Using a large cohort of 14185 patients from the SEER database, we successfully used 3 independent preoperative prognostic factors to established a nomogram for predicting 5-year CSM for patients with locally advanced PCa. To our knowledge, this is the first preoperative predictive nomogram and nomogram-based risk group stratification built for patients with locally advanced PCa. At the same time, the patient samples are the latest relatively (2010-2016), which is conducive to reducing the impact of defects in pathology, surgery and laboratory testing techniques on the results. This nomogram and nomogram-based risk group stratification is not inferior or even better than the current AJCC staging system in terms of discrimination, and is more quantitative and intuitive, which is convenient for clinicians to use.
Nomograms use a variety of biological and clinical variables to graphically depict the probability that a clinical event may occur for each individual . One of their unique abilities is to estimate individualized risk based on patient and disease characteristics in a graphical and user-friendly form. Nomograms can include many continuous variables and key factors of disease into the prognosis, and taking weights of each variable into account, to make predictive models more practical [14, 15]. Compared with the current prognostic prediction system like AJCC staging system, the nomogram showed advantages in many studies on different cancers [16, 17]. They can also establish individualized risk stratification and help clinicians identify suitable patients for optimal managements, and good results had been obtained in many studies of different tumor types. [18-20].
In our study, the statistical analysis highlighted 3 preoperative prognostic factors including clinical N stage, GS, and % positive core. Among them, GS had the greatest contribution to 5-year cancer-specific death (GS shared the most proportion of the nomogram total risk score.). GS is one of the most universal prognostic factors for PCa. International Society of Urological Pathology (ISUP) proposed in 2014, GS can be divided into five groups (2-6, 7(3+4), 7(4+3), 8, ≥9) according to the differences of prognosis . In our study, the classification of GS also referred to the ISUP grading system, and the risk trend shown by the nomogram was consistent with this grading system. Some previous studies had shown the relationship between the prognosis of PCa and GS [6, 22, 23]. These studies had confirmed the increased possibility of adverse clinical events including death and biochemical recurrence in PCa patients with increasing GS. But these studies did not show what the weight of GS compared to other variables on affecting patients was. % positive core showed the obvious impact on CSS for patients in the current study, only inferior to GS. % positive core refers to the proportion of positive needles in the needle biopsy to the total number of needles. Many studies had revealed the prognostic value of % positive core for PCa [6, 22, 24, 25]. In our nomogram, the risk score of group 75%-100% increased significantly from 7 (group 50%-75%) to 30, suggesting that 75%-100% positive core can more significantly affect the patient's prognosis. This finding was similar to the results of a study of 195 patients with high-risk PCa who undergoing RP. In that study, 70% positive was shown to have significant prognostic discrimination.  From the nomogram, regional lymph node metastasis (N1) was another important independent prognostic factor for CSS of locally advanced PCa patients. Previous studies had demonstrated that lymph node metastasis in most patients with PCa was associated with progressive disease, which was important for the prognosis of patients [26, 27]. A retrospective study of 229 patients indicated that the 5-year disease-free survival of rate of patients with regional lymph node metastasis could be decreased from 85% to approximately 50% .
Relatively, the 8th AJCC staging system is currently a general PCa risk stratification system, not aim at specific disease stages and treatment therapies . We compared our nomogram risk group stratification with AJCC staging system, and the results showed that our stratification had advantages in C-index (0.779 versus 0.763 for training cohort, and 0.773 versus 0.745 for validation cohort). At the same time, the nomogram risk stratification's CIF curve showed a better separation trend in both the training cohort and the validation cohort. There may be several reasons. First, our nomogram risk stratification considered the effect of % positive core on CSS of locally advanced PCa patients after RP, compared to AJCC staging system. AJCC staging system only took GS and clinical N stage into account, but this study and other studies mentioned above confirmed the importance of % positive core. Second, AJCC staging system was not quantitative, and the difference between the effects of GS and clinical N stage on prognosis was not reflected. In our nomogram risk stratification, we used different risk scores to reflect the weight of different prognostic factors, showing the obvious importance of GS among the three prognostic factors. At the same time, such quantitative and graphical tool is also easy for clinicians to use as manuals. In addition, the patients included in our study were more targeted than those in AJCC staging system after filtered by locally advanced PCa and RP. By classifying the prognosis of each patient after surgery, clinicians could better choose whether to perform surgery for each patient.
The main significance of the risk stratification proposed by us is to reveal the potential value of the three prognostic factors for the selection of locally advanced patients to receive RP and the influence weight of each factor on the prognosis of these patients. According to the latest guidelines, radical treatments including RP and RT are still the first-line treatments for locally advanced PCa. Only those patients who are unwilling or unable to receive any form of local treatment, or whose PSA growth is fast, or whose tumor differentiation is poor, are offered androgen deprivation therapy (ADT) monotherapy. The guidelines suggest that RP should be used in highly selective locally advanced PCa patients.  But at present, there is no high-level evidence on the appropriate selection criteria of RP or RT and the survival benefits comparison between RP and RT for locally advanced PCa patients. In our study, the risk stratification proposed by the nomogram provided a reference for the selection criteria of RP. Considering the difference of postoperative CSS, patients in high-risk group may be not suitable for RP, while patients in low-risk group may be the appropriate population for RP. For those who are not suitable for RP, the RT + ADT therapy may bring more survival benefits. In fact, previous studies have shown that RT + ADT therapy can bring significant benefits to locally advanced PCa patients with pN+ [31, 32]. However, they didn’t compare RP and RT directly. More clinical studies are needed to explore the most reasonable criteria for the selection of locally advanced PCa patients receiving RP or RT, especially the double arm randomized controlled studies of RP and RT, such as the SPCG-15 currently in progress .
There are still several limitations to our study. First, our research is constructed by retrospective data. Therefore, there may be some undetected potential bias factors in the study. Second, due to the SEER database’s limitations, we lack functional data and disease progression data for further research. Third, we also lack additional independent external validation sets, and this is our important work goal in the future.