Armando Vidal Gonzalez Cancino grupo10413 Background There is growing interest in the role of infections in the aetiology of acute myocardial infarction (AMI). We undertook a large, population-based study to explore the association between risk of AMI and recent acute respiratory-tract infection. Methods We used data from general practices in the UK (General Practice Research Database). Potential cases were people aged 75 years or younger, with no history of clinical risk factors, who had a first-time diagnosis of AMI between Jan 1, 1994, and Oct 31, 1996. Four controls were matched to each case on age, sex, and the practice attended. The date of the AMI in the case was defined as the index date. For both cases and controls the date of the last respiratory-tract infection before the index date was identified. We also did a case-crossover analysis of cases who had an acute respiratory-tract infection either before the index date or before an arbitrarily chosen date (1 year before AMI). Findings In the case-control analysis of 1922 cases and 7649 matched controls, significantly more cases than controls had an acute respiratory-tract infection in the 10 days before the index date (54 [2.8%] vs 72 [0.9%]). The odds ratios, adjusted for smoking and body-mass index, for first-time AMI in association with an acute respiratory-tract infection 1-5, 6-10, 11-15, or 16-30 days before the index date (compared with participants who had no such infection during the preceding year) were 3.6 (95% CI 2.2-5.7), 2.3 (1.34.2), 1.8 (1.0-3.3), and 1.0 (0.7-1.6); (test for trend p<0.01). The case-crossover analysis showed a relative risk of 2.7 (1.6-4.7) for AMI in relation to an acute respiratory-tract infection in the 10 days before the index date. Interpretation Our findings suggest that in people without a history of clinical risk factors for AMI, acute respiratory-tract infections are associated with an increased risk of AMI for a period of about 2 weeks. We cannot, however, completely exclude the possibility of misdiagnosis bias, if prodromal symptoms of AMI were mistaken for respiratory-tract infection. Introduction Death from cardiovascular diseases, particularly acute myocardial infarction (AMI), is commoner in winter than summer.[ 1] Although many factors could contribute to this association, the greater number of respiratory infections in winter months is a possible causal link. [ 1, 2] Death rates from cardiovascular diseases increase during epidemics of influenza, [ 3] and anecdotal reports as well as a few small studies suggest that acute respiratory-tract infections shortly before AMI could be a risk factor.[ 4-6] Furthermore, respiratory-tract infections have also been associated with an increased risk of cerebral infarction.[ 7] There is growing interest in the role of acute and chronic infections in the pathogenesis of arteriosclerosis and coronary heart disease.[ 2, 8-12] Two small randomised trials of macrolide antibiotics in the secondary prevention of coronary heart disease have reported possible benefits,[ 13, 14] and larger trials are in progress to test those claims. A better understanding of the role of chronic and acute infections in the aetiology of AMI may result in new strategies for its prevention and treatment. We conducted a large population-based study on the UK General Practice Research Database (GPRD) to explore the association between acute respiratory-tract infections and the risk of AMI, in both case-control and case-crossover analyses. Methods Study population and data source Data for the study came from general practices in the UK. General practitioners (GPs) from more than 400 practices record medical information and supply it, without identifiers, to provide data for researchers. Demographic data and characteristics of patients (height, weight, smoking status), symptoms, diagnoses (by OXMIS codes), referrals, hospital admissions, and all drug prescriptions are routinely recorded. These GPs replaced the previous hand-written records by computer records in the late 1980s. Currently, more than 4 million patients are actively enrolled on this database. On request, copies of hospital-discharge and referral letters are available for review, with identifiers removed, to validate the diagnoses recorded on the database. The GPRD is currently owned by the UK Department of Health. The accuracy and comprehensiveness of the data recorded in the GPRD has been documented. [ 1517] GPRD data have been used before in studies of patients with AMI.[ 18, 19] Case identification and validation Potential cases were selected on the basis of a first-time diagnosis of AMI between Jan 1, 1994, and Oct 31, 1996. We restricted the study to patients who were 75 years of age or younger at the date of diagnosis of AMI (the index date), and who had no history of known metabolic or cardiovascular conditions predisposing to AMI.We excluded all patients with a history of AMI, angina pectoris, unexplained chest pain, cardiac arrhythmias, congestive heart failure, stroke, intermittent claudication, venous thromboembolism, chronic renal disease, hypertension, hyperlipidaemia, diabetes mellitus, connective- tissue disorders, or cystic fibrosis more than 60 days before the AMI. All cases had to be registered on the database for at least 3 years before the index date. In two studies of GPRD data,[ 18, 19] we validated the computer-recorded diagnosis of a first AMI for more than 400 patients by reviewing hospital-discharge letters. A high percentage of computer-identified cases (>90%) were confirmed by at least two of the following documented diagnostic criteria: characteristic chest pain, characteristic changes in the electrocardiogram, characteristic serial rises in the concentrations of cardiac enzymes, an arteriogram documenting a recent coronary occlusion, or fibrinolytic therapy. In addition to this previous validation study, we reviewed hospital-discharge letters for a random sample of 40 cases in the current study.Again, a high percentage (>90%) of the diagnoses from the computer data could be confirmed (37 confirmed, three uncertain); we therefore decided to include all the potential cases identified initially from computerised patient records. Study design and analysis
