One-year health status outcomes of unstable angina versus myocardial infarction: a prospective, observational cohort study of ACS survivors

Study population and design

The Investigation oF Outcomes from acute coRonary syndroMes (INFORM) registry is a prospective, observational cohort study of consecutively hospitalized ACS patients at two Kansas City hospitals, the Mid America Heart Institute and Truman Medical Center, to identify determinants of health status outcomes among ACS survivors. It was powered for minimal detectable differences of >5 points in SAQ quality of life (disease perception) and angina frequency scores. In total, 10,911 consecutive hospitalized patients who had a troponin blood test performed at either hospital between March 2001 and October 2002 were prospectively screened for a possible ACS. 1199 patients with confirmed ACS (see definitions below) were enrolled in the registry (Figure 1) and underwent detailed interviews and chart abstractions to obtain their socio-demographic, health status, clinical, and treatment characteristics. All data elements conformed to the standards established in the American College of Cardiology Task Force on Clinical Data Standards[16]. Since our investigation focused on those who survived to hospital discharge, those patients who died during hospitalization (n = 7; 3 STEMI, 3 NSTEMI, and 1 UA) were excluded from the analyses. One year after their index hospitalization, follow-up phone interviews were conducted to collect health status and rehospitalization information. Each participant signed an informed consent to participate in this study and Institutional Review Board approval from both institutions (Saint Luke's Health System Institutional Review Board and the University of Missouri-Kansas City Adult Health Sciences Institutional Review Board) was obtained prior to the conduct of the study.

Acute coronary syndrome classification

Standard definitions were used to confirm patients' ACS diagnosis [17]. STEMI patients presented with suggestive cardiac symptoms, diagnostic electrocardiogram (EKG) changes (ST segment elevation or new-onset left bundle branch block [LBBB]), and a positive troponin blood test (during the course of this investigation, the following assays and thresholds were used by the enrolling centers: cTroponin-I assay >0.15 ng/mL, Advia-Centaur, Bayer Diagnostics, Tarrytown, NY; cTroponin-I assay >0.05 ng/mL, Dade Dimension RXL, Dade Behring Diagnostics, Deerfield, IL; cTroponin-I assay >1.9 ng/mL, Abbott Labs, Abbott Park, IL). NSTEMI patients presented with suggestive cardiac symptoms and/or EKG changes (e.g. ST segment depressions and/or T wave changes), and a positive troponin blood test. UA patients presented with suggestive cardiac symptoms, as defined by at least one of the following: new onset angina (<2 months) of at least Canadian Cardiovascular Society Classification (CCSC) class III, prolonged (>20 minutes) rest angina, recent (<2 months) worsening of angina, or angina that occurred within 2 weeks of a previous MI [18]. Although EKG changes were not a requirement for diagnosis, nearly one-half of UA patients had ischemic EKG changes on admission (LBBB 4%, ST-elevations 9%, ST-depressions 12%, T-wave inversions 22%). By definition, all UA patients had a negative troponin blood test. To further increase the specificity of the unstable angina diagnosis, those patients with a diagnostic study that excluded obstructive coronary disease, cardiac perfusion defects, or segmental wall motion abnormalities (e.g. coronary angiography, nuclear or echocardiographic stress testing) (n = 125) or confirmed an alternative explanation for their presentation (e.g. esophagogastroduodenoscopy) were excluded. Three physicians reviewed the charts of all patients for whom diagnostic uncertainty (n = 45) remained and attained consensus on the final diagnosis.

Health status, rehospitalization, and mortality assessment

After excluding expired patients by querying the Social Security Death Master File, surviving patients underwent follow-up telephone interviews to obtain one-year rehospitalization rates and health status outcomes. Health status outcomes were measured using the Seattle Angina Questionnaire (SAQ) and the Short Form-12 Version 2 (SF-12). The SAQ is a 19-item disease-specific health status measure for patients with coronary artery disease (CAD) that has well-established validity, reliability, sensitivity to clinical change, and prognostic value [13, 19–22]. Five domains are assessed: anginal frequency, anginal stability, physical limitation, treatment satisfaction, and disease perception. The scales used in these analyses range from 0–100, where higher scores indicate fewer symptoms and higher quality of life. The SF-12 is a generic health status measure that is converted into physical and mental component scores [23]. A score of 50 reflects the United States' population mean, and a deviation of 10 points reflects 1 standard deviation from that mean. Higher scores indicate better physical and mental functioning.

