in Coronary artery disease. Lancet. 2000;356:9–16
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (113 KB)
    • CrossRef
16. Mehta SR, Cannon CP, Fox KA, et al. Routine vs selective invasive strategies in patients with acute coronary syndromes: a collaborative meta-analysis of randomized trials. JAMA. 2005;293:2908–2917
    • View In Article
    • CrossRef
17. Van de Werf F, Gore JM, Avezum A, et al. Access to catheterisation facilities in patients admitted with acute coronary syndrome: multinational registry study. BMJ. 2005;330:441
    • View In Article
    • CrossRef
18. Cannon CP, Weintraub WS, Demopoulos LA, et al. Compari son of early invasive and conservative strategies in patients with unstable coronary syndromes treated with the glycoprotein IIb/IIIa inhibitor tirofiban. N Engl J Med. 2001;344:1879–1887
    • View In Article
    • MEDLINE
    • CrossRef