Risk factors for primary ventricular fibrillation during acute myocardial infarction: a systematic review and meta-analysis

European Heart Journal (2006) 27, 2499 2510 doi:10.1093/eurheartj/ehl218 Review Risk factors for primary ventricular fibrillation during acute myocardial infarction: a systematic review and meta-analysis Peter J. Gheeraert 1 *, Marc L. De Buyzere 1, Yves M. Taeymans 1, Thierry C. Gillebert 1, Jose P.S. Henriques 2, Guy De Backer 3, and Dirk De Bacquer 3 1 Department of Cardiology, University Hospital, De Pintelaan 185, 9000 Ghent, Belgium; 2 Department of Cardiology, Academic Medical Centre, Amsterdam, The Netherlands; and 3 Department of Public Health, Ghent University, Ghent, Belgium Received 15 May 2006; revised 11 July 2006; accepted 17 August 2006; online publish-ahead-of-print 4 September 2006 KEYWORDS Acute myocardial infarction; Meta-analysis; Primary ventricular fibrillation; Risk factors Introduction Primary ventricular fibrillation (PVF) during acute myocardial infarction (AMI) is conventionally defined as ventricular fibrillation that is not preceded by signs or symptoms of heart failure or cardiogenic shock. It occurs in more than 10% of patients during the first hours of chest pain related to AMI. 1 8 PVF is the most frequent cause of death related to AMI because it often occurs before monitoring. The risk factors for this complication are still unknown. Factors that were inconsistently associated with PVF include younger age, 9 12 male gender, 11,13 smoking, 11,12 absence of history of diabetes, 12,14 lower serum potassium concentration, 11,15 18 and inferior infarct location. 10,11,19,20 Other previously studied factors could have been not significant due to lack of statistical power. Such factors include history of angina, history of * Corresponding author. Tel: þ32 9 240 44 05; fax: þ32 9 240 49 99. E-mail address: peter.gheeraert@uzgent.be Aims To evaluate potential risk factors for primary ventricular fibrillation (PVF) during acute myocardial infarction (AMI) by a systematic review and meta-analyses. Methods and results We searched PubMed for English articles on humans published between 1964 and January 2006 using a validated combination of MESH terms. Twenty-one cohort studies describing 57 158 patients with AMI were analysed. Patients with validated PVF (n ¼ 2316) were characterized by an earlier admission (weighted mean difference 22.62 h), male gender [odds ratio (OR 1.27)], smoking (OR 1.26), absence of history of angina (OR for history of angina 0.84), lower heart rate at admission (weighted mean difference 24.02 b.p.m.), ST-segment elevation on admission ECG (OR 3.35), AV conduction block before PVF (OR 2.02), and lower serum potassium at admission (weighted mean difference 20.27 meq/l). Patients with validated PVF developed a larger enzymatic infarct size (standardized mean difference 0.74, P, 0.00001). PVF was not associated with a history of myocardial infarction or hypertension. Conclusion Patients who developed a validated PVF presented with characteristics of both abrupt coronary occlusion and early hospital admission. This review provides no evidence for risk factors for PVF other than ST-elevation and time from onset of symptoms. To find new risk factors, studies should compare validated PVF patients with non-pvf patients who have no signs of heart failure and comparable time delay between onset of symptoms and medical attendance. myocardial infarction, history of hypertension, heart rate and blood pressure before PVF, AV-block, and QTc-interval. Reliable risk factors for PVF can optimize the premonitoring management of AMI, including the recommendations on the management of acute chest pain given to the general public and patients. Identification of low vs. high-risk patients could also have impact on the modes of inter-hospital transports for primary angioplasty. New prospective studies on risk factors for PVF during coronary occlusion are limited by early reperfusion strategies. A systematic review and meta-analysis can provide an overview of the potential risk factors studied in humans, uncover associations that were not found in underpowered studies, quantify the strength of associations, and can also help to understand the differences across studies. We aimed to consolidate all published reports in English literature that reported risk factors for PVF and to fulfil the most recent methodological recommendations for meta-analysis of observational studies. & The European Society of Cardiology 2006. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

2500 P.J. Gheeraert et al. Methods Search strategy During the last 7 years, we gradually constructed a library of articles related to ventricular fibrillation during ischaemia or AMI by searching manually and contacts with authors in this field. For the aim of the study, we constructed a systematic search strategy in PubMed by gradually including MESH terms and combined sets that were sensitive enough to retrieve at least all references of our personal library. We searched PubMed for English articles on humans and published between 1964 and January 2006 using six MESH terms and combined sets: myocardial ischaemia, ventricular fibrillation, tachycardia, ventricular, death, sudden, cardiac, heart arrest, and autopsy. One author (P.G.), who has more than 10 years of clinical experience with PVF patients, read the titles and abstracts of all 5288 references and selected manuscripts that possibly described patients with AMI and ventricular arrhythmias. A total of 325 full papers were studied in detail to select those that specified patients with a validated PVF within 48 h of onset of AMI. PVF was considered validated if it occurred in patients