To the Editor:
We read with great interest the article entitled «Total Concentrations of Plasma Homocysteine in Puerto Rican Patients With Ischemic Heart Disease» by Rodriguo et al.1 published in the December issue of the Revista. Given the enormous pervasiveness of the subject, we would like to make some comments.
Firstly, in the Introduction, the authors comment that in Spain studies of ischemic heart disease have been focused more on the theory of the increase in cholesterol, and that there are no studies of homocysteine values in this population. There is a Spanish study2 on this topic that reported that 26% of patients with heart disease proved to have hyper-homocysteinemia.
Secondly, the authors did not determine vitamin B6, B12, and folic acid values in cases in which deficits thereof could be a nutritional cause of hyperhomocysteinemia. It has been suggested that approximately 60% of hyperhomocysteinemia is due to inadequate levels of 1 or more of these vitamins in the blood.3 Similarly, they did not comment on the dietary habits and condition of the study population, and this is probably why there was no finding of an association between heart disease and homocysteine concentration as a side effect of long- and short-term dietary variations.4 Various retrospective and prospective studies have shown the possibility that a load test would improve the ability of a fasting homocysteine measurement.5 to predict the risk of heart disease.
Thirdly, the results are expressed in an unclear manner. In Table 2, the distribution of homocysteine is grouped by age, sex, smoking habits, diabetes and arterial hypertension. The authors express homocysteine concentrations for the entire population, instead of placing them in 2 categories--those with normal coronary arteries (n=10) and those that had some degree of occlusion (n=60). In Table 3, which lists the univariate and multivariate models for the different parameters that can accelerate artery occlusion, we would like to note that recent studies have concluded that using logic regression analysis in the context of small samples requires the use of exact tests (for example, the Cytel software statistical program6). The exact test, as is well known, decreases type 1 errors associated with the conflicting theories such as those described in the study. Especially when multivariate models are used in clinical studies, the precaution must be taken to maintain a balance between the number of predictors and the number of patients in the sample. In this study, there are 70 patients with a preliminary diagnosis of heart disease with only 10 in the control group (normal coronary arteries) making it impossible to use more than 1 or 2 variables at the most in prediction for or classification of the patients.
Fourthly, the sample size that the authors present is 60 patients (all with some degree of arterial occlusion) and 10 controls (normal coronary arteries). It is well known that it is complicated to determine with coronary angiography that controls are without ischemic heart disease; nevertheless, the control group might be amplified with patients who have a negative stress test. This contradicts the conclusions of the study since the impossibility of reaching conclusions is in the actual study design itselfit lacks sufficient statistical power.