Early Detection and Diagnosis
Free PSA isoforms and intact and cleaved forms of urokinase plasminogen activator receptor in serum improve selection of patients for prostate cancer biopsy
Article first published online: 4 JAN 2007
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 120, Issue 7, pages 1499–1504, 1 April 2007
How to Cite
Steuber, T., Vickers, A., Haese, A., Kattan, M. W., Eastham, J. A., Scardino, P. T., Huland, H. and Lilja, H. (2007), Free PSA isoforms and intact and cleaved forms of urokinase plasminogen activator receptor in serum improve selection of patients for prostate cancer biopsy. Int. J. Cancer, 120: 1499–1504. doi: 10.1002/ijc.22427
- Issue published online: 30 JAN 2007
- Article first published online: 4 JAN 2007
- Manuscript Accepted: 7 SEP 2006
- Manuscript Received: 24 MAR 2006
- Deutsche Forschungsgemeinschaft. Grant Numbers: Gz Ha3168 1/1, Ste 1429/1-1
- National Cancer Institute. Grant Number: P50-CA92629
- The Swedish Cancer Society. Grant Number: 3555
- The Department of Defense. Grant Number: W81XWH-05-1-0124
- Fundacion Federico SA, The Danish Cancer Society
- The European Union. Grant Numbers: LSHC-CT-2004-503011, QLK4-CT-1999-51464, QLK3-CT-2002-02136, LSHC-CT-2003-503297
- prostate cancer diagnosis;
- prostate-specific antigen;
- nicked PSA;
- multivariable models;
- circulating biomarkers;
- human glandular kallikrein 2
Clinicians currently use simple cut-points, such as serum prostate-specific antigen (PSA) ≥≥4 ng/ml, to decide whether to recommend further work-up for prostate cancer (PCa). As an alternative strategy, we evaluated multivariable models giving probabilities of a PCa diagnosis based on PSA and several circulating novel biomarkers. We measured total PSA, free PSA (fPSA), fPSA subfractions (single-chain fPSA-I and multichain fPSA-N), total human glandular kallikrein 2 (hK2) and full-length and cleaved forms of soluble urokinase plasminogen activator receptor (suPAR) in pretreatment serum from 355 men referred for prostate biopsy. Age and total PSA were combined in a “base” regression model to predict biopsy outcome. We then compared this base model to models supplemented by various combinations of circulating markers, using concordance index (AUC) to measure diagnostic discrimination. PCa prediction was significantly enhanced by models supplemented by measurements of suPAR fragments and fPSA isoforms. Addition of these markers improved bootstrap-corrected AUC from 0.611 for a cut-point and 0.706 for the base model to 0.754 for the full model (p = 0.005). This improved diagnostic accuracy was also seen in subanalysis of patients with PSA 2–9.99 ng/ml and normal findings on DRE (0.652 vs. 0.715, p = 0.039). In this setting, hK2 did not add diagnostic information. Measurements of individual forms of suPAR and PSA isoforms contributed significantly to discrimination of men with PCa from those with no evidence of malignancy. © 2006 Wiley-Liss, Inc.
Annual prostate-specific antigen (PSA) testing combined with the use of digital rectal exam (DRE) has revolutionized our ability to detect prostate cancer (PCa) at an early and potentially curable stage. Nevertheless, the clinical application of PSA as a means for early PCa detection remains controversial. Some data show that the positive predictive value of finding PCa in patients in the total PSA (tPSA) range 4–10 ng/ml and a nonsuspicious DRE is only 12–32%.1, 2 As a result, using a tPSA of 4 or more as an indication for biopsy leads to a large number of unnecessary biopsies. Statistical tools giving probabilities of a PCa diagnosis represent an alternative to use of a simple cut-point (e.g. PSA ≥ 4 ng/ml) to guide decisions about further work-up of patients who present with a suspicious PSA level. Incorporation of demographic,3 clinical4 and laboratory5 data into multivariable regression models has been proven to increase diagnostic accuracy compared to the use of a single marker and cut-point.
Several prostate-related biomarkers have been identified in the circulation during the last decade. This has led to the development of highly specific antibodies for selective immunodetection of free (fPSA) and complexed PSA,6, 7, 8 human kallikrein-related peptidase 2 (hK2 or KLK2)9, 10 and various fPSA-subfractions.11, 12 Clinical testing of free and complexed PSA has significantly enhanced cancer specificity over the testing of tPSA alone. Commercialized immunoassays for free and complexed PSA are widely used in current clinical practice. Serum levels of hk2 and various fPSA-subfractions have been reported to differ significantly in benign vs. malignant prostatic disease.9, 10, 11, 12 Moreover, it is implicated that a combination of these measurements with free or tPSA show significant potential to further improve discrimination of cancer from noncancer subjects in referral and screening cohorts.
