Abstract
Background: The ROCkeTS study aimed to identify the best diagnostic test for Ovarian cancer in postmenopausal women. ROCkeTS conducted a head-to-head comparison of risk prediction models Risk of Malignancy index ( RMI), International Ovarian Tumor analysis models Assessment of Different NEoplasias in the ADNEXa ( ADNEX), Simple Rules, SRRisk ultrasound (USS) , Risk of Malignancy algorithm (ROMA) and Ovarian Adnexal and data reporting system (ORADS) ultrasound model in symptomatic women in a ‘real world’ setting.
Methods: Population – newly presenting postmenopausal women between 16 - 90 years with non-specific symptoms and raised CA125 and/or abnormal imaging were eligible. Women with normal CA125 and with imaging showing simple ovarian cysts < 5 cm, pregnancy, active non ovarian malignancy or previous ovarian malignancy or who declined trans-vaginal scan were ineligible. Recruitment was through rapid access, elective clinics or emergency presentations from 23 hospitals in the United Kingdom. Participants donated blood for ROMA testing and underwent USS performed mainly by NHS sonographers who recorded IOTA model variables. Sonographers achieved certification in IOTA terminology prior to participation and underwent quality assurance.
Index tests evaluated were IOTA ADNEX model at thresholds of 3% and 10% , ROMA at multiple thresholds, RMI1 at 200, IOTA sRisk model at thresholds of 3% and 10%, IOTA simple rules, CA125 at 35 iu/ml. Posthoc, ORADS at 10% was derived from IOTA ultrasound variables using established methods as ORADS was described after completion of recruitment. Tests were conducted blinded to outcomes.
The Comparator was RMI1 at 250 threshold. The Reference standard was tissue histology/cytology within 3 months or follow-up at 12 months. Primary outcome was diagnostic accuracy in primary invasive Ovarian Cancer versus benign/normal.
Sensitivity, specificity, c-statistic, area under Receiver operating characteristic (ROC) curve), Positive and Negative Predictive values and calibration plot were analysed. ISRCTN17160843
Findings: 1242 postmenopausal women were recruited from 23 hospitals of whom 215 had primary OC (17%) between 13th July 2015 to 30th November 2018 . Compared to RMI 1 at 250, sensitivity 82.9% (95% CI: 76.7 to 88.0), specificity 87.4% (95% CI: 84.9 to 89.6), IOTA ADNEX at 10% was more sensitive (difference of 13.9, 95% CI: 7.6 to 20.2, p<0.001) but less specific (difference of 28.5, 95% CI: 24.7, 32.3, p<0.001). ROMA at 29.9% had comparable sensitivity (difference of 3.6, 95% CI: -1.9 to 9.1, p=0.2379) with lower specificity (difference of 5.2, 95% CI: 2.5 to 8.0, p=0.0001). RMI1 at 200 had comparable sensitivity ( difference of 2.1, 95% CI -4.7 to 0.5, p=0.1250) with lower specificity ( difference of 3.0, 95% CI 1.7-4.3, p< 0.001). IOTA sRisk model at 10% had comparable sensitivity ( difference of 4.3, 95% CI -11.0 -2.3, p=0.2295) and lower specificity ( difference of 16.2, 95% CI 12.6 -19.8, p<0.001). IOTA Simple rules had both comparable sensitivity (difference of 1.6, 95% CI -9.3, 6.2, p=0.8238) and specificity ( difference of 2.2, 95% CI -5.1 -0.6, p=0.1354). CA125 at 35 IU/ml had comparable sensitivity ( difference of 2.1, 95% CI -6.6 - 2.3, p=0.4240) and higher specificity ( difference of 6.7, 95% CI 4.3 - 9.1, p<0.001).
In posthoc analysis, ORADS achieved comparable sensitivity ( difference of 2.1, 95% CI -8.6 - 4.3, p=0.5966) and lower specificity ( difference of 10.2, 95% CI 6.8-13.6, p<0.001).
Interpretation: Compared to RMI 250, IOTA ADNEX at 10% had higher sensitivity but significantly lower specificity whilst ROMA at manufacturer recommended threshold (29.9%) and ORADS at 10% had comparable sensitivity but reduced specificity. Gains in sensitivity are preferred over reduced specificity in postmenopausal women. IOTA ADNEX at 10% should be considered new standard of care diagnostic in OC for postmenopausal women.
