The STHLM3 prostate cancer diagnostic study: calibration, clarification, and comments

Using a Modified Delphi Approach to Gain Consensus on Relevant Comparators in a Cost-Effectiveness Model: Application to Prostate Cancer Screening

OBJECTIVE

Challenges can exist when framing the decision question in a cost-effectiveness analysis, particularly when there is disagreement among experts on relevant comparators. Using prostate cancer screening and recent developments in risk stratification, early-detection biomarkers, and diagnostic technologies as a case study, we report a modified Delphi approach to handle decision-question uncertainty.

METHODS

The study involved two rounds of anonymous online questionnaires to identify the prostate cancer screening strategies that international researchers, clinicians and decision makers felt important to consider in a cost-effectiveness model. Both purposive and snowball sampling were used to recruit experts. The questionnaire was based on a review of the literature and was piloted for language, comprehension and ease of use prior to dissemination. In Round 1, respondents indicated their preferred screening strategy (including no screening) through a series of multiple-choice questions. The responses informed a set of 13 consensus statements, which respondents ranked their agreement with on a 9-point Likert scale (Round 2). Consensus was considered reached if > 70% of participants indicated agreement and < 15% indicated disagreement.

RESULTS

Twenty participants completed Round 1 and 17 completed Round 2. Consensus was shown towards comparing no formal screening, age-based, and risk-based strategies. The risk-based approaches included screening only higher-risk men, using shorter screening intervals for higher-risk men, screening higher-risk men at an earlier age, and tailoring screening intervals based on prostate-specific antigen (PSA) level at a previous test. There was agreement that inclusion of MRI in the pathway should be considered, but disagreement on the inclusion of new biomarkers.

CONCLUSION

In disease areas where technologies are rapidly evolving, a modified Delphi approach provides a useful tool to identify relevant comparators in an economic evaluation.

Biomarker discrimination and calibration with MRI-targeted biopsies: an analysis with the Stockholm3 test

BACKGROUND

The validated Stockholm3 test is used to improve PC detection. Stockholm3, however, was developed using systematic biopsies. We aimed to assess Stockholm3 operating performance when using MRI-targeted biopsies for PC detection.

METHODS

A prospective cohort of 532 men was considered for prostate biopsy during 2016-2017. All men underwent Stockholm3 testing and MRI before biopsy. All PIRADs ≥3 lesion underwent targeted biopsy; all men underwent systematic biopsy. The primary outcome was ISUP Grade Group ≥2 (GG ≥ 2) PC. Detection strategies included: (1) systematic biopsies alone, (2) targeted biopsies alone, (3) targeted with associated systematic biopsies for MRI+, and (4) all biopsies in all men. For each strategy, the Stockholm3 operating characteristics were assessed with discrimination, calibration, and decision curve analysis (DCA).

RESULTS

Median age was 65 years, median PSA was 6.2 ng/mL, median Stockholm3 score was 16.5%, and overall detection of GG ≥ 2 PC was 36% (193/532). Stockholm3 showed accurate discrimination for separating GG ≥ 2 cancer from benign and GG1, with an area under the curve of 0.84-0.86 depending on the biopsy strategy. Calibration analysis showed that Stockholm3 underestimated risks for GG ≥ 2 PC risk using MRI-targeted biopsies: there was a net benefit over biopsies in all men for Stockholm3 at risk thresholds varying from >3% in systematic biopsies to >15% in targeted with systematic biopsies in MRI+ men. When using a Stockholm3 score of >10% cutoff, a range of 32-38% of biopsies could be avoided while missing 5-11% of GG ≥ 2 PC and 0-3% of GG ≥ 3 PC.

CONCLUSIONS

Stockholm3 shows high discriminatory performance in an MRI-targeted biopsy setting, however risks are underpredicted due to MRI-targeted biopsies being more sensitive than the systematic biopsies for which Stockholm3 was developed. Stockholm3, along with any risk prediction model developed for systematic prostate biopsy decisions, will need recalibration for optimal use in an MRI-driven biopsy setting.

Improving the evaluation and diagnosis of clinically significant prostate cancer in 2017

PURPOSE OF REVIEW

To provide an overview of the current state of the evidence and highlight recent advances in the evaluation and diagnosis of clinically significant prostate cancer, focusing on biomarkers, risk calculators and multiparametric MRI (mpMRI).

RECENT FINDINGS

In 2017 there are numerous options to improve early detection as compared to a purely prostate-specific antigen (PSA)-based approach. All have strengths and drawbacks. In addition to repeating the PSA and performing clinical work-up (digital rectal examination and estimation of prostate volume), additional tests investigated in the initial biopsy setting are: %free PSA, Prostate Health Index, 4-kallikrein score, SelectMDx, and Michigan Prostate Score and in the repeat setting: %free PSA, Prostate Health Index, 4-kallikrein score, Prostate Cancer Antigen 3, and ConfirmMDx. Risk calculators are available for both biopsy settings and incorporate clinical data with, or without, biomarkers. mpMRI is an important diagnostic adjunct.

SUMMARY

There are numerous tests available that can help increase the specificity of PSA, in the initial and repeat biopsy setting. All coincide with a small decrease in sensitivity of detecting high-grade cancer. Cost effectiveness is crucial. The way forward is a multivariable risk assessment on the basis of readily available clinical data, potentially with the addition of PSA subforms, preferably at low cost. MRI in the prediagnostic setting is promising, but is not ready for 'prime time'.

Balancing Overdiagnosis and Early Detection of Prostate Cancer using the Stockholm-3 Model

The S3M (Stockholm-3 Model) test improves discrimination for high-grade (Gleason score ≥ 7) prostate cancer compared with prostate-specific antigen (PSA) testing. Published results from the Stockholm-3 study represent a snapshot of possible outcomes for prostate cancer detection using the S3M test. In this brief report, we show how the full range of cancer detection rates and percent saved biopsies depend on the chosen S3M cut-off for recommending prostate biopsy. Using data from the Stockholm-3 validation cohort (n=47 688), we calculated the cancer detection rates and percent saved biopsies for various S3M test cut-offs in men with PSA ≥1ng/ml. Cancer detection rates decline and fewer prostate biopsies have to be performed with increasing test cut-offs. Primarily, S3M test values between 7% and 14% can be considered for biopsy decision cut-offs (10% risk of Gleason score ≥ 7 corresponds to PSA=3 ng/ml); the exact cutoff can be chosen to fit different healthcare systems and, indeed, individual men. PATIENT SUMMARY: The Stockholm-3 Model test improves the detection of high-grade prostate cancer compared with prostate-specific antigen. In this brief report, we show how the cut-off used for recommending prostate biopsies affects the number of detected cancers and performed biopsies. The exact cut-off used can be chosen to fit different individuals and healthcare systems.