Elevated levels of soluble urokinase-type plasminogen activator receptor forms in blood have been linked to prediction of patient prognosis in colon cancer and breast cancer.13, 14 Recently, specific immunoassays were developed that enabled quantification of the individual forms of soluble urokinase plasminogen activator receptor (suPAR): full-length suPAR [suPAR(I–III)], suPAR(I–III) + cleaved suPAR domains II + III [suPAR(I–III) + suPAR(II–III)] and uPAR domain I [uPAR(I)].15, 16 In a first clinical evaluation, they used a cohort of men referred for prostate biopsy to assess whether measurements of intact and cleaved suPAR forms in blood might discriminate patients found with PCa from men with no evidence of malignancy. This study showed significant differences in levels of uPAR(I) and suPAR(II–III) in the men with PCa compared to men with no evidence of PCa. Moreover, selective detection of suPAR forms suggested enhanced discrimination compared to tPSA and %fPSA.16
In the current study, we evaluated comprehensively whether demographic data (e.g. age) and novel circulating biomarkers such as prostate specific tissue kallikreins (tPSA, fPSA isoforms, hK2) and various suPAR components could enhance diagnostic accuracy of multivariable models for early PCa detection. We further developed a regression model using the most informative variables to predict biopsy outcome. Predictive accuracy of this model was validated among men with tPSA < 10 ng/ml and negative DRE, a scenario most commonly seen in a screening setting.
Material and methods
Blood was drawn from 355 men, referred in 1999 and 2000 to Department of Urology at UKE in Hamburg, Germany, for transrectal ultrasound (TRUS)-guided bilateral sextant biopsy of the peripheral zone triggered by an elevated PSA level (≥4 ng/ml) or abnormal findings on DRE. Overall, in 75 patients (29 men with and 46 without PCa), the PSA level in the prebiopsy sample was <4 ng/ml, thus did not meet the cut-off criteria for biopsy. However, 68 of those men were referred for an external PSA test result of > 4 ng/ml, which was not reproduced in our prebiopsy sample. Significant interassay and intraindividual variability of PSA testing may have contributed to this discrepancy.17 The remainder 7 patients with a PSA < 4 ng/ml were biopsied for a positive DRE.
Samples were collected before prostatic manipulation. The blood was allowed to clot, centrifuged to allow serum to be separated from the blood cells and then immediately stored frozen at −80°C until analysis.
PCa was diagnosed among 236 men (66%). This high PCa detection rate might result from the fact that this cohort represents referral patients since PSA testing for PCa screening is not officially recommended in Germany. The expected detection rate at the study institution in consecutive men is up to 40%.18 The retrospective design used in this study with high rate of men undergoing repeat biopsy (50%) and availability of high quality prebiopsy samples may have further contributed to an even higher rate of patients with evidence of cancer at biopsy.
All 236 PCa cases had clinically localized tumors and were subsequently counseled to undergo radical prostatectomy at our institution. Clinical characteristics of the PCa cohort are displayed in Table I. Thirty-one percent of the detected cancers had a positive DRE result (clinical stage T2 disease), which is consistent with the recently reported 37.4% incidence of T2 PCa among 4,277 consecutive men treated with radical prostatectomy at the study institution between 1992 and 2005.19 The expected rate of insignificant PCa in our radical prostatectomy series of referral patients is about 8%.
|Characteristic||Men with prostate cancer on needle biopsy (n = 236)|
|Prebiopsy PSA (tPSA)|
|Clinical stage (1997 TNM)|
|Biopsy Gleason score|
The control group was 119 patients (34%) with no evidence of PCa on needle biopsy specimens. Among biopsy-negative patients, 59 (50%) had only a single biopsy; repeat biopsy with negative outcome for a persistently elevated PSA level was performed once, twice or three times in 35, 14 and 11 patients, respectively. The blood samples used in this evaluation were those taken at the time of the positive biopsy or last negative biopsy, except in 12 patients with blood draws at the initial negative biopsy and subsequent positive biopsies within 1–4 years. These 12 patients were counted as positive for cancer in the analysis.