Funding: ROCkeTS was funded by National Institute of Heath Research.
Methods: Population – newly presenting postmenopausal women between 16 - 90 years with non-specific symptoms and raised CA125 and/or abnormal imaging were eligible. Women with normal CA125 and with imaging showing simple ovarian cysts < 5 cm, pregnancy, active non ovarian malignancy or previous ovarian malignancy or who declined trans-vaginal scan were ineligible. Recruitment was through rapid access, elective clinics or emergency presentations from 23 hospitals in the United Kingdom. Participants donated blood for ROMA testing and underwent USS performed mainly by NHS sonographers who recorded IOTA model variables. Sonographers achieved certification in IOTA terminology prior to participation and underwent quality assurance.
Index tests evaluated were IOTA ADNEX model at thresholds of 3% and 10% , ROMA at multiple thresholds, RMI1 at 200, IOTA sRisk model at thresholds of 3% and 10%, IOTA simple rules, CA125 at 35 iu/ml. Posthoc, ORADS at 10% was derived from IOTA ultrasound variables using established methods as ORADS was described after completion of recruitment. Tests were conducted blinded to outcomes.
The Comparator was RMI1 at 250 threshold. The Reference standard was tissue histology/cytology within 3 months or follow-up at 12 months. Primary outcome was diagnostic accuracy in primary invasive Ovarian Cancer versus benign/normal.
Sensitivity, specificity, c-statistic, area under Receiver operating characteristic (ROC) curve), Positive and Negative Predictive values and calibration plot were analysed. ISRCTN17160843
Findings: 1242 postmenopausal women were recruited from 23 hospitals of whom 215 had primary OC (17%) between 13th July 2015 to 30th November 2018 . Compared to RMI 1 at 250, sensitivity 82.9% (95% CI: 76.7 to 88.0), specificity 87.4% (95% CI: 84.9 to 89.6), IOTA ADNEX at 10% was more sensitive (difference of 13.9, 95% CI: 7.6 to 20.2, p<0.001) but less specific (difference of 28.5, 95% CI: 24.7, 32.3, p<0.001). ROMA at 29.9% had comparable sensitivity (difference of 3.6, 95% CI: -1.9 to 9.1, p=0.2379) with lower specificity (difference of 5.2, 95% CI: 2.5 to 8.0, p=0.0001). RMI1 at 200 had comparable sensitivity ( difference of 2.1, 95% CI -4.7 to 0.5, p=0.1250) with lower specificity ( difference of 3.0, 95% CI 1.7-4.3, p< 0.001). IOTA sRisk model at 10% had comparable sensitivity ( difference of 4.3, 95% CI -11.0 -2.3, p=0.2295) and lower specificity ( difference of 16.2, 95% CI 12.6 -19.8, p<0.001). IOTA Simple rules had both comparable sensitivity (difference of 1.6, 95% CI -9.3, 6.2, p=0.8238) and specificity ( difference of 2.2, 95% CI -5.1 -0.6, p=0.1354). CA125 at 35 IU/ml had comparable sensitivity ( difference of 2.1, 95% CI -6.6 - 2.3, p=0.4240) and higher specificity ( difference of 6.7, 95% CI 4.3 - 9.1, p<0.001).
In posthoc analysis, ORADS achieved comparable sensitivity ( difference of 2.1, 95% CI -8.6 - 4.3, p=0.5966) and lower specificity ( difference of 10.2, 95% CI 6.8-13.6, p<0.001).
Interpretation: Compared to RMI 250, IOTA ADNEX at 10% had higher sensitivity but significantly lower specificity whilst ROMA at manufacturer recommended threshold (29.9%) and ORADS at 10% had comparable sensitivity but reduced specificity. Gains in sensitivity are preferred over reduced specificity in postmenopausal women. IOTA ADNEX at 10% should be considered new standard of care diagnostic in OC for postmenopausal women.
Funding: ROCkeTS was funded by National Institute of Heath Research.
| Original language | English |
|---|---|
| Journal | The Lancet Oncology |
| Publication status | Accepted/In press - 18 Jul 2024 |