To estimate a potentially increased risk of first-time AMI in relation to an acute respiratory-tract infection, we did two separate analyses: case-control and case-crossover. For the case-control analysis, we matched four controls to each case on age (same year of birth), sex, and the practice they attended. We controlled for calendar time, and thereby for confounding by seasonal fluctuations, by using the same index date for the matched controls as for the cases. The same exclusion criteria were applied to controls and cases: they had to have a recorded history on the GPRD of at least 3 years, and all individuals with circulatory or metabolic diseases predisposing to AMI more than 60 days before the index date were excluded. For both cases and controls, we identified the date of the last respiratory-tract infection before the index date that led to a practice visit (non-specific acute upper-respiratory infection, bronchitis, pneumonia, chesty productive cough). We did a matched analysis (conditional logistic-regression model) with SAS (release 6.12). We obtained odds ratios of a first-time diagnosis of AMI in relation to an acute respiratory-tract infection in predefined intervals of 1-10, 11-30, 31-90, and 91-365 days before AMI. The reference group consisted of those patients who did not have a respiratory-tract infection in the year before the index date. The potential confounding factors for each case and control-body-mass index (<25.0 kg/m2, 25.0-29.9 kg/m2, greater than or equal to 30.0 kg/m2, unknown), and smoking status (never, ex, current, unknown)-were assessed from the patients' profiles. By doing stratified regression analyses, we further assessed how potential effects were modified by age, sex, smoking status, body-mass index, history of asthma, calender year, and whether or not the patient died from AMI. To explore whether a greater risk of developing AMI is related to an acute respiratory-tract infection, as opposed to another type of infection, we also evaluated whether there were any acute urinary-tract infections among cases and controls (cystitis, urinary-tract infection, renal infection, pyelonephritis). We used the same intervals of 1-10, 11-30, 31-90, and 91-365 days before the index date. The case-crossover design can be used to study the effect of a brief and transient exposure period on the risk of an acute outcome, such as AMI. This technique [ 20,21] has been used to evaluate the effect of acute short-term risk factors ("triggers") of myocardial ischaemia.[ 22-24] For a case-crossover analysis, no controls are needed because each case acts as his or her own control. We assessed for each case whether the index date, and another date chosen arbitrarily (which we call the artificial date), was immediately preceded by an acute respiratory-tract infection. To reduce the risk of confounding by seasonal fluctuations, we chose the artificial date to be exactly 1 year before the date of the AMI (index date minus 365 days). We calculated relative risks for discordant pairs only (cases who had a respiratory infection either before the index date or before the artificial date). Concordant pairs (cases who had-or did not have-an acute respiratory-tract infection at both the index date and the artificial date) were not analysed. 95% CI and two-sided p values were calculated by exact methods (Fisher's confidence limits), based on the binomial distribution.[25] Results Initially we identified 2441 patients who developed an idiopathic first-time AMI in 1994-96. We eliminated 519 people because of poor documentation of the AMI diagnosis or uncertain index date, before we knew whether or not there had been a recent acute respiratory-tract infection. The final dataset consisted of 1922 cases; 1441 (75%) were male, and 1170 (61%) were between 60 and 75 years of age at the date of the AMI. For 285 (14.8%), the AMI diagnosis was based on necropsy because the patient did not reach the hospital alive (table 1). In the years 1994 and 1995 (when there were data for the full 12 months), there were more cases of AMI in winter than in summer. The highest number (175) was in January, the lowest (110) in June. Case-control analysis Among the 1922 cases and 7649 matched controls, 54 (2.8%) cases and 72 (0.9%) controls had an acute respiratory-tract infection that led to a visit to the practice within the 10 days preceding the index date (table 1). The odds ratio of AMI for patients who had an acute respiratory-tract infection in the 10 days before the index date (with patients who had no such infection during the year before the index date as the reference group) was 3.0 (95% CI 2.1-4.4), adjusted for body-mass index and smoking (table 2). There was no substantial increase in risk for patients who had a respiratory-tract infection 11-30 days before the index date, and no increase in risk for those with a history of respiratory-tract infections 31-90 or 91-365 days before the index date (table 2). Current smoking (compared with non-smoking) and body-mass index of 30.0 kg/m2 or higher (compared with <25.0 kg/m2) were independent risk factors for AMI (table 2). We further stratified into four subgroups those cases and controls who had a recorded respiratory-tract infection in the 30 days before the index date. The adjusted odds ratios for AMI for patients with a recorded acute respiratory-tract infection (compared with patients who had no infection during the year before the index date) decreased with increasing time between infection and index date (test for trend p<0.01, figure).