Volumetric analysis of the prostate gland
Determination of total prostate volume was performed by using 7 Mhz TRUS. The prostate was measured in three dimensions and volume was automatically calculated in cubic centimeter using a modification of the prostate ellipsoid formula.20
Imunodetection of analytes
tPSA and fPSA.
To detect tPSA and fPSA, we used the commercialized version of a previously reported dual-label assay (DELFIA ProStatus PSA free/total kit, PerkinElmer Life Sciences) that measures tPSA and fPSA on an equimolar basis.21 Detection limits were 0.04 ng/ml for fPSA (coefficient of variation [CV] 3.7% at 0.44 ng/ml and 17.9% at 0.10 ng/ml), and 0.05 ng/ml for tPSA (CV 5.0% at 2.32 ng/ml and 13.9% at 0.34 ng/ml).
“Intact” free PSA (fPSA-I) was measured by a three-step immunofluorometric sandwich assay that has been extensively described in previous studies.12 It uses a highly specific detection antibody that binds to single-chain fPSA, but unable to bind the fPSA internally cleaved at Lys145-Lys146 (called nicked PSA or fPSA-N). The analytical detection limit of the assay is 0.035 ng/ml (CV of 8.8%, SD 8.9%). The remainder fPSA subfraction not detected by the assay (fPSA-N) was determined by subtracting fPSA-I from fPSA (fPSA-N= fPSA − fPSA-I).
Total hK2 was measured according to a previously reported 3-step assay.22 It uses a capture MAb and europium-labelled detection MAb, which recognize both PSA and hK2, while a combination of PSA-specific blocking monoclonal antibodies (MAbs) enables selective detection of hK2 as this procedure eliminates crossreaction of PSA to <0.001%. The analytical detection limit is 0.3 ng/l and functional detection limit (i.e. CV <20%) is 4 ng/l (0.004 ng/ml).
Full-length uPAR and uPAR fragments.
The assay design, performance characteristics and measurements of serum-levels using three uPAR immunoassays, TR-FIA 1, 2 and 3, designed for the highly sensitive and specific measurement of suPAR(I–III), suPAR(I–III) + suPAR(II–III) and suPAR(I) have been previously described in very extensive detail.15, 16
Variables thought to be potentially predictive for PCa were first entered into univariate logistic models. On the basis of the univariate findings, a subset of variables was chosen for multiple regression. Among the models constructed were a “base” model consisting of the predictors, which were readily available such as tPSA and patient age, while a full model also included free PSA, fPSA-subfractions, hK2 and suPAR components. The multivariable analysis was conducted for the entire study cohort (n = 355) and for a subgroup of 208 men with a negative DRE and a PSA of at least 2 but less than 10 ng/ml (119 with cancer [57%] and 89 without cancer [43%]), as there is no widely accepted consensus as to whether biopsy is indicated for all or only a subset of these patients. Biomarkers were used as continuous variables.
Diagnostic accuracy of models was assessed by area under the receiver operating characteristic curve (AUC). AUC ranges from 0.5 (chance or a coin flip) to 1.0 (perfect ability to rank). For this analysis, tPSA was entered as restricted cubic splines with knots at the tertiles. This was not attempted for other variables because of the limited degrees of freedom. AUCs were then compared using an algorithm reported by DeLong et al.23 AUCs were corrected using bootstrap methods. The purpose of the bootstrap is to correct estimates of predictive accuracy. This is required because the models are created and evaluated on the same data set. To estimate what would have occurred had the models been externally validated, an external data set is simulated by creating a new data set by random sampling from the study cohort. All statistics were performed using STATA software (Version 8.0, StataCorp. LP, College Station, TX).
Descriptive statistics for the study cohort are given in Table II. Levels of tPSA, hK2, uPAR(I) and suPAR(II–III) were higher in cancer patients; prostate volume, free-to-tPSA ratio (%fPSA) and nicked-to-tPSA ratio (fPSA-N/tPSA) were lower. The univariate p values for these associations are shown in Table III. For the multivariable model, we started by choosing tPSA and age (the “base” model). Percent fPSA and fPSA-N/tPSA were better univariate predictors than fPSA and fPSA-N; however, the ratios are essentially interaction terms so we chose the latter for the model. hK2 was predictive in the full cohort, but not among patients in the diagnostic grey-zone (PSA 2–9.99 ng/ml and negative DRE). Therefore, hK2 was not chosen for further analysis. We initially planned to include all urokinase receptor forms. However, as the combined amount of suPAR(I–III) + suPAR(II–III) measured by TR-FIA 2 was highly collinear with suPARI–III it was excluded from the model.