We measured the health status outcomes of symptoms, physical functioning, and health-related quality of life with the SAQ angina frequency scale, the SF-12 physical component scale, and the SAQ disease perception scale, respectively. The SAQ scales of treatment satisfaction and angina stability were less informative to our research question, and thus not analyzed. In addition, the SF-12 physical functioning scale was used instead of the SAQ physical limitation scale because the SF-12 scale was more representative of activities that registry patients performed and the disease-specific SAQ created more missing data for physical function than did the SF-12. One-year cardiac rehospitalization rates were collected by patient self-report at the follow-up interview. Hospitalizations for chest pain, heart failure, myocardial infarction (MI), cardiac revascularization (percutaneous coronary intervention [PCI], or coronary artery bypass graft [CABG]) were defined as cardiac. Hospitalizations for other reasons were coded as non-cardiac. Two-year mortality data was determined by querying the Social Security Administration Death Master File. A final query of the Social Security Death Master File, hospital and outpatient records for vital status and hospitalization data was performed in October 2004.

Statistical analysis

After categorizing patients by their ACS presentation (STEMI, NSTEMI and UA), baseline clinical characteristics were compared. Categorical data were reported as frequencies, and differences were assessed using chi-square tests. Continuous data were reported as means ± standard deviations and differences were assessed using analysis of variance.

To evaluate the independent association between ACS presentation and outcomes, multivariable models were created to adjust for all other differences in socio-demographic (age, race, sex, and insurance status), hospital (Mid-America Heart Institute or Truman Medical Center), clinical (prior MI, prior PCI, prior CABG, congestive heart failure, hypertension, diabetes, hyperlipidemia, cerebrovascular accident/transient ischemic attack, renal failure, anemia, and tobacco use), baseline physical function or quality of life (SF-12 physical functioning scores or SAQ quality of life scores) and treatment (both revascularization (PCI, CABG, or thrombolysis) and discharge medications (angiotensin converting enzyme [ACE] inhibitors, lipid lowering agents, beta-blockers, calcium channel blockers, nitrates, and aspirin)) characteristics. Anemia was defined as hemoglobin values less than the fifth percentile of the sex, race, and age-matched population [24]. Categories of variables (socio-demographic, site, clinical, baseline health status, and treatment) were entered sequentially into the model and presented accordingly, in order to illustrate the additive effects of each variable group on the various outcomes.

SAQ angina frequency scores were modeled as a dichotomous outcome of any angina vs. no angina at one year after index ACS hospitalization because of the skewed nature of the measure and our clinical goal of seeking to completely eliminate patients' angina [25]. Since one-year angina in our study was a relatively frequent event, we estimated adjusted relative risks directly using a modified Poisson regression model, rather than using logistic regression which estimates adjusted odds ratios [26]. SF-12 physical component and SAQ quality of life scores were modeled as continuous variables using multivariable linear regression. Model diagnostics were conducted using residual and normal probability plots.

Unadjusted differences in mortality and rehospitalization rates between ACS classes were described graphically with Kaplan-Meier survival plots and compared using the log-rank test. Adjusted differences in mortality and rehospitalization rates between ACS classes, using the same covariates as the health status models, were compared using multivariable Cox proportional hazards regression models. Cox proportional hazards assumptions were tested using Schoenfeld residuals.

All analyses were performed with SAS version 9.3.1 (SAS Institute, Inc., Cary, NC) and R version 2.1.1 (R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria). Statistical significance for all analyses was assumed when the p value was less than 0.05. Scheffe correction techniques for multiple comparisons were used for pair-wise comparisons.

Missing data

Among the 1192 patients who survived until hospital discharge, 77 (6.5%) died within 12 months of follow-up. Health status outcomes were not imputed for those who died. Thus, our health status results characterize those who survived for at least one year after discharge. Among survivors, 210 patients (17.6%) were not interviewed at one year, because they could not be contacted (73% of non-respondents), or refused to complete the interview (27% of non-respondents). Among those lost-to-follow-up patients, there were no significant differences by ACS presentation.

