By Mark Stovsky, MD
Cleveland Clinic is a non-profit academic medical center. Advertising on our site helps support our mission. We do not endorse non-Cleveland Clinic products or services Policy
- The prostate-specific antigen test has relatively poor specificity for prostate cancer and cannot differentiate indolent from aggressive disease. Newer tests have only modestly improved diagnostic accuracy.
- The IsoPSA™ test, developed by Cleveland Clinic in collaboration with Cleveland Diagnostics, Inc., identifies molecular structural changes in the PSA protein biomarker as opposed to PSA concentration.
- IsoPSA interrogates the entire PSA isoform distribution in a single assay and can reliably discriminate structural changes in the PSA protein that correlate with the presence or absence of cancer. It also has the potential to differentiate patients at risk for high-grade disease.
The prostate specific antigen (PSA) has arguably been one of the most successful biomarkers in medical history — defining screening, early detection and surveillance paradigms for prostate cancer for a generation of physicians worldwide.
While the PSA protein, a serine protease, is uniquely associated with the prostate gland, screening and early-detection strategies using serum PSA have come under increased scrutiny due to a relatively poor specificity for cancer as well as an inability to differentiate indolent from aggressive disease, which limit the clinical utility of the test.
To improve diagnostic precision, attention has turned to the use of analogues such as PSA density, PSA velocity and age-specific PSA, with only modest improvement in the overall performance of the test.
Other diagnostic tests of either urine or serum have been developed, based generally on the quantification of cellular mRNA or combinations of serum proteins, in an attempt to overcome the limitations of standard PSA testing.
While these new tests typically display improved diagnostic accuracy compared to standard PSA testing, the net improvement in statistical performance has been modest — especially when their cost is considered.
Seeking novel diagnostic tests
At Cleveland Clinic, in collaboration with colleagues at Cleveland Diagnostics, Inc., and in clinical research centers across the country, we have taken an “outside the box” approach to the development of novel diagnostic tests for cancer and other diseases, based on the identification of molecular structural changes in protein biomarkers as opposed to their concentration in serum or other body fluids.
For prostate cancer, our initial research indicates that this approach can differentiate the presence or absence of malignancy and also characterize the relative biochemical aggressiveness of disease with improved diagnostic accuracy.
We recently examined interim clinical data from an ongoing multicenter prospective trial designed to assess the diagnostic performance of IsoPSATM (Cleveland Diagnostics, Inc.) for the detection of prostate cancer based on the ability of the assay to differentiate the myriad structural isoforms of PSA in blood.
The study’s primary endpoint was an assessment of the ability of IsoPSA to distinguish patients with and without biopsy-confirmed evidence of cancer. Secondarily, we evaluated the test’s precision in differentiating patients with high-grade (Gleason ≥ 7) cancer from those with low-grade (Gleason = 6) disease and benign findings after standard ultrasound-guided biopsy of the prostate.
For the trial, the IsoPSA reagent kit and the Solvent Interaction Analysis assay platform (Cleveland Diagnostics, Inc.) were used to prospectively examine the performance of a single reported test parameter, K, to characterize the isoform composition of complexed PSA in serum.
The assay kit relies on one-step isoform partitioning from serum or plasma between two aqueous-based liquid phases in a single test tube. This step is followed by conventional ELISA-based PSA quantification within each phase and calculation of the composite isoform partitioning as the ratio of the complexed PSA concentrations in the two phases to yield the test parameter, K.
The test parameter K represents a unique composite structural index that provides orthogonal information independent of the absolute serum PSA expression level and which can be used alone or in combination with other parameters in probabilistic models based on the same basic IsoPSA assay.
Performance details for IsoPSA
In this interim analysis, we report the performance of the IsoPSA assay as a stand-alone test as well as within a multivariate statistical logistic regression model comprised of K, age, total PSA, complexed PSA and prostate volume on a training data set.
One hundred thirty-two plasma samples were obtained from multiple clinical sites. They were collected within 30 days prior to prostate biopsy from patients with serum total PSA between 2 and 26 ng/mL.
