Development and Validation of a Preoperative Nomogram for Predicting Survival of Patients with Locally Advanced Prostate Cancer after Radical Prostatectomy

Abstract

Background: For selected locally advanced prostate cancer (PCa) patients, radical prostatectomy (RP) is one of the first-line treatments. We aimed to develop a preoperative nomogram to identify what kinds of patients can get the most survival benefits after RP.

Methods: We conducted analyses with data from the the Surveillance, Epidemiology, and End Results (SEER) database. Univariable and multivariable Cox regression analyses were used to reveal prognostic factors. A nomogram was built by these factors and validated by concordance index (C-index) and calibration curves. Risk stratification was established based on the nomogram.

Results: We studied 14185 patients. N stage, Gleason Score, and percent of positive cores were the independent prognostic factors used to construct the nomogram. For validating, in the training cohort, the C-index was 0.779 (95% CI 0.736–0.822), and in the validation cohort, the C-index was 0.773 (95% CI 0.718–0.808). Calibration curves showed that the predicted survival and actual survival were very close. The nomogram performed better over the American Joint Committee on Cancer (AJCC) staging system (C-index 0.779 versus 0.763 for training cohort, and 0.773 versus 0.745 for validation cohort). The new stratification of risk groups based on the nomogram also showed better discrimination than the AJCC staging system.

Conclusions: The preoperative nomogram can provide favorable prognosis stratification ability to help clinicians identify patients who are suitable for surgery.

Background

Prostate cancer (PCa) is one of the most common malignant tumors of males around the world, and it is estimated that there will be 174,650 new male cases in the United States in 2019 [1]. The European Association of Urology defined PCa with cT3–4 or N+ as locally advanced PCa [2]. Patients with locally advanced PCa have an increased risk of disease progression and cancer-specific death [2]. Due to many factors like the American national guidelines advising against PSA screening, increasing applications for active surveillance, etc., the proportion of men diagnosed with locally advanced PCa has increased [3, 4]. Nowadays, the standard treatment of locally advanced PCa is still unclear due to the absence of level 1 evidence. For selected locally advanced PCa patients, radical prostatectomy (RP) is one of the first-line treatments recommended by the guidelines [2, 5]. However, it is still controversial as to which types of patients could get the most survival benefits from RP.

Some studies about the risk stratification of patients with high-risk PCa (including locally advanced PCa) after RP had been reported. Previous researchers performed a retrospective analysis of 315 high-risk PCa patients after RP in their hospital. They selected biochemical progression as the primary endpoint, and reported that the risk of biochemical progression of high-risk PCa after RP could be stratified by Gleason Score (GS) at biopsy (≥ 8) and % positive core (≥ 30%) [6]. Another research team studied 813 high-risk patients undergoing RP, and found 3 preoperative criteria including stage cT2 or greater, prostate specific antigen (PSA) >20 ng/mL, and GS >8. In their study, the number of meeting preoperative criteria was significantly predictive for recurrence-free survival (RFS), and overall survival (OS) [7]. Although these studies had proposed their own risk stratification models for high-risk PCa patients undergone RP, the patients involved in these studies were not fully compliant with the current more common definition of locally advanced PCa, the impact of prognostic factors on patient survival outcome was not quantitative, the weight between the various prognostic factors was not clear enough and the cohorts were not large enough.

To circumvent these defects, we evaluated the prostate cancer-specific survival (CSS) at a large cohort to assess potential preoperative prognostic factors, and developed a preoperative nomogram. This nomogram could be used to predict the CSS of patients with locally advanced PCa after RP and to identify what kinds of patients can get the most survival benefits from RP.

Methods

Patient selection

The Surveillance, Epidemiology, and End Results (SEER) database is a free dataset made up of 18 population-based cancer registries. It releases cancer patients’ general information annually and has almost covered 25% population of the United States [8].

From the SEER database, patients with a diagnosis of adenocarcinoma of the prostate (International Classification of diseases-O–3 code: C61.9) between 2010 and 2016 were selected. The TNM stages were assessed by the 7th edition of American Joint Committee on Cancer [AJCC] Cancer Staging Manual [9]. Inclusion and exclusion criteria were shown in the flowchart in detail (Figure 1). All the included patients were randomly divided into the training cohort and validation cohort with a ratio of 7:3.