without signs of overt heart failure or shock during an AMI defined by WHO criteria. Studies were included if PVF patients were compared with non-pvf patients in a cohort or case control study design. We excluded studies with composite outcomes such as ventricular fibrillation (n ¼ 73) without differentiating between primary and secondary ventricular fibrillation, cardiac arrest (n ¼ 16), or ventricular tachyarrhythmias (n ¼ 35) not allowing data abstraction for patients with PVF. We identified 21 original cohort studies (Table 1) and six original case control studies (Table 2) that compared PVF patients with non-pvf patients. We checked the 167 full articles that were included in the references of these studies. No additional complying study was found. This systematic review and meta-analysis were therefore restricted to these 27 original studies. The only outcome variable was PVF within 48 h of onset of AMI. Data abstractions and classification of studies We abstracted dichotomous and continuous variables that represented patient or infarction characteristics compatible with the definitions summarized in Table 3. Infarction site was the only categorical variable. We abstracted infarction site data in three mutually exclusive categories: anterior (including antero-lateral), inferior (including infero-posterior), and not-localized. The category anterior was described in all studies and the category inferior in all studies except one, 13 which was excluded from this analysis. The category not-localized was used when the infarction was not located specifically as anterior or inferior. In most studies, this category was described as non-st-elevation infarction, non-q-wave, subendocardial, not-localized, or uncertain. Two studies had also few patients in a category described as combined, 12,21 which we also abstracted in the category not localized. Missing data on infarction location were also abstracted in the category not localized. The continuous variables were abstracted only if mean values were available in both PVF patients and non-pvf patients. The cohort studies (Table 1) were classified according to seven well-defined objective criteria for subgroup or sensitivity analysis to assess heterogeneity. These subgroups were defined as (i) exclusion of patients with heart failure (Killip class. 1) in the non-pvf patients, (ii) mean admission delay after onset of symptoms more/ less than 4 h, (iii) use of fibrinolytics in more/less than 50% of patients, (iv) mean age of study population more/less than 60 years, (v) analysis of risk indicators for PVF being/not being an aim of the observational study, (vi) single/multicentre study, and (vii) published before/after 1987 (which was median of publication years). One author (P.G.) classified the studies and abstracted the data. We decided not to duplicate study classification and data abstraction by another independent person, as it was done according to easily defined criteria independent of subjective judgement. Statistical methods We extracted data to construct two-by-two contingency tables for every dichotomous variable reported in the retained studies. Abstracted data were entered into the Cochrane Review Manager (version 4.2.7) software. We pooled data using the odds ratio (OR) with two sided 95% confidence intervals (95% CI), expressing the strength of the association in a random effect model. The significance of discrepancies in the estimates of association from the different reports was assessed by means of Cochrane s test for heterogeneity. The two-sided significance for heterogeneity was conventionally set at P-value, 0.10. In order to combine continuous measures, weighted mean differences were calculated according to the inverse-variance method in a random effect model. To compare enzymatic infarction size, the standardized mean differences were calculated according to the inverse-variance method in a random effect model. To study possible publication bias, we evaluated funnel plots generated by the Cochrane Review Manager (version 4.2.7) software. Results Twenty-one cohort studies (Table 1) and six case control studies (Table 2) described 18 potential risk factors (Table 3) for PVF in patients hospitalized with AMI. Owing to the fundamental differences in study design, meta-analysis was restricted to cohort studies, except for the variables AV block and QTc, as these were only reported in case control studies (Table 4). One variable (diastolic blood pressure) was not analysed by meta-analysis, as it was reported in only one cohort and one case control study. All studies used validated criteria for diagnosis of AMI and PVF. More specific inclusion criteria are presented in Tables 1 and 2. The observational studies differed in some potentially important characteristics or qualities and were therefore categorized according to seven well-defined criteria (Tables 1 and 2). Each category included a sufficient number of studies for subgroup analysis in case of heterogeneity of results across all studies. Almost half of the cohort studies used no reperfusion therapy, six studies excluded patients with heart failure, and in 10 studies, the aim was to find risk indicators for PVF. The results of the meta-analyses are summarized in Table 5. Variables significantly associated with PVF with homogeneity across all studies Patients who developed PVF were characterized by male gender (OR 1.27), absence of history of angina (OR for history of angina 0.84), lower heart rate at admission (weighted mean difference 24.02 b.p.m.), presence of ST-segment elevation on admission ECG (OR 3.35), development or presence of AV conduction block before PVF (OR 2.02), and lower serum potassium at admission (weighted mean difference 20.27 meq/l). Patients with PVF developed a significantly larger enzymatic infarction size (standardized mean difference 0.74, P, 0.00001). Variables significantly associated with PVF with heterogeneity across studies PVF is associated with earlier hospital admission (weighted mean difference 22.62 h) and smoking (OR 1.26). The heterogeneities were caused by differences in strength of associations and not by differences of direction of

