Limitations
This is a retrospectively designed study and is therefore prone to bias.
We attempted to reduce bias by matching based on age, sex, race, body
surface area, and hypertension. We also found no significant difference
between cases and controls in many relevant areas, including smoking
status (P =0.121), connective tissue disease (P =0.619), and
bicuspid aortic valve (P >0.999). However, the two
groups were not entirely similar, as there was a significantly greater
prevalence of both diabetes and active malignancy in the control group.
In a sensitivity analyses, adjusting for these two variables did not
eliminate any of the significant effects we reported.
Blinding the researchers to case/control status was impractical since
dissection pathology was obviously apparent in image analysis. In some
cases, the dissection pathology contributed to greater difficulty in
identifying landmarks and conducting measurements. This was particularly
problematic in cases where root dilation resulted in an ambiguous STJ,
which was noted in seven cases (10.4%). However, we did not eliminate
these cases from analysis because it would complicate the case-control
matching process and likely introduce a selection bias.
The major limitation of this study is the effect of dissection pathology
on aortic length. Pre-dissection CT scans were not available for
patients in our study, thus, it should be understood that the
measurements reported here likely differ from the actual pre-dissection
values.