What's known on the subject? and What does the study add?
A significant proportion of patients diagnosed with prostate cancer do not require immediate treatment and could be managed by active surveillance, which usually includes serial measurements of prostate-specific antigen (PSA) levels and regular biopsies. The rate of rise in PSA levels, which could be calculated as PSA velocity or PSA doubling time, was previously suggested to be associated with the biological aggressiveness of prostate cancer. Although these parameters are obvious candidates for predicting tumour progression in active surveillance patients, earlier studies that examined this topic provided conflicting results.
Our analysis showed that PSA velocity and PSA doubling time calculated at different time-points, by different methods, over different intervals, and in different sub-groups of active surveillance patients provide little if any prognostic information. Although we found some significant associations between PSA velocity and the risk of progression as determined by biopsy, the actual clinical significance of this association was small. Furthermore, PSA velocity did not add to the predictive accuracy of total PSA.
To study whether prostate-specific antigen (PSA) velocity (PSAV) and PSA doubling time (PSADT) are associated with biopsy progression in patients managed by active surveillance.
Patients and Methods
Our inclusion criteria for active surveillance are biopsy Gleason sum <7, two or fewer positive biopsy cores, ≤20% tumour present in any core, and clinical stage T1–T2a. Changes in any of these parameters during the follow-up that went beyond these limits are considered to be progression.
This study included 250 patients who had at least one surveillance biopsy, an available PSA measured no earlier than 3 months before diagnosis, and at least one PSA measurement before each surveillance biopsy.
We evaluated the association between PSA kinetics and progression at successive surveillance biopsies in different sub-groups of patients by calculating the area under the curve (AUC) as well as sensitivity and specificity of different thresholds.
Over a median follow-up of 3.0 years, the disease of 64 (26%) patients progressed.
PSADT was not associated with biopsy progression, whereas PSAV was only weakly associated with progression in certain sub-groups.
However, incorporation of PSAV in models including total PSA resulted in a moderate increase in AUC only when the entire cohort was analysed. In other sub-groups the predictive accuracy of total PSA was not significantly improved by adding PSAV.
Our findings confirm that PSA kinetics should not be used in decision-making in patients with low-risk prostate cancer managed by active surveillance.
Regular surveillance biopsies should remain as the principal method of monitoring cancer progression in these men.
area under the receiver operating characteristic curve
A considerable proportion of prostate cancers currently diagnosed by PSA screening and extended biopsies would never have become clinically evident if left undetected. As a result, active surveillance (AS) has emerged as a way to reduce over-treatment in low-risk prostate cancer patients. The aim of AS is to defer potentially curative treatment until higher-risk tumour features are found. The use of AS is supported by a growing body of evidence that has resulted in its inclusion in guidelines as a management option for low-risk prostate cancer [1, 2].
One of the major determinants of the efficacy of AS is accurate and timely detection of unfavourable tumour changes that necessitate treatment. Most current protocols rely on repeated biopsies to detect such changes. However, biopsies are invasive procedures with potential morbidity and this limits their use. Furthermore, the ability of the current extended transrectal biopsy to accurately identify tumour characteristics is limited. Although alternative techniques such as transperineal and saturation biopsies may provide more accurate information , they are too invasive to be routinely used in AS patients. Therefore, there is a pressing need for less invasive tests that can predict an increased risk of tumour progression and so lead to a confirmatory biopsy.
Measures of PSA kinetics, such as PSA velocity (PSAV) and PSA doubling time (PSADT), are obvious candidates for prediction of tumour progression because they theoretically reflect dynamic changes in prostate histology. The value of these parameters in predicting repeat biopsy results in AS patients was analysed in several studies that provided conflicting results [4-8]. The clinical value of PSA kinetics is known to depend on several factors including the time interval over which PSA values are obtained and the number of PSA measurements used in the calculations. These factors may affect the ability of PSA kinetics to predict progression in AS patients. In an effort to further analyse this issue, we performed a detailed analysis of the association between PSAV, PSADT and biopsy progression in our AS cohort.
Materials and Methods
Our inclusion criteria for AS are biopsy Gleason sum <7, two or fewer positive biopsy cores, ≤20% tumour present in any core, and clinical stage T1–T2a. From October 1994 to August 2011, 277 men with prostate cancer were deemed eligible for AS and were enrolled in our programme. Patients included in our analysis had at least one surveillance biopsy, an available PSA measured no earlier than 3 months before diagnosis, and at least one PSA measurement before each surveillance biopsy. Ten men who have not yet had a surveillance biopsy were excluded from the study, as were 11 men who did not have PSA before each surveillance biopsy. We subsequently excluded five men who commenced treatments potentially affecting PSA levels (such as 5α-reductase inhibitors, exogenous testosterone or surgical treatment for benign prostatic hyperplasia) while on AS, as well as one patient who did not have a PSA measurement within 3 months of the diagnostic biopsy. The remaining 250 men comprised the study cohort. The study received institutional review board approval.