Table 1 Overview of cohort studies and their characteristics Study Source Location Observation Number number period of non-pvf patients Number Cumulative of PVF incidence patients (%) Diagnostic criteria for PVF All studies 1966 2002 54 842 2316 All excluded heart failure before VF 1 Wyman et al. 14 San Pedro, California 1970 2002 4 133 121 2.8 Absence of pulmonary oedema, shock, or previous known heart failure Specific AMI selection criteria No: 15 studies, 40 124 patients; Yes: six studies, 17 034 patients Exclusion of heart failure in non-pvf patients Yes: six studies, n ¼ 24 461; No: 15 studies, n ¼ 32 697 Admission delay (h + SD),4 h: eight studies, n ¼ 33 324; 4 h: four studies, n ¼ 2907 Fibrinolytic therapy 50%: four studies, n ¼ 32 762;,50%: 16 studies, n ¼ 20 142 Age (mean + SD) 60 years: five studies, n ¼ 10 277;.60 years: 11 studies, n ¼ 36 790 Primary aim: risk indicators for PVF during AMI Yes: 10 studies, n ¼ 32 446; patients No: 11 studies, n ¼ 24 712 patients Setting 57 158 patients with confirmed AMI No No,4 h: 74% NA 67 No Single centre, 2 Ruiz-Bailén et al. 12 Spain 1995 2000 17 026 735 4.1 Killip 1 No No 3.9 + 4.4 52% 65.2 + 12.3 Yes Multicentre registry, 119 public hospitals 3 Thompson et al. 13 Massachusetts, USA 1975 97 4 832 240 4.7 Killip 1 No Yes NA 16% 64.8 No Multicentre registry, 16 hospitals 4 Madias et al. 29 New York, USA 1986 89 474 43 8.3 Killip 1 No No NA,20% NA No Single centre, 5 Volpi et al. 11 Italy,1990 7 320 302 4.6 Killip 1 First ST-elevation MI and no Yes,4 h 100% 60 + 11 Yes Multicentre thrombolytic trial contraindications to thrombolytic treatment 6 Brezins et al. 23 Nahariya, Israel 1990 94 1 533 57 3.6 Killip 1, absence of AV block III No No NA 21% 64 + 12 Yes Single centre, 7 Kyriakidis et al. 25 Athens, Greece,1992 240 26 9.8 Killip 1 Definite Q-wave AMI and survived to coronary angiography Yes 2.2 h 75% 53 + 9.1 No Single centre, 8 Higham et al. 18 Newcastle, UK 1982 83 417 9 2.1 PVF not specified in detail No No 6.7 + 10.2 0% 63.1 + 10 Yes Single centre, 9 Behar et al. 22 Israel 1981 83 3 707 122 3.1 Killip 1 No Yes NA 0% 61 No Multicentre registry 10 Molstad 17 Hamar, Norway 1982 84 598 20 3.2 Killip 1 No No NA 0% 67.1 + 10.9 Yes Single centre, 11 Volpi et al. 21 Italy 1985 87 6 781 332 4.7 Killip 1 Absence of contraindications to thrombolytic treatment 12 Nordrehaug et al. 16 Bergen, Norway 1985 87 1019 55 5.1 Killip 1 and Excluded patients above BP. 100 mmhg 75 years of age Yes NA 50%,65 years: 65% No Multicentre thrombolytic trial No Median: 1 h 0% NA Yes Single centre, 13 Dubois et al. 24 Liége, Belgium 1977 80 1 169 61 5.0 Killip 1 No No 8.4 + 16.2 0% 55.1 + 16.1 Yes Single centre, Oslo, Norway,1985 1 133 55 4.6 Absence of UK, Ireland,1983 545 14 2.5 Killip 1 Absence of AV block or 14 Skjaeggestad and No No 3 h 0% 66.2 Yes Single centre, Arnesen 51 heart failure 15 Campbell Yes 3.4 h 0% 54.9 No Multicentre randomized et al. 20 heart rate, 60 b.p.m. trial (Tocainide) 16 Carruth and Atlanta, USA 1972 78 1 172 30 2.5 Killip 1 No No,6 h: 57.6% 0% 62.9 No Single centre, Silverman et al. 19 17 Hulting 15 Stockholm, Sweden 1979 81 521 16 3.0 Killip 1 No No NA 0% NA Yes Single centre, 18 El-Sherif et al. 44 Miami, USA 1972 74 430 20 4.4 Killip 1 No No NA 0% 63 No Single centre, 19 Lie et al. 10 Amsterdam, The Netherlands 1972 73 382 18 4.5 Absence of heart failure Definite Q-wave AMI No 4 0% 64 Yes Single centre, 20 Dhurandhar et al. 46 Toronto, Canada 1967 71 831 20 2.3 Killip 1 No No 9 0% NA No Single centre, 21 Lawrie et al. 9 Edinburgh, UK 1966 68 580 20 3.3 Absence of heart failure and systolic pressure. 100 mmhg No No,4 h: 58% 0% 57.1 No Single centre, NA, not available. nloaded from https://academic.oup.com/eurheartj/article-abstract/27/21/2499/2887167 by guest on 16 October 2018 Risk factors for PVF during AMI 2501

Table 2 Study number Overview of case control studies and their characteristics Source Location Observation period Number of non-pvf patients Number of PVF patients Diagnostic criteria for PVF Spain 1995 167 79 Killip 1 or 2, Spain 1993 187 85 Killip 1 or 2, 22 Fiol M et al. 52 PVF within 12 h of admission 23 Fiol M et al. 47 PVF within 12 h of admission Specific AMI selection criteria Exclusion of AF or advanced conduction block Exclusion of heart failure in non-pvf patients Admission delay (h + SD) Fibrinolytic therapy Age (mean + SD) Primary aim: risk indicators for PVF during AMI Setting Yes NA NA NA Yes Single centre No Yes NA NA 57 Yes Single centre 24 Potasman I et al. 37 Israel 1979 84 80 46 Killip 1 No No NA NA 61.5 + 9.2 Yes Single centre 26 Campbell R et al. 30 Israel 1975 79 28 22 Absence of 25 Flugelman M et al. 53 cardiogenic shock or terminal event, complete pre- PVF data Newcastle- upon- Tyne, UK 1977 79 21 17 Killip 1, no antiarrhythmics or diuretics in preceding 72 h, no AV block Stockholm 1981 33 15 No cardiogenic 27 Forssell and Orinius 54 shock, no pacing Chest pain or collapse with diagnostic enzymes No antiarrhythmics or diuretics in preceding 72 h, no AV block No complete heart block nloaded from https://academic.oup.com/eurheartj/article-abstract/27/21/2499/2887167 by guest on 16 October 2018 Yes 4.2 + 5 0% 59+ 11 Yes Single centre Yes All, 3 h 0% 56 + 8 Yes Single centre No NA 0% 61 Yes Single centre 2502 P.J. Gheeraert et al.

