Treatment with curative intent and survival in men with high-risk prostate cancer. A population-based study of 11 380 men with serum PSA level 20–100 ng/mL


Correspondence: Sam Ladjevardi, Department of Urology, University Hospital, 751 85 Uppsala, Sweden.



What's known on the subject? and What does the study add?

  • There are two randomized controlled trials showing that radiotherapy can be beneficial for men with locally advanced prostate cancer. The present study confirms the importance of curative treatment for men with high-risk prostate cancer.


  • To investigate the influence of curative treatment on cause-specific mortality in men diagnosed with prostate cancer (PCa) with serum prostate-specific antigen (PSA) levels between 20 and 100 ng/mL.

Materials and Methods

  • Patients with PCa (T1–4, N0/N1/NX, M0/MX), PSA 20–100 ng/mL and age ≤75 years were identified in the National Prostate Cancer Register of Sweden.
  • Data on co-morbidity diagnoses were obtained from the National Patient Register and cause of death from the Cause of Death Register.
  • Following adjustment for age at diagnosis, co-morbidity burden, Gleason score, T-category, PSA level and cause-specific mortality in relation to treatment were estimated using Cox regression analysis.


  • A total of 11 380 men were diagnosed with PCa between 1996 and 2008 and fulfilled the inclusion criteria.
  • The cumulative 10-year PCa-specific mortality was 36% for patients receiving only palliative treatment and 13% for those treated with curative intent.
  • For the 8462 (74%) patients with PSA levels from 20 to 50 ng/mL at diagnosis, the hazard ratio for death from PCa was 0.23 (95% confidence interval 0.19–0.27) for those treated with curative intent compared with those given palliative treatment after adjusting for age, co-morbidity, T category, PSA level and Gleason score. The corresponding hazard ratio was 0.22 (95% confidence interval 0.17–0.30) for patients with PSA levels from 51 to 100 ng/mL.


  • Treatment with curative intent for men with high-risk PCa was associated with reduced cause-specific mortality and should be considered even when serum PSA exceeds 20 ng/mL.

prostate cancer


National Prostate Cancer Register




radical prostatectomy


androgen deprivation therapy


The widespread use of PSA testing has led to a significant migration in stage and grade of prostate cancer (PCa), with >90% of men in the current era diagnosed with clinically localized disease [1]. Despite the trends towards lower-risk PCa, 20–35% of patients with newly diagnosed PCa are still classified as high risk, based on either PSA >20 ng/mL, Gleason score 8–10, or an advanced clinical stage [2]. Nevertheless, despite the high incidence of high-risk tumours, the mortality rate associated with PCa has declined in Europe since the 1990s [3]. This trend has also been seen in a Swedish study based on the National Prostate Cancer Register (NPCR) of Sweden [4].

The treatment of men with high-risk PCa remains controversial because of the lack of conclusive well-controlled or randomized studies comparing outcomes with palliative treatment, radiotherapy (RT), and radical prostatectomy (RP). Whereas the Swedish Prostate Cancer Group 4 trial has provided firm support for treatment with curative intent for men with localized PCa [5], the benefit from RT or RP is more uncertain for men with tumours that could be expected to have spread beyond the prostate capsule. There are two randomized controlled trials on men with high-risk PCa that have shown longer overall survival rates for those treated with RT in combination with hormonal treatment, than the ones on hormonal treatment as sole therapy [6, 7].

The outcome after surgery for men with PSA level >20 ng/mL is crucial in the treatment decision for men with high-risk tumours. A threshold PSA measurement of 20 ng/mL is used in prognostic grouping of PCa in the latest edition of the TNM-classification [8]. Men with PSA level >20 ng/mL but without signs of distant metastases, judged by bone scan, may, on the one hand, have too extensive local growth to guarantee beneficial outcome from RT or RP. On the other hand, considering the inevitably poor prognosis of these men if left untreated, denying them at least a small chance of cure is unethical. Besides the randomized controlled trials mentioned above [6, 7] there are no other studies that compare overall or cancer-specific survival in relation to treatment in patients with high-risk PCa. Moreover randomized studies are time consuming and cost consuming, and there is normally a significant delay before initiation of the study and outcome of the results. Because of this, the best evidence in this clinically important group is provided by large population-based studies.

