To report the initial results from Sweden of a large population-based randomized study of screening using prostate-specific antigen (PSA) to detect prostate cancer, as the efficacy of such screening to decrease prostate cancer mortality has not yet been proven.
From the population registry men aged 50–66 years were randomized to screening (9973) and to future controls (9973). Men randomized to screening were invited to have their serum measured for free PSA (fPSA) and total PSA (tPSA) in serum using the Prostatus® f/tPSA assay (Perkin-Elmer, Turku, Finland). Men with a tPSA of < 3.0 ng/mL were not further investigated, while those with a tPSA of ≥ 3.0 ng/mL were investigated with a digital rectal examination (DRE), transrectal ultrasonography (TRUS) and sextant biopsies.
Of those invited, 60% accepted PSA testing and 11.3% had a tPSA of ≥ 3.0 ng/mL. Altogether 145 cancers were detected (positive predictive value, PPV, 24%); none were stage M1, two were stage N+ and 10 stage T3–4. Most (59%) cancers were impalpable and 39% were both impalpable and invisible on TRUS. At biopsy, 7% were Gleason score 2–4, 71% 5–6, 19% 7 and 2% Gleason score 8–10. A threshold tPSA of ≥ 4.0 ng/mL would have detected 109 cancers in 366 biopsied men (PPV 30%) while cancer detection would have been 14% higher with a PPV of 36% using a threshold tPSA of ≥ 3.0 ng/mL combined with a f/tPSA threshold of ≤ 18%.
PSA screening detects early-stage low-grade prostate cancer. Both the sensitivity and specificity can be increased by incorporating f/tPSA with a tPSA threshold of < 4 ng/mL.
Despite the slow growth of most prostate cancers progression to advanced stages usually occurs in patients followed for a long period and those offered only hormonal therapy often die from the disease [1–3]. Improved treatment is urgently needed as prostate cancer has become a major cause of death from malignant disease and is associated with severe morbidity. Diagnosis at an early stage and curative treatment have become options to improve survival in prostate cancer, in a strategy similar to those used for other solid cancers . The introduction of serum PSA measurements provided the major breakthrough, particularly in the USA . However, the specificity of PSA level for cancer is low at < 10 ng/mL. At total PSA (tPSA) levels of 4–10 ng/mL three of four men will have other diagnoses than cancer causing the higher level of tPSA in serum , which generates many unnecessary and costly biopsies, and causes anxiety in men who do not have prostate cancer. However, measuring the proportion of free PSA (fPSA) in relation to that of complexed PSA (cPSA) or tPSA (i.e. tPSA ≈ fPSA + cPSA), as first shown by Stenman et al. and Christensson et al., can more efficiently discriminate subjects with BPH alone from those with cancer than can tPSA levels alone . In the present study we evaluated the efficacy of serum levels and proportions of fPSA and tPSA as a means to detect clinically localized prostate cancer in a large randomly selected population-based group of men aged of 50–66 years who had previously not been screened.
SUBJECTS AND METHODS
This study was based on the population of men born from 1 January 1930 to 31 December 1944 (32 298) in Göteborg, Sweden, at 1 January 1995. Of these, 10 000 were randomized to screening and 10 000 as controls. Men previously diagnosed with prostate cancer (56) were excluded from the study, using the regional cancer registry. This resulted in a screening population of 9973 men and 9973 men as controls ( Fig. 1); the numbers may not be definitive.
Men selected for screening were invited by letter for PSA testing. Blood was collected by venepuncture from men who agreed by informed consent. Men with tPSA levels of < 3 ng/mL were informed and not further evaluated. Men with tPSA levels of ≥ 3.0 ng/mL were invited for an examination by a urologist including a DRE, TRUS and sextant biopsies. DRE and TRUS were performed with the subject lying on his left side. Palpable abnormalities were characterized according to UICC 1992 . TRUS was conducted using a 3535 ultrasound system with a 7-MHz biplanar probe model 8551 (B&K Medical A/S, Naerum, Denmark). Hypoechoic lesions were registered and prostate volume calculated using the formula of the ellipsoid (width × height × length × π/6). TRUS-guided transrectal sextant biopsies were taken using 1.2 mm Trucut needles (Collibri Medical, Helsingborg Sweden) mounted in a spring-loaded biopsy gun, and directed as described by Stamey . Each biopsy was numbered and handled separately. Cancers were graded according to the Gleason system. Radionuclide bone scans were recommended in patients with cancer and tPSA levels of > 20 ng/mL.
