Cancer Therapy Vol 2, 441-454, 2004

1Department of Surgery, Uppsala Academic Hospital, Sweden;

__________________________________________________________________________________

Key words: Prostate Cancer, Mass screening, PSA

Abbreviations: a 2-macroglobulin, (A2M); a 1-antichyomotrypsin, (ACT); prostatic intra-epithelial neoplasia, (PIN); prostate specific antigen, (PSA); World Health Organisation, (WHO)

Prostate cancer is one of the most common forms of cancers amongst men in the Western world, and a major cause of cancer death. Despite this, there is still no consensus about how prostate cancer mortality should be reduced. The only primary prevention that has been prove to be effective so far is Finasteride (Thompson, 2003). Treatment with Finasteride, however, may adversely affect the sexual function. Furthermore, although it prevents or delays the appearence of prostate cancer, it may increase the risk of poorly differentiated cancers. Secondary prevention, i.e., screening, has been the subject of intense debate after the introduction of Prostate-Specific Antigen (PSA) in clinical practice in the end of the 1980's. The benefit of screening for prostate cancer, however, still remains to be proved. Large-scale randomised controlled trials on prostate cancer screening have been initiated in the United States (Gohagan, 1995) as well as in Europe (Schröder, 1999). Hopefully, these studies will provide definite results in 2005-2008 (de Koning, 2002). A population-based trial of screening has been performed in Quebec, showing a significant reduction in prostate cancer mortality in those men who fulfilled the screening (Labrie, 2004). Although this study has been criticised for the poor compliance in the group allocated to screening, the death rate from prostate cancer ten years after the study was initiated was reduced by 62% in screened men versus unscreened. This analysis was based on the 24% of the invited men who fulfilled the screening. Since no shift of men at greater risk of dying from prostate cancer from the screened group towards the non-compliers, the risk of bias in this study is probably small.

Performing a study aimed at showing mortality reduction after the introduction of a screening programme requires extremely large resources. The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial of the National Cancer Institute in the US was planned on the assumption that 74 000 men had to be recruited to achieve sufficient power to show a reduction in prostate cancer mortality after ten years (Gohagan, 1995). Furthermore, these studies must continue for a long period of time since prostate cancer progresses slowly.

One way of testing the benefit of screening for reducing prostate cancer mortality is to compare two similar populations with different screening intensity, although such comparisons are confounded by differences in environmental factors, such as diet, lifestyle and regional differences in how health care is provided. In a study from the Federal State of Tyrol in Austria a significant shift towards detection lower stages of cancer was seen after the introduction of a screening programme (Bartsch, 2001). Mortality rates also declined more rapidly in the Tyrol than in other parts of the country where mass screening was not practised. However, the outcome of this study has been questioned since the decline in mortality was too rapid to be explained by early detection of prostate cancer alone. In a study with similar design, comparing cohorts from two different regions with different prostate cancer screening intensities, no difference in prostate cancer mortality was seen (Lu Yao, 2002).

Targeted screening in men with a family history of prostate cancer is already well established (Bock, 2003) and less controversial, since the higher prevalence, earlier onset and more aggressive natural course of prostate cancer in this group gives a relatively better improvement in prostate cancer-specific survival with screening (Valeri 2002). Screening may also have a beneficial psychological effect for these men (Bratt, 2003). The conditions for targeted screening are thus somewhat different than for the whole population of men and therefore not further discussed here.

While awaiting the outcome of the large-scale randomised controlled studies, we have to rely on studies on a smaller scale addressing surrogate endpoints to obtain temporary evidence regarding the effectiveness of prostate cancer screening. Ten criteria have been identified by WHO, which can be used as prerequisites for the introduction of mass screening programmes (Wilson, 1986). The ten criteria address the following issues:

1. Is the disease under study an important health problem?
2. There must be effective treatment for patients suffering from localised disease.
3. Facilities for further diagnosis and treatment must be available.
4. There must be an identifiable latent or early symptomatic stage of the disease.
5. The technique to be used for screening must be effective.
6. The tests must be acceptable to the screened population.
7. The natural history of the disease, including development of the latent phase, into clinical disease must be sufficiently known.
8. There must be a generally accepted strategy allowing determination of which patients should be treated and which should remain untreated.
9. The cost must be acceptable

Management of the disease in it's early stages must have a favourable impact on prognosis.

