European Urology

2.1. PSA distribution in men aged 55–74

Data resulting from the first and second rounds of screening of ERSPC Rotterdam are shown in Table 1. During the first round, 1014 cancers were diagnosed (detection rate, 5.1%); men who turned 75 yr, refused to participate, or died were not eligible for the second screen. Only very minor differences in PSA distribution between the two screens are seen, which are due to the changes in the populations. This information is relevant for future consideration of PSA cut-off values. It is important to realize that with any cut-off values used, a majority of men can be given the reassuring message that no biopsy was necessary. Many similar studies of PSA distribution in clinical settings are available. Table 1 is different; it relates to the general population of men aged 55–75 yr. In particular, the large proportion of men with PSA >10.0 ng/ml is missing.

2.2. Using PSA cut-off values

Most guidelines for early detection of PCa of professional organizations use cut-off values of PSA to indicate a biopsy. Recommendations vary between PSA values of 2.0 and 4.0 ng/ml. Recent data from the control group of the Prostate Cancer Prevention Trial (PCPT) suggest however that the relationship between a positive biopsy and PSA is a continuous one [6]. In the control group setting in which 5587 men were screened and biopsied upon indication or at the end of the study, it was shown that conventional cut-off values, including the lowest ones in use, will miss substantial numbers of cancers and also many of those with aggressive characteristics by Gleason score. The resulting data were used to calculate sensitivity and specificity derived from biopsying all men participating in the control group of PCPT. The data are shown in Table 2. No cut-off points of similarly acceptable sensitivities and specificities can be identified [7]. The classical cut-off value of 4.0 ng/ml misses more than 75% and a PSA cut-off of 2.0 ng/ml would miss more than 45% of biopsy detectable cancers. As will be shown, in the PSA ranges below 3.0 ng/ml, this entails a very high risk of overdiagnosis and overtreatment as well as unnecessary biopsies. It is therefore likely that cut-off values of PSA will continue to be used until a better alternative has been found. The authors recommend not to use a PSA cut-off but to allow men to determine their own risk by using a risk calculator based on the data and to decide whether or not to undergo a biopsy (see www.uroweb.org).

Within ERSPC Rotterdam a biopsy was initially recommended if a man presented with a PSA value ≥4.0 ng/ml and/or an abnormal rectal examination (DRE) or a hypoechogenic lesion on transrectal ultrasound (TRUS). The resulting distribution of men within given PSA ranges, numbers and proportions of biopsies, cancer detection, positive predictive value (PPV), and numbers of biopsies to detect one cancer are given in Table 3. The reader should keep in mind, however, that the information given is based on one PSA determination. Biological and test variations are not taken into account.

3.1. Identification of indolent PCa

Indolent PCa has been defined by Kattan et al [17] as a subgroup of those men who can be identified prior to making treatment decisions as having clinically insignificant PCa according to well-established but arbitrary criteria [18]. In applying these criteria to 1022 radical prostatectomy specimens from two major centers, 20% of men could be retrospectively identified as having “insignificant disease.” These cancers were clinically detected. When the same selection criteria were applied to 247 screen-detected cases classified as having T1c or T2a cancers from the ERSPC Rotterdam who were all treated by radical prostatectomy, a percentage of insignificant cases of 49% resulted [19]. This data set was used to validate the nomogram designed by Kattan et al. No additional significant variables were identified, the model was confirmed, and the discriminative ability was almost identical with an area under the curve in receiver operator characteristic analysis of 0.76. If the resulting probabilities for indolence are applied to all screen-detected PCas, depending on the cut-off value of the probability of indolence, 20–45% of all detected cancers can be safely classified as potentially indolent. These men may be offered active surveillance with probabilities of only 6–15% of including potentially aggressive cancers in this selection [19], and [20].