Variables and endpoint

Data for each patient were extracted from the SEER database including age at diagnosis, marital status, race, AJCC 7th TNM stage, PSA, Gleason biopsy score (GS), percent of positive cores at biopsy (% positive core), and follow-up information. Age was categorized as ≤ 49 years, 50–59 years, 60–69 years, and ≥70 years. PSA was categorized as ≤10ng/mL, 10–20 ng/mL (not including 10), and >20 ng/mL. GS was classified as ≤6, 7(3+4), 7(4+3), 8, and ≥9. % positive core was classified as 0–25%, 25%–50% (not including 25%), 50%–75% (not including 50%), and 75%–100% (not including 75%). In addition to follow-up information, all clinical factors were obtained before undergoing RP. CSS was used as the primary endpoint. CSS was measured by all deaths caused by prostate cancer, complications of treatments, or unknown processes in patients with active prostate cancer. Follow-up time was defined as the time between the first treatment and the patient’s death or last follow-up.

Statistical analysis

A ² test was used to expose the difference between the training cohort and validation cohort in categorical variables, and the results were presented as the frequency with its proportion. Kaplan–Meier method and log-rank test were used to expose each potential prognostic variable. Variables in the univariate analysis with p-value <0.05 were chosen for multivariate Cox proportional hazard regression to identify the independent prognostic factors.

Based on clinical significance and p-value <0.05 in multivariate Cox proportional hazard regression, these independent prognostic factors were selected to construct a nomogram to predict 5‐year cancer-specific survival probabilities of patients.

The performance of the nomogram was validated by measuring discrimination and calibration in both training set and validation set. The discrimination of the nomogram was assessed by C-index (concordance index). The value of C-index ranged from 0.5 (no discrimination) to 1 (perfect discrimination), and higher C-index showed better discrimination of the prognostic model [10]. By comparing the plot of predicted probabilities from the nomogram with the plot of actual probabilities of CSS, the calibration curve was used to compare the predicted survival using nomogram with the actual observed survival, with 1000 resamples of bootstrapping. In addition, we conducted a risk stratification of the nomogram total risk score. The cut-off values were determined using X-tile software and by the minimal p-value approach [11]. The discrimination abilities between the nomogram and AJCC staging system were compared by C-index and Kaplan–Meier curve.

The statistical software packages R (The R Foundation) and EmpowerStats (http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA) were used in the above statistical analyses. A p-value < 0.05 was considered statistically significant.

Results

Patient characteristics

A total of 14185 eligible patients were included in the current analysis as the primary cohort. Among all patients, 9932 patients were placed within the training cohort, and 4523 patients were placed within the validation cohort randomly. Detail patients’ characteristics were shown in Table 1. There were no statistically significant differences in baseline characteristics between the training cohort and the validation cohort.

Determination of independent preoperative prognostic factors

Multivariate Cox proportional hazard regression was performed to demonstrate the independent preoperativeprognostic factors in the training cohort. Table 2 showed the detail univariate and multivariate analysis results. In the univariate analyses, age, N stage, PSA, GS, % positive core biopsy were variables with significant impact on CSS.

In the multivariate Cox proportional hazard regression, N stage, GS, % positive core biopsy remained significantly, while age and PSA did not show significant impact on CSS. These significant variables were thought as independent prognostic factors of CSS for locally advanced PCa patients after RP.

Construction and validation of the nomogram

Using the independent prognostic factors including N stage, GS, and %positive core biopsy, a nomogram was constructed to predict each locally advanced PCa patient’s probability of CSS at 5 years after RP. (Figure 2)

The nomograms were validated using C-index and calibration curve. In the training cohort, the C-index was 0.779 (95% CI 0.736–0.822) in CSS. In the validation cohort, the C-index was 0.773 (95% CI 0.718–0.808) in CSS. The C-index values indicated good discrimination of the nomogram (>0.7). The calibration curves showed good agreement between prediction by nomogram and actual 5‐year CSS in both training cohort and validation cohort (Figure 3).

Definition of nomogram risk group stratification

The score corresponding to each nomogram variable was listed in detail in Table 3. By summing the score for each nomogram variable, we got the nomogram total risk score for each patient in both the training cohort and the validation cohort. The patients were divided into three nomogram risk groups for CSS: a low-risk group with 0–77 points, a middle-risk group with 79–108 points, a high-risk group with no less than 112 points.

Comparison of nomogram with AJCC staging system

We compared the nomogram with the 8th edition AJCC staging system. In the training cohort, the C‐index of AJCC staging system was 0.763 (95% CI, 0.736‐0.822) for CSS, and the C‐index of the nomogram was 0.779 (95% CI 0.736–0.822) for CSS. In the validation cohort, the C‐index of AJCC staging system was 0.745 (95% CI,0.672‐0.818) for CSS, and the C‐index of the nomogram was 0.773 (95% CI 0.718–0.808) for CSS. In both cohorts, the nomogram performed better in C-index, indicating that the nomogram had a better discrimination ability than AJCC staging system.