We explored potential selection biases in the rehospitalization and health status outcomes through the use of sensitivity and propensity model analyses. Sensitivity analysis was performed by imputing poor health status scores to those patients (n = 22) who were too ill to participate in the follow-up interview. Specifically, patients were assigned a SAQ Angina Frequency score indicating presence of angina, an SF-12 physical functioning score of 25, and a SAQ quality of life score of 50; the lowest deciles for each outcome. Regression analyses were repeated using these imputed values and no statistically or clinically significant differences in results were noted (results not shown).

As an alternative approach for handling missing health status data from patients who refused 1-year interviews or could not be contacted, propensity scores were computed using non-parsimonious multivariable logistic regression models to predict the likelihood of unsuccessful follow-up [27]. Predictor variables included all available demographic, socio-economic and lifestyle factors, clinical characteristics, vital signs and laboratory studies, disease severity, baseline health status, medication, acute, and non-acute treatments received during patients' initial ACS hospitalization. From these models, a probability of success for completing an interview was calculated. The reciprocal of this probability was then assigned to those patients' scores in the multivariable regression analyses in order to assess for potential observable bias from those lost to follow-up by weighting patients that are similar to those with missing data more heavily [27]. These analyses also demonstrated that the missing patients did not impact our primary findings. In light of the consistency of our findings with both imputation and propensity approaches for handling missing data, we have presented only the primary results.

Patient characteristics by ACS presentation

Treatments also differed by ACS classification. STEMI patients had the highest revascularization rates and UA patients the lowest (86.3% and 42.0%, respectively). Significant differences in discharge medication prescriptions were also observed. STEMI and NSTEMI patients were more likely to be treated with beta-blockers at discharge, while UA patients received higher rates of ACE-inhibitors, calcium-channel blockers, and nitrates (Table 1).

Mortality and rehospitalization outcomes

UA patients had the lowest two-year mortality rates, but similar one-year rehospitalization rates. At two years, 25 (7.9%) STEMI patients and 45 (12.8%) NSTEMI patients had died as compared with only 35 (6.7%) UA patients (log-rank p-value 0.006; Figure 2a). After multivariable analysis controlling for socio-demographic, site, clinical, and treatment characteristics, UA patients had significantly lower mortality rates compared to STEMI and NSTEMI patients (UA vs. STEMI hazard ratio (HR) = 0.51; 95% confidence interval (CI) [0.28, 0.95]; UA vs. NSTEMI HR = 0.40; 95% CI [0.24, 0.65]) (Figure 2b). At one year, 102 (26.6%) UA patients were rehospitalized for cardiac causes, as compared to 43 (17.6%) STEMI patients and 60 (23.3%) NSTEMI patients (log-rank p-value 0.035: Figure 3a). After multivariable adjustment, UA patients had similar one-year cardiac rehospitalization rates to STEMI and NSTEMI patients (UA vs. STEMI HR = 1.31; 95% CI [0.86, 1.99]; UA vs. NSTEMI HR = 1.03; 95% CI [0.73, 1.47]; Figure 3b).

Health status outcomes

One year after discharge, UA patients had worse unadjusted angina rates, physical component scores, and quality of life scores than either STEMI or NSTEMI patients (Figure 4a–c). After sequential multivariable adjustment for all covariates, UA patients were more likely to experience angina than STEMI patients (UA vs. STEMI angina relative risk (RR) = 1.42; 95% CI [1.06, 1.90]). Adjusted angina rates between UA and NSTEMI were similar (UA vs. NSTEMI angina RR = 1.10; 95% CI [0.85, 1.42]) (Figure 4a), as were the remainder of health status scores between all three ACS classes (UA vs. STEMI adjusted mean physical component score difference = -0.05 points, 95% CI [-2.41, 2.30]; UA vs. NSTEMI adjusted mean physical component score difference = -1.91 points, 95% CI [-4.01, 0.18]; UA vs. STEMI adjusted mean quality of life score difference = -1.39 points, 95% CI [-5.63, 2.85]; UA vs. NSTEMI adjusted mean quality of life score difference = -0.24 points, 95% CI [-4.01, 3.54]) (Figure 4b,c).