Patients are followed with a PSA and rectal examination that are performed at least once a year. The first surveillance biopsy is performed within 1 year of the diagnosis. Further surveillance biopsies are performed every 1–2 years. All prostate biopsies were performed under transrectal ultrasound guidance and 10–12 cores were usually taken. Biopsy progression is defined as any of the following: Gleason 4/5 cancer, more than two positive cores, or more than 20% involvement of any core.
The PSAV was calculated by linear regression on untransformed PSA values. We also repeated all analyses using PSAV values calculated by regressing log-transformed PSA values as well as using the running average of the rate of change over time to verify that the method of calculation did not affect the results. PSADT was calculated using the log-slope method, as DT = ln(2)/m, where m is the slope of the regression of ln(PSA) over time.
All calculations of PSAV and PSADT were performed using only PSA measurements that were obtained before the surveillance biopsy, the result of which was used to assess the endpoint (i.e. progression or no progression). PSA kinetics in patients with and without progression were compared using the Mann–Whitney U test. We examined the accuracy of PSA kinetics in predicting biopsy progression by calculating the area under the receiver operating characteristic curve (AUC). We also analysed the sensitivity and specificity of different PSAV and PSADT thresholds (0.35, 0.5 and 0.75 ng/mL/year for PSAV and 3, 4 and 5 years for PSADT). This was initially carried out in the entire cohort. Several additional analyses were also performed:
The AUA PSA Best Practice Statement suggests that a correct measurement of PSAV requires at least three PSA measurements over a time period of at least 18 months after the diagnosis (i.e. four total PSA measurements) . Therefore, we separately analysed patients who fitted these criteria.
A short interval between PSA measurements may negatively affect the accuracy of PSA kinetics calculations . To investigate whether the time period over which a patient is on AS (and over which the PSA is measured) affects the predictive accuracy of PSA kinetics, we repeated our analysis separately for the first, second, third and further surveillance biopsies.
National Comprehensive Cancer Network guidelines state that the predictive power of PSAV may be decreased if measurements taken over more than a 24-month period are used for calculations . In patients who had at least three biopsies, we separately analysed the predictive value of PSA kinetics calculated using the PSA measures obtained over the 2 years before the last biopsy.
The PSAV is known to depend on the initial PSA , which may affect its predictive value in patients with certain PSA levels. Therefore all analyses involving PSAV were repeated in groups of patients with PSA at diagnosis of <4 ng/mL and ≥4 ng/mL.
To test possible non-linear relationships between PSA kinetics and the risk of biopsy progression, all AUC calculations were repeated with PSAV and PSADT entered as restricted cubic splines with knots at the tertiles.
We also repeated all analyses defining the biopsy progression as only the presence of Gleason 4/5, because some studies have shown correlation between PSA kinetics and more aggressive cancers . In these calculations increases in tumour volume were not counted as progression.
For subgroups in which the PSAV or PSADT were found to be predictive of progression (i.e. AUC was significantly different from 0.5), we calculated and compared AUC values for two different models. The first model included the last total PSA taken before a particular biopsy. The second model was composed of a combination of PSA and PSA kinetics values for the same biopsy. This was carried out to determine if the predictive value of PSA kinetics is independent of that of PSA.
All tests were two-sided with P ≤ 0.05 considered significant. Stata® version 11.0 was used for all data analysis.
Demographic and clinical characteristics of the study cohort are listed in Table 1. Sixteen patients (6%) had a PSA level ≥ 10 ng/mL and 90 (36%) had a PSA density ≥ 0.15 at diagnosis. In most patients the diagnostic prostate biopsy included at least 10 cores.
Table 1. Characteristics of the study cohort at diagnosis
Over a median follow-up of 3.0 (interquartile range (IQR) 1.9–4.1) years, patients had a mean of 2.3 surveillance biopsies. Sixty-four patients showed unfavourable changes consistent with progression at one of the surveillance biopsies (Fig. 1). These changes included the presence of Gleason 4/5 cancer in 44 (67%) of these patients. Of the 64 patients who progressed by our criteria, 61 (95%) were also not eligible for further AS by Epstein criteria .