The impact of prostate cancer on general health in most Western countries leaves no doubt about its importance as a public health problem. It is the second most common cancer in the Western world (Ferlay, 2001), and causes more deaths than any other cancer in men aged 55-74 (World Health Organisation, 1996). An increase in incidence as well as mortality was seen in most industrialised countries during the 1980's (Hsing, 2000), but the mortality rate became stable during the 1990's (Oliver, 2001). Although quality of life is not profoundly affected for the majority of men with prostate cancer (Sandblom, 2001), it causes a considerable reduction when the cancer reaches an advanced stage, especially if associated with pain from skeletal metastases (Litwin, 2001; Melmed, 2002; Sandblom, 2004a).

With the introduction of nerve-sparing radical prostatectomy in the beginning of the 1980's, a new method of radical treatment for localised prostate cancer without excessive blood loss and decreased risk for urinary incontinence and potency became available (Walsh, 1983). Radical prostatectomy can either be performed as retropubic prostatectomy, which is the more common approach today, or as perineal prostatectomy. Reports from specialised centres with high competence have shown very favourable results after radical prostatectomy (Han, 2001), but the same results have not been reproduced when the technique is spread to the hospital community at large (Lu Yao, 1996).

An alternative to radical prostatectomy is radiotherapy with curative intent. Although no large-scale randomised studies comparing the outcome of radical prostatectomy with radiotherapy have been reported, it is often considered that radiotherapy is the treatment of choice for older men and men with tumours with suspected extracapsular growth (T3a (UICC, 1997)). A problem with radiotherapy is the side-effects, such as incontinence, erectile dysfunction and rectal bleeding. Although such side-effects are typically associated with radical prostatectomy, they may also appear after radiotherapy. With advances in radiotherapy over the last decade the dose delivered to the organs surrounding the prostate has been reduced by the introduction of conformal radiotherapy (Dearnaley, 1999). This has made it possible to escalate the dose, without intestinal side-effects. Brachytherapy has evolved as an alternative treatment to radical prostatectomy with less severe consequences for men with small tumours (Schellhammer, 2000). Radiotherapy may also be given as salvation therapy to men with local recurrence after radical prostatectomy (Stephenson, 2004). For men with high-risk disease, the combination of brachytherapy and external radiotherapy is an alternative (Stone, 1999). The outcome after radiotherapy can also be improved by combining it with androgen ablative therapy (Bolla, 1997; Lawton, 2001; Pilepich, 2001; Ataman, 2004). The results from these studies are so convincing that it is now generally accepted that men who receive radiotherapy for locally advanced prostate cancer should always have at least short-term adjuvant hormone therapy. There are, however, still controversies regarding the duration of the hormone therapy (Hanks, 2000).

The diagnosis of prostate cancer requires access to transrectal ultrasound with the possibility of core biopsy, as well as units where radical treatment as presented above can be provided. In most Western countries where screening with PSA is practised or considered, this is usually present. However, whereas a PSA test and digital rectal examination does not require large resources, a screening programme should not be initiated in a population where socioeconomic conditions do not allow follow-up of a positive test. In South Africa, only 19% of screened black patients with an elevated PSA undergo prostate biopsy, usually due to economic problems or because they did not have a postal address or telephone number (Heyns, 2003).

It is generally accepted that prostate cancer can be treated radically as long as the primary tumour is confined to the prostate and there are no regional or distant metastases, and that the prognosis becomes much poorer if the tumour has spread beyond the capsule of the prostate (Epstein, 1993; Aus, 2003). Exploration of pelvic lymph nodes is usually done before treatment with curative intent is attempted, since local spread results in a poorer prognosis what ever treatment is given (Aus, 2003). All screening programmes are aimed at detecting prostate cancer whilst still confined to the prostate and has neither spread locally nor to the skeleton (T1-2, N0, M0 (UICC, 1997))

High-grade prostatic intra-epithelial neoplasia (PIN) is sometimes considered a pre-malignant lesion of the prostate (Häggman, 1997). However, there is not sufficient proof that PIN may develop into manifest prostate cancer to warrant screening directed at PIN detection.