Risk factors for PVF during AMI 2503 Table 3 Definitions of study variables Variables Definitions Dichotomous variables Man (vs. woman) History of MI (vs. no) History of angina (vs. no) Diabetes (vs. no) Smoking (vs. no) History of hypertension (vs. no) ST-elevation (vs. no) Anterior site (vs. inferior) Anterior (vs. other) AV block II or III (vs. other) Continuous variables Age (years) Admission delay (h) Potassium (meq/l) Heart rate (b.p.m.) Systolic blood pressure (mmhg) Diastolic blood pressure (mmhg) QTc (ms) Enzymatic infarct size associations which can be visually appreciated in Figure 1. One study 17 was excluded from this analysis because smoking was completely absent in the PVF group, which was regarded as an exceptional finding that the authors could not explain. Variables not associated with PVF with homogeneity across the studies History of AMI and history of hypertension were not associated with PVF. Five cohort studies 11,13,14,22,23 reported on the relation between history of hypertension and PVF (8243 with and 14 124 without history of hypertension). No individual study described a significant association. Pooled data of these five studies indicated homogeneity (test of heterogeneity: x 2 ¼ 5.62; four degrees of freedom, P ¼ 0.23) and had an overall OR of 0.98 (95% CI 0.82 1.17, P ¼ 0.81). Seven cohort studies 13,14,17,20,22,24,25 reported on the relation between history of AMI and PVF (4061 with and 11 767 without history of AMI). The individual ORs varied between 0.77 and 1.34. No study described a significant association between history of AMI and PVF. Pooled data of these seven studies indicated no heterogeneity (x 2 ¼ 3.88; six degrees of freedom, P ¼ 0.69) and had an OR of 1.06 (95% CI 0.88 1.28, P ¼ 0.51). Variables not associated with PVF but with heterogeneity across studies For this category of variables, there was no overall effect in the pooled analysis, but the effects across the studies were heterogeneous, indicating that associations might depend on other study or population characteristics. Therefore, seven dichotomous study variables were used to explore the heterogeneities across studies (Tables 1 and 2). MI before hospitalization for index infarction Any history of angina before hospitalization for index infarction Any type of diabetes known before admission Active smoking (including or not ex-smoking) any time before admission Treated or non-treated hypertension before admission Any systematic measurement in all patients but before onset of PVF Any anterior vs. inferior site criteria. Patients without anterior or inferior or with both sites were excluded in this analysis Any anterior site criteria compared with all others without these criteria On ECG or during monitoring but before PVF Mean age and SD at admission for index infarction Mean time between onset of pain (related to index infarction) and hospital admission Any measurement before onset of PVF At admission and before PVF At admission and before PVF At admission and before PVF Any systematic technique for measurement or correction but measured before PVF Any enzymatic estimation of size during hospitalization for index infarction (peak CK or peak transaminase) The variables systolic blood pressure and QTc were studied in only three studies. None of them allowed meaningful study subgroup analysis. Subgroup analysis for the variables age, diabetes, and infarction site are presented in Figure 2. None of the study characteristics could divide the cohort studies in homogeneous groups (Figure 2), except for the association with DM. Studies with mean admission delay, 4 h and studies that used fibrinolytic therapy in the majority of patients showed homogeneously that non-diabetic patients had a higher risk of PVF compared with diabetic patients (Figure 2). The borderline significant (P ¼ 0.07) and inverse association between PVF and age became significant in studies that included heart failure in the control group and in studies with older patients (mean age. 60 years). Discussion We performed a systematic review and meta-analysis of published English literature on risk factors for PVF. Twenty-one cohort and six case control studies covered 57 968 patients with confirmed AMI, including 2580 patients with validated PVF. We found that PVF is significantly associated with earlier hospital admission (weighted mean difference 22.62 h), male gender (OR 1.27), smoking (OR 1.26), absence of history of angina (OR for history of angina 0.84), lower heart rate at admission (weighted mean difference 24.02 b.p.m.), ST-segment elevation on admission ECG (OR 3.35), AV conduction block before PVF (OR 2.02), and lower serum potassium at admission (weighted mean difference 20.27 meq/l). PVF was not associated with anterior myocardial infarction, history of hypertension, or history of myocardial infarction. Patients with PVF developed a significantly larger enzymatic infarction size.

Table 4 Variables Summary of data extractions Pooled cohort studies Pooled case control studies Cohort studies Case control studies 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Dichotomous variables Man (vs. woman) 14 X X X X X X X X X X X X X X [X] [X] [X] [X] History of MI (vs. no) 7 X X X X X X X [X] [X] [X] History of angina (vs. no) 6 X X X X X X [X] [X] Diabetes (vs. no) 6 X X X X X X [X] [X] Smoking (vs. no) 6 X X X X X [X] X [X] [X] [X] History of hypertension (vs. no) 5 X X X X X [X] [X] ST-elevation (vs. no) 2 X X Anterior site (vs. inferior) 15 X X X X X X X X X X X X X X X [X] [X] Anterior (vs. other) 16 X X X X X X X X X X X X X X X X [X] [X] [X] AV block II or III (vs. other) 2 X X Continuous variables Age (years) 14 X X X X X X X X X X X X X X [X] [X] [X] [X] Admission delay (h) 3 X X X [X] [X] [X] Potassium (meq/l) 5 [X] X X X X X [X] Heart rate (b.p.m.) 2 X X [X] Systolic blood pressure (mmhg) 2 X X [X] Diastolic blood pressure (mmhg) [X] [X] QTc (ms) 3 X X X Enzymatic infarct size 3 X X X [X] [X] Multivariate analysis [X] [X] [X] [X] [X] X, data extracted; [X], data available but not extractable or not used in pooled analysis. nloaded from https://academic.oup.com/eurheartj/article-abstract/27/21/2499/2887167 by guest on 16 October 2018 2504 P.J. Gheeraert et al.