A total of 1432 PSA measurements were used in the PSA kinetics calculations in the entire cohort, with a median of five measurements per patient (IQR 3–7 measurements, mean 5.728). Table 2 compares PSA kinetics in patient subgroups with respect to progression or the presence of Gleason 4/5 at the last surveillance biopsy. No significant differences in PSADT were found whereas PSAV was higher in the Gleason 4/5 subgroup, as well as in patients with three PSA measurements performed over at least 18 months and initial PSA ≥ 4 ng/mL. PSAV values were also significantly increased in patients who progressed at the fourth surveillance biopsy or later.
Table 2. Prostate-specific antigen kinetics in different patient subgroups with respect to biopsy progression and presence of Gleason 4/5 cancer
Gleason 4/5 absent
Gleason 4/5 present
aPSA kinetics calculated using PSA values obtained during the 24 months before the biopsy. PSAV, prostate-specific antigen velocity; PSADT, prostate-specific antigen doubling time; IQR, interquartile range; PSA, prostate-specific antigen.
Patients with PSA ≥ 4 ng/mL at diagnosis (n = 180)
PSAV (IQR), ng/mL/year
PSADT (IQR), years
The predictive accuracy of PSA kinetics in different groups is shown in Tables 3 and 4 lists sensitivities and specificities of different PSAV and PSADT thresholds. Several tendencies are evident from the data listed in these tables. Both PSADT and PSAV were at best weakly associated with biopsy progression, although the second parameter was relatively more predictive. However, PSAV correlated better with the presence of Gleason 4/5 in biopsy than with progression by any criteria. This difference was slightly more pronounced in patients with higher initial PSA values. Non-linear transformation of PSAV generally improved its association with the outcomes.
Table 3. Predictive accuracy of prostate-specific antigen kinetics for biopsy progression and presence of Gleason 4/5 cancer in different groups of patients
Area under the curve
Restricted cubic splines
Presence of Gleason 4/5
Presence of Gleason 4/5
†Calculated using only positive PSADT values; ‡PSA kinetics calculated using PSA values obtained during the 24 months before the biopsy. PSAV, prostate-specific antigen velocity; CI, confidence interval; PSADT, prostate-specific antigen doubling time.
The ability of PSAV to predict biopsy progression slowly increased with the number of surveillance biopsies, the duration of AS, the time interval over which PSA measurements were made, and the number of these measurements. In patients who had at least four surveillance biopsies, PSA kinetics were strongly associated with progression: all men with unfavourable biopsy changes had PSAV of more than 0.75 and four out of five had PSADT of less than 5 years (Table 4). However, these calculations are based on a few patients. Repeated analyses using PSAV values calculated by alternative methods provided similar results (data not shown).
Table 5 presents predictive accuracy of models based on the last PSA before biopsy or a combination of the last PSA and PSAV before the same biopsy. This analysis was performed when PSAV was shown to predict progression. A moderate but significant enhancement of the predictive accuracy resulting from adding PSAV to total PSA was found only in the entire cohort (AUC for prediction of presence of Gleason 4/5 increased by 0.04). In other groups the predictive accuracy of total PSA was not significantly improved by adding PSAV.
Table 5. Predictive accuracy of models based on the last PSA before biopsy or a combination of the last PSA with PSA velocity before the same biopsy in different subgroups of patients
PSA is the most widely used serum marker for detecting and monitoring prostate cancer. However, its accuracy in predicting the risk of cancer and its correlation with pathological features of the tumour is limited. This is thought to result from the fact that PSA can be produced by both malignant and non-malignant prostate tissue in untreated patients . The concept of PSA kinetics was developed to overcome this limitation. Theoretically, the growth of malignant tissue is much more dynamic compared with that of benign tissue, which should result in a higher rate of PSA increase over time in patients with prostate cancer. The two most widely used PSA kinetic parameters are PSAV and PSADT, which differ in the mathematical model of PSA change over time that is used for its calculations (linear for the former and exponential for the latter).
Several retrospective studies have shown that pretreatment PSA kinetics provide prognostic information regarding the risk of treatment failure [15-17] and death from prostate cancer [16, 18]. Preoperative PSA kinetics are associated with aggressive cancer features in patients undergoing radical prostatectomy [15, 18]. The potential ability of PSA kinetics to estimate the dynamic changes in tumour aggressiveness makes them an appealing monitoring tool in AS patients. However, the studies performed to date have provided conflicting results. While earlier analyses have shown an association between PSAV, PSADT and adverse histology on repeat biopsies [4-6], two recent studies from major centres [7, 8] failed to show any prognostic value of PSA kinetics in predicting biopsy progression.