Since PSA evolved as a screening tool in the late 1980's, screening for prostate cancer has spread dramatically. PSA has become the major tool for early prostate cancer detection. As the first step in a screening programme it is used to select those men who should undergo further diagnostic procedures. It may be performed in combination with digital rectal examination and transrectal ultrasound, although this adds little to sensitivity or specificity and is often considered unnecessary (Schröder, 1998). Regular screening programmes and uncontrolled opportunistic screening has lead to a dramatic increase in the detection of prostate cancer, although evidence supporting the benefits of screening are still lacking. In a retrospective review of men undergoing radical prostatectomy, it was found that widespread early detection programmes for prostate cancer resulted in a downward stage migration (Han, 2004), but the outcome from studies of this design are difficult to interpret and may be biased by other circumstances. Although PSA is one of the tumour markers with the highest predictive values available, it is still not optimal for screening purposes. In men with PSA below 4 ng/ml enrolled in the Prostate Cancer Prevention Trial in the United States, 15% were found to have prostate cancer (Thompson, 2004). Several strategies have therefore been suggested to increase the sensitivity and specificity of PSA. An improved specificity is desirable in order to avoid unnecessary anxiety associated with suspected cancer and the risk for complications from prostate biopsy (Crundwell, 1999). A high sensitvity reduces the number of tumours missed at screening, although the risk for overdiagnosis with too high a sensitivity must be taken into account (Etzioni, 2002).

One attempt to distinguish between the elevation of PSA caused by benign prostatic hyperplasia and that caused by prostate cancer is to estimate the free-to-total PSA. The majority of PSA is complexed with a 1-antichyomotrypsin (ACT) or a 2-macroglobulin (A2M), while a smaller fraction remains unbound. A low quote of free PSA to the total amount of PSA is indicative of prostate cancer (Luderer, 1995; Catalona, 1998; Aus, 2004).

Whereas men with benign prostatic hyperplasia may have a constantly elevated PSA, a rise in PSA over time is associated with the presence of a cancer. If PSA increases by 0.75 ng/ml per year there is a greater risk for prostate cancer than in men with a slower rise in PSA (Carter, 1992). PSA velocity is most effective for detecting tumours with a relatively low initial PSA level, preferably below 4.0 ng/ml.

Even if a tumour is not detected at biopsy or the PSA level does not exceed the indication threshold for biopsy, the risk for developing cancer before the next screening occasion is related to the PSA. By taking this into consideration when deciding on screening intervals, the specificity can be improved (Hugosson, 2003; Ito, 2004b). In a study based on a screening programme in Japan, it was suggested that the rescreening interval was set at 3-5 years with a PSA level below 1.0 ng/ml, 1-2 years with a PSA level of 1.1-2.0 ng/ml and 1 year with a PSA of 2.1-4.0 ng/m (Ito, 2004a). A screening programme with the intervals adapted to the baseline PSA levels may be designed as in Figure 1.

On the assumption that the equivalent volume of a prostate tumour causes a greater elevation in PSA than benign prostatic hyperplasia, PSA density has been suggested as method of increasing the sensitivity (Benson, 1992). PSA density is defined as serum PSA divided by the prostate volume. This requires transrectal ultrasound, which limits the usefulness of PSA density.

A cut-off level of 3-4 ng/ml is commonly set as the level triggering further investigations, representing a balance between the risk for false negative and false positive results. If the natural increase in PSA with time is taken into consideration, the number of unnecessary biopsies may be reduced (Oesterling, 1993). Whereas for men below 50 years of age a PSA level of 2.5 ng/ml may be the optimal cut-off level, the cut-off level is gradually increased to 6.5 ng/ml for men older than 70 years in order to avoid false positive results. Although this also leads to an increasing number of false negative results, sensitivity and specificity is maintained at a reasonably stabile level through all age groups.

There are natural fluctuations in PSA levels which may result in false positive results if an isolated elevation in PSA is considered sufficient to trigger further investigations. This can be avoided by confirming the result a few weeks later before proceeding with biopsy (Eastham, 2003). One way of doing this is shown in Figure 2.

Pro prostate-specific antigen is a precursor form of PSA having higher levels in tumour tissue compared to benign prostate tissue. It provides a more specific serum marker for prostate cancer than the combination of free and total PSA. It has been shown to give the best improvement in specificity in the PSA interval 2-4 ng/ml, but may be useful with PSA levels up to 10 ng/ml (Catalona, 2003). Pro prostate-specific antigen in combination with free prostate-specific antigen can be used for screening as shown in Figure 3.

