| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Cancer Diagnostics |
Departments of1
Surgery (Urology), 2
Epidemiology and Biostatistics, 3
Clinical Laboratories, and 4
Medicine (GU-Oncology), Memorial Sloan-Kettering Cancer Center, New York, NY.
5 Department of Urology, University Hospital Hamburg-Eppendorf, Hamburg, Germany.
6 Department of Biotechnology, University of Turku, Turku, Finland.
7 Department of Laboratory Medicine, Division of Clinical Chemistry, Lund University, University Hospital, Malmö, Sweden.
aAddress correspondence to this author at: Memorial Sloan-Kettering Cancer Center, Departments of Clinical Laboratories, Urology, 1275 York Ave., Box 213, New York, NY 10021. Fax 212-422-2379; e-mail liljah{at}mskcc.org.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: We retrospectively measured total PSA (tPSA), free PSA (fPSA), and total hK2 (thK2) in preoperative serum samples from 867 men [and assessed free hK2 (fhK2) measured in 577 of these men] treated with radical prostatectomy for clinically localized prostate cancer. Associations between biomarker concentrations and extracapsular extension, seminal vesicle invasion, and biochemical recurrence (BCR) were evaluated. A subset of patients with PSA 
10 µg/L, the group most commonly seen in clinical practice in the US, was analyzed.
Results: thK2 was the strongest predictor of extracapsular extension and seminal vesicle invasion (areas under the ROC curve [AUC], 0.662 and 0.719, respectively), followed by tPSA (AUC, 0.654 and 0.663). All biomarkers were significant predictors of BCR. hK2 forms, but not PSA forms, remained highly significant for predicting BCR in the low-PSA group. Combining tPSA, fPSA, and thK2 in a multivariable model improved prediction compared with any biomarker used individually (AUC, 0.711, 0.755, and 0.752 for this combination predicting extracapsular extension, seminal vesicle invasion, and BCR, respectively; P <0.001 for all).
Conclusions: Increased concentrations of hK2 in the blood are significantly associated with unfavorable features of prostate cancer, and thK2 is predictive of locally advanced and recurrent cancer in patients with PSA 10 µg/L. Independent of tPSA and fPSA, hK2 predicts unfavorable prognosis.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Human kallikrein 2 (hK2), the product of the KLK2 gene, is a serine protease with 80% sequence identity to PSA. The enzymes share the property of being expressed chiefly in the prostate under androgen regulation (7).
Various tissue studies have documented increases in the ratio of hK2 expression to PSA expression during carcinogenesis and PCa progression (8)(9)(10). Thus, it was hypothesized that hK2 might be a useful biomarker for PCa, particularly for advanced disease. It is, however, unclear whether protein concentrations in tissue correlate with those in circulation; PSA and hK2 concentrations are up to 106-fold higher in tissue than in blood (11). Nevertheless, recent studies have demonstrated that hK2 concentrations in serum are significantly associated with extracapsular extension (ECE) of PCa and with the volume of PCa in prostatectomy specimens (12)(13)(14). Although these studies have provided indications that serum hK2 may be a predictor of advanced PCa, definitive evidence is lacking.
To compare prostate-specific kallikreins in blood for the differentiation of favorable from unfavorable PCa, we assessed free and total PSA (tPSA), total hK2 (thK2), and free hK2 (fhK2) in a large series of 867 patients treated with radical prostatectomy for clinically localized PCa. We assessed the degree of overlap in the prognostic information from free and total forms of hK2 and PSA to gain insight as to whether increased concentrations of these kallikreins are attributable to similar biology or whether they reflect different aspects of the malignant process. Pretreatment concentrations of biomarkers were tested for the capacity to reflect the presence of ECE or seminal vesicle invasion (SVI). Both ECE and SVI are commonly accepted as adverse prognostic factors (15)(16)(17). Biochemical recurrence (BCR) after radical prostatectomy is an unequivocal indicator of eventual clinical progression (18) and thus was chosen as an additional study endpoint. This study is the first clinical evaluation that includes selective measurements of fhK2.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
).
|
All patients treated with radical prostatectomy were scheduled for an annual follow-up visit at our institutional outpatient clinic. Of the 867 patients, 784 patients had at least 1 follow-up evaluation and hence were eligible for analysis of BCR. Informed consent was obtained from all of the participating patients, and the protocol was approved by the institutional review board.
pathologic evaluation
All radical prostatectomy specimens were surface-inked and processed using serial transverse sections at 3-mm intervals according to the Stanford protocol (19). Histological tumor grading was performed according to the Gleason grading system (20). Pathologic stage was defined according to the 1997 American Joint Committee on Cancer staging classification (21). Tumors infiltrating the seminal vesicles were defined as SVI. Our definition of ECE included cancers with or without SVI.
definition of bcr
BCR was defined as postoperative concentrations of tPSA 
0.40 µg/L. The selection of this cut point was based on a study that demonstrated that a significant proportion of patients with a lower PSA (e.g., 0.2 or 0.3 µg/L) did not experience further PSA increases (22). None of the patients received adjuvant therapy before evidence of cancer recurrence.
measurements of biomarkers
tPSA and Free PSA (fPSA).