BLOOD SAMPLING AND PSA MEASUREMENT
Blood (7 mL) was collected in Vacutainer® tubes and allowed to clot; serum was separated from blood cells by centrifugation at 3000 g for 20 min and stored frozen at −20 °C within 3 h from sampling. PSA in the serum was analysed within 2 weeks from sampling and < 3 h from thawing the samples at the Department of Clinical Chemistry, University Hospital UMAS, Malmö, Sweden. We used a commercial dual-label assay for simultaneous measurements of fPSA and tPSA (DELFIA Prostatus® total/free PSA-Assay, Perkin-Elmer, Turku, Finland). The assay uses monoclonal capture antibody (mAb) H117 for fPSA and tPSA, samarium-labelled mAb H50 as a tracer antibody for tPSA, and europium-labelled mAb 5A10 as tracer antibody for fPSA [17,18]. mAbs H117 and H50 provide equimolar detection of free and complexed forms of PSA with a correlation of r= 0.97 with the Tandem-R PSA assay (Hybritech Inc., San Diego, CA) [18,19]. mAb 5A10 detects fPSA with < 0.2% cross-reaction from PSA-≈α-chymotrypsin [10,11]. The detection limit for tPSA is 0.05 ng/mL (5% coefficient of variation at 2.3 ng/mL), and 0.04 ng/mL for fPSA (5.9% at 0.25 ng/mL).
Comparisons of median values and statistical significance were tested with nonparametric tests (Mann–Whitney U-test) as tPSA and fPSA did not fulfil the criterion of a normal (Gaussian) distribution. The chi-squared test was used for comparing positive predictive values (PPVs) for different tests.
The study started in January 1995 and all men invited for PSA testing and those with a tPSA of ≥ 3.0 ng/mL had been biopsied by end of December 1996. Invitation errors occurred in 162 of the 9973 randomized men. In all, 5853 of the remaining 9811 men (59.7%) agreed to venepuncture for measurements of fPSA and tPSA in serum. The distribution of tPSA levels was skewed, with a median of 1.0 ng/mL (Table 1). tPSA levels were ≥ 3.0 ng/mL in 11.3% (660) subjects (Fig. 1) of whom 611 (92.6%) agreed to a further examination with DRE and TRUS, and TRUS-guided biopsies (sextant biopsies in 559 and other types of biopsies in 52).
Table 1. Characteristics of 611 biopsied men with serum tPSA levels of ≥ 3.0 ng/mL; NS, not significant.
Prostate cancer was diagnosed in 145 men (PPV 23.7%), whereas the remainder (466) had benign tissue alone in their biopsies. The overall cancer detection rate was 2.5%; 1.3% at 50–55 years, 2.6% at 56–60 and 3.8% at 61–66; the detailed characteristics are given in Table 2. Forty of 44 men with cancer and a tPSA of > 20 ng/mL were investigated with radionuclide bone scans; none was diagnosed with bone metastases. Two patients were diagnosed with lymph node metastasis (stage N1) while 66 of 68 men who had lymph node dissection were stage N0 (Table 3).
Table 2. Numbers of PSA-tested men, detected cancers and percentage of palpable cancers stratified for various tPSA intervals; NA, not assessed.