Human kallikrein 2, like PSA, is secreted by the prostate epithelial cells with the function of converting the precursor form of PSA into active PSA. It is very similar to PSA and has been suggested as a measure used in combination with PSA to predict extraprostatic extension of the cancer (Haese, 2000). In men with PSA 3.0 ng/ml or greater with a negative prostate biopsy, measurement of human kallikrein 2 has been shown to be useful in combination with total and free PSA, giving a higher specificity than PSA alone (Becker, 2003). Human kallikrein could thus be used in follow-up screening to help select men who should undergo repeated biopsies if PSA continues to be elevated (Figure 4).

Figure 1. Screening with differentiated intervals. Reproduced from Ito et al, 2004 with kind permission from International Journal of Cancer.

Although the second step in the diagnostic process, i.e. transrectal ultrasound and core biopsy, has a limited sensitivity, especially for larger tumours (Rietbergen, 1998; Basillote, 2003; Ung, 2003), most efforts to improve screening effectiveness were previously focused on the first steps in the process. However, in recent years several reports on how biopsy technique can be improved have been published. Using computer-simulation models of the prostate, the sensitivity has been shown to increase from 66%, applying traditional sextant biopsies, to 91% (Kawata, 2003). Another way of increasing sensitivity is to take biopsy samples via through the transperineal approach and not the transrectal approach, which is the more common technique (Emiliozzi, 2003). With an extended biopsy scheme, including laterally directed sextant biopsies, the yield can be increased and the variance in prostate-specific and age-related cancer rates minimised (Presti Jr, 2003). The higher detection rate with laterally directed biopsies has also been confirmed in other studies (Fink, 2003; de la Taille, 2003; O'Connell, 2004)C:\GetARef\Refs\review prostatacancer, sorterad 3 040615.ref #96; C:\GetARef\Refs\review prostatacancer, sorterad 3 040615.ref #101;. Furthermore, tumours detected in the lateral zone also seem to have more malignant (Kawata, 2003). Another way of increasing the sensitivity is to direct the biopsies towards the transitional zone (Damiano, 2003; de la Taille, 2003).

Although PSA sampling itself is widely accepted in the screened population, the anxiety raised by a false positive result has to be considered. Transrectal ultrasound with directed biopsies of the prostate is a safe and simple procedure to confirm the diagnosis in the case of a positive if PSA (Applewhite, 2001). Although core biopsies carry the risk of complications such as infections and bleeding, it is usually accepted by the population (Aus, 1993). However, a study from South Africa has shown poor compliance regarding further investigation after a positive PSA result (Heyns, 2003), possibly reflecting a reluctance to further investigations even when the suspicion of cancer has been raised. This emphasises the importance of adequate information before the first steps of screening are initiated in order to avoid interrupting the investigation when only halfway towards the diagnosis of a cancer.

Autopsy series show a high prevalence of prostate cancer in men without symptoms. At the age of 50 years, approximately every third man has occult prostate cancer (Sakr, 1993). As the mortality in prostate cancer does not reach this level, it is thought that the majority of these tumours are indolent tumours that progress very slowly or remain stable for very long periods of time. On the other hand, there is a small number of men with highly malignant prostate cancer which progresses rapidly and eventually kills the host if not treated in time (Albertsen, 1998; Sandblom, 2000). So far, no studies have reliably identified which of the early stage tumours eventually develop into lethal cancer. The total population of men with prostate cancer seems to comprise a continuum from men with practically stationary tumours to those with rapidly progressing cancer.

The natural history of a tumour can be divided into two phases: the interval from the point where the tumour becomes detectable until it presents with clinical symptoms (sojourn time); and the interval from the point where the tumour presents with clinical symptoms until death. As with most malignant diseases, our knowledge of the two phases above is insufficient and partly dates from the time when the natural progress of the cancer was not suppressed by the methods of treatment used today. Androgen ablation has been practised since the 1950's (Huggins, 1941), which obscures our knowledge about the second phase.

The purpose of screening is to detect a tumour as early as possible during the first phase. A major problem is that the natural course of the disease probably differs between different tumours. Whereas a well differentiated tumour with little malignant potential probably has a long sojourn time, it may also have a long interval between onset of clinical symptoms and reaching the stage when it may kill the host. This makes it easily detected in a screening programme, but reduces the relative benefit of early curative treatment. The opposite is the case for poorly differentiated, highly malignant tumours. This is reflected in the classic statement of Whitmore "Is cure possible in those for whom it is necessary, and is cure necessary in those for whom it is possible" (Whitmore Jr, 1990). However, although this may lead to a passive, fatalistic attitude to prostate cancer, it should not be ignored that it is always a potential threat to its host.

