The outcome of prostate cancer patients treated with curative intent strongly depends on survival after metastatic progression

Background Five-year survival in patients with localized prostate cancer (PCa) is nearly 100%, but metastatic disease still remains incurable. Clinical management of metastatic patients has become increasingly complex as novel therapeutic strategies have emerged. This study aims at evaluating the impact of the first metastatic progression on the outcome of PCa patients treated with curative intent. Methods The analysis was conducted using data of 913 cases of localized PCa diagnosed between 2000 and 2014. All patients were treated with curative surgery (N = 382) or radiotherapy (N = 531) with or without adjuvant therapy. All metastases were radiologically documented. The prognostic impact of the first site of metastasis on metastasis-free survival (MFS) and PCa-specific survival (PCaSS) was investigated by univariate and multivariate analyses. Results One hundred and thirty-six (14.9%) patients developed a metastatic hormone-sensitive PCa and had a median PCaSS of 50.4 months after first metastatic progression. Bone (N = 50, 36.8%) and LN or locoregional (N = 52, 38.2%) metastases occurred more frequently with a median PCaSS of 39.7 and 137 months respectively (p < 0.0001). Seven patients developed visceral metastasis only (5.1%; liver, lung, brain) and 27 (19.9%) concurrent metastases; this last group was associated with the worst survival with a median value of only 17 months. Thus, each subgroup exhibited a survival after metastasis significantly different from each other. In multivariate analysis the site of the first metastasis was an independent prognostic factor for PCaSS along with Gleason score at diagnosis. The correlation between survival and first site of metastasis was confirmed separately for each therapy subgroup. Median metastasis-free survival from primary diagnosis to first metastasis was not correlated with the first site of metastasis. Conclusions In non-metastatic PCa patients treated with curative intent, the PCa-specific survival time depends on the time after metastatic progression rather than the time from diagnosis to metastasis. Moreover, the site of first metastasis is an independent prognostic factor for PCaSS. Our data confirm that the first metastatic event may confer a differential prognostic impact and may help in identifying patient at high risk of death supporting the treatment-decision making process following metastatic progression. Electronic supplementary material The online version of this article (10.1186/s12885-017-3617-6) contains supplementary material, which is available to authorized users.

In the last years several studies have suggested the prognostic importance of the site of metastasis in men with de novo metastatic PCa [8,11] or metastatic castration-resistant prostate cancer (mCRPC) [5,13]. To our knowledge, few studies [9,12,14,18] have assessed the impact of location of metastatic disease on the outcome of men with PCa after receiving curative treatment. Shao et al. [18] firstly demonstrated that primary treatment may make a difference with regard to survival time after metastasis; both Nini et al. [12] and Moschini et al. [14] found that nodal and local recurrence have a more favorable prognosis compared with skeletal and visceral metastases in pN+ patients treated with radical prostatectomy. Local and nodal site were the most frequent primary location of metastasis in patients treated with both radiotherapy [9] and prostatectomy [12,14].
In this work we sought to address this issue by using a cohort of non-metastatic primary PCa patients who underwent prostatectomy or radiotherapy with curative intent with the aim of evaluating the impact of the first metastatic event on PCa-specific survival in order to respond to the need to improve the treatment-decision making process following metastatic progression.

Patients
An observational analysis was performed by using a database of 1364 patients diagnosed with PCa from 2000 to 2014 at Oncology Institute of Southern Switzerland (IOSI) and Urology Unit of San Giovanni Hospital (OSG). Clinical, pathological, and demographic data were registered. Agreement was obtained from the Ethics Committee of Canton Ticino to collect and analyze data without disclosing patient identifiers. Follow-up data were collected through August 2015.

Definitions of metastasis subgroups
Patients developing a metastatic hormone-sensitive disease, defined as a tumor responding to hormone therapy, were categorized into one of the following subgroups according to the site of the first metastatic progression of the disease after curative treatment: 1) presence of exclusive bone metastasis (Bone-only subgroup); 2) presence of LN or locoregional metastasis, alone or concomitant (Locoregional/LN-only subgroup); 3) presence of exclusive visceral disease (Visceral-only subgroup) and 4) presence of multiple sites of metastasis (Multiple site subgroup). Visceral disease was defined as metastatic disease to liver, lung, brain and other organ sites. Patients with multiple metastatic sites were further stratified in one of the following categories: a) patients with bone metastasis with LN or locoregional disease and b) patients with visceral disease with bone, nodal or locoregional involvement. All metastases were radiologically documented (MRI or CT scan or Choline-PET scan or bone scintigraphy). Followup visits and imaging examinations were performed according to standard clinical practice or in case of symptomatic disease.

