Abstract 1122‐000046: A Latin American Model of the PHASES Score

Abstract

Introduction : The PHASES score was developed to predict the 5‐year risk of rupture for intracranial aneurysms (IAs). However, only populations from North America, Europe, and Japan were included in the original study. As the population of origin is an item in the score, it has yet to be applied in a Latin American population. We aimed to determine the best approximation to employ this model in this previously unstudied population. Methods : We extracted the data of 848 Peruvian patients with ruptured (n = 486) and unruptured (n = 362) IAs from 2010 to 2020. According to the PHASES score, the North American and European (other than Finish), Japanese, and Finnish populations are rated with 0, 3 and 5 points, respectively. Therefore, we developed three PHASES‐derived models in which our Peruvian population is rated with 0 (Model A), 3 (Model B), and 5 (Model C) points. We compared the observed probability of each model to the expected probability reported by the original PHASES score using a scatter plot. We then compared the goodness‐of‐fit of each model using the Hosmer‐Lemeshow test in STATA version 14. Results : Nineteen percent of the patients were female. Hypertension was found in 34% of patients and 15% were >70 years. Fifty‐four percent of the aneurysms were smaller than 7mm, 25% ranged between 7 and 9.9mm, 18% were between 10 and 19.9mm, and 3% were larger than 20mm. Previous subarachnoid hemorrhage was found in 4%. The location of the aneurysms was the internal carotid artery in 4%, the middle cerebral artery in 4%, and arteries of the anterior and posterior circulation (including the anterior and posterior communicating artery) in 92%. When Model A was applied, 63% of the patients among the ruptured subgroup have an estimated 5‐year risk of rupture of <3% while 77% of the patients have an estimated risk of <3% in the unruptured subgroup. When Model B was applied, 30% of the patients among the ruptured subgroup have an estimated 5‐year risk of rupture of <3% and 42% of patients among the unruptured subgroup have an estimated risk of <3%. When Model C was applied, 96% of the patients among the ruptured subgroup have an estimated 5‐year risk of rupture of >3% while in the unruptured subgroup an estimated risk of <3% was observed only in 4% of the patients. When comparing observed to expected frequencies, model B presented a better calibration to the values reported by the original PHASES score. Additionally, the Hosmer‐Lemeshow showed Model B to have improved goodness‐of‐fit, compared to other models, although all presented adequate fit. Conclusions : We found that rating the Peruvian population with 3 points was the best approximation to the estimated risk calculated by the PHASES score to predict the 5‐year risk of rupture for IAs.