Even if androgen deprivation therapy (ADT) before and during RT and RP may give a longer cancer-free and overall survival than RP and RT as mono-therapy, the crucial question is whether or not treatment with curative intent gives better survival than ADT as mono-therapy. Even the risk for recurrence after RT and RP in men with high PSA levels is high, and if overall survival is prolonged substantially, any harm from treatment is acceptable. The purpose of the present study was to investigate treatment modalities in relation to cause-specific mortality in a large unselected population-based cohort of men with PCa and serum PSA levels between 20 and 100 ng/mL.

Materials and Methods

The present study is based on the NPCR of Sweden [9] with additional information retrieved by record-linkage to the National Cancer Register, and the Cause of Death Register, using the unique 10-digit personal registration number assigned to every resident in Sweden [10].

The NPCR of Sweden was established in 1996 to provide a base for quality assurance of PCa care in Sweden. It covers 98% of all PCa cases registered in the National Cancer Register since 1998 [11].

By 31 December 2008, the NPCR had assembled data on more than 90 000 incident PCa cases, with information on mode of detection, tumour stage, tumour differentiation, serum PSA levels and treatment initiated within 6 months of diagnosis [9]. The National Patient Register includes information on hospital admissions and discharges from all national health-care hospitals in Sweden. Each inpatient record contains dates of hospital admissions and up to eight discharge diagnoses, coded according to the International Classification of Diseases. Information on date and the underlying cause of death, coded according to the International Classification of Diseases, was obtained from the Cause of Death Register. The number of non-reported events in the Cause of Death Register was estimated to be 0.7% of all deaths in 2006 [12].

The Charlson Co-morbidity Index was used to assess the burden of concomitant disease for each PCa patient [13] and has been shown in a study based on NPCR to have an impact on management and survival [14].

The number of men who were diagnosed with PCa was 133 per 100 000 in 1996 and 193 per 100 000 in 2008 [15].

A total of 98 924 men diagnosed with PCa were identified in the NPCR of Sweden. The inclusion criteria in the present study were; age 75 years or less at date of diagnosis, any primary tumour category (T1–T4), any lymph node category (NX/N0/N1), without evidence of distant metastases (MX/M0) and serum PSA level between 20 and 100 ng/mL, date of diagnosis between 1 January 1996 and 31 December 2008.

In time-to-event analysis, the outcome of interest was PCa-specific mortality (C61 in International Classification of Diseases, version 10). Survival time was defined as the interval between the date of PCa diagnosis and the date of outcome, emigration or end of follow-up 31 December 2009. To see if the treatment groups differed in PCa-specific survival, we estimated the cumulative probability of PCa death using Kaplan–Meier estimates. Cox regression with relative risks expressed as hazard ratios (HR) and 95% CI were calculated to assess PCa-specific mortality in relation to treatment, adjusting for age at diagnosis, T category, serum PSA, co-morbidity and Gleason Score. All analyses were performed with the R statistical software package (R Foundation for Statistical Computing, Vienna, Austria).

The study was approved by the Central Research Ethics Committee and the Research Ethics Committee at Uppsala University.


The cohort assembly is shown in Fig. 1. The final study population in the present study consisted of 11 380 men. Of these, 8462 (74%) men were diagnosed with a serum level of PSA 20–50 ng/mL and 2918 (26%) had PSA level between 51 and 100 ng/mL (Table 1). Among these patients, 10.6% were younger than 60 years and most (74%) of the patients had palpable tumours. Of all patients, 44% were diagnosed after local symptoms, whereas 17.3% of patients were diagnosed in an asymptomatic stage after PSA testing.

Figure 1.

