We performed an observational longitudinal prospective study of patients diagnosed with STEMI in the coronary care units of Hospital Universitario Virgen de la Arrixaca (Murcia, Region of Murcia, Spain) and Hospital Universitario de Santa Lucía (Cartagena, Region of Murcia, Spain). The patients were recruited between January 1998 and January 2008.

The STEMI was defined as typical chest pain of ≥ 30min duration and/or elevated myocardial necrosis markers with ST-segment elevation in ≥ 2 precordial leads > 0.2mm in V1, V2, or V3 and > 0.1mm in lateral leads (aVL, I) or inferior leads (II, III, and aVF) or presumed new-onset left bundle branch block. The study was approved by the ethics committee and patients gave their written consent.

Atrial fibrillation was defined as any electrocardiographically documented irregular ventricular rhythm in a 12-lead recording without a clear definition between the p and f waves (flutter). Previous AF was defined as a previously documented diagnosis of AF (electrocardiogram or medical history) and new-onset AF was defined as AF that appeared during hospitalization without a prior diagnosis of AF. New-onset AF was subdivided according to the time of onset: within 24h of hospitalization (≤ 24h) or after more than 24h of hospitalization (> 24h).

Severe or major bleeding complications were defined as bleeding in the brain, retroperitoneum, or any other location leading to hemodynamic deterioration and/or the need for transfusion of whole blood or blood components.

Angioplasty during hospitalization was defined as the composite of primary angioplasty and angioplasty performed during hospitalization, ie, delayed angioplasty or angioplasty performed the day after successful fibrinolysis. Heart rupture was defined as the composite of free wall rupture, ventricular septal rupture, or ruptured mitral chordae tendineae.

All patients underwent long-term postdischarge follow-up (median, 7.2 years) by telephone contact, review of medical records, visits to outpatient clinics, and review of death records. In-hospital mortality was excluded for the purpose of this analysis. The follow-up rate was 98%.

Multivariate binary logistic regression was performed to analyze the factors associated with in-hospital mortality and new-onset AF during hospitalization. Odds ratios (OR) and 95% confidence intervals (95%CI) were calculated. The chi-square test was used to assess the significance of each variable in the prediction model of new-onset AF. Survival analysis was performed using the Kaplan-Meier method and the Mantel-Haenszel test. Hierarchical Cox multivariate regression models were used to study postdischarge mortality and the hazard ratio (HR) and the 95%CI were estimated. Non-Gaussian variables were transformed to their base-decimal logarithm. The model was adjusted by using the enter method with all covariates entered as blocks, including those considered relevant in the literature: age, sex, body mass index, diabetes mellitus, hypertension, current smoking, dyslipidemia (block 1); chronic obstructive pulmonary disease, newly diagnosed tumors (< 1 year), chronic kidney failure, previous ischemic heart disease, previous stroke, previous New York Heart Association functional class ≥ II (block 2); heart rate and systolic blood pressure on arrival at the emergency department, Killip class, left ventricular ejection fraction on hospitalization (block 3); revascularization with angioplasty and fibrinolysis (block 4); and HF during hospitalization (> 24h) (block 5). The log-linear and proportional hazards assumptions were checked by graphical methods. Final model discrimination and calibration were determined using the C-statistic and the Hosmer-Lemeshow test, respectively. The Harrell C-statistic was calculated for the Cox models.

The linear trends test was used to analyze trends and 5 two-year periods were established according to the time of recruitment: period 1 (1998-1999), period 2 (2000-2001), period 3 (2002-2003), period 4 (2004-2005), and period 5 (2006-January 2008). Missing values per variable were generally < 2% for most variables (99%). A P value of < .05 was used as a cutoff for statistical significance. Analyses were conducted using PASW, version 20 (IBM, Unites States) and STATA 9.1 (College Station, Texas, United States).

Table 1 shows the baseline characteristics of the study sample (n=4284). The mean age was 64 years; 24.0% were women. In total, 3.2% of the patients had previous AF and 9.8% developed new-onset AF (60.0% in the first 24h). After 24h, new-onset AF occurred in 80.8% before the end of the third day. Postdischarge AF persisted in 8.7% of the patients with new-onset AF who survived hospitalization. Table 2 shows hospital treatment including reperfusion therapy and treatment at discharge. Table 3 shows in-hospital complications.

Compared with patients without previous AF, those with previous AF were significantly older, mainly women, and more of them had diabetes and hypertension; however, there were fewer patients with dyslipidemia in this group and fewer were current smokers. Comorbidity was markedly higher in this group of patients, more of whom had a history of ischemic heart disease, stroke, peripheral artery disease, chronic kidney failure, chronic obstructive pulmonary disease, and New York Heart Association ≥ II (Table 1). Regarding hospital treatment, this group underwent fewer angioplasty procedures during hospitalization and were more often treated with angiotensin-converting enzyme inhibitors and diuretics. Patients in this group, however, were less often treated with beta blockers and statins. These patients also had worse left ventricular ejection fraction (Table 2).

