Objectives: To evaluate the performance characteristics of seven predetermined imaging features on pretreatment computed tomography (CT) in identifying extranodal extension (ENE) in cervical lymph node metastases from human papillomavirus-positive oropharyngeal carcinoma (HPV-OPC).Study Design: Retrospective study. Methods: Seventy-three patients with HPV-OPC who underwent primary surgery and cervical lymph node dissection were included. Preoperative contrast-enhanced CT (cCT) imaging was evaluated by two radiologists blinded to pathological results. Each cCT was scored for seven imaging features of interest: 1) indistinct capsular contours, 2) irregular nodal margins, 3) perinodal fat stranding, 4) perinodal fat planes, 5) nodal necrosis, 6) intranodal cysts, and 7) nodal matting. Logistic regression was employed to determine radiologist-specific odds ratios (OR) of predicting ENE for each imaging feature and radiologistspecific receiver operating characteristics (sensitivity [Sn], specificity [Sp], area under the curve [AUC], positive predictive value [PPV], negative predictive value [NPV]) for each imaging feature.Results: Thirty-two (44%) patients had ENE-positive lymph nodes. The presence of irregular margins (OR A = 12.3, 95% confidence interval [CI] A = 2.3-65.9; OR B = 7.0, 95% CI B = 1.4-36.3) and absence of perinodal fat plane (OR A = 6.8, 95% CI A = 2.0-23.3; OR B = 14.2, 95% CI B = 1.7-120.5) were significantly associated with ENE for each radiologist. Irregular nodal margin status was most specific for ENE (Sn A = 45%, Sp A = 94%, AUC A = 69%, PPV A = 82%, NPV A = 73%; Sn B = 28%, Sp B = 95%, AUC B = 61%, PPV B = 80%, NPV B = 64%). Absence of perinodal fat plane was most sensitive for ENE (Sn A = 87%, Sp A = 50%, AUC A = 69%, PPV A = 59%, NPV A = 62%; Sn B = 96%, Sp B = 34%, AUC B = 65%, PPV B = 53%, NPV B = 63%).Conclusions: Of the seven imaging features hypothesized to be associated with ENE-status, the presence of irregular nodal margins and absence of perinodal fat plane were the most specific and sensitive features, respectively.