Table 5 Comparison Summary of meta-analyses Studies included Participants Statistical method in random effect model Effect size (95% CI) Test for overall effect Test for heterogeneity Overall significant Homogeneous (P. 0.1) Man (vs. woman) 14 42 448 OR 1.27 (1.12, 1.43) 0.0001 0.68 History of angina (vs. no) 6 22 594 OR 0.84 (0.71, 0.99) 0.04 0.46 Potassium (meq/l) 5 10 286 WMD 20.27 (20.35, 20.19) 0.00001 0.25 Heart rate (b.p.m.) 2 8181 WMD 24.02 (26.06, 21.97) 0.0001 0.96 ST-elevation (vs. no) 2 22 015 OR 3.35 (2.43, 4.62) 0.00001 0.6 AV block II or III (vs. other) 2 398 OR 2.02 (1.09, 3.72) 0.02 0.92 Enzymatic infarct size 3 18 453 SMD 0.74 (0.56, 0.93) 0.00001 0.27 Heterogeneous (P, 0.1) Smoking (vs. no) 6 36 286 OR 1.26 (1.04, 1.53) 0.02 0.06 Admission delay (h) 3 19 550 WMD 22.62 (24.70, 20.55) 0.01 0.00001 Overall not significant Homogeneous (P. 0.1) History of MI (vs. no) 7 15 828 OR 1.06 (0.88, 1.28) 0.53 0.71 History of hypertension (vs. no) 5 22 367 OR 0.98 (0.82, 1.17) 0.81 0.23 Heterogeneous (P, 0.1) Age (years) 14 42 481 WMD 21.19 (22.50, 0.12) 0.07 0.00001 Diabetes (vs. no) 6 36 917 OR 0.84 (0.61, 1.17) 0.3 0.02 Systolic blood pressure (mmhg) 2 8 181 WMD 23.15 (215.84, 9.55) 0.63 0.07 Anterior site (vs. inferior) 15 39 604 OR 0.91 (0.75, 1.11) 0.37 0.001 Anterior (vs. other) 16 54 022 OR 1.03 (0.88, 1.21) 0.7 0.006 QTc (ms) 3 359 WMD 24.36 (218.78, 67.49) 0.27 0.00001 SMD, standardized mean difference; WMD, weighted mean difference. nloaded from https://academic.oup.com/eurheartj/article-abstract/27/21/2499/2887167 by guest on 16 October 2018 Risk factors for PVF during AMI 2505

2506 P.J. Gheeraert et al. Figure 1 Meta-analysis of the associations between PVF and time (h) from onset of symptoms to hospital admission (A) and smoking (B). WMD, weighted mean difference in random effect model. These data suggest that the substrate for developing PVF includes a first AMI, no prior angina, and ST-segment elevations findings in the aggregate that would support those with an abrupt, or acute, coronary artery occlusion without time for pre-conditioning or collaterals to form. This hypothesis can also explain the larger enzymatic infarction size independent of anterior or inferior location. Large myocardial infarction registries in the USA and Europe have shown that patients who were admitted within 2 h of onset of symptoms are also characterized by male gender, 26 28 smoking, 26,27 absence of history of angina, 26 28 more frequent ST-segment elevation, 26 lower heart rate, 27 and more frequent AV-block. 26 Patients with early admission for AMI also present with lower potassium concentrations compared with patients admitted later. 29 To appreciate this resemblance with PVF more quantitatively, we pooled the data of three infarction registries 26 that compared patients admitted within 2 h of onset of symptoms with patients admitted later. We then compared the profiles of patients with PVF with the profiles of patients with hospital admission within 2 h of symptom onset (Figure 3). 27,28 Patients who develop PVF during hospitalization for AMI presented with characteristics comparable with those of patients who were admitted early to hospital for AMI and who did not develop PVF. ST-elevation was stronger associated with PVF than with early admission. Therefore, the previously reported clinical and infarction characteristics other than ST-elevation are probably no risk factors for PVF but are associates confounded by early admission. Previous studies on patients with validated PVF and AMI inevitably selected PVF patients with early medical attendance because more than half of all PVF events occur within 2 h of onset of symptoms. In future analysis, validated PVF patients have to be compared with non-pvf patients with comparable admission delay to find associates that are independent of admission delay. To our knowledge, no previous cohort study had performed a multivariate analysis that included admission delay into the model, and only one small case control study 30 matched for admission delay. It is expected that patients with ST-elevation have a higher risk for PVF compared with patients with no ST-elevation infarction. A recent large study of patients with acute coronary syndromes confirmed that patients with no ST-segment elevation had a lower cumulative incidence of VF (1.3%) 31 compared with the cumulative incidence of PVF in patients with confirmed AMI ranging from 2.1 18 to 9.8%. 8 The finding of lower serum potassium before PVF was very consistent. Although this association can also be related to early admission, 29 it deserves more discussion. Five cohort studies independently showed that patients with PVF had a lower level of potassium before the event compared with patients without PVF. Pooled analysis demonstrated that the weighted mean difference was small (20.27 meq/l, 95% CI 20.35 to 20.19). Other clinical and experimental data indicated a causal and concentration-dependent risk between hypokalaemia and ventricular arrhythmias during AMI independently of diuretic usage. 15,32 36 We additionally analyse systematically all studies on PVF that reported on diuretic usage at admission. Diuretic usage at admission was specified in two cohort 17,24 and one case control 37 studies and there were no associations with PVF, suggesting that the relation between low potassium and PVF cannot, or only partially, be explained by diuretic usage. Other leading reviews on potassium during AMI 35,38 also conclude that the associations with ventricular arrhythmias during AMI were independent of diuretic usage. The association between low potassium and ventricular arrhythmias was also confirmed in other clinical settings 35 and in individual studies

Risk factors for PVF during AMI 2507 Figure 2 Exploration of heterogeneities across studies by grouping studies according to seven dichotomous characteristics. The number of studies in the subgroups is indicated; bars represent the OR or WMD with 95% CI; open square is proportional to the number of patients; filled square indicates homogeneous results across studies (test for heterogeneity: P 0.10). of patients resuscitated from out-of-hospital VF during AMI. 39 However, potassium levels were also inversely correlated with catecholamine concentrations during AMI. The catecholamine surge during AMI shifts potassium intracellularly especially through muscular beta 2 -receptor stimulation of the sodium potassium ATPase. 40 It cannot be excluded that the relatively small decrease of potassium before PVF would only be a marker of catecholamine surge. The association of lower level of potassium and PVF does not imply that immediate corrections of these small dips would lower the risk of ventricular fibrillation. The association between PVF and lower heart rate is puzzling especially if one of the mechanisms cited to explain low potassium is a catecholamine surge. These seemingly paradoxical observations deserve some further comments. We believe that abrupt coronary occlusion (without time for pre-conditioning or collaterals to form) is indeed the aggregate of most of the described risk factors including the higher catecholamine surge and the resulting lower potassium concentration. However, abrupt coronary occlusion is also associated with more pronounced local cardiac autonomic reflexes 41 independent of peripheral catecholamine levels. The resulting heart rate then largely depends on the site of infarction. Patients with abrupt inferior infarction have a more pronounced vagal reflex compared with those with anterior infarction. 