Flow chart of cohort assembly recorded in the National Prostate Cancer Register of Sweden between 1996 and 2008.

Table 1. Clinical characteristics of patients diagnosed with primary prostate cancer between 1996 and 2008, by level of prostatic-specific antigen (PSA) and planned initial treatment
 PSA 20–50 ng/mLPSA 51–100 ng/mLPSA 20–100 ng/mL
n (%)n (%)n (%)n (%)n (%)n (%)
  1. CCI, Charlson Co-morbidity Index.
All3387 (100.0)5075 (100.0)517 (100.0)2401 (100.0)3904 (100.0)7476 (100.0)
Calendar period      
1996–2000875 (25.8)2116 (41.7)162 (31.3)981 (40.9)1037 (26.6)3097 (41.4)
2001–2003864 (25.5)1313 (25.9)115 (22.2)623 (25.9)979 (25.1)1936 (25.9)
2004–20081648 (48.7)1646 (32.4)240 (46.4)797 (33.2)1888 (48.4)2443 (32.7)
Age of diagnosis      
<60676 (20.0)241 (4.7)124 (24.0)168 (7.0)800 (20.5)409 (5.5)
60–692021 (59.7)1678 (33.1)298 (57.6)899 (37.4)2319 (59.4)2577 (34.5)
70–75690 (20.4)3156 (62.2)95 (18.4)1334 (55.6)785 (20.1)4490 (60.1)
Mode of detection*      
Screening956 (35.0)638 (18.7)110 (28.8)261 (16.0)1066 (34.2)899 (17.8)
Symptoms1400 (51.2)2241 (65.9)227 (59.4)1159 (70.9)1627 (52.2)3400 (67.5)
Other reason206 (7.5)285 (8.4)15 (3.9)95 (5.8)221 (7.1)380 (7.5)
Missing173 (6.3)239 (7.0)30 (7.9)120 (7.3)203 (6.5)359 (7.1)
Clinical T-stage      
T07 (0.2)39 (0.8)1 (0.2)13 (0.5)8 (0.2)52 (0.7)
T1ab75 (2.2)212 (4.2)11 (2.1)47 (2.0)86 (2.2)259 (3.5)
T1c1035 (30.6)1039 (20.5)117 (22.6)273 (11.4)1152 (29.5)1312 (17.5)
T21376 (40.6)1768 (34.8)176 (34.0)698 (29.1)1552 (39.8)2466 (33.0)
T3865 (25.5)1780 (35.1)209 (40.4)1166 (48.6)1074 (27.5)2946 (39.4)
T412 (0.4)172 (3.4)2 (0.4)171 (7.1)14 (0.4)343 (4.6)
TX17 (0.5)65 (1.3)1 (0.2)33 (1.4)18 (0.5)98 (1.3)
N02220 (65.5)200 (3.9)349 (67.5)63 (2.6)2569 (65.8)263 (3.5)
N188 (2.6)463 (9.1)20 (3.9)146 (6.1)108 (2.8)609 (8.1)
NX/Missing1079 (31.9)4412 (86.9)148 (28.6)2192 (91.3)1227 (31.4)6604 (88.3)
M02839 (84.3)2671 (53.1)439 (85.1)1351 (56.8)3278 (84.4)4022 (54.3)
MX/Missing529 (15.7)2360 (46.9)77 (14.9)1026 (43.2)606 (15.6)3386 (45.7)
Gleason Score      
GS 2–61277 (37.7)1708 (33.7)136 (26.3)552 (23.0)1413 (36.2)2260 (30.2)
GS 71503 (44.4)1995 (39.3)257 (49.7)1045 (43.5)1760 (45.1)3040 (40.7)
GS 8–10580 (17.1)1215 (23.9)119 (23.0)751 (31.3)699 (17.9)1966 (26.3)
Level of PSA, mean (sd)29.0 (8.1)31.5 (8.7)65.8 (12.6)71.9 (14.6)33.9 (15.3)44.5 (21.8)
Comorbidity Index      
CCI 02905 (85.8)3703 (73.0)466 (90.1)1760 (73.3)3371 (86.3)5463 (73.1)
CCI 1340 (10.0)803 (15.8)36 (7.0)397 (16.5)376 (9.6)1200 (16.1)
CCI 2+142 (4.2)569 (11.2)15 (2.9)244 (10.2)157 (4.0)813 (10.9)

