Background: We sought to identify PSA density (PSAD) and Prostate Imaging-Reporting and Data Systems (PI-RADS) category cut-offs that would allow deferring biopsy in men with suspicion for clinically significant prostate cancer (csPCa). Methods: Our institution’s prostate MRI registry (n = 1718) was queried for patients who had MRI-guided biopsy (MRI-GB) and/or systematic biopsy (SB) performed after prostate MRI between January 2013 and October 2018 (n = 676). Patients in the diagnostic group (either biopsy naïve or with prior negative biopsy) and patients with PCa on active surveillance (AS) were considered eligible. PSA, PSAD, and PI-RADS category were entered into logistic regression models for predicting csPca (grade group [GG] ≥ 2) at biopsy. Receiver operating characteristic (ROC) analysis was performed to assess model accuracy and results were stratified by biopsy indication and PI-RADS categories. Results: Logistic regression models that combined PSAD and PI-RADS categories had the highest ROC’s in both the diagnostic and AS groups (AUC=0.830 and 0.778, respectively). For diagnostic group patients with PSAD ≤0.15, csPCa was found in 6/89 (6.7%) of negative MRI patients (i.e. PI-RADS ≤ 2), 4/90 (4.4%) of PI-RADS 3 patients, 59/159 (37%) of PI-RADS 4-5 patients. If a PSAD cutoff of ≤ 0.15 and PI-RADS category ≤ 3 MRI were used in combination as criteria for biopsy deferral, only 10/526 (1.9%) of patients would have had csPCa missed on subsequent biopsy. Among patients in the AS group with a negative MRI, 0/22(0%) and 3/8 (37.5%) had csPCA if the PSAD was ≤0.15 and >0.15, respectively. Conclusions: For the diagnostic group of patients undergoing prostate biopsy, PSAD cut off ≤0.15 is useful for deferring biopsy only in patients with a PI-RADS ≤ 3. Confirmatory biopsy in patients should be strongly considered before enrolling patients in AS even in the setting of a negative MRI if the PSAD is > 0.15.