42,43 We therefore believe that abrupt coronary occlusion resulting in PVF will be associated with lower heart rate in patients with inferior AMI and with higher heart rates in patients with anterior AMI, both independent of the systemic plasma catecholamine level. The two studies that evaluated heart rate had in their PVF group predominantly patients with inferior infarction resulting in lower mean heart rate. The question then remains why in some studies PVF was associated with inferior MI and in others associated with anterior MI. We could document this heterogeneity of associations with infarction site (Table 5) but could not explain it even after subgroup analysis (Figure 2). Limitations This meta-analysis does not allow multivariate analysis because it includes only group-data of observational studies. Individual data could not be obtained because we aimed to cover 40 years of research. In this first approach, we aimed to review all current data available, including older studies up to the early periods of monitoring patients

2508 P.J. Gheeraert et al. Figure 3 Associates of PVF (open rhombi) compared with associates of hospital admission within 2 h (filled squares, pooled data from infarction registries 26 28 ); bars represent OR and 95% CI. with AMI in the early 1960s. These older observational cohort studies present unique data of relatively long observations of cohorts without reperfusion therapy. By including these older studies (that have only group data available) in meta-analysis, it was possible to evaluate if the observed associations were independent of reperfusion therapy and, in case of heterogeneity, to perform additional analysis according to whether or not reperfusion therapy was applied (Figure 2). The meta-analysis was probably not influenced by publication bias, as more than half of the reports had another aim than finding risk indicators (Table 1) and did not show different trends compared with those with the aim of finding risk factors. This was also confirmed by funnel plots, which did not show trends towards publication bias. One common flaw uncovered in most studies is that history of angina, hypertension, diabetes, and smoking were not well defined. For instance, history of angina can theoretically include pre-infarction angina (within 72 h before the AMI) as well as chronic angina. The effect of preinfarction angina, which is more equivalent to experimental preconditioning, could therefore have been underestimated by all cohort studies. Accordingly, smoking and DM were not uniformly defined, which could have created heterogeneities between studies. We did not provide a quality score for the individual studies because our inclusion criteria ensured a minimum standard of quality, and studies were subsequently categorized according to seven well-defined study characteristics. We did not include warning ventricular arrhythmias because of extremely varying definitions in too few studies on PVF. 14,37,44 49 Only one cohort study 25 and one case control study 50 compared left ventricular ejection fraction (EF) in patients with confirmed PVF with patients without PVF and without signs or symptoms of heart failure. In both studies, the EF was measured more than 1 week after onset of AMI, and results were conflicting, showing, respectively, no significant difference and lower EF in PVF patients. One study 25 evaluated the extent of coronary artery disease (CAD) in survivors of PVF and concluded that PVF patients had more extensive CAD when measured with the Gensini coronary arteriographic score and by number of diseased vessels. Treatment before admission was not clearly separated from treatment on admission or treatment before PVF. 13,17,24,25,37 The effects of early treatment with thrombolytics, beta-blockers, lidocaine, and other antiarrhythmics should preferentially be studied in specific meta-analyses of randomized control trials, which is beyond the scope of this study. Clinical implications This meta-analysis shows that both ST-segment elevation and the first hours of symptoms dramatically elevate the risk of PVF. Clinicians should not rely on the traditional risk factors for atherosclerosis (such as gender and age), infarction-related characteristics (such as anterior vs. inferior site), or on the medical history to estimate the risk of the event. Time is muscle is important to reduce the final infarct size, time is life would better incorporate the strong and important association between PVF and time. Implications for future research This systematic review confirmed that after 40 years of research, a significant portion of the risk of PVF remains

Risk factors for PVF during AMI 2509 unexplained and may be unrelated to traditional risk factors for atherosclerosis and myocardial infarction. PVF must be related to unrecognized temporarily acquired functional and structural changes and to genetic and environmental interactions that more directly influence arrhythmic susceptibility. Additional cohort and case control studies directed at discovering these risk factors and their interactions would appear to have considerable potential for reducing cardiovascular mortality. This meta-analysis provides a framework to design prospectively or retrospectively well-matched case control or cohort studies. These future studies should compare validated PVF patients with non-pvf patients who have no signs of heart failure and have comparable time delay between onset of symptoms and medical attendance. These studies could include cardiac autonomic function (e.g. heart rate variability and baroreflex sensitivity), early abnormalities in cardiovascular function (e.g. left ventricular size, EF, early remodelling, diastolic function), electrocardiographic variables (e.g. late potentials, T-wave alternans, QRS duration, dispersion of repolarization), metabolic factors (e.g. circulating non-esterified fatty acids), and genetic factors (e.g. mutations of ion channel-encoding genes, polymorphisms of coagulation factors or beta-adrenergic receptors). Conflict of interest: none declared. References 1. The pre-hospital management of acute heart attacks. Recommendations of a Task Force of the European Society of Cardiology and the European Resuscitation Council. Eur Heart J 1998;19:1140 1164. 2. Chambless L, Keil U, Dobson A, Mahonen M, Kuulasmaa K, Rajakangas AM, Lowel H, Tunstall-Pedoe H. Population versus clinical view of case fatality from acute coronary heart disease: results from the WHO MONICA Project 1985 1990. Multinational MONItoring of Trends and Determinants in CArdiovascular Disease. Circulation 1997;96:3849 3859. 