Article Text
PDF
Abstract
Introduction/Background Clinical guidelines for pelvic SLNB in endometrial cancer (EC) do not address the need for evaluation of the aortic region. Isolated aortic involvement in EC is rare. However, in selected groups, the incidence increases, nearby 25%. Moreover, >50% of the cases with pelvic involvement also exhibit aortic involvement. The objective of this study is to develop a prediction model for aortic involvement to guide SLNB, based on preoperative risk factors.
Methodology We evaluated the area under the ROC curve of a prediction model for aortic lymph node involvement using logistic regression, constructed with 376 women who underwent surgery for EC at the University Hospital Donostia (August 2014 - July 2022).
Results The prediction model demonstrated good discrimination, with a c-index of 0.82, and explained 29.33% of the variability in aortic lymph node involvement.
To assess the clinical utility of the model, a decision curve analysis was conducted. Firstly, the net benefit graph was created, not performing aortic lymph node assessment in any patient. It can be observed that the strategy of performing aortic BSGC based on the risk predicted by the prediction model is superior to performing it only in patients with preoperative risks. The use of the model is also superior for the majority of the probability ranges, until the match at 3%. This is because 3% is the minimum predicted probability by the model, so its results are the same as performing BSGC in all cases. Morover, the net true negatives graph was created, using the strategy of performing aortic BSGC in all patients, as is done at the University Hospital Donostia.
Conclusion The graph demonstrates that using the prediction model to restrict aortic lymph node assessment to patients with a predicted risk above a certain threshold would result in a significant reduction of unnecessary evaluations.
Analysis of decision curves. Net benefit curves (A) and net reduction in explosions (B). The abscissa axis represents the threshold probability, the limit risk from which the performance of the SLNB would be considered. The net benefit is equivalent to the proportion of true positives in absence of false positives. Thus, for example, the prediction model has a net benefit of around 0.03 at the threshold probability of 10%, which would be equivalent to detecting 3 patients with aortic lymph node involvement without indicating any necessary SLNB for every 100 patients. The maximum value of the net benefit is equal to the prevalence, which occurs when the threshold risk is 0, or at whatever threshold in which classification is perfect (no false positives or false negatives). The net reduction in burst on the other hand is equivalent to the proportion of true negatives in the absence of false negatives. For example, at a threshold probability of 10%, performing SLNB based on the risk estimated according to the prediction model is equivalent to a strategy that reduces the rate of aortic SLNB by around 62% without overlooking any affected aortic SLN.
Disclosures .