Men whose prostate MRI scans show a PI-RADS 3 lesion face an awkward clinical middle ground: the finding is neither clearly benign nor clearly malignant, yet current practice often pushes them straight to biopsy. A retrospective study published in BMC Urology proposes a way to sort these patients before any needles are used, by splitting them according to whether they have had a previous prostate biopsy and where the suspicious lesion sits in the gland. The researchers report that this pre-stratified approach, combined with a statistical model built from routine clinical variables, could have deferred biopsy in a large share of patients while missing only a minority of clinically significant prostate cancers.
The study included 919 men with PI-RADS 3 lesions on multiparametric MRI who underwent systematic prostate biopsy between January 2022 and June 2025. The cohort was divided into a training set of 684 patients and a validation set of 235.
The subgroup for which the authors developed a model was biopsy-naive patients whose lesions were not in the peripheral zone. In this group, multivariable logistic regression identified three independent predictors of clinically significant cancer. Each additional year of age raised the odds by a modest margin, with an odds ratio of 1.07. A higher ratio of free to total prostate-specific antigen, or f/tPSA, was strongly protective, and prostate-specific antigen density, or PSAD, which adjusts PSA levels for gland volume, carried an odds ratio of 5.34, meaning higher density substantially increased the odds of significant cancer.
Combining age, f/tPSA and PSAD produced a model with an area under the receiver operating characteristic curve of 0.839 in the training cohort, a level of discrimination the authors describe as excellent for this setting. The practical value of the model lies in how it was applied. Using a 10 percent probability threshold, the optimized biopsy strategy would have averted 44.6 percent of immediate biopsies while missing only 7.6 percent of clinically significant cancers, with a sensitivity of 86 percent and a specificity of 72.5 percent. In other words, nearly half of the needles in this subgroup could have been avoided at the cost of a small fraction of missed diagnoses.
Performance declined, as expected, when the model was applied to the validation cohort it had never seen. The area under the curve dropped to 0.746, and the optimized strategy would have deferred biopsy in 43.8 percent of patients while missing 14.7 percent of clinically significant cancers. The authors frame the strategy as risk-adapted rather than a replacement for biopsy, reserving immediate biopsy for patients whose calculated risk crosses the threshold.
The clinical problem the study addresses is well recognized. By pre-stratifying patients according to biopsy history and lesion location before applying a risk model, the study focused its model on the biopsy-naive, non-peripheral-zone subgroup. The authors conclude that the approach would have deferred nearly half of unnecessary biopsies while maintaining a low rate of missed clinically significant cancer diagnoses.
Several caveats temper the findings. The study is retrospective and single-cohort in design, drawing on men who were biopsied between 2022 and 2025, so the model has not yet been tested prospectively or in a multi-center setting. The validation results, with a higher missed-cancer rate of 14.7 percent, show that real-world performance may fall short of the training data.
Subject of Research: A risk prediction model for clinically significant prostate cancer in PI-RADS 3 MRI lesions
Article Title: Development and validation of a pre-stratified strategy for predicting clinically significant prostate cancer in PI-RADS 3 lesions
Article References: Chen, H., Zhu, B., Li, J., Wu, J., Zheng, X., Liu, B., Xie, L., & Wang, X. (2026). Development and validation of a pre-stratified strategy for predicting clinically significant prostate cancer in PI-RADS 3 lesions. BMC Urology. https://doi.org/10.1186/s12894-026-02364-2
Image Credits: AI Generated
DOI: 10.1186/s12894-026-02364-2
Keywords: prostate cancer, PI-RADS 3, mpMRI, prostate biopsy, PSA density, risk model, logistic regression, retrospective study, BMC Urology, decision curve analysis, biopsy-naive, peripheral zone
News Source: Ophelia Keating. (October 28, 2026). New risk model may help spare nearly half of men with PI-RADS 3 prostate lesions from immediate biopsy. Scienmag.