3. Sans S, Kesteloot H, Kromhout D. The burden of cardiovascular diseases mortality in Europe. Task Force of the European Society of Cardiology on Cardiovascular Mortality and Morbidity Statistics in Europe. Eur Heart J 1997;18:1231 1248. 4. Norris RM. Fatality outside hospital from acute coronary events in three British health districts, 1994 5. United Kingdom Heart Attack Study Collaborative Group. BMJ 1998;316:1065 1070. 5. Lowel H, Lewis M, Hormann A. Prognostic significance of prehospital phase in acute myocardial infarct. Results of the Augsburg Myocardial Infarct Registry, 1985 1988. Dtsch Med Wochenschr 1991;116:729 733. 6. Adgey AA, Allen JD, Geddes JS, James RG, Webb SW, Zaidi SA, Pantridge JF. Acute phase of myocardial infarction. Lancet 1971;2: 501 504. 7. Sayer JW, Archbold RA, Wilkinson P, Ray S, Ranjadayalan K, Timmis AD. Prognostic implications of ventricular fibrillation in acute myocardial infarction: new strategies required for further mortality reduction. Heart 2000;84:258 261. 8. O Doherty M, Tayler DI, Quinn E, Vincent R, Chamberlain DA. Five hundred patients with myocardial infarction monitored within one hour of symptoms. BMJ (Clin Res Ed) 1983;286:1405 1408. 9. Lawrie DM, Higgins MR, Godman MJ, Oliver MF, Julian DG, Donald KW. Ventricular fibrillation complicating acute myocardial infarction. Lancet 1968;2:523 528. 10. Lie KI, Wellens HJ, Durrer D. Characteristics and predictability of primary ventricular fibrillation. Eur J Cardiol 1974;1:379 384. 11. Volpi A, Cavalli A, Santoro L, Negri E. Incidence and prognosis of early primary ventricular fibrillation in acute myocardial infarction results of the Gruppo Italiano per lo Studio della Sopravvivenza nell Infarto Miocardico (GISSI-2) database. Am J Cardiol 1998;82:265 271. 12. Ruiz-Bailen M, Aguayo D, Ruiz-Navarro S, Issa-Khozouz Z, Reina-Toral A, Diaz-Castellanos MA, Rodriguez-Garcia JJ, Torres-Ruiz JM, Cardenas- Cruz A, Camacho-Victor A. Ventricular fibrillation in acute myocardial infarction in Spanish patients: results of the ARIAM database. Crit Care Med 2003;31:2144 2151. 13. Thompson CA, Yarzebski J, Goldberg RJ, Lessard D, Gore JM, Dalen JE. Changes over time in the incidence and case-fatality rates of primary ventricular fibrillation complicating acute myocardial infarction: perspectives from the Worcester Heart Attack Study. Am Heart J 2000; 139:1014 1021. 14. Wyman MG, Wyman RM, Cannom DS, Criley JM. Prevention of primary ventricular fibrillation in acute myocardial infarction with prophylactic lidocaine. Am J Cardiol 2004;94:545 551. 15. Hulting J. In-hospital ventricular fibrillation and its relation to serum potassium. Acta Med Scand Suppl 1981;647:109 116. 16. Nordrehaug JE, Johannessen KA, von der Lippe G. Primary ventricular fibrillation associated with hypokalaemia in acute myocardial infarction. Mater Med Pol 1987;19:254 257. 17. Molstad P. Primary ventricular fibrillation in acute myocardial infarction. J Intern Med 1989;226:107 111. 18. Higham PD, Adams PC, Murray A, Campbell RW. Plasma potassium, serum magnesium and ventricular fibrillation: a prospective study. Q J Med 1993;86:609 617. 19. Carruth JE, Silverman ME. Ventricular fibrillation complicating acute myocardial infarction: reasons against the routine use of lidocaine. Am Heart J 1982;104:545 550. 20. Campbell RW, Hutton I, Elton RA, Goodfellow RM, Taylor E. Prophylaxis of primary ventricular fibrillation with tocainide in acute myocardial infarction. Br Heart J 1983;49:557 563. 21. Volpi A, Maggioni A, Franzosi MG, Pampallona S, Mauri F, Tognoni G. In-hospital prognosis of patients with acute myocardial infarction complicated by primary ventricular fibrillation. N Engl J Med 1987;317: 257 261. 22. Behar S, Goldbourt U, Reicher-Reiss H, Kaplinsky E. Prognosis of acute myocardial infarction complicated by primary ventricular fibrillation. Principal Investigators of the SPRINT Study. Am J Cardiol 1990;66: 1208 1211. 23. Brezins M, Elyassov S, Elimelech I, Roguin N. Comparison of patients with acute myocardial infarction with and without ventricular fibrillation. Am J Cardiol 1996;78:948 950. 24. Dubois C, Smeets JP, Demoulin JC, Pierard L, Foidart G, Henrard L, Tulippe C, Preston L, Carlier J, Kulbertus HE. Incidence, clinical significance and prognosis of ventricular fibrillation in the early phase of myocardial infarction. Eur Heart J 1986;7:945 951. 25. Kyriakidis M, Petropoulakis P, Antonopoulos A, Barbetseas J, Georgiakodis F, Aspiotis N, Kourouclis C, Toutouzas P. Early ventricular fibrillation in patients with acute myocardial infarction: correlation with coronary angiographic findings. Eur Heart J 1993;14:364 368. 26. Ottesen MM, Kober L, Jorgensen S, Torp-Pedersen C. Determinants of delay between symptoms and hospital admission in 5978 patients with acute myocardial infarction. The TRACE Study Group. Trandolapril cardiac evaluation. Eur Heart J 1996;17:429 437. 27. Newby LK, Rutsch WR, Califf RM, Simoons ML, Aylward PE, Armstrong PW, Woodlief LH, Lee KL, Topol EJ, Van de Werf F. Time from symptom onset to treatment and outcomes after thrombolytic therapy. GUSTO-1 Investigators. J Am Coll Cardiol 1996;27:1646 1655. 28. Goldberg RJ, Yarzebski J, Lessard D, Gore JM. Decade-long trends and factors associated with time to hospital presentation in patients with acute myocardial infarction: the Worcester Heart Attack study. Arch Intern Med 2000;160:3217 3223. 29. Madias JE, Shah B, Chintalapally G, Chalavarya G, Madias NE. Admission serum potassium in patients with acute myocardial infarction: its correlates and value as a determinant of in-hospital outcome. Chest 2000;118:904 913. 30. Campbell RW, Murray A, Julian DG. Ventricular arrhythmias in first 12 h of acute myocardial infarction. Natural history study. Br Heart J 1981;46: 351 357. 31. Al Khatib SM, Granger CB, Huang Y, Lee KL, Califf RM, Simoons ML, Armstrong PW, Van de Werf F, White HD, Simes RJ, Moliterno DJ, Topol EJ, Harrington RA. Sustained ventricular arrhythmias among patients with acute coronary syndromes with no ST-segment elevation: incidence, predictors, and outcomes. Circulation 2002;106:309 312. 32. Nordrehaug JE, von der Lippe G. Serum potassium concentrations are inversely related to ventricular, but not to atrial, arrhythmias in acute myocardial infarction. Eur Heart J 1986;7:204 209. 33. Nordrehaug JE, Johannessen KA, von der Lippe G. Serum potassium concentration as a risk factor of ventricular arrhythmias early in acute myocardial infarction. Circulation 1985;71:645 649. 34. Hohnloser SH, Verrier RL, Lown B, Raeder EA. Effect of hypokalemia on susceptibility to ventricular fibrillation in the normal and ischaemic canine heart. Am Heart J 1986;112:32 35.