3.2. Preliminary evidence around overdiagnosis

Overdiagnosis is defined as the diagnosis of a disease entity, which in the absence of screening would not be diagnosed during the lifetime of its carrier [21]. If overdiagnosis leads to treatment of this disease, the term “overtreatment” is applied. Overdiagnosis is determined by disease and host factors, specifically life expectancy. Within ERSPC, with regard to the age group of 55–74 yr and a 4-yr screening interval, mean lead time has been estimated to amount to 10.3 yr and overdetection to 54% [21]. Overdiagnosis, in addition to the parameters mentioned, depends on age and the screening methodology. If screening reduces PCa mortality, a certain amount of overdiagnosis may be acceptable. The degree is up for ethical considerations. For the time being, however, it still remains uncertain which screen-detected PCas can be earmarked as “overdetected” and, in case treatment is omitted, what the outcome in terms of metastatic progression and PCa mortality in this group of men might be. Data of ERSPC Rotterdam do offer opportunity to shed more light on this issue.

3.3. Characteristics and outcome of cancers found in low PSA ranges

There is evidence from the PCPT control group and from ERSPC that the proportion of aggressive cancers increases with increasing PSA values at the time of diagnosis and that, on the other hand, the proportion of indolent cancers is inversely related to PSA levels. ERSPC offered the opportunity to study the prevalence of potentially insignificant (minimal) tumors according to [22] in 550 radical prostatectomy specimens of cancers detected during the first and second rounds of screening (Table 4). The authors realize that these definitions are arbitrary to a large degree because their untreated natural history is poorly defined. Also, it must be noted that a substantial proportion (30–40%) of patients in [22] have the characteristics of aggressive disease. In four specimens the biopsy-proven PCa could not be retrieved. In both rounds the proportion of minimal disease was strongly dependent on PSA values, amounting to 67% and 56% for PSA below 3 ng/ml in round 1 and round 2, respectively, and overall to 33% versus 43% if all cancers were considered [22]. This finding is much in line with the findings obtained in [19]. The nomogram presented in this study is available for the identification of potentially indolent cancers, which are best treated by active surveillance (www.uroweb.org).

ERSPC Rotterdam offers an opportunity to evaluate with an 8-yr follow-up period the outcome of those 15,853 men who participated in the screening arm and initially presented with PSA values <3.0 ng/ml. All cancers detected by screening or interval cancers are included. The total number of men biopsied in the second round because of progression to PSA ≥3.0 ng/ml was 1090. Of the total of 462 cancers, 11 were diagnosed with potentially advanced disease, and 5 died of PCa; all 11 were in the group of 66 interval cancers (Schröder et al, submitted 2007).

The resulting overall detection rate of 2.9% in 15,853 men over an 8-yr period in ERSPC is in sharp contrast with the 21.9% detection rate in 5587 men in PCPT who were all biopsied during an 8-yr period. If PCPT procedures had been used in the ERSPC cohort, 21.9% of men would have been found to have PCa, a total number of 3472 cases. In fact 462 were found. From the available data it can be concluded that 3010 PCa cases were missed and did not show up as screen detected or interval cancers during the 8-yr interval. Although the follow-up period of 8 yr may be too short, the data are reassuring for those men who present with PSA values below 3.0 ng/ml and who do not undergo a biopsy. There seems to be no need to identify all biopsy detectable cancers at a first screen in the PSA range below 3.0 ng/ml (Schröder et al, submitted 2007).

4.1. Verification bias

The application of a new test to a cohort of men with PCa diagnosed by the use of different tests is subject to verification bias (attribution bias, assignment bias). Verification bias arises because biopsies are generally indicated only with elevated PSA values or other abnormal clinical tests such as rectal examinations and not by the new test under study. If not all test negatives undergo a biopsy, the numbers of true positives and true negatives are wrongly estimated, and the accuracy of the test cannot be determined. Sensitivity is calculated by dividing the number of true positives by the sum of the true positives and false negatives (Fig. 1). Determining sensitivity seems easy. However, since we know that there are many biopsy detectable cancers present in the low PSA ranges, the number of men who would test positive and have cancer and the true number of men who test negative and do not have cancer remain unknown. This means that sensitivity and specificity cannot be properly calculated. In light of these limitations, it has been suggested to use the terms “relative sensitivity” and “relative specificity” [23].

