Abstract
OBJECTIVE: To systematically review the literature on the Bosniak classification system in CT to determine its diagnostic performance to diagnose malignant cystic lesions and the prevalence of malignancy in Bosniak categories. METHODS: A predefined database search was performed from 1 January 1986 to 18 January 2016. Two independent reviewers extracted data on malignancy rates in Bosniak categories and several covariates using predefined criteria. Study quality was assessed using QUADAS-2. Meta-analysis included data pooling, subgroup analyses, meta-regression and investigation of publication bias. RESULTS: A total of 35 studies, which included 2,578 lesions, were investigated. Data on observer experience, inter-observer variation and technical CT standards were insufficiently reported. The pooled rate of malignancy increased from Bosniak I (3.2 %, 95 % CI 0-6.8, I(2) = 5 %) to Bosniak II (6 %, 95 % CI 2.7-9.3, I(2) = 32 %), IIF (6.7 %, 95 % CI 5-8.4, I(2) = 0 %), III (55.1 %, 95 % CI 45.7-64.5, I(2) = 89 %) and IV (91 %, 95 % CI 87.7-94.2, I(2) = 36). Several study design-related influences on malignancy rates and subsequent diagnostic performance indices were identified. CONCLUSION: The Bosniak classification is an accurate tool with which to stratify the risk of malignancy in renal cystic lesions. KEY POINTS: • The Bosniak classification can accurately rule out malignancy. • Specificity remains moderate at 74 % (95 % CI 64-82). • Follow-up examinations should be considered in Bosniak IIF and Bosniak II cysts. • Data on the influence of reader experience and inter-reader variability are insufficient. • Technical CT standards and publication year did not influence diagnostic performance.