One of the most important problems to be resolved in screening for prostate cancer is the absence of reliable predictive factors which indicate those tumours detected at early stage which eventually may develop into highly malignant cancers. In a study of specimens from cystoprostatectomy for bladder cancer in men with a mean age of 60 years without clinical signs of prostate cancer, latent tumours were found in 44% of the step-sectioned glands (Hautmann, 2000). When systematic core biopsy and fine needle aspiration biopsy samples were taken from these specimens in the same fashion as in normal clinical practice, cancer was detected in about 5% of the cases. Although the tumours detected by biopsy in this way in general had larger volumes than those remaining undetected, there was no clear threshold volume at which cancer was found. Whether or not a latent tumour is found or remains undetected after systematic core biopsy or fine needle biopsy, seems thus to be a random process, with the probability of detecting cancer increasing with the tumour volume and diagnostic activity. Accordingly, the more you seek, the more you find, but the prognosis remains unpredictable. As a result of this, no conclusion regarding the progress and survival of tumours detected at an early stage can be made without taking into consideration how the sample was obtained. Although a study based on the Tyrol PSA screening project has indicated that early stage tumours exhibit potentially malignant features, including heterogeneity in differentiation, multifocality, ploidy and proliferation index (Horninger, 2004), the clinical significance of these findings remain to be established. Inevitably screening results in the detection of a large number of small indolent tumours, with a natural course that cannot be predicted.

The time gained by detecting the tumour by screening as opposed to waiting until the tumour becomes symptomatic, i.e. the lead time, has also to be defined. The lead time depends on the methods used at screening as well as the screened population. The sojourn time can be divided into the interval from the inception of preclinical disease to the point at which it is detected at screening (delay time) and the interval between the point of detection at screening until clinical symptoms appear. Several attempts have been made to estimate the sojourn time and the lead time.

In a study from Stockholm, the cumulative incidence of prostate cancer over a twelve year period in a population undergoing screening consisting of a combination of digital rectal examination, PSA and transrectal ultrasound was determined. By comparing the incidence with that of an age-matched population, the median lead time was estimated to be 4.5 years in men with serum levels >3.0 ng/ml (Tornblom, 2004). Based on survival data from the Surveillance, Epidemiology and End Results database, the mean lead time was estimated to be 9 years (Nicholson, 2002). Based on the European Randomised Study of Screening for Prostate Cancer, the mean lead time was estimated to be 12.3 years with a single screening test at the age of 55 years and 6.0 years if screening was performed at 75 years (Draisma, 2003). The diverging estimates in these studies may be explained by different age distributions and screening strategies.

In the majority of cases prostate cancer has a protracted course and mostly affects older men. Competing mortality from intercurrent diseases is therefore high. Since essentially all men with advanced cancer receive hormonal treatment, at least if the cancer reaches the stage that it may be lethal, the true natural course of the disease can not be determined. When analysing the survival of men managed conservatively, the assumption is that they receive hormonal treatment at some stage in the disease before the cancer kills the host.

A few studies on survival of men with localised prostate cancer managed conservatively have been published. By following men who would be potential candidates for treatment with curative intent, i.e. those with tumours confined to the prostate, the natural course of cancer in men who would otherwise be treated with radical prostatectomy or radiotherapy may be determined. In a population-based Swedish study, 223 men with localised prostate cancer were managed by watchful waiting, i.e. they were left untreated at the time of diagnosis and received hormonal treatment when the cancer showed signs of progress (Johansson, 1997). The patients in this group had a survival close to that of age-matched men without prostate cancer in the same population. The high age at diagnosis of this cohort has raised the concern that a high competing mortality may have resulted in a relatively minor effect on the overall mortality from the cancer, and that an analysis of the prostate cancer specific mortality would have been less favourable. Even so, this study has caused debate about the benefit of curative treatment as well as early detection. In a recent publication based on the same cohort, it was shown that the probability of progression to a more aggressive and lethal phenotype increased after 15 years (Johansson, 2004). In another Swedish study a steadily increasing prostate cancer-specific mortality was likewise seen for men with prostate cancer when followed after ten years had elapsed since diagnosis (Hugosson, 1995).

