Wine Consumption and Cerebrovascular Disease Mortality in Spain
To the Editor:
Artalejo et al1 2 3 have studied the association between various diet and lifestyle habits and the risk of cerebrovascular disease (CVD) and ischemic heart disease (IHD) mortality in Spain. The CVD mortality showed considerable variation within the 50 Spanish provinces, and high wine intake was observed in regions with low CVD mortality rates.
In one article,1 it was shown that the correlation coefficient between wine intake and standardized mortality rate (SMR) from CVD was −0.26 (P=0.061). In subsequent multiple correlation analyses (including illiteracy, hypertension, sedentary lifestyle, wine, high wine consumption, blond [light?] cigarettes, diabetes, and body mass index of ≥30), wine and high wine consumption had a partial correlation coefficient of −0.12; (P=0.43) and 0.14 (P=0.37), respectively. When the analysis was repeated after exclusion of illiteracy because there was correlation with body mass index and blond cigarettes, the partial correlation coefficient for wine and high wine consumption was −0.32 (P=0.034) and 0.28 (P=0.065), respectively. The authors claimed in the beginning of the “Discussion”1 that lower wine consumption might explain the higher CVD mortality in the southern and eastern parts of the country. In the last paragraph of the same article they stated that IHD mortality showed the same pattern, ie, wine consumption was associated with a “higher” mortality. Actually, the article about IHD mortality3 showed that wine consumption had a partial correlation coefficient of −0.31 (P=0.038), while high wine consumption had a partial correlation coefficient of 0.35 (P=0.037). However, in neither of these 2 articles did the authors clearly define “high wine consumption.” The crude correlation coefficients seemed to express a straightforward beneficial effect of increasing intake of wine. Was “high wine consumption” defined according to criteria before initiation of the study, or was it chosen after seeing a scatterplot of wine intake and SMR, a plot that would have been very valuable to the readers?
In an article published in Stroke in August 1998, Artalejo et al2 reported that an increase in fruit and a decrease in wine consumption from 1964 to 1980 may have contributed to the decline in CVD mortality in Spain during 1975–1993, using the same material.2 They based this on the findings of a partial correlation coefficient for percentage change in wine consumption of 0.30 (P=0.04) in a multiple linear regression analysis, including fruit, fish, vegetables, tobacco, and illiteracy. The authors stated that the data were consistent with those in a previous study,1 in which they found that excess wine consumption was associated with higher CVD mortality across regions in Spain.
The conclusion that the percentage change in wine consumption is responsible for the decline in CVD mortality does not seem to be warranted. Changes in percentage of consumption will depend on the existing intake, and a 10% decrease in a high-intake region is very different from a 10% decrease in a region with low intake, in terms of absolute values. This becomes even more interesting, as the authors have found that at least some wine intake was associated with a reduced risk of CVD mortality.1 While the authors in their previous 2 articles estimated the effect of an undefined “high wine intake” group, it did not appear in the latest article.2 It would have been interesting to see whether CVD changes followed the changes in percentage wine consumption equally for regions with high, medium, and low wine intakes.
- Copyright © 1999 by American Heart Association
Artalejo FR, Guallar-Castillón P, Gutiérrez F, Banegas JR, Calero JR. Socioeconomic level, sedentary lifestyle, and wine consumption as possible explanations for geographic distribution of cerebrovascular disease in Spain. Stroke.. 1997;28:922–928.
Artalejo FR, Guallar-Castillón P, Banegas JR, Manzano BA, Calero JR. Consumption of fruit and wine and the decline in cerebrovascular disease mortality in Spain (1975–1993). Stroke.. 1998;29:1556–1561.
Artalejo FR, Banegas JR, Colmenero CG, Calero JR. Lower consumption of wine and fish as a possible explanation for higher ischaemic heart disease mortality in Spain’s Mediterranean region. Int J Epidemiol.. 1996;25:1196–1201.
We thank Drs Truelsen and Gronbaek for their interest in our work on the determinants of the geographic distribution of ischaemic heart disease and cerebrovascular disease (CVD) mortality in Spain.R1 R2 R3 They raise three main issues. The first is whether the model regressing CVD mortality on wine consumption was specified before or after examining the data. The second issue concerns the form of measuring changes in CVD and its determinants, and the third refers to the consistency between the cross-sectionalR2 and the longitudinalR3 analysis of the data.
Regarding the first issue, model specification was theory driven and was carried out before watching the data. Most information on the relationship between alcohol consumption and cerebrovascular disease suggest that the relationship is U- or J-shaped.R4 Therefore, the model should initially include a “wine” term to describe the left part of the relationship, and a “wineR2 ” term to describe the right part. Once the model was fitted, we checked its appropriateness against the data by examining scatterplots and by residual analysis. Finally, we also checked that a polynomial model was a better description of the data than a model without a quadratic term. The correct interpretation of the wine and wineR2 terms is not low and high wine consumption, but rather lower and higher consumption across the provincial distribution of wine consumption in Spain. The limit between lower and higher wine consumption in our data is approximately 2 dL/person per day (approximately 24 g/person per day of alcohol, assuming that alcohol volume in wine is 12%). However, this figure should be valued with caution because of the limited number of observations involved (Figure⇑).
Regarding the second issue, we have used relative measures of change to correct for baseline values of the study variables. This is common practice in etiologic studies, in the same way as exposure effects are measured by relative as opposed to absolute measures.R5 Obviously, we agree that a fixed relative change can correspond to different absolute changes, depending on the baseline values. However, the existence of only 50 provinces (observations) precludes any meaningful analysis stratified by categories of baseline wine consumption.
Regarding the third issue, there is not a quadratic term in the model of the longitudinal analysis. It was removed from an initial model after we verified that in this case the quadratic term did not contribute to the description of the data.
There is no inconsistency in the results of the cross-sectionalR2 and longitudinalR3 analyses. In the period 1964–1980, there has been a decline in wine consumption in Spain. This decline affected to a greater extent those provinces with higher consumption in 1964. Therefore, the decline has been more important in the provinces situated in the right part of the wine distribution, having a favorable effect on the CVD mortality of those provinces and therefore contributing to the decline in CVD mortality in Spain overall.
Rodríguez Artalejo F, Banegas JR, García Colmenero C, del Rey Calero J. Lower consumption of wine and fish as a possible explanation for higher ischemic heart disease mortality in Spain’s Mediterranean region. Int J Epidemiol.. 1996;25:1196–1201.
Rodríguez Artalejo F, Guallar-Castillón P, Gutiérrez-Fisac JL, Banegas JR, del Rey Calero J. Socioeconomic level, sedentary lifestyle, and wine consumption as possible explanations for geographic distribution of cerebrovascular disease mortality in Spain. Stroke.. 1997;28:922–928.
Rodríguez Artalejo F, Guallar-Castillón P, Banegas Banegas, JR, de Andrés Manzano B, del Rey Calero J. Consumption of fruit and wine and the decline in cerebrovascular disease mortality in Spain (1975–1993). Stroke.. 1998;29:1556–1561.
Camargo CA Jr. Moderate alcohol consumption and stroke: the epidemiologic evidence. Stroke.. 1989;20:1611–1626.
Greenland S, Rothman KJ. Relative versus absolute measures. In: Greenland S, Rothman KJ. Modern Epidemiology. 2nd ed. Philadelphia, Pa: Lippincott-Raven; 1998:51–52.