TY - JOUR
T1 - Impact of clinical history on maximum PI-RADS version 2 score
T2 - A six-reader 120-case sham history retrospective evaluation
AU - Shankar, Prasad R.
AU - Kaza, Ravi K.
AU - Al-Hawary, Mahmoud M.
AU - Masch, William R.
AU - Curci, Nicole E.
AU - Mendiratta-Lala, Mishal
AU - Sakala, Michelle D.
AU - Johnson, Timothy D.
AU - Davenport, Matthew S.
N1 - Publisher Copyright:
© RSNA, 2018.
PY - 2018
Y1 - 2018
N2 - Purpose: To assess the impact of clinical history on the maximum Prostate Imaging Recording and Data System (PI-RADS) version 2 (v2) score assigned to multiparametric magnetic resonance (MR) imaging of the prostate. Materials and Methods: This retrospective cohort study included 120 consecutively selected multiparametric prostate MR imaging studies performed between November 1, 2016, and December 31, 2016. Sham clinical data in four domains (digital rectal examination, prostate-specific antigen level, plan for biopsy, prior prostate cancer history) were randomly assigned to each case by using a balanced orthogonal design. Six fellowship-trained abdominal radiologists independently reviewed the sham data, actual patient age, and each examination while they were blinded to interreader scoring, true clinical data, and histologic findings. Readers were told the constant sham histories were true, believed the study to be primarily investigating interrater agreement, and were asked to assign a maximum PI-RADS v2 score to each case. Linear regression was performed to assess the association between clinical variables and maximum PI-RADS v2 score designation. Intraclass correlation coefficients (ICCs) were obtained to compare interreader scoring. Results: Clinical information had no significant effect on maximum PI-RADS v2 scoring for any of the six readers (P = .09–.99, 42 reader-variable pairs). Distributions of maximum PI-RADS v2 scores in the research context were similar to the distribution of the scores assigned clinically and had fair-to-excellent pairwise interrater agreement (ICC range: 0.53–0.76). Overall interrater agreement was good (ICC: 0.64; 95% confidence interval: 0.57, 0.71). Conclusion: Clinical history does not appear to be a substantial bias in maximum PI-RADS v2 score assignment. This is potentially important for clinical nomograms that plan to incorporate PI-RADS v2 score and clinical data into their algorithms (ie, PI-RADS v2 scoring is not confounded by clinical data).
AB - Purpose: To assess the impact of clinical history on the maximum Prostate Imaging Recording and Data System (PI-RADS) version 2 (v2) score assigned to multiparametric magnetic resonance (MR) imaging of the prostate. Materials and Methods: This retrospective cohort study included 120 consecutively selected multiparametric prostate MR imaging studies performed between November 1, 2016, and December 31, 2016. Sham clinical data in four domains (digital rectal examination, prostate-specific antigen level, plan for biopsy, prior prostate cancer history) were randomly assigned to each case by using a balanced orthogonal design. Six fellowship-trained abdominal radiologists independently reviewed the sham data, actual patient age, and each examination while they were blinded to interreader scoring, true clinical data, and histologic findings. Readers were told the constant sham histories were true, believed the study to be primarily investigating interrater agreement, and were asked to assign a maximum PI-RADS v2 score to each case. Linear regression was performed to assess the association between clinical variables and maximum PI-RADS v2 score designation. Intraclass correlation coefficients (ICCs) were obtained to compare interreader scoring. Results: Clinical information had no significant effect on maximum PI-RADS v2 scoring for any of the six readers (P = .09–.99, 42 reader-variable pairs). Distributions of maximum PI-RADS v2 scores in the research context were similar to the distribution of the scores assigned clinically and had fair-to-excellent pairwise interrater agreement (ICC range: 0.53–0.76). Overall interrater agreement was good (ICC: 0.64; 95% confidence interval: 0.57, 0.71). Conclusion: Clinical history does not appear to be a substantial bias in maximum PI-RADS v2 score assignment. This is potentially important for clinical nomograms that plan to incorporate PI-RADS v2 score and clinical data into their algorithms (ie, PI-RADS v2 scoring is not confounded by clinical data).
UR - http://www.scopus.com/inward/record.url?scp=85049027301&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85049027301&partnerID=8YFLogxK
U2 - 10.1148/radiol.2018172619
DO - 10.1148/radiol.2018172619
M3 - Article
C2 - 29664338
AN - SCOPUS:85049027301
SN - 0033-8419
VL - 288
SP - 158
EP - 163
JO - RADIOLOGY
JF - RADIOLOGY
IS - 1
ER -