In the United State, the age at diagnosis is lower than that in Sweden. A study on men in Connecticut aged 55 to 74 years at diagnosis has shown a relatively small risk for death in prostate cancer in men with a Gleason score of 6 or less, whereas the prognosis was much poorer for those with Gleason scores of 7-10 (Albertsen, 1998). This pattern, with a relatively favourable prognosis for the majority of prostate cancer patients but a smaller group with rapidly progressing disease has been repeated in several other studies (Chodak, 1994; Lu Yao, 1997; Sandblom, 2000).

The favourable natural course during the first 15 years shown for men with localised prostate cancer in these observational studies implies that there is not much room for improvement in survival by radical prostatectomy or radiotherapy unless the patient is very young and has a long expected survival. As a paradoxical contradiction to these longitudinal studies of men with localised prostate cancer, prostate cancer is still one of the leading causes of cancer death in Sweden as well as most other Western countries. One way of explaining this is that the majority of men dying of prostate cancer have rapidly progressing cancers that escape all efforts of early detection and treatment while they are still curable. Accordingly, screening results in the detection of less malignant tumours that do not benefit from curative treatment. A contrary explanation is that the cohorts of men with localised cancer studied in these observational studies do not have the same natural course as those who are detected in populations where screening is practised more actively (Walsh, 1997). Screening would in that case result in the detection of cancer in men who would have a greater number of years lost due their cancer than those followed in these studies, although this is difficult to assess due to the lead time bias of early tumour detection. The actual truth is probably somewhere between these two explanations.

In a decision model of the cost of screening based on a randomised study in Sweden, including cost of administration, loss of patient time, diagnostic measures and management strategies, it was estimated that the incremental cost per extra detected localised cancer was 168 000 SEK (18.400 Euro) and per curatively treated cancer 356 000 SEK (40.000 Euro) (Sennfält, 2004). The cost is probably similar in other European countries, although price levels in general are quite high in Sweden. This cost should be compared with the life-time cost of palliative treatment of a man with advanced prostate cancer, which was estimated to be 198 400 SEK in the same population (Sennfält, 2003). If improved survival from early detection is eventually proved, a cost of 356 000 SEK may considered reasonable, although this has to be seen the context of other prioritisations made in the health-care budget.

The survival of men with localised prostate cancer if left without curative management is one of the most crucial questions that must be resolved before mass screening for prostate cancer can be recommended. The favourable prognosis of men with localised cancer managed conservatively mentioned in Paragraph G has raised doubt about the potential benefit of radiotherapy and radical prostatectomy. Overdetection, i.e. the percentage of tumours diagnosed at screening that otherwise would not be detected within the patient's lifetime, gives a minimum measure of how many patients do not benefit from early detection and curative treatment. Based on the European Randomised Study of Screening for Prostate Cancer and the assumption of annual screening of men aged 55 to 67 years with PSA, overdetection was estimated to be 50% (Draisma, 2003). By performing a computer simulation on a hypothetical cohort aged 60-84 years in the United States, the rate of overdiagnosis was estimated to be 29% for whites and 44% for blacks (Etzioni, 2002). Another problem in interpreting survival of patients with screen-detected tumours is the risk for misclassification of the cause of death, which may cause a bias in favour of screening (Feuer, 1999).

Despite the concern raised by the slow natural course of prostate cancer shown in several studies, the only randomised and sufficiently powered study comparing radical prostatectomy with conservative management published so far showed a significantly better prostate cancer-specific survival in the men undergoing radical treatment (Holmberg, 2002). No significant difference in overall survival was seen in this study, although a survival benefit may become apparent in the future as further years of follow-up pass by. However, this study was not based on a sample of patients with screen-detected tumours. As mentioned in Paragraph G, screen-detected tumours may have another natural course which could limit the benefit of aggressive treatment. No large studies comparing radiotherapy and conservative management have been published.

