Researchers at the University Medical Center Groningen (Groningen, The Netherlands) and the University of Twente (Enschede, The Netherlands) have developed a prediction model that they claim could assist in the preoperative identification of patients who are unlikely to achieve sufficient endograft apposition after endovascular aneurysm repair (EVAR).
Willemina van Veldhuizen, MD, and colleagues from the University Medical Center Groningen, including members of the Virtual Stenting study group, shared their work in a paper published online ahead of print in the European Journal of Vascular and Endovascular Surgery (EJVES).
By way of background, the authors first underline an association between challenging infrarenal aortic neck characteristics, such as short apposition (less than 10mm circumferential shortest apposition length) on the first postoperative computed tomography angiography (CTA), and an increased risk of type Ia endoleak after EVAR. It was the study team’s aim to develop a model to predict postoperative shortest apposition length in patients with an abdominal aortic aneurysm (AAA) based on the preoperative shape.
Van Veldhuizen et al note that they developed a statistical shape model to obtain principal component scores, with the dataset comprising 147 patients in total: 93 treated with standard EVAR without complications and 54 treated with EVAR who went on to experience a late type Ia endoleak.
In terms of methods, the authors detail that they obtained the infrarenal shortest apposition length from the first postoperative CTA, which they subsequently binarised between those under 10mm in length and those equal to or longer than 10mm. The researchers then used the principal component scores that were statistically significant between the shortest apposition length groups as input for five classification models, and subsequently determined area under the receiver operating characteristics curve (AUC), accuracy, sensitivity and specificity for each classification model.
Van Veldhuizen and colleagues report in EJVES that, of the 147 patients included in the study, 24 had an infrarenal shortest apposition length of less than 10mm and 123 had a shortest apposition length of 10mm or more.
The authors share that the gradient boosting model resulted in the highest AUC of 0.77 and that, using this model, 114 (78%) patients were correctly classified, with sensitivity (less than 10mm apposition was correctly predicted) and specificity (equal to or longer than 10mm apposition was correctly predicted) at 0.7 and 0.79, respectively.
Van Veldhuizen et al summarize that the model they present can predict the binarised shortest apposition length of an endograft into the infrarenal aortic neck for treatment of an AAA on the first postoperative CTA scan, with an accuracy of 78%. Given that a shortest apposition length of less than 10mm has been associated with higher risks of type Ia endoleak, the researchers posit that their model “could help vascular specialists in the preoperative phase to accurately identify patients who are unlikely to achieve sufficient apposition after EVAR.”
In the discussion of their findings, the authors acknowledge a series of limitations to their study, including the fact that the data they used came from different datasets, which may have created a selection bias.
Looking ahead, van Veldhuizen and colleagues suggest that their model “should be externally validated with a consecutive patient series that includes patients with an early type Ia endoleak before it can be used as a patient-specific virtual stenting tool in clinical practice.”