To measure tPSA and fPSA, we used the commercial version of a previously reported dual-label assay (DELFIA Prostatus Dual Assay, Perkin-Elmer) that measures tPSA and fPSA on an equimolar basis (23). Detection limits were 0.04 µg/L for fPSA (CV, 3.7% at 0.44 µg/L and 17.9% at 0.10 µg/L) and 0.05 µg/L for tPSA (CV, 5.0% at 2.32 µg/L and 13.9% at 0.34 µg/L). The percentage of fPSA (%fPSA) was calculated as %fPSA = fPSA/tPSA · 100.
thK2 and fhK2.
PSA- and hK2-specific monoclonal antibodies were used in solid-phase, 2-site immunofluorometric assays to detect fhK2 and thK2 (24). The thK2 assay used PSA-specific antibodies to block nonspecific signals. The capture antibody of the fhK2 assay did not cross-react with PSA. The functional detection limit (defined as the concentration at which the intraassay CV was <15%) in serum was 0.003 µg/L for thK2 and 0.01 µg/L for fhK2. The thK2 assay imprecision ranged from 2.8% for low (0.01 µg/L) to 1.6% for high (4.28 µg/L) hK2 concentrations. Assay imprecision for fhK2 ranged from 3.7% for low (0.01 µg/L) to 2.0% for high (3.34 µg/L) hK2 concentrations. In 290 patients, we lacked sufficient sample to measure fhK2. Concentrations of fhK2 are therefore available for 577 patients of the 867 who were eligible for final analysis.
statistical methods
To facilitate comparisons between biomarkers in univariate and multivariable analyses, we standardized each marker by dividing by its SD within the group under consideration. The coefficient in the regressions can thus be interpreted as the change in log odds/hazard for a 1 SD increase in the predictor. Analyses were performed using Stata 8.2 (Stata Corp.) and S-PLUS software with the Design library (version 6.2, Insightful Corp.).
Risk Assessment for ECE and SVI.
Univariate logistic regression analysis was performed to assess the association of all biomarkers with the presence of ECE or SVI. Diagnostic accuracy of each variable was quantified by the area under the ROC curve (AUC), which ranges from 0.5 (chance or a coin flip) to 1.0 (perfect ability to rank). Multivariable regression models were generated to assess whether prediction would be improved by combining markers; the AUC was adjusted by bootstrap methods (25).
Risk Assessment for BCR.
Cox proportional hazards regression was used to determine the association between each biomarker and BCR after radical prostatectomy. We constructed multivariable models including a combination of all biomarkers, again using bootstrap methods to correct for overfit. Predictive accuracy was defined in terms of the concordance index (c-index). In brief, the c-index is comparable to the AUC and can be used to quantify discrimination for a single-variable or multivariable model for survival time data.
Comparison of Predictive Accuracy.
We sought to determine whether a single biomarker or combination of biomarkers enhanced predictive accuracy compared with tPSA alone. We used tPSA alone as the comparator, because it is the marker most frequently used for prediction of ECE, SVI, and BCR. Significance tests for differences in AUC and c-index were conducted, respectively, using the roccomp command in Stata and the rcorrp.cens function from the Design library of S-PLUS.
fhK2 measurements were available for only 577 patients. Consequently, fhK2 and %fhK2 were evaluated separately. These 577 patients, however, did not differ significantly in clinical and pathologic characteristics from the whole cohort of 867 patients. For the sake of simplicity, the fhK2 and %fhK2 results are displayed in the same tables as analysis of thK2, tPSA, and fPSA for all 867 patients. For comparison of predictive accuracy, the comparator for fhK2 and %fhK2 was the AUC/c-index calculated for tPSA, excluding patients for whom fhK2 was missing.