In 24.9% of the patients the disease was categorized as N0, in 69% as NX and in 6.3% as N1. According to clinical practice in Sweden, classification of N-category was almost exclusively based on a limited pelvic lymphadenectomy. Of men who were included in this study, 64% had their cancer classified as M0. More than 90% of all patients had a Charlson Co-morbidity Index of 0 or 1 at the time of diagnosis, and only 8.5% of men had a Charlson Co-morbidity Index of 2 or more. A total of 32% of patients presented with Gleason score 6 or less, 42.2% with Gleason score 7 and 25.8% with Gleason score 8–10.

As shown in Table 1 almost two-thirds of all patients in the study only received palliative treatment. Treatment distribution is shown in Table 2. Table 3 shows a multivariable analysis of the study cohort. It shows a significant difference in cancer-specific survival between ‘curative’ and ‘non-curative’ treatments after adjusting for age, co-morbidity, Gleason score, T category and PSA level at the time of diagnosis. Age and co-morbidity were not, as expected, associated with cancer-specific survival. Gleason score, PSA and T category, however, had significant impact. The 10-year cumulative probability of PCa death is shown in Fig. 2.

Figure 2.

Prostate cancer-specific mortality.

Table 2. Distribution of treatment in the group of men receiving treatment with curative intent
 PSA 20–50 ng/mLPSA 51–100 ng/mLTotal
n (%)n (%)n (%)
  1. PSA, prostate-specific antigen.
Radical prostatectomy (RP)1028 (30.4)71 (13.7)1099 (28.2)
RP + radiotherapy (RT)61 (1.8)7 (1.4)68 (1.7)
RT brachy332 (9.8)46 (8.9)378 (9.7)
RT external1384 (40.9)297 (57.4)1681 (43.1)
RT brachy and external509 (15.0)86 (16.6)595 (15.2)
Other73 (2.2)10 (1.9)83 (2.1)
Total3387 (100.0)517 (100.0)3904 (100.0)
Table 3. Cause-specific survival estimated by multivariate proportional hazard models with hazard ratios (HR) and 95% confidence intervals (CI)
 PSA 20–50 ng/mLPSA 51–100 ng/mLPSA 20–100 ng/mL
HR (CI 95%)HR (CI 95%)HR (CI 95%)
  1. CCI, Charlson Co-morbidity Index; PSA, prostate-specific antigen.
  2. Subgroup 1. Level of PSA 20–50, M0/MX.
  3. Subgroup 2. Level of PSA 51–100, M0/MX.
  4. Subgroup 3. Level of PSA 20–100, M0/MX.
Initially planned treatment   
Conservative1.00 ref.1.00 ref.1.00 ref.
Curative0.28 (0.24–0.33)0.30 (0.22–0.41)0.29 (0.20–0.26)
Age at diagnosis (years)   
<601.00 ref.1.00 ref.1.00 ref.
60–690.84 (0.68–1.05)0.83 (0.64–1.08)0.84 (0.71–0.99)
70–750.76 (0.62–0.95)0.87 (0.67–1.13)0.81 (0.69–0.96)
Log PSA (ng/mL)1.14 (0.92–1.41)1.68 (1.17–2.41)1.45 (1.32–1.61)
Charlson Co-morbidity Index   
CCI 01.00 ref.1.00 ref.1.00 ref.
CCI 10.91 (0.77–1.09)0.89 (0.72–1.11)0.94 (0.84–1.05)
CCI 2+1.17 (0.96–1.43)1.20 (0.93–1.56)1.19 (1.04–1.36)
Gleason Score   
GS 2–61.00 ref.1.00 ref.1.00 ref.
GS 71.72 (1.46–2.03)1.46 (1.18–1.81)1.63 (1.43–1.85)
GS 8–103.59 (3.04–4.25)2.66 (2.15–3.31)3.21 (2.82–3.67)
Clinical T-stage   
T1c1.00 ref.1.00 ref.1.00 ref.
T1ab1.71 (1.21–2.42)1.93 (1.12–3.32)1.74 (1.31–2.33)
T21.60 (1.29–1.98)1.31 (0.96–1.79)1.52 (1.27–1.82)
T3–T42.57 (2.09–3.16)1.84 (1.37–2.49)2.32 (1.96–2.75)

