Prostate cancer (PC) is the most common male malignancy and the second most common cause of cancer mortality in Norway [1]. It is presently the most prevalent cancer and there has been an increasing incidence until 2007 where after a reduction was noted [1]. When PSA was widely introduced in Norway in the early 1990s, an increasing number of men were diagnosed and treated with curative intent (33% in 2001) [2]. Although declining, Norway has the highest mortality rate of all western countries [3, 4]. There has been an increasing use of radical prostatectomy (RP), but a majority of patients diagnosed with PC will not have symptomatic disease or die of the disease as they have non-lethal P [5]. The reduced risk of prostate cancer specific mortality after a radical prostatectomy compared to watchful waiting has been estimated to range between 0-25% depending on tumor characteristics [6]. The only RCT evidence for a reduction in prostate cancer mortality is the SPCG-4 trial [7].

Disease-specific survival after RP has mainly been reported from a number of single institutions, but a few studies have reported on larger cohorts and three studies have described survival in nationwide cohorts in Europe [8–10]. Some of these larger studies lack sufficient follow-up, relevant prognostic parameters and clinically relevant end-points in the analysis.

Clinicopathological variables for predicting disease outcome after RP are numerous with Gleason grade and score, pTNM-stage, preoperative PSA and surgical margins as the most widely adopted ones [11, 12]. User-friendly predictive tools like look-up tables, risk classifications and nomograms have been developed [11, 13]. Some variables like lymphovascular invasion [14], tumor volume [15], pT2 subclassification [16] and tertiary Gleason grade [17] are conflicting or have insufficient supporting data yet.

For patients with biochemical failure (BF) there are several treatment options including continuous, intermittent or deferred androgen deprivation, salvage irradiation of the prostate bed and trial participation. To choose the optimal therapy for these patients it is crucial to understand the risk factors for a subsequent clinical failure or death of PC, especially since management strategy remains controversial [18–20]. After BF, PSA kinetics, or more specifically PSA-Doubling Time (PSA-DT), has emerged as a prognostic variable [21–25]. However, as a pretreatment marker, PSA-DT has not found its position [26, 27].

The objective of the present study was to describe disease outcome data for patients operated in the PSA era in three urological centers in two major health regions in Norway, and to examine the impact of post-prostatectomy PSA-DT on clinical outcomes.

This is the first large Scandinavian multicenter study presenting the impact of prognostic variable information regarding BF, CF and PCD in Scandinavia in the PSA era. We found post RP PSA-DT to be a strong predictor of CF and PCD, even outperforming Gleason score. In addition we found PNI to be an independently strong predictor of all event-free survivals. Otherwise we found the prognostic factors in this material to be mostly consistent to previously published larger series [34–36].

The strength of this paper is the unselected study population from Central and Northern Norway. It is reasonable to assess that more than 95% of men diagnosed with PC in this geographical area, were operated in hospitals that participated in this study. Moreover, the study has a relatively long FU, only patients in the PSA era were included, all tissues have been re-evaluated by two experienced pathologists, few patients had missing PSA-DT information, we have included relevant prognostic factors, and adjuvant treatment after RP was rare in the timeframe of this study. Weaknesses of the study are the retrospective design, the probable impact of salvage radiotherapy on the risk of time to events and events are low at longer follow-up times, contributing to low precision and large CIs (Tables 2 and 3). In addition, a number of tissue blocks were missing from one center, thereby reducing representativity.

Like other studies, we have found that the time from RP to CF and death of PC to be extensive, even in patients with a BF [21, 37]. Prognostic factors to stratify patients for risk-adapted follow-up, treatment regimens or clinical trials are crucial since the majority of operated patients will not have BF. Furthermore an even greater majority will not experience symptoms of their disease and only a very few will die of PC. On the other hand, it will be important to identify those patients who otherwise will have symptoms from the recurrent disease or die. Another interesting observation is that the involved surgeons seem very capable to select patients for surgery with a long expected survival as only 8% patients died of other causes during follow up.

