Younger age and married status were associated with prolonged OS
Previous studies based on the SEER database have reported that the median age of localized and locally advanced PCa patients are 61 and 65 years, respectively [
15,
16]. Our study found that PCa patients with bone metastases were significantly older than those with localized and locally advanced PCa patients, with a median age of 71 years. Meanwhile, older age (age > 70 years) was associated with a poor prognosis, consistent with previous reports [
12,
17,
18].
Studies based on the SEER database have reported that the married proportion of localized and locally advanced PCa patients are 77% and 71%, respectively [
15,
16]. In our study, the proportion of married PCa patients with bone metastases was lower than that of localized and locally advanced PCa patients (54% vs 77% and 71%). Conceivably, this is because PCa patients with bone metastases were older and more likely to be widowed than younger patients. Meanwhile, an unmarried status was also associated with a worse survival, possibly due to married patients more easily to receive social and financial support from their families and to choose a proactive treatment. As reported in previous studies, marital status is a prognostic factor for the better survival in PCa patients [
19,
20].
In addition, our study indicated that high pathological grades, high PSA levels (> 50 ng/ml), and high Gleason scores (Gleason score ≥ 8) of the tumor were associated with a worse OS. These findings are in line with those of former studies [
12,
17,
21].
Asian or Pacific Islander ethnic background was associated with favorable OS
Several studies have reported that for localized and locally advanced PCa patients, Black men consistently have a higher mortality rate than do men of other ethnicities. One possible reason for this finding is that Black men are less likely to be treated with a curative intent than are White men [
22,
23]. However, in our study, Black men presented survival times similar to those of White men. This phenomenon might be explained by the fact that in the metastatic stage, there are few proactive treatments available to significantly change the oncological outcome of these patients. Interestingly, our study demonstrated that patients with an Asian or Pacific Islander background were associated with a better OS than White men, which is consistent with previous reports showing that Asian men have superior survival in de novo metastatic PCa than do men of other races [
24]. Therefore, genomic diversities rather than treatments are most accountable for the different survival times between races [
24]. It is known that allelic imbalance at 13q14 and 13q21 is significantly higher in Japanese patients than in Caucasians [
25]. The frequency of the TMPRSS2-ERG fusion in Chinese patients is significantly lower than those patients from Western countries [
26]. Furthermore, differences in diet (e.g., soy food consumption, which is popular in Asians and associated with a 25% to 30% reduced risk of PCa) and other lifestyle factors may also affect survival [
24,
27].
Lung was the most frequent site of synchronous metastases
Our study shows that in PCa patients with bone metastases, approximately 10% of patients have synchronous distant metastases at other sites. Among these sites, the lung was the most frequently involved extra-skeletal organ, followed by the liver, whereas brain metastasis was relatively uncommon. These results are in line with those of former studies [
18,
28]. A previous autopsy study of 1589 PCa patients also demonstrated that lung was the second most frequent site of involvement, following bone [
29]. However, there is very few published researches on the mechanisms of lung metastasis in prostate cancer. Several biological processes, such as hemodynamics, bone-specific signaling interactions, and the “seed and soil” hypothesis, were attributable to the bone metastasis of PCa [
30]. Lung metastasis of breast cancer, another primary of high incidence and metastatic rate, is often explained by the “seed and soil” hypothesis [
31]. To elucidate the mechanism of lung metastasis from PCa remains a challenge, which need further study to understand its pathogenesis and prognosis.
As described in the results section, we observed significant associations between the sites of metastases and patients’ survival. In particular, patients with liver metastases showed a poorer prognosis than those with other types of metastases. It is important to understand which of the two factors, the number of metastatic sites and the specific site of metastases, has a greater impact on the prognosis of patients. For this purpose, we analyzed PCa patients with lung-only, brain-only, and liver-only metastases using Kaplan–Meier survival curves and log-rank tests and found that patients with liver metastases showed the worst prognosis (Additional file
1: Fig. S1A). Kaplan–Meier survival curves and log-rank tests showed that as long as a patient had liver metastases, no matter how many other sites of metastases were present, the survival time was expected to be similarly short (Additional file
1: Fig. S1B). However, this phenomenon was not seen in patients with lung or brain metastases, as the number of metastatic sites significantly affected prognosis (Additional file
1: Fig. S1C and D).
It is known that PCa patients with liver metastases often have a lower Karnofsky Performance Status (KPS) scores [
32]. Patients with low KPS scores cannot tolerate intensive systemic treatment. Another potential reason is that liver metastases from PCa tend to have a propensity for neuroendocrine differentiation, which is independently associated with poor survival outcomes [
30,
33]. The neuroendocrine differentiation propensity in liver metastasis patients was also observed in our cohort. In PCa patients with bone-only metastases, neuroendocrine differentiation consisted of only 0.7%, whereas in bone and liver metastases, the rate rose up to 7% of the patients (Additional file
2: Table S1). These results suggest that we need to carefully evaluate the synchronous metastases of extra-skeletal organs prior to treatment in PCa patients with initial bone metastases.