In a recent systematic review, Vickers et al.  questioned the ability of pretreatment PSAV or PSADT to provide predictive information that cannot be obtained by PSA alone. Analysis of the literature by Vickers et al.  revealed only one article in which the predictive accuracy of models with and without PSA kinetics was formally compared. Furthermore, many studies were found to have serious methodological shortcomings. Based on their findings, the authors suggested several recommendations for future research.
In the current analysis, we tried to follow the recommendations previously made by Vickers et al. (Table 6) . We also sought to study the effect of the length of the AS period on the predictive accuracy of PSA kinetics. This was based on two considerations. First, the accuracy of PSA kinetics in reflecting cancer growth may depend on the time over which the measurements were made as well as the number of measurements . For instance, King et al.  have shown that sensitivity of PSAV improves with the length of the interval over which it is calculated.
Table 6. Recommendations for research on PSA kinetics suggested by Vickers et al.  and corresponding methodological aspects of the current analysis
All PSA measurements included in the analysis were performed before the last assessment of the outcome (i.e. surveillance biopsy). The biopsy schedule was not based on the PSA kinetics.
Endpoint and the marker should be independent
PSA or its derivatives was not used as a criterion of progression.
PSA kinetics should be entered into the analysis as continuous variables
PSA kinetics values were entered into the analysis as continuous variables.
PSA kinetics should be modelled using non-linear terms
All predictive accuracy analyses were repeated with PSAV and PSADT entered as restricted cubic splines.
Presence of additive predictive power of PSA kinetics when added to existing clinical information should be confirmed
Predictive accuracy of the models based on last total PSA and PSA velocity was compared with that of total PSA alone.
Secondly, the unfavourable changes found at repeat biopsies in AS patients may result from either true tumour progression or better sampling of the prostate. The latter is likely to be more common in early biopsies . As some of these patients may not show significant changes in tumour grade and volume after diagnosis, the PSA levels may also remain relatively stable. Therefore, PSA kinetics may be more accurate in predicting progression in patients managed by AS for an extended period of time. In addition, the practical importance of PSA kinetics may also be higher in such patients. In our experience, patients tend to become more reluctant towards yearly surveillance biopsies after the first 3–4 years of management. At this stage a non-invasive test indicating the necessity of repeat biopsy may be particularly useful.
We performed the analysis in a cohort of AS patients that differ from most other cohorts of similar size, including those in which earlier studies of the PSA kinetics were carried out. In contrast to most other institutions, we do not use PSA or its derivatives as an AS inclusion criterion. This leads to a wider range of PSA, PSA density and, potentially, PSA kinetics values in our cohort. Another potential difference of our cohort results from the fact that compared with most other protocols, our criteria are more conservative with regards to the biopsy findings.
The results from our study showed that PSAV and PSADT calculated at different time-points, by different methods, over different intervals, and in different subgroups of AS patients provide little if any prognostic information. Although some of the comparisons were significant, their actual clinical significance was small. Furthermore, PSAV did not add an important level of predictive accuracy to PSA. For example, in the entire cohort the increase of AUC for prediction of high-grade cancer in repeat biopsy was significant when PSAV was added to total PSA, but with an absolute increase of just 0.04. Whereas some prognostic potential of the PSAV was noted in patients who had at least four biopsies, this subgroup was particularly small. Nevertheless, this finding should be analysed further.
Our analysis has several limitations. The study was retrospective and data were obtained from a single institution. Measurements of PSA used different detection assays, which may have affected the accuracy of PSA kinetics calculations . However, we used the results from one laboratory for most of the studied patients. As mentioned before, our AS criteria differ from those used by others; though most of our patients would have been eligible for other protocols at the time of diagnosis and ineligible at the time of progression. We assessed prostate cancer progression using biopsy results, which are known to be limited in their ability to conclusively describe true characteristics of prostate cancer. Therefore, using biopsy results as a sole measure of progression may not be accurate in describing the absence of disease changes in some of the patients. Finally, as mentioned above, the power of many analyses was limited by the small number of patients included.
The results of this study confirm that PSA kinetics should not be used in decision-making in patients with low-risk prostate cancer managed by AS. Although we have found weak association of PSAV with biopsy progression in some subgroups of the patients, it was not clinically meaningful. Therefore, regular surveillance biopsies should remain as the principal method of monitoring cancer progression in these men.
The authors thank the Center for Urologic Research, Education, and Diseases (CURED) and Mr Vincent Rodriguez.
Conflict of Interest
None declared. Source of funding: Department of Urology, University of Miami Miller School of Medicine, FL, USA.