Risk of AMI in relation to timing of previous acute respiratory-tract infection Vertical bars=95% CI. To detect the possibility of effect modification, we stratified cases and controls into further subgroups. The strength as a risk factor for AMI of an acute respiratory-tract infection in the 10 days before the index date increased with age: the odds ratio was 1.6 (95% CI 0.7-3.6) in patients below the age of 60, 2.7 (1.0-7.1) for patients aged 60-64, and 4.0 (2.5-6.5) for those aged 65 and older, adjusted for smoking and body-mass index (test for interaction, with age-categories <60 and greater than or equal to 60 years, p=0.07). There was no evidence for effect modification for sex (odds ratio 3.0 for men, 3.1 for women), smoking status (nonsmokers and past-smokers 3.0, current smokers 2.1), body-mass index (<25.0 kg/m2 3.1, greater than or equal to 25.0 kg/m2 4.1), history of asthma (no 3.2, yes 5.2), or whether the patient died from AMI at the index date (survived 2.6, died 4.3). Stratification by calendar year resulted in similarly consistent odds ratios for each year (1994, 2.9; 1995, 2.6; 1996, 4.3). Aspirin use was low in this study population, as recorded on the computer, and did not affect the results. We found no increased risk of AMI in relation to acute urinary-tract infections at any time in the year before the index date (table 2; reference group patients with no urinary-tract infection during the year before the index date). To find out whether differential morbidity between cases and controls could lead to a substantial detection bias, we assessed the total number of practice visits and the number of drug prescriptions for all cases and controls in the 3 years before the index date. The median number of practice visits was ten for both cases (range 0-93) and controls (range 0-128). Case-crossover analysis Among the 1922 cases, there were 1848 concordant pairs who either had no respiratory-tract infection in the 10 days before the index date and the artificial date, or had an acute respiratory-tract infection in the 10 days before the index date and the artificial date. There were 74 discordant pairs: 54 cases had a respiratory-tract infection in the 10 days before the date of the AMI, but not before the artificial date, and 20 cases had a respiratory-tract infection in the 10 days before the artificial date, but not before the date of the AMI. The relative risk of AMI for cases with infection before the AMI index date was 2.7 (1.6-4.7). Discussion The findings of this large study of 9571 patients suggest that an acute respiratory-tract infection is a risk factor for AMI. The risk of AMI was three times higher among patients who had an acute respiratory-tract infection in the previous 10 days than among those who did not. The risk decreased with increasing time between the acute respiratory-tract infection and the AMI, and was absent after about 2 weeks. The findings of both the case-control and the case-crossover analysis were similar. No increase in AMI was found in relation to urinary-tract infections, another common type of infection that we chose arbitrarily for comparison. We also reanalysed the datasets from two previous case-control studies from the GPRD, including a total of 394 otherwise healthy hypertensive patients with a first-time AMI and 1196 hypertensive controls matched for age, sex, practice, and time.[ 18, 19] The findings were similar, showing a more than two-fold increase in the risk of AMI in relation to a recent acute respiratory-tract infection (odds ratio 2.1 [95% CI 0.9-4.7]). Our findings are consistent with a few previously published hospital-based studies from the 1980s, which used smaller groups of patients.[ 4-6] We matched controls to cases on age, sex, and practice, and we adjusted for smoking status and body-mass index in the multivariate regression analysis. Substantial confounding by these variables is therefore unlikely to explain our findings. We cannot
rule out subtle confounding by other variables that are difficult to measure or not available to us (such as socioeconomic status, household heating, diet, alcohol consumption, physical activity, stress, job activities). If such variables are to lead to positive confounding, they have to increase both the risk of AMI and the risk of acquiring respiratory-tract infections. Substantial confounding by such variables seems unlikely because any such differences between cases and controls would have led to an increased infection rate among future AMI cases in general, not only during a short period of time (2 weeks) immediately preceding the index date. Furthermore, the case-crossover analysis also showed an increased risk of AMI in association with an acute respiratory-tract infection. Confounding by stable characteristics, such as socioeconomic status or diet, can hardly explain this finding because the analysis involved the same case at two points in time, but not controls. A limitation of this observational study is that we were not able to validate the clinical diagnosis of the respiratory-tract infection, because sputum cultures, blood cultures, and serological tests are rarely done in general practice for community-acquired