A subgroup analysis was performed on men with moderately increased tPSA concentrations (10 µg/L), who are more typical of PCa patients in the US and other countries where tumors are generally detected by PSA screening.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
. This cohort is reasonably representative of a contemporary referral population (26). ECE or SVI was observed in 249 (29%) and 87 (10%) of patients, respectively. In subanalysis of the 627 men with pretreatment PSA 10 µg/L, ECE or SVI was present in 142 (23%) and 43 (7%) of patients, respectively. Median concentrations of tPSA, fPSA, thK2, and fhK2 increased with increasingly adverse prostate pathology: concentrations were lowest in the overall cohort and highest in patients with SVI (Table 2
). Conversely, %fPSA and %fhK2 decreased with increasingly adverse pathology (Table 2
).
|
prediction of ece and svi
The associations of biomarkers with ECE and SVI are shown in Tables 3
and 4
. thK2 was a strong discriminator of ECE from non-ECE. However, compared with tPSA alone, thK2 alone was not a significantly superior classifier of ECE in the entire cohort (AUC = 0.654 for tPSA and AUC = 0.662 for thK2; P = 0.6; Table 3
) The combination of tPSA, fPSA, and thK2 into a single multivariable model significantly enhanced discrimination of ECE compared with the use of tPSA alone for the whole sample (P = 0.001). This model, however, was not enhanced by adding fhK2 or any ratio of single markers (data not shown).
|
|
thK2 measurements were strong predictors of SVI (Tables 3
and 4
), but thK2 was not a significantly superior classifier of SVI compared with tPSA alone in the whole sample (AUC = 0.719 and AUC = 0.663, respectively; P = 0.09). In the low-PSA subgroup, however, thK2 remained highly significant. These results suggest that thK2 is a strong predictor of SVI for men with tPSA 10 µg/L. As with ECE, combining tPSA, fPSA, and thK2 (but not fhK2) into a single multivariable model significantly enhanced predictive accuracy compared with the use of tPSA alone for the entire sample (P = 0.0005).
prediction of bcr
Median follow-up for patients without BCR was 36 months. Among 784 patients eligible for analysis of BCR, there were 119 cases of BCR. The 3- and 5-year recurrence-free probabilities for the study cohort were 83% [95% confidence interval (CI), 80%–86%] and 74% (95% CI, 68%–79%), respectively. BCR occurred in 47 patients with tPSA 10 µg/L (n = 556). The 3- and 5-year recurrence-free probabilities for those men were 91% (95% CI, 87%–93%) and 84% (95% CI, 77%–88%), respectively.
Associations between biomarkers and BCR are displayed in Table 5
, A and B. hK2 forms were very strong predictors of BCR in the whole sample. This was also reflected in the men with tPSA concentrations 10 µg/L, in whom both the free and thK2 forms were highly significant. Combining tPSA, fPSA, and thK2 into a single multivariable model significantly enhanced predictive accuracy compared with the use of tPSA alone for the whole sample (P <0.0005). Of note, in multivariable analyses for men with tPSA 10 µg/L, thK2 remained highly statistically significant (P <0.0005).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
10%–50% that of PSA transcripts (27), which might indicate that these closely related proteases may differ in regard to translation rates or protein stability. The initial work on hK2 as a potential tumor marker was performed in tissue studies by Darson et al. (8). They found more intense hK2-specific immunostaining in lymphatic metastases and in high-grade tumors compared with well-differentiated tumors and benign tissue, whereas contrary findings were reported for PSA (8). Herrala et al. quantified expression by a conceptually different approach; in situ hybridization revealed that hK2 was expressed at higher concentrations in PCa tissue compared with benign prostate tissue (P <0.0005), whereas PSA expression had the reverse pattern (P = 0.06) (9). Herrala et al. further reported that the hK2 gene (KLK2) but not the PSA gene (KLK3) was amplified in PCa tissue, one possible explanation for relative changes in their protein expression. Lintula et al. recently reported confirmatory findings obtained using RT-PCR to quantify relative concentrations of hK2 and PSA mRNA in benign tissue and PCa (10). In their study, the ratio of hK2:PSA mRNA was significantly higher in cancer than in benign tissue (P = 0.03) and higher still in high-grade PCa (P = 0.006) (10). This finding suggests that changes in relative expression of hK2 vs PSA may be associated with carcinogenesis and progression, hence providing further evidence that hK2 might be a useful biomarker for PCa and, in particular, advanced disease.
How do these observations from tissue studies translate into protease concentrations in the circulation? The tight compartmentalization of PSA and hK2 in the healthy prostate is altered in prostatic disease. The disintegration of the continuous layer of basal cells, a characteristic early feature of carcinogenesis, leads to loss of the normal glandular architecture and allows substantial leakage of various proteins into circulation. The covariance of hK2 and PSA concentrations in blood has been determined to be generally <60%, suggesting that hK2 might be useful as an independent biomarker for PCa (28).
Several retrospective studies have investigated pretreatment serum hK2 in patients who have subsequently undergone radical prostatectomy for clinically localized PCa. Increasing hK2 was significantly associated with cancer volume, presence of ECE, and high-grade PCa (12)(13)(14).
