Improved predictability is needed for calculators used to preoperatively determine the etiology of splenic masses in dogs: an external validation study of the HeLP score and T-STAT

Abstract

Abstract OBJECTIVE To assess the predictability of the hemangiosarcoma likelihood prediction (HeLP) score and the Tufts Splenic Tumor Assessment Tool (T-STAT) for hemangiosarcoma and malignancy, respectively. ANIMALS 261 dogs undergoing splenectomy for a splenic mass. METHODS Medical records were retrospectively reviewed; variables for the HeLP score and T-STAT were collected, and scores were assigned. Area under the curve (AUC) was calculated for each score. RESULTS The HeLP score included 141 dogs; hemangiosarcoma was diagnosed in 87 (61.7%) dogs. The median cumulative HeLP score was 51 (range, 17 to 82; IQR, 39 to 58) for dogs with hemangiosarcoma and 28 (range, 0 to 70; IQR, 17 to 41) for dogs without hemangiosarcoma. The categorical HeLP score was low (28; 32.2%), medium (31; 35.6%), and high (28; 32.2%) for dogs with hemangiosarcoma and was low (41; 75.9%), medium (9; 16.7%), and high (4; 7.4%) for dogs without hemangiosarcoma. The AUC of the cumulative and categorical HeLP scores for diagnosis of hemangiosarcoma were 0.79 (95% CI, 0.71 to 0.86) and 0.73 (95% CI, 0.65 to 0.82), respectively. The T-STAT included 181 dogs. Lesions were benign in 95 (52.5%) and malignant in 86 (47.5%) dogs. The median T-STAT score was 62% (range, 5% to 98%; IQR, 36% to 77%) for dogs with malignant lesions and 38% (range, 5% to 91%; IQR, 24% to 59%) for dogs with benign lesions. The T-STAT had an AUC of 0.68 (0.60 to 0.76) for diagnosis of malignancy. CLINICAL RELEVANCE The HeLP score had acceptable performance, and the T-STAT had poor performance for diagnosis prediction. A tool with excellent or outstanding discrimination is needed to more reliably predict the presence of hemangiosarcoma or a malignant lesion preoperatively.