There is a need for improved case selection for shockwave lithotripsy (SWL) treatment. Many studies claim to predict SWL success with a high degree of accuracy but these methods have not been replicated in clinical practice. This study aims to use robust statistical methods and 2 high quality clinical databases to investigate which variables (clinical and imaging) are important factors affecting SWL outcome, and how this may be used pragmatically by urologists for SWL case selection.


This study analyzed data on 20 variables considered to affect SWL efficacy from consecutive SWL cases at a single center in the UK (459 cases) and Beijing, China (148 cases). CT texture analysis (CTTA), a novel method of assessing stone heterogeneity, used histogram-based statistical analysis to derive the Entropy, Skewness, Kurtosis and Total number of pixels in the largest slice of the stone. Uni- and multivariable analyses of these variables were performed by expert statisticians.


In the UK data, SWL success rate, defined as completely stone free, was 46%. Mean (SD) stone volume was 261 (333) mm3. Variables associated with a significantly (p<0.05) lower SWL success rate were: increasing age, female sex, larger stone size, higher mean and SD of the Hounsfield units, 2 or more stones in the same location, vesico-ureteric junction location, and higher CTTA calculated Entropy and Total. Increasing skin-to-stone distance (SSD), presence of a stent and a lower pole position did not result in a significantly lower SWL success rate. Multivariable analysis using the Least Absolute Shrinkage and Selection Operator approach, and checked using the Partial Least Squares method, produced a model that predicted for cases that were most likely to fail SWL with negative predictive value of 84%, when using a predicted probability of success cut-off of 0.29. A model to predict for SWL success in general had an AUC of 0.63. The data from China produced a model of similar predictive ability.


This study has identified the significant factors of patient age, stone size and texture features, and the lack of significance of SSD and lower pole position for predicting SWL success, which is supported by the more recent evidence. Many variables are co-dependent and therefore do not add additional value. Even with expert statistical analysis, it has not been possible to produce a highly predictive model for SWL success. However, the analysis has produced a way to identify those at highest risk of SWL failure, which has been integrated into a software tool for further evaluation.