Although many questions remain unanswered, several of the prerequisites for screening may be fulfilled. If the effectiveness in detecting tumours with screening is proved and the benefit of treatment while the tumour is still localised holds true, the randomised studies initiated in the United States may eventually confirm an improvement in survival with screening. The Prostate, Lung, Colorectal and Ovary trial in the United States and the European Randomised Screening for Prostate Cancer trial in Europe have the power to show definite results in 2005-2008 (de Koning, 2002). However, even if reliable evidence supporting screening is provided, several problems still remain to be resolved. More than 20 years after results from the first large-scale randomised studies on mammography were presented, clearing the way for mass screening programmes, screening for breast cancer still remains controversial and a the subject of intense debate (Miettinen, 2002; Jatoi, 2003; Green, 2003; Retsky, 2003). In a report from the Cochrane Collaborative, the validity of all randomised controlled studies on breast cancer screening were questioned on several points, in particular regarding differential exclusion of women with breast cancer and misclassification of cause of death (Olsen, 2001). Similar concerns may arise with the prostate cancer screening studies if careful steps are not taken in advance to avoid analogous problems.

Even if future evidence of improved survival provides support for screening, the optimal screening schedule has still to be established. Which age groups should be included? Which methods of detection should be used to reach optimal cost-effectiveness, sensitivity and specificity? Is a high sensitivity always desirable or does a too effective a screening programme result in unnecessary overdiagnosis of tumours that otherwise would have remained undetected throughout the man's lifetime?

Several reports in recent years have supported a screening interval longer than one year, typically 2-4 years (Ross, 2000; Yao, 2001; Draisma, 2003; Hugosson, 2003; van der Cruijsen, 2003; Postma, 2004; Sandblom, 2004b).

Most screening programmes have included men in the range from 50-55 years to 70-75 years. The low prevalence of prostate cancer in men younger than 50 years makes it meaningless to screen earlier than this. Because of the slow natural course (see Paragraph G), radical treatment usually does not result in essential improvement of overall survival if the expected survival is not longer than 10 years. Men older than 75 years are therefore not included in most screening programmes.

As new techniques of early detection and methods of treatment evolve, the design of screening programmes must also be adjusted. This causes problems, since the slow progress of prostate cancer makes it impossible to evaluate an intervention before at least a decade has passed. Accordingly, most studies presenting data on survival often refer to principles of management that are not up to date. When debate about screening for prostate cancer started at the end of the 1980's, digital rectal examination was the main method of detection. PSA appeared later as a complement to digital rectal examination. The first screening programmes therefore usually included digital rectal examination , PSA and sometimes transrectal ultrasound. Examples of such programmes are that in Norrköping, Sweden (Figure 5), the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial in the United States (Figure 6) and the European Randomised Study of Screening for Prostate Cancer (Figure 7). Transrectal ultrasound has now been abandoned in the European Randomised Study of Screening for Prostate Cancer (de Koning, 2002). In recent years, screening has become increasingly dominated by PSA in combination with other serum markers. Several approaches have been used to optimise sensitivity, specificity, such as differentiated screening intervals (Figure 1), repeated PSA measures (Figure 2), using Pro Prostate-Specific Antigen (Figure 3), and free PSA or kallkrein-2 (Figure 4) as a complement to PSA. A common approach is to use an additional marker to decide on biopsy in intermediate PSA ranges. It may be that a combination of several of these approaches may prove useful in the future.

One of the most important questions that remains to be answered is not how sensitivity should be improved, but rather which level of sensitivity is optimal to obtain a balance between overdiagnosis on the one hand and the risk for not detecting tumours that eventually kill the host (Zappa, 1998) on the other hand. Can a screening programme be designed to efficiently filter out the most malignant tumours and leaving the remaining pool of latent tumours undetected? The ideal for that purpose would be marker that is more sensitive for malignant cancers than latent tumours at the same cut-off level. It may be that human kallikrein 2 could act in this way (Haese, 2003), although this remains to be proved in clinical practice. Even if the randomised studies in Europe and the United States will eventually show a significantly increased overall or prostate cancer-specific survival, reflecting the fact that death in prostate cancer was avoided for a part of those screened, there will inevitably be a large group of men who have their tumours detected and treated early although the tumour would never have killed them, and another group of men who die of prostate cancer despite all efforts to avoid it. The natural course of prostate cancer will always tend to concentrate men with prostate cancer to the second and third groups.

Figure 5. The Screening programme in Norrköping, Sweden, from 1993. Reproduced from Sandblom et al, 2004b with kind permission from the European Urology.

Figure 7. The Rotterdam branch of the European Randomised Study of Screening for Prostate Cancer. Reproduced from Postma et al, 2004 with kind permission from Cancer.