We did not find any differences in cancer-specific mortality between N0 and NX patients. Cancer-specific mortality was, however, significantly higher for patients with MX compared with M0.

Primary treatment is shown in Table 1. The 10-year cause-specific mortality for patients with PSA level of 20–50 ng/mL was 36% for patients treated without and 13% for patients treated with curative intent. For patients with PSA level of 50–100 ng/mL, the 10-year cause-specific mortality was 55% for palliative treatment and 20% for patients treated with curative intent. The same survival benefit from treatment with curative intent rather than palliative treatment was seen for men with PSA levels of >50 ng/mL and for those with levels <50 ng/mL.


Our study indicates that treatment with curative intent is associated with better cancer-specific survival even for patients with PSA levels of 20–100 ng/mL (Fig. 3). The cumulative 10-year PCa-specific mortality was 36% for patients treated without, and 13% for those treated with curative intent. Androgen deprivation therapy is often the first line of treatment for men with PSA levels >20 ng/mL, but according to these results ADT alone is not sufficient for these men. Either RP or RT in combination with ADT should therefore be considered for men without evidence of distant metastases, even with PSA levels close to 100 ng/mL.

Figure 3.

Hazard ratio for death from prostate cancer for men undergoing treatment with curative intent, with adjustment for age at diagnosis, T category, serum prostate-specific antigen (PSA) level, co-morbidity and Gleason Score, and men receiving only palliative treatment as reference group. The shaded field indicates 95% confidence interval.

Although there are no universally accepted criteria for defining risk categories in PCa, one of the most stringent and reproducible criteria for high-risk PCa is PSA level [16]. Even if PSA has poorer sensitivity as well as specificity for identifying locally advanced tumours, men with PSA levels >20 ng/mL undoubtedly constitute a high-risk group [8].

In a recently published study based on the PCBaSe of Sweden, men with locally advanced PCa (T3 tumours and/or PSA level of 20–100 ng/mL) receiving palliative treatment had high PCa-specific mortality [17], suggesting under-treatment of this group. The PCBaSe study, although having different inclusion criteria and no comparison of treatment strategies, supports the findings of the current study. These studies indicate that we should not withhold curative treatment from men with high-risk PCa for fear of treatment failure. The gain in survival after radical treatment of men with high-risk PCa may be even more pronounced than that for men with localized tumours.

Inman et al. [18] studied long-term outcomes of RP in men with a preoperative PSA level ≥50 ng/mL and found that 25% of patients with serum PSA values ≥50 ng/mL had Gleason score 8–10 tumours. Furthermore, nearly half of the patients had extraprostatic extension of their cancer, and a third had pelvic lymph node involvement. The cancer-specific survival for patients with PSA levels between 50 and 99 ng/mL was 94%, 79% and 72% at 5, 10 and 15 years, respectively, i.e. survival rates similar to those in the present study.

Androgen deprivation therapy alone has been considered the treatment of choice for men with high-risk PCa [18], though ultimately most tumours become hormone refractory and require second-line or third-line treatments. However, this treatment alternative has been questioned in recent years because of controversy over the outcome of men with locally advanced disease. Furthermore, the morbidity as well as mortality associated with endocrine treatment should not be neglected. A recent study based on data from the NPCR of Sweden showed a twofold increased risk for thromboembolic diseases in men on endocrine therapy [19]. Furthermore, increased relative risks for non-fatal and fatal cardiovascular disease have been reported, especially in patients treated with endocrine therapy [20].

