SEARCH

SEARCH BY CITATION

To the Editor

  1. Top of page
  2. To the Editor
  3. Reference

The paper ‘Is the use of cholesterol in mortality risk algorithms in clinical guidelines valid? Ten years prospective data from the Norwegian HUNT 2 study’, by Petursson et al. [1] raises a number of questions concerning the analysis of longitudinal data in observational studies, but we were particularly surprised to see that there were no associations between total cholesterol and cardiovascular mortality in the HUNT 2 study [1]. However, there seem to be inconsistencies in the results. In Petursson et al. [1], table 1 gives these figures in the four cholesterol groups when dividing the number of deaths with the number of persons: 1.0, 1.6, 2.5 and 3.5% for cardiovascular disease among men. This translates into the following relative risks: 1.0 (Reference), 1.6, 2.5 and 3.5, which are quite different from the reported ones: 1.00, 0.80, 0.87 and 1.05. This discrepancy can hardly be explained by the adjustment. Similar differences apply to all three cause groups and both sexes. For instance, all cause mortality in women gives these figures: 1.3, 2.8, 5.2 and 7.1%, which give relative risks of: 1.0 (reference), 2.2, 4.0 and 5.5, which reveals quite the opposite trend of what is presented in the table. How is the analysis time (age) handled in the Cox analyses?

Reference

  1. Top of page
  2. To the Editor
  3. Reference