RESUMO
STUDY OBJECTIVE: To develop a risk prediction model for occult uterine sarcoma using preoperative clinical characteristics in women undergoing hysterectomy for presumed uterine leiomyomata. DESIGN: Cases of uterine sarcoma were identified from the electronic medical records. Age/race-matched controls were selected at a 2:1 ratio (controls:cases) from a cohort of 45 188 women who underwent hysterectomy for uterine leiomyomata or abnormal bleeding during the same time interval. Unadjusted conditional logistic regression was performed to identify risk factors for occult uterine sarcomas, defined as no preoperative suspicion for malignancy. A risk prediction model was developed using a weighted logistic regression model, and the performance of the model was assessed using the receiver operator characteristic curve and corresponding area under the curve. SETTING: A large integrated health care system in California PATIENTS: Women 18 years of age and older who underwent a hysterectomy and were diagnosed with a uterine sarcoma and matched controls from 2006 to 2013. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: There were 117 cases of occult uterine sarcomas that met inclusion criteria during the study period. The final risk prediction model included age, race/ethnicity, number of myomas, uterine weight, uterine size increase, degree of pelvic pain, and recent history of blood transfusion. The risk prediction model showed high accuracy based on the receiver operating characteristic curve method (area under the curveâ¯=â¯0.83; 95% confidence interval, 0.77-0.90); however, the positive predictive values were low (0.048 or less) at all risk thresholds. CONCLUSION: Multiple clinical features are associated with the presence of a uterine sarcoma, but when incorporated into a prediction model, they fail to provide significantly more information about women who may have an unrecognized sarcoma and only marginally improve the certainty about women who are not likely to have sarcoma.
