Fate of the unoperated ascending thoracic aortic aneurysm—patient selection and the importance of the denominator

This commentary refers to ‘Fate of the unoperated ascending thoracic aortic aneurysm: three-decade experience from the Aortic Institute at Yale University’, by J. Wu et al.,https://doi.org/10.1093/eurheartj/ehad148 and the discussion piece ‘Intricacies of aortic risk assessment studies: not a simple matter’, by M. A. Zafar, et al.,https://doi.org/10.1093/eurheartj/ehad795. We congratulate and thank the authors for their continued analysis of the Yale thoracic aortic aneurysm (TAA) database to better delineate natural history,1 particularly when so many patients undergo surgery and are removed from the denominator. Among unoperated patients, they found an increased relative risk of adverse events when the aneurysm surpassed 5.0 cm, with overall slow aortic growth. The authors’ conclusions deserve caution, however, since the sample represents selected high-risk patients, which cannot capture the ‘true’ rate of aortic dissection for all comers—e.g. the patient in your exam room. Considerations include: Excluded cohorts: Out of the 3914 TAA patients, only 964 (25%) who did not undergo surgery were studied. This excludes 75% of the cohort who had neither dissection nor death before surgery, which often occurs years after diagnosis. By excluding the time-at-risk for these patients, the denominator of those without events is drastically reduced. This skews the sample towards those at highest surgical risk, which usually coincides with higher risks of dissection and death from comorbidities and biases the estimates upward. Inconsistent metrics: While TAA sizes for patients with adverse events were assigned to the most recent measurement during follow-up, for those without events it is unclear which TAA size was used. If a patients’ TAA size increased but did not have an adverse aortic event, and they are categorized with their smaller initial size, this would inflate the event rate for larger TAAs. Mixed cohort: Including connective tissue patients, who represent a different risk profile than degenerative TAA patients, clouds the picture, even though they were just 2.5% of the sample. Subjective outcomes: The outcome of ‘aortic death’ is subjective and based on methods estimating cause of death for abdominal aortic aneurysm, a different (higher risk of event) phenotype than TAA. Competing risks of death are very high and autopsies rarely done. Nearly everyone dies over the long-term when you start at advanced ages. Event rate discrepancies: Event rates are higher than other large, published cohorts (Table 1), and do not resonate with clinical practice. A significant 11% of patients in the 5.0–5.4 cm range under surveillance having aortic dissection would be widely noted by aortic specialists. Risk evaluation: Clinical-decision making should be rooted in absolute risk, not relative risk. The decision to operate must be individualized to balance operative vs. surveillance risks, as surgical risk varies greatly across patients. We appreciate the authors’ transparency that, ‘At our institution, and at many other centres, the risk of death of ascending/arch replacement is only 1.9%, and the accompanying risk of stroke is only 1.4%’.2 A patient in the 5.0–5.5 cm range facing a 1.9% surgical death risk vs. a 1.5% dissection risk over five years has a challenging decision. Comparison of aortic dissection and rupture rates across three recent natural history studies1,3,4 AD, aortic dissection. 1-year event rates using index TAA size as the benchmark are: 4–4.5 (0.15%), 4.5–5.0 (0.33%), 5.0–5.5 (0.43%), 5.5–6.0 (0.30%), > 6.0 (2.19%). aSample size reflects number of patients with TAA size at index date. 1- and 5-year risk of AD/rupture reflects time-updated TAA size over follow-up. In summary, we appreciate and are grateful for this contribution, but disagree that these data represent a hinge point warranting earlier operation at 5.0 cm for most TAA patients. We look forward to results from the TITAN trial for further insights into the most appropriate TAA size for intervention.5 All authors declare no disclosure of interest for this contribution.