Clinical utility and cost modelling of the phi test to triage referrals into image-based diagnostic services for suspected prostate cancer: the PRIM (Phi to RefIne Mri) study.
Biopsy
Cambridge prognostic groups
Prostate cancer
Prostate health index (phi)
mpMRI
Journal
BMC medicine
ISSN: 1741-7015
Titre abrégé: BMC Med
Pays: England
ID NLM: 101190723
Informations de publication
Date de publication:
17 04 2020
17 04 2020
Historique:
received:
14
12
2019
accepted:
03
03
2020
entrez:
18
4
2020
pubmed:
18
4
2020
medline:
17
12
2020
Statut:
epublish
Résumé
The clinical pathway to detect and diagnose prostate cancer has been revolutionised by the use of multiparametric MRI (mpMRI pre-biopsy). mpMRI however remains a resource-intensive test and is highly operator dependent with variable effectiveness with regard to its negative predictive value. Here we tested the use of the phi assay in standard clinical practice to pre-select men at the highest risk of harbouring significant cancer and hence refine the use of mpMRI and biopsies. A prospective five-centre study recruited men being investigated through an mpMRI-based prostate cancer diagnostic pathway. Test statistics for PSA, PSA density (PSAd) and phi were assessed for detecting significant cancers using 2 definitions: ≥ Grade Group (GG2) and ≥ Cambridge Prognostic Groups (CPG) 3. Cost modelling and decision curve analysis (DCA) was simultaneously performed. A total of 545 men were recruited and studied with a median age, PSA and phi of 66 years, 8.0 ng/ml and 44 respectively. Overall, ≥ GG2 and ≥ CPG3 cancer detection rates were 64% (349/545), 47% (256/545) and 32% (174/545) respectively. There was no difference across centres for patient demographics or cancer detection rates. The overall area under the curve (AUC) for predicting ≥ GG2 cancers was 0.70 for PSA and 0.82 for phi. AUCs for ≥ CPG3 cancers were 0.81 and 0.87 for PSA and phi respectively. AUC values for phi did not differ between centres suggesting reliability of the test in different diagnostic settings. Pre-referral phi cut-offs between 20 and 30 had NPVs of 0.85-0.90 for ≥ GG2 cancers and 0.94-1.0 for ≥ CPG3 cancers. A strategy of mpMRI in all and biopsy only positive lesions reduced unnecessary biopsies by 35% but missed 9% of ≥ GG2 and 5% of ≥ CPG3 cancers. Using PH ≥ 30 to rule out referrals missed 8% and 5% of ≥ GG2 and ≥ CPG3 cancers (and reduced unnecessary biopsies by 40%). This was achieved however with 25% fewer mpMRI. Pathways incorporating PSAd missed fewer cancers but necessitated more unnecessary biopsies. The phi strategy had the lowest mean costs with DCA demonstrating net clinical benefit over a range of thresholds. phi as a triaging test may be an effective way to reduce mpMRI and biopsies without compromising detection of significant prostate cancers.
Sections du résumé
BACKGROUND
The clinical pathway to detect and diagnose prostate cancer has been revolutionised by the use of multiparametric MRI (mpMRI pre-biopsy). mpMRI however remains a resource-intensive test and is highly operator dependent with variable effectiveness with regard to its negative predictive value. Here we tested the use of the phi assay in standard clinical practice to pre-select men at the highest risk of harbouring significant cancer and hence refine the use of mpMRI and biopsies.
METHODS
A prospective five-centre study recruited men being investigated through an mpMRI-based prostate cancer diagnostic pathway. Test statistics for PSA, PSA density (PSAd) and phi were assessed for detecting significant cancers using 2 definitions: ≥ Grade Group (GG2) and ≥ Cambridge Prognostic Groups (CPG) 3. Cost modelling and decision curve analysis (DCA) was simultaneously performed.
RESULTS
A total of 545 men were recruited and studied with a median age, PSA and phi of 66 years, 8.0 ng/ml and 44 respectively. Overall, ≥ GG2 and ≥ CPG3 cancer detection rates were 64% (349/545), 47% (256/545) and 32% (174/545) respectively. There was no difference across centres for patient demographics or cancer detection rates. The overall area under the curve (AUC) for predicting ≥ GG2 cancers was 0.70 for PSA and 0.82 for phi. AUCs for ≥ CPG3 cancers were 0.81 and 0.87 for PSA and phi respectively. AUC values for phi did not differ between centres suggesting reliability of the test in different diagnostic settings. Pre-referral phi cut-offs between 20 and 30 had NPVs of 0.85-0.90 for ≥ GG2 cancers and 0.94-1.0 for ≥ CPG3 cancers. A strategy of mpMRI in all and biopsy only positive lesions reduced unnecessary biopsies by 35% but missed 9% of ≥ GG2 and 5% of ≥ CPG3 cancers. Using PH ≥ 30 to rule out referrals missed 8% and 5% of ≥ GG2 and ≥ CPG3 cancers (and reduced unnecessary biopsies by 40%). This was achieved however with 25% fewer mpMRI. Pathways incorporating PSAd missed fewer cancers but necessitated more unnecessary biopsies. The phi strategy had the lowest mean costs with DCA demonstrating net clinical benefit over a range of thresholds.
