Knowledge of the magnitude of IHD risk, along with identification of the factors responsible and populations at greatest risk, are fundamental for the planning and assessment of preventive and health-care strategies and identification of specific excess-risk scenarios. There is a dearth of such information in Spain. This study sought to describe the epidemiology of ischaemic heart disease and acute myocardial infarction in a cohort drawn from the population attending primary-care centres.
Epidemiological studies conducted in a primary care setting can be subject to several limitations, that must be taken into account [26
]. Firstly, this cohort cannot be regarded as representative of the general population but only of those who use primary-care services, although it should be stressed that the Spanish National Health Service covers the entire population and that, according to the Spanish National Health Survey, 80% of the Spanish adult population attends such health centres at least once a year [27
]. In this study, the age and sex distribution of the cohort was very similar to that of the general population, but substantial differences in frequentation to primary care by age and sex have been well described in many studies, so it is likely that detection and exposure classification biases could be present; to minimize this potential biases the study protocol included systematic search of events and exposures during follow-up and at the end of the study period. This systematic approach explains the very low percentage of patients lost for follow-up. Anyway, we must further remark that the study population is not a random sample of general population, so any generalization of the results should be cautious. Furthermore, the cohort is not a sample but the entire population covered by the participant health centres but these were not randomly sampled, this meaning that the results are applicable only to the study population, and of course this population is not representative of other populations.
Clinical trials have demonstrated that effective medical control of vascular risk factors reduces risk, so classifying treated patients as exposed (as has traditionally been done) could have biased the results, mostly when the study was conducted in a medical setting. In the same manner, including treated patients as not exposed would have also introduce a bias because effective control is not always achieved. This potential exposure misclassification derived from medical treatment and control has been approached by taking the mean of repeated measures for arterial pressure, serum cholesterol and glucose, so treated patients were therefore classified as exposed when the mean of the repeated measures was above the cut-off values; this way, not controlled patients remained as exposed and controlled patients were classified either as exposed / not exposed depending on the degree of effective control.
In the same way, multiple imputation of missing data was used to prevent biases derived from the aggregation of missing data in different variables when considering them jointly in multivariate analyses [24
The age-adjusted acute myocardial infarction incidence rates registered in this study are similar to those of the IBERICA study for the general population registers of 7 Spanish regions as a whole, the MONICA-Catalonia, the REGICOR and the IBERICA-Murcia studies [2
]. Nevertheless, it should be borne in mind here that our study targeted an older population, i.e., whereas in the above studies the study population was aged 25–75 years, in our cohort 10% of women and 5% of men were over 75 years old. Still, when this age breakdown was taken into account, the incidence rates in our cohort were almost the same as those found in the general population, despite the fact that this was a medical-settled study. The likelihood of information on events in the cohort being lost is practically non-existent, since the researchers were the designated general practitioners of the patients enrolled, and therefore responsible for their medical follow-up and prescription of drugs in the case of cardiovascular episodes. At all events, on termination of the study all patients who had not attended the primary-care centre in the preceding 9 months were contacted by telephone.
In Spain, only two other cohort studies have been conducted to date, aimed at measuring and analysing the risk of IHD, namely the Manresa and DRECE studies, though their results are not entirely comparable vis-à-vis our study or even each other, due to differences in methodology and design [6
]. Taking the above limitations on comparability into account, it can be concluded that the incidence rates registered in the present study do not diverge from those published in Spain until now, and that any differences are attributable to design. This incidence is very high compared to Spanish incidence rates for other chronic diseases, and indeed exceeds the joint incidence of colon, rectal, lung and breast cancer combined [29
]. In the international context, the myocardial infarction incidence rate in this population is very low compared to those reported for non-Spanish populations in the MONICA study [2
] and under-65 mortality is lower than that reported in cohorts included in the SCORE project, though these cohorts were studied 20–30 years ago [15
IHD incidence in this cohort should be interpreted in the light of its level of cardiovascular risk. Compared to the general Spanish population, this cohort registered comparable prevalences of arterial hypertension and diabetes, slightly higher prevalences of overweight and hypercholesterolaemia, and a lower prevalence of tobacco use. The differences in risk profile in studies targeting persons attending primary-care centres is known and has been previously reported [9
]. However, with the sole exception of the low frequency of smoking (34% and 13% among men and women respectively versus 40% and 25% in the general population), the differences in this study are not unduly pronounced, something that would lead one to expect IHD incidence rates to be within the range of those reported by population-based studies in Spain, as is indeed the case.
Analysis of the association between cardiovascular risk factors and risk of ischaemic heart disease yields results in line with what was expected, with age, smoking, elevated LDL-cholesterol, hypertension, diabetes and overweight emerging as linked to the risk of suffering an event. This is the first ever cohort study in Spain to show these associations, since previous studies were unable to rule out the null hypothesis for overweight, dyslipaemia or diabetes [6
]. Insofar as arterial hypertension is concerned, the absence of statistical significance in the association among women may be reflecting a better medical control of AHT than among men. Information on treatment and effective control of AHT was not collected in this study, and we believe that further research is needed to identify possible gender differences in AHT awareness, treatment and control in Spain.
Lastly, this study shows the feasibility of conducting epidemiological studies in primary-care settings. Furthermore, the opportunities offered by the primary health care setting were not utilised fully in this study, i.e. treatment information was not recorded when it would have been feasible and desirable. Particular strenghths were found, as the minimization of follow-up and event losses or diagnostic misclassifications. Potential weaknesses, as selection biases due to demographic differences of primary-care frecuentation can been also minimized by an adequate, systematic approach. This systematic approach should also be used to prevent the miss of data, a possibility always present in large cohort studies, moreover with the high work load of primary-care doctors.