At discharge, patients with previous AF were more often prescribed triple therapy, angiotensin-converting enzyme inhibitors, and digoxin but were less often prescribed salicylates, thienopyridines, and beta blockers (Table 2).

A total of 418 patients developed new-onset AF during hospitalization. Unadjusted predictors of new-onset AF were age, female sex, diabetes mellitus, a history of stroke, baseline New York Heart Association ≥ II, a higher heart rate at hospitalization and Killip class, higher levels of creatine kinase MB isoform, and a higher frequency of HF during hospitalization. Protective factors were dyslipidemia, current smoking, higher systolic blood pressure, and higher left ventricular ejection fraction (Table 1 of supplementary material). The only independent predictors in the model were age (per decimal logarithm in years, OR=266; 95%CI, 42-1673), systolic blood pressure (per decimal logarithm in mmHg, OR=0.04; 95%CI, 0.01-0.11), and HF during hospitalization (OR=2.49; 95%CI, 1.88-3.31). According to the chi-square statistic, the most significant predictor was HF during hospitalization, followed by age and systolic blood pressure (Table 1 of supplementary material).

Patients with previous AF had a higher rate of HF during hospitalization (54% vs 28%; P<.001) and a greater crude in-hospital mortality rate (22% vs 12%, P<.001) than patients without previous AF; however, no significant differences were found in other in-hospital complications. Patients with new-onset AF were more likely to develop HF during hospitalization (P<.001), complete atrioventricular block (P<.001), stroke (P<.001), heart rupture (P<.001), and in-hospital mortality (P<.001) (Table 3). The 2 most important causes of in-hospital mortality were cardiogenic shock and electromechanical dissociation in patients with previous AF (46.7% and 33.3%, respectively) and in patients with new-onset AF (63.7% and 19.4%, respectively).

In a well-calibrated adjusted model (Table 4) with high discriminative power, new-onset AF during hospitalization (OR=1.55; 95%CI, 1.08-2.22) and new-onset AF within 24h (OR=2.01; 95%CI, 1.26-3.21) were independent predictors of in-hospital mortality. Previous AF (OR=0.55; 95%CI, 0.34-1.26) and new-onset AF > 24h (OR=1.17; 95%CI, 0.71-1.95) were not predictors of in-hospital mortality. The risk of death associated with new-onset AF and new-onset AF ≤ 24h was constant among patients who developed HF during hospitalization and those who did not (Table 2 of supplementary material) (interaction, P=.398 and P=.984).

During follow-up (median, 7.2 years [interquartile range, 2.7-10.3]), the long-term mortality rate was 3.5/100 patient-years: 11.11/100 patient-years among patients with previous AF and 5.35/100 patient-years among patients with new-onset AF. Figure 1 shows the survival curve for patients with previous AF or new-onset AF.

Figure 1.

Kaplan-Meier mortality survival curve showing post-discharge all-cause mortality stratified by previous atrial fibrillation and new-onset atrial fibrillation during hospitalization. AF, atrial fibrillation.

(0.08MB).

Table 5 shows that previous AF (HR=1.24; 95%CI, 0.94-1.64), total new-onset AF (HR=0.98; 95%CI, 0.80-1.21), new-onset AF ≤ 24h (HR=0.96; 95%CI, 0.73-1.28), and new-onset AF>24h (HR=1.02; 95%CI, 0.77-1.37) were not predictors of long-term mortality. The results presented did not change when stratified according to the development of HF during hospitalization (Table 3 of supplementary material) (interaction, P>.05 in all cases).

During the 10 years of recruitment (Figure 2), the rate of new-onset AF remained constant, whereas the rate of HF during hospitalization significantly decreased (period 1 vs period 5, 34% vs 27%; trend, P<.001). Angioplasty during hospitalization significantly increased (33.3% vs 88.2%; trend, P<.001). There was a substantial and significant decrease in in-hospital mortality (14.5% vs 9.6%; trend P<.001).

Figure 2.

Trends in the rate of new-onset atrial fibrillation during hospitalization, heart failure during hospitalization, and reperfusion therapy. AF, atrial fibrillation. P values are presented for the trend test. AF, atrial fibrillation. *Composite of primary angioplasty and angioplasty performed during hospitalization (delayed angioplasty or angioplasty the day after successful fibrinolysis).

(0.11MB).

The clinical profile of patients with new-onset AF>24h differed from that of patients with new-onset AF<24h. The most significant differences (all P<.05) were older age, higher systolic blood pressure at hospitalization, worse Killip class, worse left ventricular ejection fraction, more diseased coronary vessels, reduced reperfusion, and an increased rate of in-hospital complications (complicated HF, complete atrioventricular block, and angina or reinfarction) (Tables 1–3). There was no significant difference between the groups in hospital mortality (P=.321), but the patients with new-onset AF>24h had higher long-term mortality (log rank test, chi square=4.60; P=.032).

No information was available on the duration of AF during hospitalization or whether it was persistent or permanent, which may be of relevance.16

We cannot exclude the possibility that previous AF may be associated with long-term mortality and may therefore be clinically relevant, given the small number of patients with previous AF compared with the total sample, which gave rise to a type 2 error of 0.42 (HR=1.24).