|Variable||Cancer (n = 236) median (interquartile range)||No cancer (n = 119) median (interquartile range)|
|Age (y)||63 (59–66)||64 (59–69)|
|Prostate volume (cm3)||42 (32–52)||58 (43–72)|
|tPSA (ng/ml)||7.67 (5.10–11.28)||5.06 (2.98–8.08)|
|fPSA (ng/ml)||0.94 (0.69–1.41)||0.85 (0.53–1.30)|
|fPSA-I (ng/ml)||0.47 (0.30–0.71)||0.39 (0.24–0.63)|
|fPSA-N (ng/ml)||0.46 (0.29–0.72)||0.47 (0.23–0.67)|
|hK2 (ng/ml)||0.074 (0.046–0.107)||0.06 (0.036–0.093)|
|fPSA/tPSA (%)||12 (9.10–17.60)||18.60 (14.30–23.90)|
|fPSA-I/fPSA (%)||49 (40.10–58.60)||44 (35.90–53.60)|
|fPSA-N/tPSA (%)||5.80 (3.90–8.90)||9.40 (6.00–13.30)|
|uPAR(I) (pmol/L)||72.04 (48.04–100.97)||61.48 (41.09–80.81)|
|suPAR(I-III) (pmol/L)||49.38 (42.98–56.66)||48.92 (39.89–57.98)|
|suPAR(I-III) + suPAR(II-III) (pmol/L)||94.26 (81.30–108.70)||89.53 (80.53–117.61)|
|suPARII-III (pmol/L)||48.30 (39.13–61.63)||44.26 (36.36–53.54)|
|Variable||All patients||PSA 2–9.99 ng/ml, negative DRE|
|Odds ratio||95% C.I.||p value||Odds ratio||95% C.I.||p value|
|Prostate volume (cm3)||0.97||0.95–0.98||<0.0005||0.96||0.94–0.97||<0.0005|
|suPAR(I-III) + suPAR(II-III) (pmol/L)||1.28||0.94–1.73||0.12||1.06||0.75–1.50||0.7|
Table IV gives the multivariable analysis for the entire cohort, and for the subset of patients with PSA 2–9.99 ng/ml and negative DRE. Of the novel markers, suPAR(I–III) and uPAR(I) were significantly associated with cancer in the multivariable model. suPAR(II–III) and fPSA-N, however, did not prove to be independent predictors. Table V shows the increment in predictive accuracy for adding each category of predictor variable to the base model. Free PSA and fPSA-N (model 2), and suPAR components (model 3) clearly provided additional diagnostic information (see Figure 1). The full model incorporates age and all the blood markers that were selected for multivariable analysis. The AUCs of this model for the entire cohort and for the biopsy uncertain subgroup (i.e. men with PSA from 2 to 9.99 ng/ml and negative DRE) were 0.779 and 0.735, respectively (p = 0.005 and p = 0.039 compared to base model).
|Variable||All patients||PSA 2–9.99 ng/ml, negative DRE|
|Odds ratio||95% C.I.||p value||Odds ratio||95% C.I.||p value|
|Prediction based on||Study cohort||PSA 2–9.99 ng/ml, neg. DRE|
|AUC||AUC, bootstrap corrected||p value (vs. base model)||AUC||AUC, bootstrap corrected||p value (vs. base model)|
|tPSA ≥ 4 ng/ml (conventional cut-point)||0.611||–||–||0.574||–||–|
|tPSA and age (base model)||0.715||0.706||–||0.674||0.652||–|
|+ fPSA, fPSA-N (model 2)||0.748||–||0.072||0.696||–||0.3|
|+ uPAR(I), suPAR(I–III), suPAR(II–III) (model 3)||0.745||–||0.11||0.705||–||0.2|
|+ fPSA, fPSA-N, uPAR(I), suPAR(I–III), suPAR(II–III) (full model)||0.779||0.754||0.005||0.735||0.715||0.039|
We further explored whether levels of biomarkers were significantly associated with biopsy Gleason Grade. In multivariable models, there was no statistically significant association between biopsy Gleason grade (categorized as 6 or less vs. 7 or more) and either free PSA (p = 0.7), nicked PSA (p = 0.3) or suPAR forms (p = 0.15). There was some evidence that fPSA-N (p = 0.038), but not fPSA (p = 0.20) or suPAR forms (p = 1) predicted pathological stage (categorized as ≥pT3 vs. ≤pT2); although a p-value close to statistical significance should be interpreted with caution given the large number of analyses conducted. tPSA was strongly associated with biopsy Gleason grade, and pathological stage (p < 0.0005 for both).