We evaluated which marker in blood is most closely associated with locally advanced cancer or with subsequent BCR and thus may be used to identify patients who are at risk to develop metastasis. Analysis of our study population, a reasonably representative cohort of 867 men from a contemporary population treated with radical prostatectomy for clinically localized PCa at a single institution, included fPSA and a recently published method for fhK2 detection (24). Previous studies indicated that a low ratio of fPSA to tPSA (%fPSA) is linked to advanced prostate pathology. In a recently published study, Shariat et al. demonstrated that %fPSA was significantly associated with ECE, SVI, and BCR in a cohort of 402 men with a PSA <10 µg/L (29). Others, however, were not able to confirm these findings. The ratio of fhK2 to thK2 (%fhK2) has been previously measured in a set of 103 patients with PCa, in whom it ranged from 17% to 131% (mean, 81%) (24). However, no study has ever evaluated the clinical relevance of serum fhK2.
Our data demonstrate that serum tPSA concentrations significantly indicate the presence of ECE and SVI and risk of BCR in a heterogeneous population, in which tPSA ranges widely (here from 0.11 to 83 µg/L). In the subgroup of patients with tPSA 10 µg/L, however, tPSA was limited as an indicator of ECE, SVI, and BCR. Hence, our data support recent findings from other investigators, who concluded that tPSA had limited capacity to predict unfavorable features of PCa in men with tPSA <10 µg/L, the group most commonly seen in clinical practice (4)(5)(6). Intriguingly, hK2 retained significant ability to mirror ECE, SVI, and BCR in this truncated PSA range.
The clinical relevance of our findings is important considering of the high proportion of men with serum PSA 10 µg/L in contemporary radical prostatectomy series. Particularly in the US, PSA concentrations at diagnosis have fallen dramatically because of widespread use of tPSA testing for the detection of PCa (5). As an example of a European population, among patients who recently underwent radical prostatectomy (from 2002 to 2003) at University Hospital Hamburg-Eppendorf, the percentage with tPSA 10 µg/L increased to 82%. This trend indicates that the use of PSA testing may also significantly decrease the proportion of men with tPSA >10 µg/L at diagnosis in countries where PSA screening is not officially recommended.
Results from our multivariable analysis demonstrate that a combination of several markers adds to the predictive ability of a single analyte. More accurate predictions of ECE, SVI, and BCR can be obtained by combining tPSA, fPSA, and hK2 in a multivariable model. This observation may be related to a specific interaction of the various protease cascades, which presumably indicates pathologic alteration of the prostate gland before clinical signs. Thus, biological variables should probably not be considered in isolation; more information can be derived from markers in combination. Markers that may be useful for analysis of PCa include the prostate-specific tissue kallikreins or other markers of local cancer progression, such as TGF-ß1, interleukin-6 receptor (30), and the urokinase-like plasminogen activation cascade (31)(32).
The ratio of fhK2 to thK2 in our study slightly decreased from a median of 88% in organ-confined tumors to 78% in cancers with SVI. In univariate analysis, both fhK2 and %fhK2 were associated with ECE, SVI, and BCR. Thus, hK2 in blood is not consistently in the free, unbound form in PCa patients. Significant proportions (median, 15%–20%) are inferred to be stably complexed with serpin-type antiproteases, and these proportions may vary in benign and malignant prostate disease. The diagnostic benefit of selective fhK2 measurement, however, remains limited: according to our multivariable analysis, addition of fhK2 or %fhK2 provided no substantial increase in predictive accuracy for ECE, SVI, or BCR (data not shown). Further evaluations are warranted, including the application of fhK2 for differentiating cancer from noncancer before prostate biopsy.
Several limitations may influence the validity of our findings. The retrospective study design (analysis of frozen serum) may have influenced accuracy of measurements of the analytes. However, the accuracy of the methods used in this study has been demonstrated for tPSA and fPSA measured in archived serum (33). This study used patients from a single institution with large surgical volume, and it featured unusually stringent preanalytical workup and assays of biomarkers; therefore, the findings may not be representative of all clinical settings.
In conclusion, these data support findings from various tissue studies and demonstrate that hK2 concentrations in serum (independent of tPSA and fPSA) indicate PCa of unfavorable prognosis. Our finding that a model combining fPSA, tPSA, and thK2 contributes superior discrimination compared with any of these analytes measured on their own suggests that each reflects different aspects of the malignant process. Serum hK2 testing might play an important role in clinical assessment of the risk of cancer progression after local therapy in PCa patients with only moderately increased PSA, which is typical of the contemporary patient population. If possible, a biochemical serum profile of high-risk PCa should be developed from the combination of established and emerging biomarkers of local progression.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
2 Human genes: KLK3, kallikrein-related peptidase 3; KLK2, kallikrein-related peptidase 2.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