Outcomes
Study endpoints were: PCa-specific survival (PCaSS), defined as the time interval from the date of primary PCa diagnosis to the date of PCa related-death or last followup; metastasis-free survival (MFS), defined as the time interval from the date of primary PCa diagnosis to the first radiographic metastasis; PCa-specific survival after metastasis (PCaSS after metastasis), defined as the time interval from the date of the first radiographic metastasis to the date of PCa related-death or last follow-up.

Statistical analysis
Demographic characteristics of patients were reported using median and interquartile ranges for continuous variables and frequencies and proportions for categorical variables. The independent t test and the chi-square test were used to assess associations between continuous and categorical variables, respectively.
Estimates of medians, rate and 95% confidence intervals (CIs) were determined using the Kaplan-Meier method. Patients were censored if they were still alive or they were lost to follow-up. Differences in survival times were evaluated using the log-rank test. A multivariable Cox regression analysis was used to assess the prognostic impact of the first site of metastasis on PCaSS and MFS after adjusting for other covariates that might partially influence the outcome. All variables associated with univariate value of p ≤ 0.05 were included in the multivariate model.
All tests were considered statistically significant at p ≤ 0.05. Statistical analyses were carried out using software STATA software (StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP).

Patients
Of the 1364 men with PCa registered in our database, 913 patients with localized disease who underwent curative treatment were identified. Histologically all of them were acinar adenocarcinoma of the prostate. In particular, 382 patients underwent radical prostatectomy, without (N = 295) or with adjuvant radiation (N = 40) or hormonal (N = 25) therapy or both (N = 22), and 531 men were treated with external beam radiation treatment (N = 531), without (N = 99) or with hormonal therapy (N = 432). Median age at diagnosis of the entire cohort was 67 years (IQR 62.7-71.8). After a median follow-up of 5.7 years (IQR 2.9-8.8) from the date of primary tumor diagnosis, 60 patients (6.6%) have died.
Disease characteristics of the cohort according to the development of the metastatic disease and curative treatment are summarized in Tables 1 and 2, respectively. Patients who progressed to metastasis had a lower age at diagnosis (median value 65.3 vs 67.2 years old), higher PSA, elevated Gleason Score and more LN involvement than patients who did not progress ( Table 1). As expected, patients treated with prostatectomy were younger (median value 63.4 vs 70.3 years old) and had different disease characteristics compared with patients who underwent radiotherapy ( Table 2). But no significant statistically differences between the two treatment modalities were found according to the site of metastasis at first progression of the disease after curative treatment (p = 0.158).

PCa-specific survival
Five-year PCaSS rate were 100% and 85% (95% CI 77.5-90.2) for patients who did not develop and who developed metastasis, respectively (Additional file 2: Figure  S1). When patients were stratified according to the first site of metastasis, the Locoregional/LN-only subgroup had significantly higher 5 Figure S2).
PCa-specific survival after first metastasis  Fig. 3a). Whereas PCaSS after metastasis in the radiotherapy subgroup were 32.9 months (95% CI 18.4-65.4) and 15.07 (95% CI 6-nr) for men with bone-only metastases and multiple metastatic sites, respectively; men with Locoregional/LN metastasis did not reach median survival (p = 0.0006) (Fig. 3b). None of subgroups of metastasis showed statistically significant differences between the two curative treatments (Logrank test, p = 0.2054 for Bone-only, p = 0.6158 for Locoregional/LN-only, p = 0.6321 for Multiple site).

Multivariate analysis predicting PCa-specific survival
Multivariable Cox regression analysis showed that the first site of metastasis is an independent predictor of PCaSS (Table 3). Particularly, patients who progressed to boneonly and to multiple sites had a higher risk of dying from PCa compared with those developing Locoregional/LNonly disease with HR of 3.88 (p = 0.011) and 3.56, respectively (p = 0.019). Moreover, Gleason score was the only primary tumor parameter that represented an independent predictor of PCaSS. Patients with Gleason score 8-10 had a worst PCaSS compared with those having lower grade cancer (HR 2.45, p = 0.020).