Using either the single-test parameter K (N = 132) or a model combining K with other clinical elements (N = 118 due to the absence of reported prostate volume data for some of the samples), IsoPSA was evaluated against standard 12-core transrectal ultrasound (TRUS)-guided biopsy results as the gold-standard comparator. In this trial, the prevalence of all cancers and high-grade cancers in the sample cohort was 50 percent and 34 percent, respectively.
Notably, the IsoPSA assay using K as a single parameter for the composite ratio of complexed PSA isoforms yielded an all-cancer area under curve (AUC) of 0.80. At a cut-off value of K = 8.0, IsoPSA demonstrated sensitivity (SN) of 94 percent, specificity (SP) of 47 percent, negative predictive value (NPV) of 88 percent and positive predictive value (PPV) of 64 percent.
For the same patients, total serum PSA displayed inferior diagnostic performance, with an all-cancer AUC of 0.58. At a cut-off value of PSA = 4 ng/mL, total PSA demonstrated SN of 91 percent, SP of 12 percent, NPV of 57 percent and PPV of 51 percent.
Therefore, the mere substitution of the IsoPSA structure-based composite index for the standard expression-based index of serum PSA resulted in substantial improvement in all-cancer diagnostic accuracy (Table 1).
Next we explored a multivariate model for patients with Gleason ≥ 7 pathology compared to those with benign and Gleason = 6 biopsy findings using a single probability parameter, KP, which yielded an
AUC of 0.85 and two distinct regions in the ROC curve. For the probability parameter value KP < 15 percent, a cohort of patients with very low probability for Gleason ≥ 7 could be identified with SN of 97 percent, SP of 47 percent, NPV of 97 percent and PPV of 49 percent.
In the same model, KP > 64 percent identified a second patient cohort at high risk with SN of 47 percent, SP of 96 percent, NPV of 78 percent and PPV of 86 percent (Table 1).
Importantly, translating these probabilities into clinically actionable parameters in the current 118-patient cohort with 40 positive biopsies and 78 negative biopsies would have resulted in 37 avoided biopsies in the low-risk group (with 1 delayed diagnosis) and the correct identification of 19 high-risk patients for biopsy (with only 3 unneeded biopsies).
For the intermediate group with 15 percent < KP < 64 percent, even if all patients in that cohort were referred for TRUS biopsy, the net reduction in total biopsies performed based on this IsoPSA risk stratification algorithm would have been almost 50 percent, with the added benefit of explicit identification of a high-risk patient group that should clearly be referred for biopsy (Figure 1).
Surpassing the limitations of existing tests
In summary, currently available prostate cancer biomarkers rely on the correlation between the serum or urine concentration of either protein or mRNA substrate with the presence or absence of disease. Clinical information such as prostate volume and digital rectal exam findings is often added to a multivariate analysis in an attempt to improve the diagnostic performance of these tests.
However, the clinical utility of these tests will always be limited by the fact that biomarker concentrations may be affected by cancer-unrelated physiological processes such as inflammation, as well as the relative lack of specificity of these biomarkers to the cancer phenotype. Further, the histologic heterogeneity of prostate cancer will generally limit the sensitivity of tests that analyze only one or several unique biomarkers to the exclusion of others that may indicate the presence of disease.
In comparison, the elegance of the IsoPSA test is that it interrogates the entire PSA isoform distribution in a single assay to yield a unique ratiometric parameter that can reliably discriminate structural changes in the PSA protein, which then correlates with the presence or absence of cancer.
By adding standard clinical information in a multivariate analysis, the IsoPSA test has also been shown in this prospective trial to have the potential to differentiate patients at risk for high-grade disease.
We look forward to further analysis of the completed trial data to confirm our findings, as well as to future studies that assess the ability of IsoPSA to play a role in addressing other clinically important questions such as clinical surveillance of patients after treatment.
Dr. Stovsky is a staff member of Cleveland Clinic Glickman Urological & Kidney Institute’s Department of Urology. He has a leadership position (Chief Medical Officer) and investment interest in Cleveland Diagnostics, Inc.
View the Late breaking webcast of the isoPSA abstract.