Purpose: The aim of the present study is to identify predictive imaging findings and construct a diagnostic model for differentiating renal cell carcinoma (RCC) with and without sarcomatoid dedifferentiation (sRCC and non-sRCC).
Methods: This study is a single-center retrospective study. All patients had magnetic resonance imaging (MRI) with gradient-echo T1-weighted images, single-shot T2-weighted images (T2WI), and enhanced nephrographic phase images. Forty pathologically confirmed sRCCs and 80 non-sRCCs were included in this study. Control cases were selected by matching the tumor diameter and the year of MRI. Two radiologists independently evaluated the following findings: growth pattern, presence of low-intensity area on T2WI in the tumor (T2LIA), presence of non-enhancing area, local tumor stage, and presence of regional lymphadenopathy. Two radiologists measured the diameter of the tumor, T2LIA, and the non-enhancing area. Multivariable logistic regression analysis was used to identify independent predictive factors for differentiating sRCC from non-sRCC. Selected variables were entered in the logistic regression model, and the area under the curve (AUC) was calculated for each reader with 95% confidence intervals (CIs).
Results: Larger T2LIA-to-tumor diameter ratio, regional lymphadenopathy, and local tumor stage 4 were associated with sRCC, and selected for the subsequent construction of a logistic regression model. With this model, the AUCs were 0.76 (95% CI, 0.66-0.85) and 0.70 (95% CI, 0.59-0.81) for prediction of sRCC.
Conclusion: In conclusion, larger T2LIA-to-tumor diameter ratio, regional lymphadenopathy, and local tumor stage 4 are predictive findings of sRCC. As a result, the model constructed using these findings demonstrated a moderate degree of diagnostic accuracy.
Keywords: Diagnostic accuracy; MRI; Renal cell carcinoma; Sarcomatoid renal cell carcinoma.
© 2022. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.