1. UOG Journal Club: January 2013
Improving strategies for diagnosing ovarian cancer: a summary of the International Ovarian Tumor Analysis (IOTA) studies
J. Kaijser, T. Bourne, L. Valentin, A. Sayasneh, C. Van Holsbeke, I. Vergote,
A. Testa, D.Franchi, B. Van Calster, D. Timmerman
Volume 41, Issue 1, Date: January 2013, pages 9–20
Journal Club slides prepared by Ligita Jokubkiene
(UOG Editor for Trainees)

2. Correct discrimination between benign and malignant ovarian masses
Previous studies limited by:
small sample size
single-center population
different tumor types
not standardized ultrasound terms and definitions
lack of consistency in histological reports

3. Aims of the IOTA studies
Improving strategies for diagnosing ovarian cancer: a summary of the International Ovarian Tumor Analysis (IOTA) studies
Kaijser et al., UOG 2013
Aims of the IOTA studies
To develope rules and models to characterize ovarian pathology
To test the diagnostic performance of rules and models by external validation with examiners of different levels of ultrasound experience
To establish the role of CA 125 and other serum tumor markers for the diagnosis of ovarian cancer
To identify the characteristics of ovarian tumors that are difficult to classify as benign or malignant
To validate these models or rules in non-operated patients by studying the outcome of adnexal masses classified as benign

5. IOTA phase 1 Training set Test set
1066 non-pregnant women
At least one persistent adnexal mass
Nine clinical centers in five countries
Training set
754 (71%) patients
Test set
312 (29%) patients
Two logistic regressions models developed
(LR1 and LR2)
Timmerman et al, J Clin Oncol, 2005

6. Variables used in the logistic regression models
Personal history of ovarian cancer
Current hormonal therapy
Age of the patient*
Maximum diameter of the lesion
Pain during examination
Ascites*
Blood flow within a solid papillary projection*
Purely solid tumor
Maximum diameter of the solid component*
Irregular internal cyst wall*
Acoustic shadows*
Color score
LR1
(12 variables)
*LR2
(6 variables)
Timmerman et al, J Clin Oncol, 2005

7. Prospective validation of the models
IOTA phase 1b
507 consecutive women
Three centers
Prospective validation of the models
IOTA phase 2
997 patients in twelve new centers and
941 patients in seven centers from phase 1
External validation of the models
JVan Holsbeke et al, Clin Cancer Res, 2009 and 2012; Timmerman et al, UOG 2010

8. Simple ultrasound-based rules
Based on subjective assessment of ultrasound images
Rules could be applied to 77% of ovarian tumors
Classify tumors as benign, malignant or inconclusive
Included into RCOG guideline for evaluating ovarian pathology in premenopausal women
Timmerman et al, UOG, 2008

9. Features of a benign mass (B-features)
A mass is classified as benign if at least one B-feature is present and no M-features are present

10. Features of a malignancy (M-features)
A mass is classified as malignant if at least one M-feature is present and no B-features are present

11. Simple ultrasound-based rules
If the rules are inconclusive if no B/M-features are present or both B and M features are present...
... rely on subjective assessment by an expert ultrasound examiner as a second stage test

12. Diagnostic performance of the models and rules
External validation
ROC AUC Sensitivity Specificity LR LR-
Similar diagnostic performance between LR1 and LR2
LR1
cut-off 10%
% %
LR2
cut-off 10%
% %
Simples rules*
N/A 90% %
RMI
% %
* Simple rules supplemented with subjective assessment of ultrasound findings when the rules could not be applied. IOTA phase 2.

13. Diagnostic performance of the models and rules
LR1, LR2 and simple rules had similar diagnostic performance in IOTA phase 1b and phase 2 datasets
Timmerman et al, BMJ, 2010

14. Descriptors of an ovarian mass used to make a diagnosis
BD, benign descriptor; MD, malignant descriptor.

15. The role of CA 125 in diagnosing ovarian
cancer according to IOTA results
CA 125 has no significant impact on performance of logistic regression model for women at any age
Adding information on serum CA 125 level to subjective assessment of ultrasound findings does not improve diagnostic performance of experienced ultrasound examiner
Timmerman et al, J Clin Oncol, 2007; Van Calster et al, J Natl Cancer Inst, 2007,
Valentin et al, UOG, 2009

16. Diagnostic performance of the models and simple rules to detect Stage 1 ovarian cancer
LR1 and LR2 had higher detection rate of Stage 1 primary ovarian cancer than RMI
Simple rules combined with subjective assessment when rules did not apply missclassified fewer Stage 1 ovarian cancer than RMI and CA 125
JVan Holsbeke et al, Clin Cancer Res, 2012; Timmerman et al, BMJ, 2010

17. Summary of the IOTA project
Improving strategies for diagnosing ovarian cancer: a summary of the International Ovarian Tumor Analysis (IOTA) studies
Kaijser et al., UOG 2013
Summary of the IOTA project
Pattern recognition of ultrasound features of an ovarian mass by an experienced examiner is the best way to characterize ovarian pathology
A small proportion of solid tissue makes a malignant mass more likely to be a borderline tumor or a Stage 1 primary invasive epithelial ovarian cancer
CA 125 does not improve diagnostic performance of assessment by experienced ultrasonographers
Two main approaches to classify ovarian masses have been developed:
Risk prediction models – LR1 and LR2
Simple rules or ”easy descriptors”
Multiclass models have been created to distinguish between benign, borderline, primary invasive and metastatic disease

18. Recommendations for clinical practice
Improving strategies for diagnosing ovarian cancer: a summary of the International Ovarian Tumor Analysis (IOTA) studies
Kaijser et al., UOG 2013
Recommendations for clinical practice
1. IOTA simple rules can be used as a triage test in 75% of all adnexal masses for estimating the risk of malignancy 2. A two-step strategy with referral to a specialist in gynecological ultrasound of unclassifiable masses rules has excellent diagnostic performance 3. An alternative to the simple rules is the LR2 model
4. LR2 or the simple rules should be adopted as the principal test to characterize masses as benign and malignant in premenopausal women 5. Measurement of serum CA 125 marker is not necessary for characterization of ovarian pathology in premenopausal women and is unlikely to improve the performance of experienced ultrasound examiners even in postmenopausal women.

19. Different approaches to estimate risk of malignancy
Improving strategies for diagnosing ovarian cancer: a summary of the International Ovarian Tumor Analysis (IOTA) studies
Kaijser et al., UOG 2013
Different approaches to estimate risk of malignancy

20. Improving strategies for diagnosing ovarian cancer: a summary of the International Ovarian Tumor Analysis (IOTA) studies
Kaijser et al., UOG 2013
Discussion points
Does serum CA 125 level help to discriminate between benign and malignant ovarian tumors?
Which test should be used for discriminating between benign and malignant ovarian tumors by a non-expert ultrasound examiner?
Can logistic regression models better predict malignancy than the IOTA simple rules or subjective evaluation by an experienced ultrasound examiner?
Do we need to use IOTA simple rules or logistic regression models when classifying an adnexal mass as benign and malignant?
Should we use the same models and rules for both premenopausal and postmenopausal patients?
Are the IOTA logistic regression model and simple rules superior to the Risk of Malignancy Index (RMI) in discriminating between benign and malignant ovarian tumors?