Biomarker-Based Models for Preoperative Assessment of Adnexal Mass: A Multicenter Validation Study

Authors

MustRafał Watrow­ski,^1,2,† Eva Ober­mayr,^2,† Chris­ti­ne Wal­lisch,³ Ste­fa­nie Aust,² Nico­le Con­cin,⁴ Ele­na Ioa­na Braicu,⁵ Toon Van Gorp,⁶ Annet­te Hasen­burg,^7,8 Jalid Seh­ouli,⁵ Ignace Ver­go­te,⁶ and Robert Zeil­lin­ger^2,*

Simple Summary

Ova­ri­an can­cer (OC) is the dead­liest geni­tal tumor in women. In this mul­ti­cen­ter stu­dy, we deve­lo­ped three new dia­gno­stic models based on serum pro­te­ins and pati­ent age. All models were then vali­da­ted using data from cen­ters other than tho­se used for the model deve­lo­p­ment. We also com­pared the per­for­mance of the­se models with com­mon sin­gu­lar mar­kers (CA125, HE4) and algo­rith­ms (ROMA-50 and the Copen­ha­gen Index). We used modern tech­no­lo­gy (Luminex^®, Aus­tin, TX, USA) that enables the simul­ta­neous deter­mi­na­ti­on of seve­ral bio­mar­kers in a small amount of blood. A com­bi­na­ti­on of pati­ent age with four to six mar­kers per­for­med best: CA125, osteo­pon­tin, pro­lac­tin, macro­pha­ge migra­ti­on inhi­bi­to­ry fac­tor, and, even­tual­ly, HE4 and lep­tin. Our models were bet­ter than the ROMA-50 index but did not out­per­form the Copen­ha­gen Index. In post­me­no­pau­sal pati­ents, all the new­ly deve­lo­ped models per­for­med excel­lent­ly. Unfort­u­na­te­ly, none of the models tes­ted impro­ved the dia­gno­sis in pre­me­no­pau­sal pati­ents and tho­se missed due to nor­mal CA125 levels.

Abstract

Ova­ri­an can­cer (OC) is the most lethal geni­tal mali­gnan­cy in women. We aimed to deve­lop and vali­da­te new pro­teo­mic-based models for non-inva­si­ve dia­gno­sis of OC. We also com­pared them to the modi­fied Risk of Ova­ri­an Mali­gnan­cy Algo­rithm (ROMA-50), the Copen­ha­gen Index (CPH‑I) and our ear­lier Pro­teo­mic Model 2017. Bio­mar­kers were asses­sed using bead-based mul­ti­plex tech­no­lo­gy (Luminex^®) in 356 women (250 with mali­gnant and 106 with benign ova­ri­an tumors) from five Euro­pean cen­ters. The trai­ning cohort included 279 women from three cen­ters, and the vali­da­ti­on cohort 77 women from two other cen­ters. Of six pre­vious­ly stu­di­ed serum pro­te­ins (CA125, HE4, osteo­pon­tin [OPN], pro­lac­tin, lep­tin, and macro­pha­ge migra­ti­on inhi­bi­to­ry fac­tor [MIF]), four con­tri­bu­ted signi­fi­cant­ly to the Pro­teo­mic Model 2021 (CA125, OPN, pro­lac­tin, MIF), while lep­tin and HE4 were omit­ted by the algo­rithm. The Pro­teo­mic Model 2021 reve­a­led a c‑index of 0.98 (95% CI 0.96, 0.99) in the trai­ning cohort; howe­ver, in the vali­da­ti­on cohort it only achie­ved a c‑index of 0.82 (95% CI 0.72, 0.91). Adding pati­ent age to the Pro­teo­mic Model 2021 con­sti­tu­ted the Com­bi­ned Model 2021, with a c‑index of 0.99 (95% CI 0.97, 1) in the trai­ning cohort and a c‑index of 0.86 (95% CI 0.78, 0.95) in the vali­da­ti­on cohort. The Full Com­bi­ned Model 2021 (all six pro­te­ins with age) yiel­ded a c‑index of 0.98 (95% CI 0.97, 0.99) in the trai­ning cohort and a c‑index of 0.89 (95% CI 0.81, 0.97) in the vali­da­ti­on cohort. The vali­da­ti­on of our pre­vious Pro­teo­mic Model 2017, as well as the ROMA-50 and CPH‑I reve­a­led a c‑index of 0.9 (95% CI 0.82, 0.97), 0.54 (95% CI 0.38, 0.69) and 0.92 (95% CI 0.85, 0.98), respec­tively. In post­me­no­pau­sal women, the three new­ly deve­lo­ped models all achie­ved a spe­ci­fi­ci­ty of 1.00, a posi­ti­ve pre­dic­ti­ve value (PPV) of 1.00, and a sen­si­ti­vi­ty of >0.9. Per­for­mance in women under 50 years of age (c‑index below 0.6) or with nor­mal CA125 (c‑index clo­se to 0.5) was poor. CA125 and OPN had the best dis­cri­mi­na­ting power as sin­gle mar­kers. In sum­ma­ry, the CPH‑I, the two com­bi­ned 2021 Models, and the Pro­teo­mic Model 2017 show­ed satis­fac­to­ry dia­gno­stic accu­ra­ci­es, with no clear supe­rio­ri­ty of eit­her model. Nota­b­ly, alt­hough com­bi­ning values of only four pro­te­ins with age, the Com­bi­ned Model 2021 per­for­med com­pa­ra­b­ly to the Full Com­bi­ned Model 2021. The models con­firm­ed their excep­tio­nal dia­gno­stic per­for­mance in women aged ≥50. All models out­per­for­med the ROMA-50.