infections. We cannot infer whether infections with particular bacteria or viruses are associated with a higher risk of triggering AMI than others. A potential risk for some misdiagnosis exists: in theory, a patient might have had prodromal symptoms, related to the future AMI, which were misdiagnosed by the GP as acute respiratory infection. Another potential bias could have been introduced into our study if future AMI cases were generally sicker than controls, which could have led to more practice visits and to a greater likelihood of getting a diagnosis of an acute respiratory-tract infection. We cannot rule out this possibility completely, but we took several steps to reduce such a risk. We restricted the study to cases and controls who had no metabolic or circulatory disorders predisposing to AMI. The cases and controls had the same numbers of practice visits in the 3 years before the index date. This finding, however, does not imply that the two groups had similar general health for the period immediately preceding the date of the AMI. To explore further whether confounding could occur due to differential health and medical attention of cases and controls directly before the index date, we checked whether any patients had made visits to the practice for symptoms potentially related to the coming AMI (such as angina pectoris, chest pain) in the 15 days before the index date, and we counted how many visits there had been in this same period. In a further analysis, we excluded all cases (n=740) and controls (n=102) who attended the practice with prodromal symptoms. In this analysis of participants without prodromal symptoms, we adjusted for number of visits to the practice (none, one or two, and three or more) in the 15 days before the index date. The association between an acute respiratory-tract infection and the risk of AMI remained high (odds ratio 3.3 [95% CI 2.1-5.1]), which shows that the association cannot be explained simply by more visits to the practice due to prodromal symptoms of the later AMI. We also looked at urinary-tract infections, because any detection bias would presumably also apply to this infection also.There was, however, no difference between cases and controls in rates of these infections. A strength of the study is that it was not based on interviews, and thereby not prone to recall bias.[ 22, 23, 26] The GPs in the UK record data in the GPRD in the order in which the events occur, and without a study hypothesis in mind. Chance seems to be an unlikely explanation for the results; not only was the dataset large and the finding highly significant, but also there was consistency throughout the years 1994, 1995, and 1996. Furthermore, the finding was also present in the dataset of a previous study, which used a different set of cases and controls.[ 18, 19] Possible pathophysiological mechanisms by which acute infections might trigger AMI have been extensively discussed.[ 2, 9, 10, 12] Systemic inflammation related to the respiratory infection could increase concentrations of C-reactive protein; high concentrations of this protein are predictive of AMI.[ 27, 28] Systemic inflammation could also lead to leucocytosis, lower serum-albumin concentrations, alter endothelial function, or cause atheroma instability and subsequent plaque rupture.[ 2] Changes in circulating clotting factors, such as fibrinogen, could lead to an increased clotting tendency and thereby to an increased risk of thrombotic coronary occlusion.[ 1] Furthermore, acute infections could increase concentrations of inflammatory cytokines, and so inhibit the function of vasodilating nitric oxide or prostaglandins.[ 2] The absolute risk of AMI in association with acute respiratory-tract infections was low; only 4% of AMI cases in our study population had an acute infection in the 2 weeks before the AMI, the time span in which we found an increased risk of AMI. A limitation of our study is that misdiagnosis cannot be excluded with absolute certainty (ie, that prodromal symptoms of a future AMI were mistaken as a respiratory-tract infection). Further investigation of the part played by acute and chronic infections in the aetiology of AMI is important, to improve the prevention and treatment of AMI. Contributors Christoph Meier was the lead investigator who conceived the hypothesis, did the main data analysis, and wrote the first draft of the manuscript. Hershel Jick made critical contributions to the study design, the case validation, and the manuscript. Susan Jick, Laura Derby, and Catherine Vasilakis were involved in the project at all stages and made important contributions to the data analysis and the preparation of the manuscript. Acknowledgments We thank the participating general practitioners for their co-operation, and Alan Dean and his team for their generous help. The Boston Collaborative Drug Surveillance Program is supported by grants from: Astra AB, Berlex Laboratories, Bayer AG, Glaxo Wellcome, Hoffman La Roche, RW Johnson Pharmaceutical Research Institute, Novartis Pharmaceuticals, and Pfizer. Table 1: Distribution by age and sex, and characteristics of acute respiratory-tract infections (ARTI) of AMI cases and controls