Combining Prostate Health Index density, magnetic resonance imaging and prior negative biopsy status to improve the detection of clinically significant prostate cancer

Sasha C. Druskin, Jeffrey J. Tosoian, Allen Young, Sarah Collica, Arnav Srivastava, Kamyar Ghabili, Katarzyna J. Macura, H. Ballentine Carter, Alan W. Partin, Lori J. Sokoll, Ashley E. Ross, Christian P. Pavlovich

Research output: Contribution to journalArticle

Abstract

Objectives: To determine the performance of Prostate Health Index (PHI) density (PHID) combined with MRI and prior negative biopsy (PNB) status for the diagnosis of clinically significant prostate cancer (PCa). Patients and Methods: Patients without a prior diagnosis of PCa, with elevated prostate-specific antigen and a normal digital rectal examination who underwent PHI testing prospectively prior to prostate biopsy were included in this study. PHID was calculated retrospectively using prostate volume derived from transrectal ultrasonography at biopsy. Univariable and multivariable logistic regression modelling, along with receiver-operating characteristic (ROC) curve analysis, was used to determine the ability of serum biomarkers to predict clinically significant PCa (defined as either grade group [GG] ≥2 disease or GG1 PCa detected in >2 cores or >50% of any one core) on biopsy. Age, PNB status and Prostate Imaging Reporting and Data System (PI-RADS) score were incorporated into the regression models. Results: Of the 241 men who qualified for the study, 91 (37.8%) had clinically significant PCa on biopsy. The median (interquartile range) PHID was 0.74 (0.44–1.24); it was 1.18 (0.77–1.83) and 0.55 (0.38–0.89) in those with and without clinically significant PCa on biopsy, respectively (P < 0.001). On univariable logistic regression, age and PNB status were associated with clinically significant cancer. Of the tested biomarkers, PHID demonstrated the highest discriminative ability for clinically significant disease (area under the ROC curve [AUC] 0.78 for the univariable model). That continued to be the case in multivariable logistic regression models incorporating age and PNB status (AUC 0.82). At a threshold of 0.44, representing the 25th percentile of PHID in the cohort, PHID was 92.3% sensitive and 35.3% specific for clinically significant PCa; the sensitivity and specificity were 93.0% and 32.4% and 97.4% and 29.1% for GG ≥2 and GG ≥3 disease, respectively. In the 104 men who underwent MRI, PI-RADS score was complementary to PHID, with a PI-RADS score ≥3 or, if PI-RADS score ≤2, a PHID ≥0.44, detecting 100% of clinically significant disease. For that subgroup, of the biomarkers tested, PHID (AUC 0.90) demonstrated the highest discriminative ability for clinically significant disease on multivariable logistic regression incorporating age, PNB status and PI-RADS score. Conclusions: In this contemporary cohort of men undergoing prostate biopsy for the diagnosis of PCa, PHID outperformed PHI and other PSA derivatives in the diagnosis of clinically significant cancer. Incorporating age, PNB status and PI-RADS score led to even further gains in the diagnostic performance of PHID. Furthermore, PI-RADS score was found to be complementary to PHID. Using 0.44 as a threshold for PHID, 35.3% of unnecessary biopsies could have been avoided at the cost of missing 7.7% of clinically significant cancers. Despite these encouraging results, prospective validation is needed.

Original languageEnglish (US)
Pages (from-to)619-626
Number of pages8
JournalBJU International
Volume121
Issue number4
DOIs
StatePublished - Apr 2018

Keywords

  • diagnosis
  • magnetic resonance imaging
  • prostate cancer
  • prostate-specific antigen
  • protein isoforms

ASJC Scopus subject areas

  • Urology

Fingerprint Dive into the research topics of 'Combining Prostate Health Index density, magnetic resonance imaging and prior negative biopsy status to improve the detection of clinically significant prostate cancer'. Together they form a unique fingerprint.

  • Cite this