References

1. Sung H., Fer­lay J., Sie­gel R.L., Laver­san­ne M., Soer­jo­mat­aram I., Jemal A., Bray F. Glo­bal Can­cer Sta­tis­tics 2020: GLOBOCAN Esti­ma­tes of Inci­dence and Mor­ta­li­ty World­wi­de for 36 Can­cers in 185 Count­ries. CA A Can­cer J. Clin. 2021;71:209–249. doi: 10.3322/caac.21660. — DOI - Pub­Med
2. Zhou Z., Wang X., Ren X., Zhou L., Wang N., Kang H. Dise­a­se Bur­den and Attri­bu­ta­ble Risk Fac­tors of Ova­ri­an Can­cer from 1990 to 2017: Fin­dings from the Glo­bal Bur­den of Dise­a­se Stu­dy 2017. Front. Public Health. 2021;9:619581. doi: 10.3389/fpubh.2021.619581. — DOI - PMC - Pub­Med
3. Glo­bal Bur­den of Dise­a­se 2019 Can­cer Col­la­bo­ra­ti­on. Kocar­nik J.M., Comp­ton K., Dean F.E., Fu W., Gaw B.L., Har­vey J.D., Hen­rik­son H.J., Lu D., Pen­ni­ni A., et al. Can­cer Inci­dence, Mor­ta­li­ty, Years of Life Lost, Years Lived with Disa­bi­li­ty, and Disa­bi­li­ty-Adjus­ted Life Years for 29 Can­cer Groups From 2010 to 2019: A Sys­te­ma­tic Ana­ly­sis for the Glo­bal Bur­den of Dise­a­se Stu­dy 2019. JAMA Oncol. 2022;8:420–444. doi: 10.1001/jamaoncol.2021.6987. — DOI - PMC - Pub­Med
4. Nolen B.M., Loks­hin A.E. Pro­te­in Bio­mar­kers of Ova­ri­an Can­cer: The Forest and the Trees. Future Oncol. 2012;8:55–71. doi: 10.2217/fon.11.135. — DOI - PMC - Pub­Med
5. Bast R.C., Lu Z., Han C.Y., Lu K.H., Ander­son K.S., Dre­scher C.W., Skates S.J. Bio­mar­kers and Stra­te­gies for Ear­ly Detec­tion of Ova­ri­an Can­cer. Can­cer Epi­de­mi­ol. Bio­mark. Prev. 2020;29:2504–2512. doi: 10.1158/1055–9965.EPI-20–1057. — DOI - PMC - Pub­Med