Create Platinum Slide Series

Fig. 1 Definitions of test accuracy.

Establishing the true sensitivity of PSAV in diagnosing PCa needs to be done prospectively in a cohort of men who were all biopsied and underwent sequential PSA determinations following a predetermined schedule. The test characteristics of any desired cut-off value can then be determined. If a preset PSAV cut-off value is selected as a biopsy indicator again, “relative” sensitivity and specificity will result because the denominator in both equations remain unknown because the test characteristics of PSA below the chosen cut-off level still remain unknown.

4.2. PSA velocity below total PSA cut-off values

This section refers to commonly used PSA cut-off values of 2.5, 3.0, and 4.0 ng/ml. There is a need to identify aggressive cancers below this cut-off. Do PSA kinetics, specifically PSAV, help in this respect as originally suggested by Carter et al [24].

Smith and Catalona [25] studied 982 serially screened men whose initial screening was negative for cancer. All men had at least one PSA value >4.0 ng/ml and all underwent biopsy. For those men who entered the study with PSA levels <4.0 ng/ml, the PSAV cut-off point of 0.75 ng/ml/yr or more maximized sensitivity and specificity for predicting cancer with an odds ratio of 7.2 (95% confidence interval [95%CI], 4.52–11.47. The authors concluded that PSA slope is useful for serial PCa screening. These data could not be reproduced in a very similar setting [26]. Of the 588 consecutive participants of ERSPC Rotterdam studied, all presented with PSA values <4.0 ng/ml at their first screen and progressed 4 yr later to PSA values >4.0 ng/ml. None were biopsied in round 1; all were biopsied in round 2. The results are shown in Table 5. PSAV was calculated as the rise of PSA in 4 yr divided by 4. With rising PSAV the number of cancers diagnosed and the relative sensitivity decreases dramatically, and cancers detected by the PSA cut-off are missed. This result is associated with an increase of relative specificity in direct relation to increasing PSAV. Because biopsy indication is driven by PSA values ≥4.0 ng/ml, which function as the “gold standard” in this setting, it is obvious that PSAV cannot improve sensitivity (ie, exceed the number of cases found by the “gold standard biopsy indication”). Men who have a rapidly rising PSA but do not reach 4.0 ng/ml are not biopsied, which obviously introduces attribution bias. Perrin [27] pointed out this bias without identifying it by name. He correctly stated that “none of the available (retrospective) studies proves the diagnostic value of PSAV.” Even in a prospective setting, it is unlikely that PSA kinetics will be useful because of the very low proportion of men who have rapidly rising PSA values [25], [28], and [29]. However a strategy that uses PSA (but not biopsy) to identify men with rapid increasing PSA may be useful [10], [30], and [31].

It remains unknown how many PCas would have been detected if PSAV instead or in addition to the PSA cut-off had been used as a biopsy indication.

Other data from ERSPC show, in an easily understandable fashion, that a high PSAV is a relatively rare event in screening for PCa [28]. Table 6 gives the rate of PSA progression to values of ≥3.0 ng/ml 4 yr after an initial normal screen in which all 5771 men presented with a PSA <3.0 ng/ml and did not undergo biopsy. It is evident that those who had the highest PSAV are those men who presented with PSA values <2.0 ng/ml in the first round. The data confirm the findings in PLCO (Schröder et al, submitted 2007) that these men are a relatively small fraction of 4.8% of the 4880 men in this range in round 1. Most cancers are detected in those men who progress from the PSA range 2.0–2.9 to ≥3.0 ng/ml. Data in the original paper show that prognostic factors do not differ between those cancers diagnosed in men who progressed from very low PSA values as opposed to those who progressed from PSA range 2.0–2.9 ng/ml [28].

