Diagnostic and therapeutic value of predictive models in preoperative classification of adnexal masses

[Z. Vilendečić]

80 str.

Medicina - Humana reprodukcija, perinatologija i neonatologija / Medicine - Human reproduction, perinatology and neonatology

CILJEVI RADA: Analiza dijagnostičkih karakteristika ispitivanih prediktivnih modela, utvrđivanje optimalne cut-off vrednosti prediktivnih modela u ispitivanoj populaciji i analiza doprinosa biomarkera u preoperativnoj klasifikaciji adneksalnih promena. MATERIJAL I METODE: Istraživanje je sprovedeno kao prospektivna opservaciona studija preseka u periodu između januara 2019. i juna 2021. godine. U studiju su uključene pacijentkinje sa barem jednom adneksalnom promenom koja je zahtevala hiruršku procenu na osnovu kliničkih, ultrazvučnih i laboratorijskih nalaza. Dihotomna preoperativna klasifikacija adneksalnih promena je izvedena subjektivnom procenom (SA), modelom jednostavnih pravila dopunjenim subjektivnom procenom u nekonkluzivnim slučajevima (SR+SA), ADNEX, RMI i ROMA modelom. Patohistološki nalazi su smatrani referentnim standardom. REZULTATI: Prevalencija maligniteta u uzorku je iznosila 29,8% (118/396). Senzitivnost i specifičnost SA je iznosila 88,1% (95% CI: 80,9-93,4) i 92,8% (95% CI: 89,1-95,6). AUC za SA je iznosila 0,928. Senzitivnost i specifičnost SR+SA je iznosila 88,1% (95% CI: 80,9-93,4) i 94,2% (95% CI: 90,8-96,7). AUC za SR+SA je iznosila 0,912. Pri cut-off vrednosti od 10%, ADNEX model je ostvario senzitivnost od 92,4% (95% CI 86,0-96,5), dok je specifičnost iznosila 73,0% (95% CI 67,4-78,2). Vrednost AUC je iznosila 0,914. Senzitivnost i specifičnost RMI modela je iznosila 73,7% (95% CI: 64,8-81,4) 85,8% (95% CI: 81,3-89,8). Vrednost AUC za RMI model je iznosila 0,854. Senzitivnost i specifičnost ROMA modela je iznosila 69,5% (95% CI: 60,3-77,6) i 77,7% (95% CI: 72,3-82,5). AUC za ROMA model je iznosila 0,813, pri čemu je vrednost AUC za postmenopauzalnu grupu bila značajno veća u odnosu na premenopauzalnu grupu. Nije uočena razlika u AUC za modele SA, SR+SA i ADNEX, dok je AUC za RMI model bila značajno manja u odnosu na SA, SR+SA i ADNEX model. ROMA model je bolje karakteristike ostvario u postmenopauzalnom uzorku. ADNEX model je najveću preciznost ostvario pri cut-off vrednosti od 35% sa senzitivnošću i specifičnošću koje su iznosile 79,7% i 90,6%. Cut-off vrednost RMI modela je bila dobro podešena (optimalna cut-off vrednost >192). Nije uočena razlika u AUC za model SR+SA u odnosu na novokonstruisani model SR+CA125. Novokonstruisani model SR+ROMA je imao manji AUC u odnosu na SR+SA. Model SA je imao veći AUC u odnosu na nokonstrisane modele SA+CA125 i SA+ROMA. ZAKLJUČCI: Subjektivna procena, model jednostavnih pravila i ADNEX model predstavljaju pouzdane modele za preoperativnu dihotomnu stratifikaciju adneksalnih promena. Obzirom na izvodljivost i karakteristike, model jednostavnih pravila (SR) je primenjiv u inicijalnoj trijaži pacijentkinja sa dijagnostikovanom adneksalnom promenom. Uočene razlike u performansama ADNEX modela pri različitim cut-off vrednostima ukazuju na potrebu validacije ovog modela u različitim kliničkim okruženjima.

OBJECTIVES: To analyse the diagnostic characteristics of the studied predictive models, to determine the optimal cut-off values of predictive models in the studied population, and to analyse the contribution of biomarkers in the preoperative classification of adnexal masses. MATERIALS AND METHODS: A prospective observational cross-sectional study was conducted in the period between January 2019 and June 2021. The study included patients with at least one adnexal mass that required surgical evaluation based on clinical, ultrasound, and laboratory findings. Dichotomous preoperative classification of adnexal changes was performed by subjective assessment (SA), simple rules model supplemented by subjective assessment in inconclusive cases (SR+SA), ADNEX, RMI and ROMA model. Postoperatively obtained histological findings were used as a reference. RESULTS: The prevalence of malignancy in the sample was 29.8% (118/396). The sensitivity and specificity of SA was 88.1% (95%CI: 80.9-93.4) and 92.8% (95% CI: 89.1-95.6). AUC for SA was 0.928. The sensitivity and specificity of SR+SA was 88.1% (95% CI: 80.9-93.4) and 94.2% (95% CI: 90.8-96.7). AUC for SR+SA was 0.912. At a cut-off value of 10%, the ADNEX model achieved a sensitivity of 92.4% (95% CI 86.0-96.5), while the specificity was 73.0% (95% CI 67.4-78, 2). The AUC value was 0.914. The sensitivity and specificity of the RMI model was 73.7% (95% CI: 64.8-81.4) 85.8% (95% CI: 81.3-89.8). The AUC value for the RMI model was 0.854. The sensitivity and specificity of the ROMA model was 69.5% (95% CI: 60.3-77.6) and 77.7% (95% CI: 72.3-82.5). The AUC for the ROMA model was 0.813, where the AUC value for the postmenopausal group was significantly higher compared to the premenopausal group. We observed no difference in AUC for the SA, SR+SA and ADNEX models, while the AUC for the RMI model was significantly lower compared to the SA, SR+SA and ADNEX models. The ROMA model achieved better characteristics in the postmenopausal sample. The ADNEX model achieved the highest precision at a cut-off value of 35% with sensitivity and specificity of 79.7% and 90.6%. The cut-off value of the RMI model was well adjusted (optimal cut-off value >192). No difference in AUC was observed for the SR+SA model compared to the newly constructed SR+CA125 model. The newly constructed SR+ROMA model had a lower AUC compared to SR+SA. The SA model had a higher AUC compared to the unconstrained models SA+CA125 and SA+ROMA. CONCLUSIONS: Subjective assessment, Simple Rules and the ADNEX model are reliable models for the preoperative dichotomous stratification of adnexal masses. Considering its feasibility and characteristics, the Simple Rules model is suitable for the initial triage of patients with an adnexal mass. Observed differences in the performance of the ADNEX model at different cut-off values indicate the need to validate this model in different clinical settings.

adneksalna promena, ultrazvuk, prediktivni model, subjektivna procena, CA 125, rak jajnika

adnexal mass, ultrasound, prediction models, subjective assessment, CA 125, ovarian cancer

618.11/.12-006-07(043.3)

Tekst