Decision curve analysis assessing the clinical benefit of NMP22 in the detection of bladder cancer: Secondary analysis of a prospective trial

Objective To employ decision curve analysis to determine the impact of nuclear matrix protein 22 (NMP22) on clinical decision making in the detection of bladder cancer using data from a prospective trial. Patients and Methods The study included 1303 patients at risk for bladder cancer who underwent cystoscopy, urine cytology and measurement of urinary NMP22 levels. We constructed several prediction models to estimate risk of bladder cancer. The base model was generated using patient characteristics (age, gender, race, smoking and haematuria); cytology and NMP22 were added to the base model to determine effects on predictive accuracy. Clinical net benefit was calculated by summing the benefits and subtracting the harms and weighting these by the threshold probability at which a patient or clinician would opt for cystoscopy. Results In all, 72 patients were found to have bladder cancer (5.5%). In univariate analyses, NMP22 was the strongest predictor of bladder cancer presence (predictive accuracy 71.3%), followed by age (67.5%) and cytology (64.3%). In multivariable prediction models, NMP22 improved the predictive accuracy of the base model by 8.2% (area under the curve 70.2-78.4%) and of the base model plus cytology by 4.2% (area under the curve 75.9-80.1%). Decision curve analysis revealed that adding NMP22 to other models increased clinical benefit, particularly at higher threshold probabilities. Conclusions NMP22 is a strong, independent predictor of bladder cancer. Addition of NMP22 improves the accuracy of standard predictors by a statistically and clinically significant margin. Decision curve analysis suggests that integration of NMP22 into clinical decision making helps avoid unnecessary cystoscopies, with minimal increased risk of missing a cancer.