No. of men biopsied
No. with cancer (PPV)
% palpable cancers
Table 3. Distribution of clinical stage (T stage) and Gleason score in biopsies from the 145 men diagnosed with cancer
Stage or grade
T2 (palpably intracapsular)
T3–4 (palpably extraprostatic growth)
2–4 (well differentiated)
5–6 (moderately differentiated)
7 (moderately to poorly differentiated)
8–10 (poorly differentiated)
Different combinations of fPSA and tPSA were evaluated for their ability to detect prostate cancer (Tables 4 and 5). The tPSA threshold of ≥ 4.0 ng/mL detected only 109/145 cancers (75.2%) whereas 257 men with benign biopsies had tPSA levels of ≥ 4.0 ng/mL (PPV 29.8%). As many as 36 men diagnosed with cancer (24.8%) would have remained undetected with a tPSA of ≥ 4.0 ng/mL (instead of ≥ 3.0 ng/mL) as the sole indication for biopsy. Furthermore, a DRE would not detect many prostate cancers in men with a tPSA of 3.0–4.0 ng/mL, as only five of 36 of these men (14%) had a positive DRE.
Table 4. Comparisons among four different screening algorithms
tPSA, ng/mL f/tPSA, %
No. of men with:
Cancer detection, %
Table 5. Number of men with cancer vs men with benign biopsies using different thresholds of f/tPSA at various tPSA levels
In four men with benign outcomes at biopsy and one with cancer the f/tPSA was not analysed.
There were 126 of 144 (88%) men with cancer who had an f/tPSA of ≤ 18%, whereas 226 of 462 (49%) with benign biopsies had a f/tPSA of ≤ 18%, which corresponds to a PPV of 35.8% (Table 5). Eliminating 236 men from biopsy reduced false-positives by 51% at the expense of missing 18 of 144 cancers (12.5%) with an f/t PSA of ≥ 18%. However, the median gland volume of 51.0 mL (mean 53.9) in these 18 men was significantly higher (P < 0.001) than that of 30.0 mL (mean 33.0) in men with cancer and an f/tPSA of ≤ 18%. A threshold at ≤ 22% rather then ≤ 18% increased the sensitivity for detecting cancer from 88% to 93% while false-positives decreased by 31% (Table 5).
The tPSA distribution in the present study is similar to that found by Kane et al.; the present incidence of elevated tPSA in serum (11.3%) is lower than the 17.9% reported by Labrie et al. but this probably depends on the older men included in the Canadian study (45–80 years). Most men had a very low tPSA, half with levels of ≤ 1.0 ng/mL and 80% of ≤ 2 ng/mL (Table 2). The cancer detection rate of 2.5% is lower than in other studies. The higher detection rates reported by Catalona et al. of 3.1% and Labrie et al. of 3.5% are probably mostly a result of including older men and the selection bias from recruitment of study populations by advertisement . The proportion of cancers among those with a positive screening test is very similar, at 21.9% (145/660) in the present study, to the 19.3% in  and 20.1% in . Also, higher detection rates have been reported when other screening tools have been incorporated in the primary screening. Bangma et al. combined tPSA in serum with a DRE in 1726 men aged 55–76 years; they reported a cancer detection rate of 4.0% from the biopsy of men with an abnormal DRE, TRUS or tPSA of ≥ 4.0 ng/mL. Gustafsson et al. reported a detection rate of 3.6% using an abnormal DRE or TRUS, or PSA of ≥ 10 ng/mL as indications for biopsy in 1782 men aged 55–70 years.
The present study detected cancers of early stage; 93% were clinically localized, there were no distant metastases, two had lymph node metastases and 10 had locally advanced disease. This stage distribution resembles that in other screening studies; Labrie et al. reported that 8% had disseminated disease and 71% were found with localized disease. Catalona et al. reported that 94% had clinically localized disease and 63% had negative margins at radical prostatectomy. Gustafsson et al. found that 3% had disseminated disease and 62% clinically localized disease. However, our findings are significantly different from current disease demographics reported for 1995–96 in Sweden, as only 10% of the patients with cancer detected were offered curative treatment, whereas half were given palliative treatment alone .