As an exploratory analysis, we sought to determine whether our results for the subgroup with uncertain indications for biopsy depended on how we defined this group. We therefore derived several alternative definitions and assessed the AUC of our models for each. The results of this sensitivity analysis are shown in Table VI. It can be seen that the incremental value of the full model was fairly stable across definitions.
|Various subgroups of the diagnostic grey-zone||AUC: base model||AUC: full model||Increment in AUC|
|PSA 2–9.99, negative DRE (original definition)||0.674||0.735||0.061|
|PSA <10, negative DRE||0.695||0.753||0.058|
|PSA <10, any DRE||0.680||0.748||0.068|
|PSA <8, negative DRE||0.680||0.769||0.089|
|PSA <12, negative DRE||0.688||0.743||0.055|
In a further exploration, we added hK2, PSA-F/T, fPSA-N/T and ratio of uPAR(I) to suPAR(I–III) to the full model, but no variable was a significant predictor of cancer (p = 0.3 or greater) and there was no increase in AUC. However, when the data set was restricted to patients with PSA 2–9.99 ng/ml and negative DRE, PSA-F/T and uPAR(I) to suPAR(I–III) ratio both added, separately and together ∼0.005 to the AUC (basic model: 0.725; plus PSA-F/T: 0.741; plus uPAR(I) to suPAR(I–III) ratio: 0.739; plus both: 0.746).
As a final analysis, we added prostate volume to our full model. Prostate volume was associated with a large improvement in predictive accuracy compared to the base model both for all patients (bootstrap-corrected AUC increased from 0.706 to 0.824, p < 0.0005) and for the low PSA subgroup (AUC increased from 0.652 to 0.777, p < 0.0005). However, prostate volume cannot be measured noninvasively, and so its value in aiding the decision whether to refer a patient to biopsy must be questioned.
The utility of biomarkers for PCa detection may result from disease-specific activation of various tissue protease cascades in the prostate gland, which presumably translates into a distinctive profile of biomarkers in the circulation detectable before the disease is clinically evident. Current evidence suggests that no single analyte is likely to achieve the desired level of diagnostic accuracy for early PCa detection. The combination of biomarkers, clinical and demographic data has produced substantial, but not yet sufficient, progress toward the goal of optimal screening strategies for selecting appropriate patients for prostate biopsy.4, 5
Previous results have provided reason to think that serum levels of suPAR components may be linked to PCa. The plasminogen activation cascade has been reported to participate in degradation of extracellular matrix during cancer progression.13 Early studies utilizing serum measurements of suPAR for early detection of PCa produced conflicting results. McCabe et al. measured increased levels of circulating bulk suPAR in both BPH and PCa patients compared to healthy controls.24 However, they were not able to establish a statistically significant relationship between suPAR levels and the presence of PCa. A conceptually different approach was selected by Piironen et al. using immunoassays that enabled quantification of the individual forms of suPAR.15, 16 An evaluation of these assays using clinically annotated samples showed improved specificity of PCa detection when combined with tPSA and %fPSA.16 Although the total amount of suPAR forms may not be informative, levels of individual suPAR forms may reflect plasminogen activation catalyzed by urokinase plasminogen activator. Hence, the concept of measuring the individual forms of uPAR may generate a more specific biochemical profile than that obtained from the previously available polyclonal immunoassays, which measure the combined amount of suPAR forms.
Based on these findings, we decided to include these data generated by measurements of circulating suPAR forms into our present study, which (in a comprehensive manner) evaluates whether a combination of several PCa-related biomarkers may improve PCa detection in a consecutive series of patients referred for prostate biopsy at a single institution, where repeat biopsy was performed on 50% of all men with no evidence of cancer at the biopsy. Our univariate regression analysis demonstrated significant association of tPSA, %fPSA, fPSA-N/tPSA and suPAR forms with presence of PCa on biopsy. After adjusting for the effects of tPSA, only uPAR(I) and suPAR(I–III) were independent predictors of PCa. However, previous studies suggested that variables should not be judged by their multivariable statistical significance.25 Instead, variables should be assessed according to their ability to predict the outcome of interest. This is best achieved with the concordance index (here, area under the curve), which generates a quantitative measure to the predictive accuracy of a single variable or several variables. Moreover, the most objective assessment of the contribution of any given variable can be made within a multivariable model that incorporates established predictors. The incremental value of a target variable can be assessed by comparison of such a model with a multivariable model that does not contain the target variable (base model). The validity of this approach was elaborated upon by Kattan26 and later confirmed by Graefen et al.27 Using this method, our analysis showed that the predictive accuracy of a base model, which uses only age and tPSA is increased by addition of either the different suPAR forms or fPSA together with fPSA-N. The highest diagnostic accuracy is obtained when all these variables are combined. Considering the moderate size of the study cohort, we would be cautious in making premature conclusions referring to whether it may be critically important to only use 1 or 2, compared to all 3 uPAR-measures to obtain significantly higher accuracy in predictions of biopsy outcomes compared to the base-model. However, the currently reported data implicate that we may merely need 2 [uPAR(I) and uPAR(II-III)] but not 3 assays.