Discussion
In this data set analysis of men receiving curative radiotherapy or surgery for primary PCa, the site of the first metastasis was an independent prognostic factor for PCa death. In the entire cohort, men initially developing LN or locoregional metastases, which were the most frequent along with bone in our series as already reported by other authors [4,9,12,14], had the best survival followed by those with osseous metastases. Men having metastases at multiple locations exhibited the worst prognosis; moreover, a shorter survival was associated with disease involving visceral sites too. These data are consistent with other authors who showed that multiple recurrences had a poorer prognosis than a single recurrence in pN+ patients treated with radical prostatectomy [14] and a worse outcome in presence of visceral involvement in newly metastatic patients [8,10]. This trend in PCaSS was confirmed when the analysis was done separately for curative surgery or radiotherapy. The Locoregional/LN-only subgroup had the best prognosis in both therapy subgroups confirming previous findings in pN+ patients treated with radical prostatectomy [12,14]. Therefore, our study highlights that PCa-specific survival strongly depends on the time after metastasis. In fact, median survival from primary diagnosis to initial metastasis (MFS) was independent of the site of the first metastatic event; indeed, each metastatic subgroup exhibited a very similar MFS with a median value of 49.6 months, very alike to that found by Ost et al. [9] in patients treated with radiotherapy. This result was confirmed also for each therapy subgroup, separately. Our findings could be explained by the expression of different biologic characteristics that underlie the spread of PCa cells to metastatic sites. Indeed, it seems that tumor cells that spread only to LNs may have acquired specific phenotypic modifications that predispose to the preferential invasion of lymphatic vessels and access to lymph nodes [10,12,20,19]. These cells might acquire the ability to spread from lymph nodes to distant organs via blood or lymphatic channels only after subsequent neoplastic transformations [3,12,20,21]. Thus, PCa cells that disseminate to nodes harbor a less aggressive phenotype compared with those spreading systemically to other sites. This may explain the more favorable prognosis of patients with local/nodal recurrence compared with those with systemic disease, as already suggested by others [12]. On the other hand, visceral disease seems to be a very adverse prognostic factor, especially in de novo metastatic PCa [8] and mCRPC patients [5,13,15,17]. However, it was reported as having a favorable outcome in the absence of extensive bone metastases [16] and in  metastatic patients at initial diagnosis [10,11,22]. Interestingly, Pouessel et al. [22] analyzing patients with localized or locally advanced disease at diagnosis found that median overall survival was 6 months for patients who had a late diagnosis of liver metastasis and 14 months for whom liver was part of the initial pattern of metastases. This finding was consistent with Wang et al. [23] that showed that the outcome of liver metastasis was worse for patients whose liver metastasis was synchronous at primary PCa diagnosis than in those for whom the liver was a site of progressive PCa. In our series men having visceral-only disease were underrepresented and it was impossible to make a definitive conclusion on the outcome of those patients; however, we found that the presence of visceral disease worsened the outcome of men having other metastases, as previously shown by others [8,10]. Of note, Gleason score was the only baseline parameter predicting PCaSS endorsing its prognostic value for disease-specific survival [24] and further reinforcing that poorly differentiated cancers tend to have a more aggressive biological behavior, including high risk of metastasis [25]. Finally, the prevalence of metastasis to specific sites, mostly to bone and nodal/locoregional area [4,9,12,14], has been explained biologically by the interaction between metastatic tumor cells and the organ microenvironment [26], favored by a preferential homing [27], or purely by the anatomy of vascular and lymphatic drainage from the site of the primary tumor [28].
Our work is certainly limited by its observational design and institutional registry-based study; thus, nonstandardized timing for imaging, changes in treatment indication and imaging over years, inherent biases in the institution and other confounders may have been affected the results. On the other hand, being a singlecenter study it has allowed us to analyze a homogeneous curative cohort of non-metastatic PCa managed by standard local protocol and to demonstrate the prognostic role of the location of the metastatic hormonesensitive disease independently in both intervention groups. Moreover, our data clarify that the time after metastatic progression rather than the time from diagnosis to metastasis strongly impacts on patients' outcome and above all that the first metastatic event is an important factor in defining the prognosis of patients treated with curative intent and it will help in the treatment-decision making process following metastatic progression. Further studies addressing this topic are imperative to find additional parameter to stratify men with PCa into prognostic groups according to their metastatic disease. Moreover, studies investigating the biologic mechanisms underlying metastatic spread to specific locations are also needed. All these data might have important implications for the development of novel therapeutic approaches targeting the specific metastatic site that will help in selecting the optimal therapy for individual patients according to the metastatic disease they will experience.

Conclusions
In PCa patients initially treated with curative intent, the PCa-specific survival strongly depends on the time after metastatic progression and the first location of metastasis. Thus, metastatic site may confer a differential prognostic impact and may be used to identify patients at the highest risk of death. Our results provide the conceptual framework for treating patients according to the metastatic disease and advance arguments to introduce location of metastasis as a research parameter in PCa studies.