Strategies to diagnose ovarian cancer: new evidence from phase 3 of the multicentre international IOTA study
Author(s)
Type
Journal Article
Abstract
Background:
To compare different ultrasound-based international ovarian tumour analysis (IOTA) strategies and risk of malignancy index (RMI) for ovarian cancer diagnosis using a meta-analysis approach of centre-specific data from IOTA3.
Methods:
This prospective multicentre diagnostic accuracy study included 2403 patients with 1423 benign and 980 malignant adnexal masses from 2009 until 2012. All patients underwent standardised transvaginal ultrasonography. Test performance of RMI, subjective assessment (SA) of ultrasound findings, two IOTA risk models (LR1 and LR2), and strategies involving combinations of IOTA simple rules (SRs), simple descriptors (SDs) and LR2 with and without SA was estimated using a meta-analysis approach. Reference standard was histology after surgery.
Results:
The areas under the receiver operator characteristic curves of LR1, LR2, SA and RMI were 0.930 (0.917–0.942), 0.918 (0.905–0.930), 0.914 (0.886–0.936) and 0.875 (0.853–0.894). Diagnostic one-step and two-step strategies using LR1, LR2, SR and SD achieved summary estimates for sensitivity 90–96%, specificity 74–79% and diagnostic odds ratio (DOR) 32.8–50.5. Adding SA when IOTA methods yielded equivocal results improved performance (DOR 57.6–75.7). Risk of Malignancy Index had sensitivity 67%, specificity 91% and DOR 17.5.
Conclusions:
This study shows all IOTA strategies had excellent diagnostic performance in comparison with RMI. The IOTA strategy chosen may be determined by clinical preference.
To compare different ultrasound-based international ovarian tumour analysis (IOTA) strategies and risk of malignancy index (RMI) for ovarian cancer diagnosis using a meta-analysis approach of centre-specific data from IOTA3.
Methods:
This prospective multicentre diagnostic accuracy study included 2403 patients with 1423 benign and 980 malignant adnexal masses from 2009 until 2012. All patients underwent standardised transvaginal ultrasonography. Test performance of RMI, subjective assessment (SA) of ultrasound findings, two IOTA risk models (LR1 and LR2), and strategies involving combinations of IOTA simple rules (SRs), simple descriptors (SDs) and LR2 with and without SA was estimated using a meta-analysis approach. Reference standard was histology after surgery.
Results:
The areas under the receiver operator characteristic curves of LR1, LR2, SA and RMI were 0.930 (0.917–0.942), 0.918 (0.905–0.930), 0.914 (0.886–0.936) and 0.875 (0.853–0.894). Diagnostic one-step and two-step strategies using LR1, LR2, SR and SD achieved summary estimates for sensitivity 90–96%, specificity 74–79% and diagnostic odds ratio (DOR) 32.8–50.5. Adding SA when IOTA methods yielded equivocal results improved performance (DOR 57.6–75.7). Risk of Malignancy Index had sensitivity 67%, specificity 91% and DOR 17.5.
Conclusions:
This study shows all IOTA strategies had excellent diagnostic performance in comparison with RMI. The IOTA strategy chosen may be determined by clinical preference.
Date Issued
2014-08-12
Date Acceptance
2014-05-14
Citation
British Journal of Cancer, 2014, 111 (4), pp.680-688
ISSN
1532-1827
Publisher
Cancer Research UK
Start Page
680
End Page
688
Journal / Book Title
British Journal of Cancer
Volume
111
Issue
4
Copyright Statement
© 2014 Cancer Research UK
Identifier
https://www.nature.com/articles/bjc2014333
Subjects
Science & Technology
Life Sciences & Biomedicine
Oncology
LOGISTIC-REGRESSION MODELS
PROSPECTIVE EXTERNAL VALIDATION
MALIGNANT ADNEXAL MASSES
SIMPLE ULTRASOUND RULES
TUMOR-ANALYSIS
MATHEMATICAL-MODELS
PATTERN-RECOGNITION
PREDICTION MODELS
CLUSTERED DATA
BENIGN
Publication Status
Published
Date Publish Online
2014-06-17