In a literature review it was concluded that patients with lymph node metastases benefit in terms of cancer-specific survival when treated with curative intent [21]. This is in accordance with the present study, showing that patients with cancer-positive lymph nodes have longer cancer-specific survival when treated with curative intent.

Whereas there are numerous studies on men with localized tumour [5, 9, 22], the outcome after treatment of men with PSA levels between 20 and 100 ng/mL and locally advanced tumour at diagnosis is not so clear, despite the fact that this group is almost as large and has a much poorer prognosis. In a study by Han etal. [23], men with palpable tumour (T2 to T3) and PSA level >15 ng/mL show overall PSA-free survival rates at 5 years and 10 years of 56% and 40%, respectively. However, whereas early PSA relapse cannot be ruled out in the cohort of the present study, the overall, relative and cancer-specific survival benefit from treatment with curative intent strongly argues against an attitude of defeatism regarding this group, even if attempted radical cure is often followed by tumour recurrence [24].

In another study of patients with cancer-positive lymph nodes, overall and relative survival rates were higher with RP than when regional metastases were confirmed and RP was abandoned [25]. These studies suggest that treatment with curative intent is beneficial even in patients with advanced local growth and regional metastases. In the Scandinavia Prostate Cancer Group 7 trial, cancer-specific mortality and cumulative overall mortality were compared between men treated with ADT only and those with combined RT [6]. In another recently published study, overall survival for men treated with ADT and ADT in combination with RT was compared [7]. These studies also showed that treatment with curative intent can be optimized and should not be withheld from men with locally advanced growth as long as distant metastases are not present.

The crucial question is whether it is beneficial to begin curative treatment even if the PSA level raises suspicion of tumour growth beyond the boundaries of the prostate. The outcome of the present study indicates that survival may be prolonged despite the potential risk of a cancer-positive resection margin, early PSA relapse and other surrogate measures of treatment failure.

With a population of 11 380 men, the current study is to our knowledge the second largest study analysing cancer-specific mortality for men with high-risk PCa. The risk of selection bias when assessing the outcome after different treatments is, however, a well-known fact and should not be neglected. Even when adjusting for age, co-morbidity and stage of disease, PSA level and Gleason score, there may be residual bias that cannot be accounted for. The diagnoses retrieved from the NPR gave information on hospital care diagnoses only. Co-morbidity is therefore, underestimated in our study, especially for patients with minor co-morbidity, i.e. Charlson score 1. However, data regarding treatment in the NPCR are based on the first 6 months after diagnosis. This may have affected the outcome of the study because delayed treatment is not included and variations in the practice of adjuvant and salvage therapies may confound the study.

In studies on disease-specific survival, the uncertainties regarding cause of death are often a source of bias. The reliability of death certificates for patients with PCa included in the NPCR has, on the contrary, been shown be high [26]. The Charlson Co-morbidity Index has also been validated as a good indicator of disease burden in patients with PCa [9]. However, it cannot be excluded that the validity of cause-of-death information is lower in older men, in particular those with a high co-morbidity burden. Furthermore, lymph node staging and screening of distant metastases was only performed in a few members of the group. We did not find any differences in cancer-specific mortality between N0 and NX patients. Cancer-specific mortality was, however, significantly higher for patients with MX compared with M0. This may be a result of the presence of non-detected metastases in the MX group. This supports the practice of routine screening for distant metastasis in all men with PSA levels >20 ng/mL that are under consideration for treatment with curative intent.

In conclusion, this study shows that treatment with curative intent is beneficial in the group of men with PCa, absence of distant metastases and PSA levels between 20 and 100 ng/mL. Many patients with such tumours were under-treated in Sweden during the study period.

Conflict of Interest

None declared.