To our knowledge, no study has previously sought to characterize inframortality due to ischemic heart disease in Spain. Studies that have analyzed the geographical distribution of vascular mortality and other causes of death, have focused on identifying areas with excess mortality [6,7] or been targeted at studying the impact of specific factors, such as social inequalities [8-10] and environmental [11-14] or dietary factors, [15-17] on increased risk of dying. At an international level, what are now regarded as classic ecological studies [18,19] drew attention to the lower mortality in southern European countries, but analyses on the Mediterranean paradox and its link to diet and hypercholesterolemia [20-22] used population clusters that were too large and, by extension, too heterogeneous (e.g. country), and failed to take multifactorial and contextual aspects of the disease into account. Many studies have shown that the socioeconomic characteristics of the local setting have an influence on individual cardiovascular risk, even when adjustment is made for individual risk [23].

In the last decade, the development of spatial epidemiology and epidemiology of the social context has generated new theoretical and methodological frameworks, which enable the study of low coronary mortality to be addressed with a new approach of greater potential [24].

These towns were distributed across the northern half of the country, and belonged to the provinces of Burgos (10), Cuenca (5), Cantabria (3), Palencia (3), Teruel (2), Navarre (2), Pontevedra (2), Lugo (1), Orense (1), Tarragona (1), Salamanca (1), Segovia (1) and Soria (1). Only two were situated on the coast (Suances in Cantabria, and Deltebre in Tarragona).

The map plotted in Figure ​Figure11 shows that 23 of the 32 towns with inframortality were geographically grouped into four areas, namely: central Galicia (marked in the figure as area 1); provinces of Burgos and Palencia in northern Castille-Leon (area 2); province of Cuenca (area 3); and western Navarre (area 4).

In comparison with the remainder (Table ​(Table2),2), towns with coronary inframortality were located in provinces with low population density (p 0.023), and had few inhabitants, though the difference vis-à-vis the mean size of the country's remaining towns did not prove statistically significant. Insofar as socioeconomic characteristics were concerned, the towns selected displayed lower proportions of illiteracy, unemployment, aged population, population engaged in farming, and deprivation index, and a higher socioeconomic index of disposable family income. These differences were small, however, and failed to reach statistical significance, with the exception of the lower proportion of the economically active population devoted to farming (28.5% versus 36.0%, p 0.046), and with marginal statistical significance, the smaller proportion of unemployed population (11.6% versus 15.4, p 0.066).

The dietary pattern of the geographic setting of these towns was characterized by a calorie content that was slightly higher than the mean. Fish and wine consumption was much higher in towns with inframortality, 23.6% and 40% respectively (p < 0.001), as was that of alcohol in general (27.8% p < 0.001). Although the prevalences of smoking, sedentarism, obesity and diabetes were significantly lower than those of controls from a statistical standpoint, the difference was small (1%). No differences were observed in the prevalence of perceived arterial hypertension and hypercholesterolemia.

All-cause and tumor-related mortality were 20% lower than the national standard and equal to the mean SMR of the remaining towns. Longevity was longer by 2 years in such towns. Furthermore, the physician-population ratio was equal to and the hospital bed-population ratio was higher than that of control towns, and mortality due to ill-defined causes was 70% lower than the national standard.

Of the factors considered, those that went furthest to explain municipal inframortality (Table ​(Table3)3) were, at a local level, the lower proportion of the population aged 65 years and over, and at a provincial level, the lower prevalence of obesity and the higher consumption of wine and, in particular, fish. No significant associations were found with the frequencies of arterial hypertension, hypercholesterolemia or smoking. Paradoxically, frequency of leisure-time sedentarism was positively associated with coronary inframortality. The fitting sequence showed that the absence of association of some risk factors in model 2 was caused by confounding due to the effect of the protective factors.

The results of the two sensitivity analyses conducted are presented in Table ​Table4.4. First, the analysis was repeated, solely considering towns of over 1000 inhabitants (Analysis I) and adding indicators, such as the industrial index, which were exclusively available for this group of towns. The results were similar but the provincial physician-population ratio was strongly associated with municipal inframortality. Second, we conducted a sensitivity analysis modifying the selection criteria to test if selection of these towns affect the results (Analysis II). Although this new analysis was based in a quite different set of towns with different characteristics (ie, small villages with older population and higher income), the main results, namely diet as the factor most associated with low mortality, were basically unchanged.

The results suggest that coronary inframortality in the towns selected, half the national rate and as much as 5 times lower than the rate of northern European countries, [1] is not attributable to underdiagnosis resulting from lack of health care resources, poor quality in the certification of cause of death, or competitive mortality because mortality due to other causes, and to cancer in particular, is likewise low. The greater longevity of the inhabitants of these towns serves to further support the fact that coronary inframortality is real, since there is practically no possibility of there being an unregistered death in Spain. It also implies that these populations exhibit a low risk of dying for any cause, fact that may be explained by many vascular risk and protective factors being common to other diseases, in fact the most frequent ones, such as cerebrovascular disease and cancer