We used tPSA levels of ≥ 3.0 ng/mL as the biopsy indication and a DRE was not used in men with a tPSA of < 3.0 ng/mL. Standard indications for biopsy are still an abnormal DRE or a tPSA of ≥ 4.0 ng/mL. Finding that 47% of tumours had spread outside the capsule at tPSA levels of 4–10 ng/mL  suggests that men with cancer at these tPSA levels may have a higher risk of disease progression than those with lower tPSA levels . In our study, 15% of the men with a tPSA of 3.0–4.0 ng/mL had cancer. Therefore, 36 of 145 (25%) of all men detected with cancer had a tPSA of 3.0–4.0 ng/mL but only five of 36 (14%) were palpable. The DRE is less sensitive for detecting cancer at low PSA levels (Table 2) and most cancers were not DRE-positive in this interval . Furthermore, six tumours were not organ-confined and the risk of treating low-volume ‘insignificant’ cancer may be low . This is consistent with the report from Catalona et al. who found 81% of cancers with a tPSA of 2.6–4.0 ng/mL to be organ-confined and 17% to be ‘insignificant’. In that study, 59% of the cancers were impalpable. In our study the proportion of palpable cancers increased from 14% at a tPSA of 3.0–4.0 ng/mL to 75% in cancers at a tPSA of > 20 ng/mL (Table 1). This indicates that the DRE is sensitive at high but insensitive at low tPSA levels. Nor does TRUS increase the sensitivity, as 52% of cancers were invisible on TRUS and 39% were both impalpable and invisible (Table 2). The proportion of DRE-negative and TRUS-negative cancer is higher than in other reports [13–16], which is probably because of the different biopsy indications. We suggest that the DRE might be replaced as a primary screening tool by a tPSA threshold of ≥ 3.0 ng/mL to increase the detection of organ-confined cancers, as the DRE is time-consuming and insensitive. Our high proportion of impalpable TRUS-negative cancers suggests that sextant biopsies are indicated in men with a tPSA of ≥ 3.0 ng/mL.
An important disadvantage of low tPSA thresholds is the decrease in specificity. In this study the PPV at a tPSA of 3.0–4.0 ng/mL was 14.7% (Table 2), but disadvantages of low PPVs caused by low tPSA thresholds can be significantly diminished by using f/tPSA [22,23], which was measured simultaneously with tPSA using a commercially available dual-label assay. Excluding men with > 18% f/tPSA would save 236/462 (51%) with benign biopsies from further examination. Eighteen cancers (12.5%) with f/tPSA threshold would not have been detected with this algorithm, which agrees with the value reported by Bangma et al.. Cancers with > 18% f/tPSA are most common at tPSA levels of 3.0–5.0 ng/mL, and in large glands. It is possible that thresholds for f/tPSA should be higher at low tPSA levels and lower at high tPSA levels. Our data suggest that a tPSA of ≥ 3.0 ng/mL combined with f/tPSA of ≤ 18% is more sensitive and more specific as a primary screening tool than a tPSA of ≥ 4.0 ng/mL (Table 4). At < 22% f/tPSA, 11 of 145 cancers (8%) would not be detected (Tables 4 and 5), at the expense of decreased specificity (30% PPV). This study shows that low-grade low-stage cancers are detected by screening with tPSA in serum from men aged 50–66 years in a previously unscreened population. Less than 10% of cancers were too advanced for curative treatment. A primary screening programme using tPSA and fPSA testing in serum gives a positive test in 7% of men and a cancer detection rate of 2.2%. Randomized studies like the European Randomized Screening for Prostate Cancer are necessary to evaluate the benefits and hazards with screening for prostate cancer.
This work was supported by grants from the Swedish Medical Research Council (project number 7903); the Swedish Cancer Society (project number 3555); the Faculty of Medicine at Lund University; the Cancer Research Fund of the University Hospital, Malmö; Fundacion Frederico S. A., and Perkin-Elmer Life-Sciences (Wallac), Turku, Finland.