Other evaluations of PSAV in low PSA ranges include Fang et al [31] who showed in a comparison of 21 men with clinically detected cancer and 68 from the Baltimore Longitudinal Study of Ageing (BLSA) that a PSAV of 0.1 ng/ml or more was associated with a relative risk of being diagnosed with PCa of 6.53. Yu et al [32] examined the relationship between total PSA and PSAV in 1851 men who were biopsied and 894 men who were diagnosed with PCa. They showed that, in the PSA ranges <2.5 or 2.6–4.0 ng/ml, the proportion of men with PSA velocities >2.0 ng/ml/yr was 1% and 14%, respectively. They found that PSAV varies directly with total PSA levels. No multivariate comparison of total PSA and PSAV was carried out.

PSA velocity was also assessed in ERSPC [33] and the PCPT [7]. In these trials sufficient information was collected to allow a multivariate analysis determining the relative value of each of the parameters under investigation. Roobol et al [33] evaluated PSAV in a multivariate setting on data derived from the second round screening of ERSPC Rotterdam in 774 men who all underwent biopsy. Parameters studied included PSA, DRE, suspicious TRUS, PSAV, PSADT, prostate volume, and age. PSAV turned out to be insignificant with an odds ratio of 0.73 (95%CI, 0.21–2.67) [34].

Also, in a complete evaluation of the predictors of PCas in round 1 and round 2 screening in ERSPC Rotterdam in multivariate analysis of round 2 parameters, only PSA increase was significant (odds ratio, 6.9) for those men who had not been previously biopsied. A previous negative biopsy and a large prostate were significant negative predictors. Details are given in reference [34]. The database of PCPT, specifically the data related to the 5519 men biopsied in the placebo control group, also allowed a multivariate analysis of predictors of a positive biopsy [35]. PSAV was a significant predictor in a monovariate analysis but was no longer statistically significant as soon as the single predictor PSA was added to the risk equation.

4.3. Replacing PSA with PSAV

The concept of using PSAV as the biopsy indication independent of PSA cut-off values is based on the observation that PSAV (and PSADT) may selectively diagnose aggressive cancers. If proven, this concept would fulfil one of the major needs of early detection of PCa and potentially decrease the proportion of diagnosis and treatment of cases that might never come to the surface clinically during the lifetime of a patient (overdiagnosis, overtreatment). The value of PSA kinetics in diagnosing aggressive PCa has been studied. Relevant data from observational studies [10], [30], and [31] and from the only available randomized study of surgical treatment versus observation of clinically diagnosed locally confined PCa [36] have been used. As suggested by Carter et al [37], a PSA history of at least three PSA values separated by at least 3 mo should be obtained and collected over a period of 18–24 mo [37]. Then a cut-off of PSAV can be applied as a biopsy indication; 0.35 ng/ml/yr is suggested by Carter et al [10]. Limiting the biopsy indication to those men with a high PSAV in this concept would selectively identify potentially deadly PCa. A number of open questions need to be answered before this concept is applicable: How reliable are three PSA determinations separated by 3 mo and collected over 18–24 mo in estimating PSA kinetics? Does PSAV have added value once the last value is known? Is PCa still curable once this evaluation has been concluded? “Aggressive” disease must be further defined (morbidity also counts). The concept is based on small numbers of observations of clinical cases in whom PSAV was determined at 3-yr intervals. Is it applicable to the screening situation? Appropriate prospective studies will have to answer these questions.

Another still experimental concept is application of the PSA test early in life starting at age 40 yr. Fang et al [38] found on the basis of data from the BLSA that, for men aged 40–49.9 yr, the relative risk was 3.75 (range, 1.6–8.6) with a PSA >0.6 ng/ml to develop PCa 25 yr later. In men aged 50–59 yr using a PSA cut-off level of 0.71 ng/ml, a relative risk of similar magnitude was seen. Lilja et al [30] presented confirmatory findings based on data of the Malmö health care study. These studies relate to clinical cases. The value of serial early PSA determinations for early detection is still to be shown.