Our observation of post-prostatectomy PSA-DT as the strongest predictor of CF and PCD in patients with BF is consistent with numerous other studies recognizing the importance of PSA-DT as a predictor of CF [21, 22, 38] and PCD [37, 39, 40] after RP. We found the same pattern in both health regions, thereby internally validating these results. The importance of PSA-DT in patients treated with radiotherapy has also been reported [41]. In accordance with Antonarakis et al. we found patients in the two lowest PSA-DT categories (<3 months and 3–8,9 months) to have the worst prognosis with comparable Hazards ratios in the multivariable analysis. 64% of CF events and 80% of PCD deaths were in these two groups even though they collectively constituted only around 1/3 of the patients (35%). Although there was a correlation between Gleason Score and PSA-DT we saw the same trends of poor event-free survival when stratified for Gleason Score subgroups, but numbers were statistically insignificant due to the low number of patients in each subgroup. Our observation adds weight and validates the importance of PSA-DT for selecting BF-patients at high risk of developing clinically significant disease in the future.

At time of diagnosis, Gleason score has been shown to be a strong predictor of high risk PC [42], but also metastasis and PCD after RP [38, 43]. Including all patients in the analyses we consistently found patients with a Gleason sum ≥9 to have the highest risk of BF (equal to pT3b), CF and PCD. This highlights a major impact of Gleason score in risk stratification following RP. Some have, however, suggested that Gleason score loses its value after including PSA-DT in the risk-stratification [44]. Analyzing patients with BF only, the PSA-DT removed Gleason score from the step-wise multivariate analyses due to its co-variation and prognostic strength. Antonarakis et al. found both Gleason score and PSA-DT to contribute to estimate metastasis free-survival [21] although PSA-DT was the strongest predictor.

Herein, extraprostatic PNI was in addition to Gleason score the only clinicopathological variable to predict both CF and PCD. Patients with an observed extraprostatic PNI had estimated HRs of CF and PCD at around 3. The ability of PNI to predict subsequent events has been controversial. Some have found PNI to independently predict BF [45–48] while others did not found PNI to be a predictive factor for any event [49–53]. D’Amico et al. reported PNI to be an independent predictor of BF, but only for low risk PC after evaluating biopsies [54]. Besides, the evaluation of its role as a prognostic factor has been hampered as it does not seem to be included as a histopathological variable in the large series of the world e.g. the Johns Hopkins database [55]. Most studies have only addressed the correlation between PNI and BF and not the more clinical relevant endpoints of CF and PCD. In the small, but interesting study by Aumayr et al. reported that a high amount of extraprostatic nerve infiltration correlated with tumor progression [45]. PNI found in preoperative biopsies, has also been found to be a predictor of metastasis and PCD in patients treated with dose-escalated radiotherapy [56]. Our finding of extraprostatic PNI as independently significant for prognosis with respect to CF and PCD, but not for BF, is in accordance with these findings.

PSM rates in our material are high (overall 53%) and among the highest that have seen published. In a large published single-center series from Mayo by Boorjian et al. in an almost identical period of time they found a PSM rate of 31.1% with an decreasing incidence over time [57]. The explanation could be the high incidence of pT3x cancers (30%) compared to 12% in the Mayo cohort and higher Gleason grades in our material, as these are independent predictors of PSM. In addition, as our centers during this timeframe were low-volume centers by international standards, this may have contributed to the high PSM rates [58]. Refinement of surgical technique and stage migration has been documented to improve the histopatological outcomes [59, 60]. Margin location was relevant in our cohort, as non-apical PSM were associated with a poor BFFS and CFFS. Margin localization was not analyzed by Boorjian et al., but in a study from the same institution by Blute et al., they found PSM at the prostate base to be independently associated to outcome, as did Obek et al. [61, 62]. A study by Godoy et al. found the base and anterior localization of PSM to be independently associated to an increased risk of BF compared to the other margin localizations. They specifically advocate that there is over-reporting of PSM from the apex and that an observation strategy is to be adopted for the large group of patients with a apical localization of PSM [63]. A Danish study also found non-apical PSM to be independently associated to BF, whereas apical margins were insignificantly associated to BF in multivariate analysis [64].

In conclusion, for the minority of patients with a subsequent BF we found a low PSA-DT to be the strongest prognosticator for CF and PCD, recognizing its superiority in risk-stratification in this subgroup. We also re-launch PNI in the pathological specimen as a possible strong predictor of CF and PCD following RP and a thorough evaluation in larger patient series is warranted.

As most patients, even after risk-stratification, will not experience a clinically significant relapse of the disease we need new prognostic markers to identify the relevant subgroups. We have included tissues from tumor and stroma of these thoroughly described and largely unselected patients in tissue micro array blocks. Hence, this forms an excellent platform for future molecular studies which will hopefully give us some of these answers.