TITLE: Prognostic Value of Cardiac Risk Factors and Coronary Artery Calcium Screening for All-Cause Mortality
REF: Radiology 2003;228:826-833 http://radiology.rsnajnls.org/cgi/c...tract/228/3/826
PATIENTS: cohort study of 10,377 asymptomatic individuals undergoing cardiac risk factor evaluation and coronary calcium screening with electron-beam CT. Average follow-up was 5 years
STUDY TYPE: retrospective review and analysis
FINDINGS: Those with the lowest calcium scores (10 or less) had a 99% 5-year survival, compared to those with the highest calcium score (over 1000) who had a 95% 5 year survival.
CONCLUSION: The authors state calcium scoring adds incremental information to the prediction of 5-year all cause mortality
COMMENT: The value of calcium scoring still appears to be unresolved. It is unclear as to why the authors chose the scoring categories they did. There certainly is not a linear relationship between the calcium score and risk. Furthermore, the authors' statistics do not add up. The authors also use questionable statistics, as shown by their decision to use the standard error of the mean in giving the average follow-up period instead of the standard deviation.
They give the group with a calcium score of 10 or less a baseline 5 year risk of death of 1.0%. They then say that the group with a calcium score of > 1000 has a 5 year risk of death of 5.0%. Yet in another section they state that the relative risk of a >1000 score compared to a <10 score is 4.03 which would mean that the 5 year risk of death for those with a score of >1000 is 4.03% (not 5.0%).
The authors appear to be doing an exercise in statistics, and they don't appear to be doing it very well. As we all know, if you get a group large enough, there will always be "statistically significant" statistical values that pop-up.
In summary, my objections to this article:
1) the calcium scores used to group the subjects appears to be arbitrary. I would have found simple statistics much better, e.g. giving the average calcium score of those that died, and compared this with the average calcium score of those that lived. The use of confidence intervals would have even been better. The fact that the authors chose not to report these basic statistcal values, but instead grouped the patients into arbitrary calcium score ranges is very concerning. It appears that they tried very hard to come up with proper patient grouping in order to get the result they wanted.
2) The statistical values that they gave do not add up. If the lowest group had a 1.0% risk of death, and the highest group a 5.0% chance of death, then the relative risk should be 5 -- not 4.03
I am very disappointed in this paper, because calcium scoring almost certainly has value, but the authors did not help us understand what exactly that value is. It is written by prestigious authors from respected institutions. It is truly unfortunate that they did not present their data in a statistically fair manner.