This superior diagnostic accuracy was also seen in subanalyses of subjects within the so-called diagnostic grey zone, in which traditional PSA testing lacks cancer specificity. This held true when the grey zone was defined as negative DRE and PSA ranging from 2.0–9.99 ng/ml, as well as for several other definitions.
In subanalysis, we could not see any significant association with suPAR forms and biopsy Gleason grade in a multivariable evaluation (p = 0.15), or any association with suPAR forms and pathological stage (p = 1). These results, however, need to be interpreted with caution since this study is designed to predict the risk of positive biopsy in referred patients. Currently, we are, in a comprehensive manner, evaluating the association of pretreatment suPAR levels with posttherapeutic endpoints such as pathological Gleason grade, pathological stage and the risk of biochemical failure following radical prostatectomy in an extended cohort of patients.
HK2 has also been reported to be involved in the process of plasminogen activation. It has shown ability in vitro to inactivate the major plasminogen activator inhibitor (PAI-1),28 and to directly convert single-chain urokinase plasminogen activator into the active form.29 Hence, increased secretion/release of hK2 may influence tumor biology by initiating a proteolytic cascade that facilitates local invasion of PCa cells. However, in our multivariable evaluation, hK2 was not a significant predictor of cancer and there was no increase in AUC when hK2 was incorporated into the prediction model. In a previous report, Becker et al. found that hK2 levels increased twice as much over time for men with PCa (p = 0.004) compared with matched controls in a screening population.9 Thus, a static, single measurement of total hK2, as used in our present study, may not represent the optimal application of this biomarker, whereas enhanced diagnostic accuracy may result from longitudinal testing, available in prospectively designed screening trials.
Prostate volume was significantly higher in patients without evidence of PCa in our study, and the incorporation of prostate volume into our prediction model provided a substantial increase in accuracy. Prostate volume has previously been incorporated into predictive tools.30 Prostate volume is, however, usually measured when the patient is subjected to TRUS-guided biopsies, and seldom measured in populations subjected to screening. Hence, it would appear logical and intuitive that the ideal prediction tool should use only data generally available (e.g. patient age, BMI, family history) and measurements of circulating biomarkers, which require a simple blood draw.
Several limitations of our study deserve mention. We used archived serum from sample banks and not from a prospective and sequentially enrolled population. Hence, the results may not be representative of the general population. Further, the validity of our analytical data might be influenced by preanalytical bias caused by the well-documented but commonly neglected lack of ex vivo stability of fPSA after the blood draw.31 However, we meticulously processed our samples (rapid separation of serum from the blood cells, immediate freezing at −80°C and analysis of samples immediately after thawing), in order to minimize preanalytical degradation of fPSA.
Our analysis should be interpreted as a proof-of-principle study, demonstrating that measurements of circulating kallikrein-related proteases and suPAR forms help predict the results of biopsy for PCa. Further research is warranted to confirm these preliminary findings. Specifically, we recommend research on a large and representative cohort of men in which biomarkers are assessed in the same sample that is used to trigger referral to biopsy, rather than, as in the current study, a blood sample taken at the time of biopsy. Levels of biomarkers could then be combined in single multivariable prediction model, which would allow clinicians to estimate an individual patient's probability of a positive biopsy. This approach has been previously reported by other authors for the prediction of a positive biopsy outcome32 or the risk of biochemical failure subsequent to radical prostatectomy.33
We designed and internally validated a diagnostic model for superior PCa detection based on age and a combination of biomarkers: fPSA, fPSA-N and 2 suPAR forms. This model should improve our ability to counsel patients with moderately elevated PSA levels and negative DRE.