CONCLUSION
phi as a triaging test may be an effective way to reduce mpMRI and biopsies without compromising detection of significant prostate cancers.
Identifiants
pubmed: 32299423
doi: 10.1186/s12916-020-01548-3
pii: 10.1186/s12916-020-01548-3
pmc: PMC7164355
doi:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
95Subventions
Organisme : Addenbrooke's Charitable Trust, Cambridge University Hospitals
ID : 900123
Pays : International
Références
BJU Int. 2017 Jul;120(1):61-68
pubmed: 27743489
AJR Am J Roentgenol. 2019 Feb;212(2):357-365
pubmed: 30512996
J Urol. 2018 Dec;200(6):1215-1220
pubmed: 29906434
Eur Urol. 2020 Feb;77(2):180-188
pubmed: 31606332
Sci Rep. 2016 Oct 17;6:35364
pubmed: 27748407
Value Health. 2016 Mar-Apr;19(2):153-7
pubmed: 27021748
BJU Int. 2018 Jul;122(1):13-25
pubmed: 29699001
BMC Urol. 2019 Mar 18;19(1):19
pubmed: 30885200
Urol Oncol. 2019 Feb;37(2):138-144
pubmed: 30528698
Eur Urol Focus. 2020 Mar 15;6(2):267-272
pubmed: 30327280
BMC Med. 2018 Feb 28;16(1):31
pubmed: 29490658
Eur Urol. 2019 Apr;75(4):570-578
pubmed: 30477981
Urol Oncol. 2019 Oct;37(10):672-677
pubmed: 31378585
Eur Urol. 2019 Feb;75(2):310-318
pubmed: 30082150
Eur Urol. 2013 Jan;63(1):88-96
pubmed: 22902040
World J Urol. 2020 Mar;38(3):531-545
pubmed: 31399825
BJU Int. 2012 Aug;110(3):353-62
pubmed: 22077934
J Comp Eff Res. 2019 Jan;8(1):7-19
pubmed: 30525982
World J Oncol. 2019 Apr;10(2):63-89
pubmed: 31068988
BJU Int. 2019 Jul;124(1):1
pubmed: 31206999
J Urol. 2014 Jul;192(1):60-6
pubmed: 24518780
BJU Int. 2012 Dec;110(11):1661-5
pubmed: 22564540
Urology. 2018 Sep;119:115-120
pubmed: 29940232
Eur Urol. 2019 Nov;76(5):582-583
pubmed: 31303257
Eur Urol Focus. 2018 Dec;4(6):790-803
pubmed: 28753865
Int J Mol Sci. 2019 Apr 12;20(8):
pubmed: 31013716
J Antimicrob Chemother. 2013 Oct;68(10):2419-20
pubmed: 23681272
Eur Urol. 2018 Dec;74(6):722-728
pubmed: 30001824
Clin Radiol. 2019 Nov;74(11):894.e11-894.e18
pubmed: 31627803
Med Decis Making. 2006 Nov-Dec;26(6):565-74
pubmed: 17099194
Lancet Oncol. 2019 Jan;20(1):100-109
pubmed: 30470502
Eur Urol. 2017 Aug;72(2):250-266
pubmed: 28336078
Int J Mol Sci. 2019 Apr 02;20(7):
pubmed: 30986955
Scand J Urol. 2019 Oct;53(5):304-311
pubmed: 31661357
Urology. 2013 Dec;82(6):1355-60
pubmed: 24080222
Asian J Urol. 2019 Apr;6(2):137-145
pubmed: 31061799
Sci Rep. 2014 May 23;4:5012
pubmed: 24852453
Prostate. 2017 Oct;77(14):1401-1407
pubmed: 28853167
World J Urol. 2019 Mar;37(3):481-487
pubmed: 29980838
Eur Urol Oncol. 2018 Jun;1(2):109-117
pubmed: 31100233