Geographically, these towns with extremely low ischemic heart disease mortality are situated in sparsely populated provinces in the northern half of the country and, unexpectedly, reveal a trend towards geographical aggregation. Although the towns identified are located within the large areas of low mortality shown on official atlases [4,5] and the methods used favor the appearance of geographical aggegations, the possibility cannot be ruled out that grouping may be due to the presence of factors with a spatial pattern -whether of a protective nature or entailing lower exposure to risk factors- which were not taken into account in the analysis. Hence, certain environmental factors (such as climatic factors, composition of the local drinking-water supply or environmental pollution) could not be considered, since these data were unavailable for many towns in the country, including many of the towns identified. Similarly, genetic factors could not be considered and, while genetic population studies undertaken in Spain show that the Spanish population is homogeneous in terms of overall genetic structure, fine structure analyses nevertheless reveal a geographic variation that may be more evident in small, rural or isolated samples [37]. Consequently, one cannot rule out the possibility that, in the case grouping of small towns, there may be some genetic characteristic which confers low vascular risk. One example of a town having low mortality attributable to genetic features is Limone sul Garda, a small town in the north of Italy, whose inhabitants present with a mutation in the apolipoprotein A1 (ApoA1), which confers cardiovascular protection and increases longevity [38,39]. Taking all together, the results indicate that further in-depth studies of this spatial agregation are needed.

With respect to the characterization of these towns, when compared to the rest of the country, inframortality was linked to their having: a less aged population structure; a contextual dietary pattern characterized by a higher fish and wine content, lower calorie content and a lower prevalence of obesity; and, in towns of over 1,000 inhabitants, a higher physician-population ratio. In line with the results of earlier studies on social inequalities and health in Spain, [40-42] the strength of association between economic level and coronary inframortality decreased when adjustment was made for diet and vascular risk factors, a finding which might be interpreted as indicating that the latter are intermediate factors and a possible explanation for this association [43]. Nevertheless, the difference in income between these and the remaining towns is small: in other words, they are towns with an income level which, albeit higher, is not excessively so.

Low prevalence of obesity and high consumption of fish appear as the variables most closely associated with coronary inframortality. The effect of both of these on risk of coronary disease has been clearly demonstrated in prospective population studies conducted in a number of countries [44,45] and is also reflected at a population-cluster level in ecologic studies [46]. In Spain, fish consumption is very high in comparison with other western countries, with the communities of Cantabria and Castile-León registering the highest intake [35]. The high mortality of some Spanish provinces has already been associated with lower fish consumption by previous studies [17]. Insofar as obesity is concerned, the result is coherent with recent studies, which reckon that, in the Spanish population, risk of ischemic heart disease attributable to excessive weight can be assumed to be very high and even higher than that posed by the classic risk factors [47,48]. While mean calorie intake per person per day was somewhat higher (90 Kcal) in towns with inframortality than in the remaining towns, this difference might nevertheless be accounted for by a higher wine consumption. Indeed, the adjusted results show an inverse association between calorie intake and inframortality. Consumption of wine also showed an association with coronary inframortality, in agreement with many studies that have reported the protective effect of moderate wine consumption [17,49,50]. The positive association between leisure-time sedentarism and inframortality is paradoxical and can only be interpreted by uncontrolled confounding, e.g., due to physical exercise during the work day.

The explanation for the low coronary mortality of these towns does not lie in differences in the prevalence of arterial hypertension and hypercholesterolemia or diabetes. This may be real or, alternatively, it may be an artifact, due to the data having been drawn from a survey [36] and the fact that the prevalence reported is perceived, i.e., diagnosed and possibly treated. Similarly, these towns' low coronary mortality is not explained by differences in the prevalence of smoking.

Lastly, not only are the health care resources of towns with coronary inframortality no greater than those of the rest of the towns, but they actually have fewer high-tech resources, since such resources have been placed in high mortality areas. Nevertheless, when adjustment was made for the remaining factors and the analysis was confined to towns with more stable mortality figures, the physician-population ratio was observed to be positively associated with inframortality in towns with over 1,000 inhabitants.

The possibility of results being biased because the restrictive criteria used to define inframortality can be ruled out as sensitivity analysis modifying the selection criteria to include 397 towns in which there were no coronary death for 15 years yieded similar results, namely dietary factors as the factor most associated with low mortality.

The variables analyzed were based on the information available, in some cases incomplete, such as environmental data, or nonexistent, such as genetic data. While obtaining these data for all Spanish towns is not feasible, studying representative population samples of such towns and of other control towns is possible. This study enables future studies to be steered in this direction. Similarly, there is no data at a municipal level on the factors most directly implicated, such as diet or the prevalence of vascular risk factors. In such cases, provincial data were used by way of giving a description of the context in the absence of the pertinent data. The use of multilevel and spatial models minimizes the biases of this approach [51]. Furthermore, in aspects such as diet and its health-related consequences, the context not only determines individual behavior, but also has an influence on cardiovascular risk, even when individual risk is adjusted for [23,24]. However, although we used hierarchical regression models with province-specific random intercepts to adjust the association between municipal coronary mortality for between-province differences in any relevant factor, residual confounding induced by within-province variations in cardiovascular risk factors and diet cannot be ruled out. Lastly, the dietary variables, coming from the only nation-wide nutritional study available in Spain, refer to 1991 which is eight years before the beginning of the study period, and can be temporarily inadequate to mortality data. Nevertheless, dietary factors take time to cause coronary disease and to lead to death, so an eight years lag can be regarded as adequate. Moreover, the nutrition data used are not intake measures, but the composition of the dietary pattern in households, which do not change in short time. Despite these considerations, the stength of the dietary associations with low coronary mortality found in this study deserves further investigations with in depth nutritional study.