Evaluation of LR1/LR2 logistic regression models' precisely estimated probability of ovarian mass malignancy via ultrasonic examination and comparing their application to the widely used and so far suggested RMI (Doctoral thesis)

Σταύρος, Σοφοκλής

Background: Correct characterization of ovarian lesions is critical to optimize patient care. The purpose of this study is to evaluate the diagnostic performance of the International Ovarian Tumour Analysis (IOTA) logistic regression model (LR1), logistic regression model (LR2) and the Risk of Malignancy Index (RMI) for preoperative characterization of adnexal masses, when ultrasonography is performed by an non experienced examiner level I-II. Methods: A 9-year retrospective single-center study in the gynecological ultrasonography unit of Alexandra universal hospital. Transvaginal ultrasonography was performed using a standardized approach. The final outcome was the surgical findings and histological diagnosis which was the ‘gold standard’. To characterize the adnexal masses, the twelve- variable prediction model (LR1) with a cutoff of 0.1, the six-variable prediction model (LR2) with a cutoff of 0.1, and the RMI with cutoff of 200 were applied. The area under the curves (AUCs) for performance of LR1, LR2 and RMI were calculated. Diagnostic performance measures for all models assessed were sensitivity, specificity, positive and negative likelihood ratios (LR + and LR -), and the diagnostic odds ratio (DOR). Results: One thousand and thirty-two (1032) women with adnexal masses underwent transvaginal ultrasonography, whereas 792 had surgery. Prevalence of malignancy was 14, 6 %. The AUCs for LR1, LR2 and RMI for all masses were 0,886 (95% confidence interval (CI): 0,855-0,917), 0,839 (95% confidence interval (CI): 0,795-0,882) and 0,869 (95%CI: 0,810-0,928). The sensitivity for LR1, LR2, and RMI for all masses was 0,741, 0,724 and 0,667, while the specificity for LR1, LR2 and RMI were 0,854, 0,855 and 0,960. The twoIOTA models had higher sensitivity compared with the non IOTA RMI model. For all masses, the DORs for LR1, LR2 and RMI, were 16,707 (95% CI: 13,165-21,204), 15,482 (95% CI: 12,243-19,578), 47,846 (95% CI: 32,553-70,324). Conclusion: This is the first study in Greece including a large number of patients, which compare both LR1 and LR2 IOTA models with the well known RMI model. Eventually, the IOTA models LR1 and LR2 due to their higher sensitivity seem to be a pivotal tool for the ultrasonic evaluation of the nature of ovarian formations for the majority of cliniciangynecologists in order to identify the adnexal pathology and especially malignancy. The non IOTA RMI model remains an equal important tool for the discrimination of adnexal pathology.
Institution and School/Department of submitter: Πανεπιστήμιο Ιωαννίνων. Σχολή Επιστημών Υγείας. Τμήμα Ιατρικής
Subject classification: Ωοθήκες -- Καρκίνος
Keywords: Ωοθηκική βλάβη,υπερηχογράφημα,λογισμικά μοντέλα παλινδρόμησης,ovarian lesion,ultrasonography,logistic regression models
URI: https://olympias.lib.uoi.gr/jspui/handle/123456789/29630
http://dx.doi.org/10.26268/heal.uoi.9632
Appears in Collections:Διδακτορικές Διατριβές

Files in This Item:
File Description SizeFormat 
Δ.Δ. ΣΤΑΥΡΟΣ Γ. ΣΟΦΟΚΛΗΣ 2019.pdf7.13 MBAdobe PDFView/Open


 Please use this identifier to cite or link to this item:
https://olympias.lib.uoi.gr/jspui/handle/123456789/29630
  This item is a favorite for 0 people.

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.