Chemotherapy improved OS in specific patient subpopulations
With regard to treatment, our study revealed that surgery, RT, and chemotherapy were uncommonly used in de novo PCa patients with bone metastases, with rates of only 9.6%, 22.6%, and 13.3%, respectively. This phenomenon was related to the fact that ADT had been the first line of treatment for metastatic PCa patients for many years [
34].The addition of chemotherapy to impact PCa survival was first examined in castration-resistant prostate cancer (CRPC) in 2004. Since then, additional studies have described the role of chemotherapy increasingly earlier in disease presentation [
35]. In recent years, several studies have supported the upfront use of chemotherapy at the initial diagnosis of metastatic prostate cancer. For example, in a large phase III trial, CHAARTED, the addition of 6-cycle docetaxel chemotherapy to ADT in men with newly diagnosed hormone-naive PCa resulted in a 14-month improvement in median survival (57.6 vs. 44 M) [
36]. And the phase III study of ADT alone or with docetaxel in non-castrate metastatic prostate cancer (GETUG-AFU 15) found that, the addition of 9-cycle docetaxel chemotherapy to ADT in radiologically confirmed metastatic PCa resulted in modest OS benefits (60 vs. 54 M) [
37]. Therefore, only in recent years has chemotherapy been recommended by the NCCN and EAU guidelines for certain metastatic PCa patients who are fit enough to receive the combination treatment at initial diagnosis [
10,
11]. In our study, similarly, when all patients were pooled for multivariate Cox analysis, chemotherapy had a positive role in reducing the risk of mortality, with an HR of 0.848.
Interestingly, landmark analysis showed that, compared with bone-only metastases patients receiving no chemotherapy, patients receiving chemotherapy had a significantly longer OS before the landmark point of 35 months. However, significantly shorter OS after 35 months was otherwise observed. The reasons may include (1) the Kaplan–Meier analysis was not adjusted for age, marriage, race, PSA, Gleason score, etc., whereas the results of the multivariate Cox regression were adjusted for the above variable and showed that chemotherapy prolonged OS; and (2) as recommended by the NCCN and EAU guidelines, patients with a low-volume metastatic disease have less certain benefit from early-delivered chemotherapy. Therefore, we might assume that the majority of patients receiving chemotherapy tend to be of high-volume disease, which may affect patients’ long-term survival. In addition, it is noted that more than 93% of chemotherapy patients deceased before 35 months, while only less than 7% of chemotherapy patients stayed alive after 35-month. Therefore, the power of Kaplan–Meier analysis may be limited.
Another interesting result was that, although chemotherapy was an independent prognostic factor in patients with only bone, bone and lung, and bone and liver metastases, it was not in patients with bone and brain, or bone metastases and ≥ 2 other sites (including some brain metastases) of metastases. These results indicate that chemotherapy only improves OS in patients without brain metastases, which might be explained by the effect of the blood–brain barrier on the efficacy of chemotherapy [
38]. Experimentally, van der Veldt et al. [
39] once reported very low uptake of docetaxel in the brain of PCa patients receiving chemotherapy.
RP and BT improved OS in patients with bone-only metastases
In our study, when all patients were pooled for multivariate Cox analysis, neither surgery nor RT had an effect in reducing the risk of mortality, with HRs of 1.006 and 0.955, respectively. Therefore, all patients were stratified according to the sites of metastases and treatment modalities for further evaluation of survival benefits. Although, in our study, in the group of PCa patients with bone-only metastases, only 206 and 94 patients underwent RP and BT, respectively. Interestingly, RP and BT both presented significant benefits in reducing the risk of mortality, with HRs of 0.339 and 0.567, respectively. However, similar to PCa patients with bone-only metastases, RP and BT were rarely used in PCa patients with bone metastases and other sites of metastases. Due to the small amount of data, multivariate Cox analysis was not possible. In 2014, Culp et al. [
12] reported that local therapy with RP and BT appeared to confer a survival benefit for metastatic PCa. Our study, therefore, provided meaningful support to the study by Culp et al., as we pointed out that the benefits of RP and BT were mainly observed in PCa patients with bone-only metastases. Due to the small sample size, further investigation is warranted to validate these findings.
EBRT to the primary tumor was strongly discouraged for metastatic PCa patients in the NCCN and EAU guidelines in the versions issued prior to 2019. EBRT was often used to palliate metastatic bone pain, spinal cord compression, and obstructive symptoms due to the primary tumor [
40,
41]. However, in the most recent version of the NCCN and EAU guidelines, EBRT combined with ADT was recommended for metastatic PCa patients who had a low metastatic burden according to the CHAARTED criteria (a high metastatic burden was defined as four or more bone metastases with one or more metastases outside of the vertebral body or pelvis, or visceral metastases, or both; all other assessable patients were considered to have a low metastatic burden) [
10,
11]. This updated recommendation was based on the STAMPEDE phase 3 randomized trial, which was reported in 2018 and suggested that prostate RT improves OS for men with metastatic PCa who have a low metastatic burden but not for all patients [
42]. Since the SEER database does not provide information on the sites of RT, we assumed that BT was RT for the primary lesion, while EBRT may have been RT for the primary or metastatic lesion. This finding may explain why the BT group showed a longer OS, while the EBRT group showed no OS advantage in de novo PCa patients with bone-only metastases in our study.
Similar to other studies that have utilized the SEER database as their data source, our study did have certain limitations. First, only four distant sites of metastases were assessed, and details on the metastases, such as the sizes and exact metastatic lesion quantity in specific organs, were lacking. Second, all information on metastases was from the initial diagnosis, and there was a lack of follow-up information. Third, although the SEER data included information regarding the use of surgery, RT, and chemotherapy, the details of these therapies (i.e., surgical margins, radiation dose, radiation to the primary tumor, chemotherapy regimen and chemotherapy sequence) were not recorded in the database. Fourth, some other treatment information that might be important for prognosis, such as ADT and abiraterone acetate, was lacking.