2510 P.J. Gheeraert et al. 35. Macdonald JE, Struthers AD. What is the optimal serum potassium level in cardiovascular patients? J Am Coll Cardiol 2004;43:155 161. 36. Yano K, Hirata M, Matsumoto Y, Hano O, Mori M, Ahmed R, Mitsuoka T, Hashiba K. Effects of chronic hypokalemia on ventricular vulnerability during acute myocardial ischemia in the dog. Jpn Heart J 1989;30: 205 217. 37. Potasman I, Abinader EG, Oliven A, Cohen A. Primary ventricular fibrillation complicating acute myocardial infarction: an assessment of predictability and prognosis. Isr J Med Sci 1984;20:491 495. 38. Gettes LS. Electrolyte abnormalities underlying lethal and ventricular arrhythmias. Circulation 1992;85:I70 I76. 39. Thompson RG, Cobb LA. Hypokalemia after resuscitation from out-of-hospital ventricular fibrillation. JAMA 1982;248:2860 2863. 40. Brown MJ, Brown DC, Murphy MB. Hypokalemia from beta2-receptor stimulation by circulating epinephrine. N Engl J Med 1983;309: 1414 1419. 41. Airaksinen KE. Autonomic mechanisms and sudden death after abrupt coronary occlusion. Ann Med 1999;31:240 245. 42. Webb SW, Adgey AA, Pantridge JF. Autonomic disturbance at onset of acute myocardial infarction. BMJ 1972;3:89 92. 43. Mark AL. The Bezold Jarisch reflex revisited: clinical implications of inhibitory reflexes originating in the heart. J Am Coll Cardiol 1983;1: 90 102. 44. El Sherif N, Myerburg RJ, Scherlag BJ, Befeler B, Aranda JM, Castellanos A, Lazzara R. Electrocardiographic antecedents of primary ventricular fibrillation. Value of the R-on-T phenomenon in myocardial infarction. Br Heart J 1976;38:415 422. Clinical vignette ARVC with left ventricular involvement in a young woman 45. Lie KI, Wellens HJ, Downar E, Durrer D. Observations on patients with primary ventricular fibrillation complicating acute myocardial infarction. Circulation 1975;52:755 759. 46. Dhurandhar RW, MacMillan RL, Brown KW. Primary ventricular fibrillation complicating acute myocardial infarction. Am J Cardiol 1971;27:347 351. 47. Fiol M, Marrugat J, Bayes de Luna A, Bergada J, Guindo J. Ventricular fibrillation markers on admission to the hospital for acute myocardial infarction. Am J Cardiol 1993;71:117 119. 48. Tye KH, Samant A, Desser KB, Benchimol A. R on T or R on P phenomenon? Relation to the genesis of ventricular tachycardia. Am J Cardiol 1979;44:632 637. 49. Bekheit S, Turitto G, Fontaine J, El Sherif N. Initiation of ventricular fibrillation by supraventricular beats in patients with acute myocardial infarction. Br Heart J 1988;59:190 195. 50. Dewhurst NG, Hannan WJ, Muir AL. Ventricular performance and prognosis after primary ventricular fibrillation complicating acute myocardial infarction. Eur Heart J 1984;5:275 281. 51. Skjaeggestad O, Arnesen H. Primary ventricular fibrillation complicating acute myocardial infarction: incidence in subgroups. J Oslo City Hosp 1985;35:107 109. 52. Fiol M, Marrugat J, Bergada J, Guindo J, Bayes de Luna A. QT dispersion and ventricular fibrillation in acute myocardial infarction. Lancet 1995;346:1424. 53. Flugelman MY, Hasin Y, Tur-Caspa I, Friedlander Y, Gotsman MS. Prediction of in-hospital ventricular fibrillation from admission data in acute myocardial infarction. Clin Cardiol 1983;6:156 162. 54. Forssell G, Orinius E. QT prolongation and ventricular fibrillation in acute myocardial infarction. Acta Med Scand 1981;210:309 311. doi:10.1093/eurheartj/ehi874 Online publish-ahead-of-print 19 April 2006 Thomas Thouet 1 *, Julia J. Krueger 2,3, Ingo Paetsch 1, Sebastian Kelle 1, and Eike Nagel 1 1 Department of Internal Medicine/Cardiolgy, German Heart Institute Berlin, Augustenburger Platz 1, 13353 Berlin, Germany; 2 Department of Congenital Heart Disease, German Heart Institute Berlin, Berlin, Germany; and 3 Department of Pediatric Cardiology, German Heart Institute Berlin, Berlin, Germany * Corresponding author. Tel: þ49 30 45932400; fax: þ49 30 45932458. E-mail address: thouet@dhzb.de A 22-year-old patient was sent to our institution with recurrent palpitations and echocardiographic findings of an isolated right ventricular (RV) dilation. Holter-ECG demonstrated frequent, polymorphic ventricular ectopia and an asymptomatic nonsustained ventricular tachycardia. Twelve-lead ECG showed inverted T-waves in right precordial leads. Cardiac MR imaging revealed pronounced thinning of the dilated RV myocardium and regional akinesia of the basal and equatorial RV free wall with a global RV function of 25%. No direct visualization of intramyocardial fatty tissue was possible despite high-resolution MRI most likely due to the thinning of the free wall. However, fibrofatty tissue was found in the inferolateral wall of the left ventricle (LV) using a series of T1 (Panel A) and T2 weighted images with and without fat saturation pre-pulse (Panel B) and an inversion recovery technique 15 min after Gd-DTPA injection (Panel C). Hypoperfusion during dynamic perfusion imaging could be observed (Panel D). We diagnosed an arrhythmogenic right ventricular cardiomyopathy (ARVC) with left ventricular involvement according to the modified McKenna criteria. The present case shows the necessity of a vigilant and detailed observation of the LV myocardium in patients with suspected ARVC, especially because histopathological studies have shown LV involvement in 76% of patients. Panel A. The T1-weighted blackblood images revealed two localized areas of hyperintense signal in the inferolateral LV-wall. Panel B. Using a T1-weighted blackblood sequence with fat saturation prepulse (STIR) the hyperintense signal in the LV wall was suppressed and identified as fatty tissue. Panel C. 15 minutes after administration of Gd-DTPA (0.2 mmol/kg/bodyweight) fibrous tissue with hyperintnese signal in both areas could be detected using an inversion recovery technique. Panel D. During injection of Gd-DTPA, hypoperfusion of both areas could be observed using a balanced turbo field echo imaging sequence dedicated for dynamic first pass perfusion.