Impossible statistics
I read with some amusement an article by Gansera et al entitled “Does bilateral ITA grafting increase perioperative complications?: Outcome of 4462 patients with bilateral versus 4204 patients with single ITA bypass” in the August 2006 edition of the European Journal.
It was an article highlighted for CME, probably due to educational value and clinical importance, but certainly not on the quality of statistical reporting. It is surprising that an article of such standing reached the printing press with P=0.00 reported throughout the manuscript. It suggests that the authors and reviewers were unaware of the implications.
A P value is the probability of observing the given results or more extreme values given that there is no true difference (ie the “the null hypothesis is true”). Therefore P=0.00 means that the probability of observing the results or more extreme values is zero given that there is no true difference, which in itself is impossible. What the authors really meant was that the probability was less than 1% (ie P<0.01).
The tables in the manuscript were confusing, as they reported the percentages without actual numbers. Furthermore, the categories in some tables were subdivided by disease (eg 30-day mortality, table 7), therefore and the reported percentages had nothing to do with the denominators listed at the top of the table (BITA n=4462 and SITA n=4204).
The authors used multivariate instead of multivariable. A multivariable (correct term) analysis refers to one dependant variable and many explanatory variables, as per the standard logistic regression. A multivariate analysis means that there were more than one dependant variable that was analysed simultaneously, this is rarely used in medical literature, and consists of techniques such as principal component analysis and factor analysis.
Many other questionable aspects of the statistical reporting aside, I couldn’t help but feel that the article was reported in a manner biased towards BITA usage. The conclusion included:
“CABG using both ITAs can be performed routinely with good clinical results and low mortality. Compared with single ITA grafting, sternal and bleeding complications were slightly increased...”
which, after reading the paper I would rephrase as…
“CABG using both ITAs can be performed routinely (for perceived benefits from non-randomised observational studies that have usually performed BITA on selected healthier, lower risk patients) if you are willing to accept a longer duration of surgery, bypass time, cross clamp time, more bleeding, 4.8 fold increase in re-thoracotomy rate (why re-thoracotomy instead of re-sternotomy?), 2 fold increase sternal dehiscence and 3.5 fold increase in mediastinitis, with low mortality ”