The AJCC staging could divide the whole eligible patients in current study into 3 groups including IIIB, IIIC, and IVA. Relatively, as mentioned above, the risk groups based on the nomogram included low-risk group, middle-risk group, high-risk group. For risk groups based on the nomogram, the degree of separation of Kaplan-Meier curves of CSS between groups was more obvious than AJCC staging system in both training cohort and validation cohort (Figure 4). The Kaplan–Meier method still demonstrated the new stratification of risk groups based on the nomogram had better prognostic discrimination than AJCC staging system.

Discussion

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 CSS 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 [12]. 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 [12, 13]. Compared with the current prognostic prediction system like AJCC staging system, the nomogram showed advantages in many studies on different cancers [14, 15]. 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. [16–18].

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 [19]. 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, 20, 21]. 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, 20, 22, 23]. 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 with 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. [22] 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 [24, 25]. A retrospective study of 229 patients indicated that the 5-year disease-free survival of rate of patients with regional lymph node metastasis could decreased from 85% to approximately 50% [26].

Relatively, the 8th AJCC staging system is currently a general PCa risk stratification system, not aim at specific disease stages and treatment therapies [27]. 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 Kaplan–Meier 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.

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.

Conclusions

By analyzing a large cohort of 14185 patients, we successfully revealed 3 independent preoperative prognostic factors for CSS of locally advanced PCa patients after RP. Using these 3 factors, we constructed and validated a preoperative prognostic nomogram that could predict the probability of CSS at 5 years of this type of patients after undergoing RP. Based on the nomogram, a risk stratification system was established to help clinicians better identify patients suitable for surgery.

List of Abbreviations

PCa: Prostate cancer; CI: confidence interval; RP: radical prostatectomy; RFS: recurrence-free survival; OS: overall survival; CSS: cancer-specific survival; PSA: prostate specific antigen; GS: Gleason score; SEER: the Surveillance, Epidemiology, and End Results; AJCC: American Joint Committee on Cancer

Declarations

Acknowledgments

The authors acknowledge the efforts of the Surveillance Research Program in the establishment and maintenance of the SEER database.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors

Availability of data and materials

All analyzed data are publicly available at the SEER website (http://www.seer.cancer.gov), and should be requested under the approval of the SEER Program administration.

Authors’ contributions

XHZ, QYN, XLM are responsible for writing, collecting data, analysis, interpretation, revision and final approval of present article. KJ, TZ are responsible for analysis and revision.

All authors have read and approved the final manuscript.

Ethics approval and consent to participate

This study used a general dataset from the Surveillance, Epidemiology, and End Results database built by a public health program, and therefore did not require institutional review board approval and the patients’ data were atomized

Consent for publication

All authors agreed to publish the manuscript titled “Development and Validation of a Preoperative Nomogram for Predicting Survival of Patients with Locally Advanced Prostate Cancer after Radical Prostatectomy” in the journal BMC Cancer.

Competing interests

The authors declare that they have no competing interests.

References

1.Siegel RL, Miller KD, Jemal A. Cancer Statistics, 2019. CA Cancer J Clin. 2019; 1: 7–34.

2.Mottet N, Bellmunt J, Bolla M et al. EAU-ESTRO-SIOG Guidelines On Prostate Cancer. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol. 2017; 4: 618–629.

3.Huland H, Graefen M. Changing Trends in Surgical Management of Prostate Cancer: The End of Overtreatment? Eur Urol. 2015; 2: 175–178.

4.Fletcher SA, von Landenberg N, Cole AP et al. Contemporary National Trends in Prostate Cancer Risk Profile at Diagnosis. Prostate Cancer Prostatic Dis. 2019.

5.Mohler JL, Antonarakis ES, Armstrong AJ et al. Prostate Cancer, Version 2.2019, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 2019; 5: 479–505.

6.Nagao K, Matsuyama H, Matsumoto H et al. Identification of Curable High-Risk Prostate Cancer Using Radical Prostatectomy Alone: Who are the Good Candidates for Undergoing Radical Prostatectomy Among Patients with High-Risk Prostate Cancer? Int J Clin Oncol. 2018; 4: 757–764.

7.Ploussard G, Masson-Lecomte A, Beauval JB et al. Radical Prostatectomy for High-Risk Prostate Cancer Defined by Preoperative Criteria: Oncologic Follow-Up in National Multicenter Study in 813 Patients and Assessment of Easy-To-Use Prognostic Substratification. Urology. 2011; 3: 607–613.

8.Cronin KA, Ries LA, Edwards BK. The Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute. Cancer-Am Cancer Soc. 2014; 3755–3757.

9.Cuccurullo V. AJCC Cancer Staging Handbook: From the AJCC Cancer Staging Manual (7Th Edition). European Journal of Nuclear Medicine & Molecular Imaging. 2011; 2: 408.

10.Weiss A, Chavez-Macgregor M, Lichtensztajn DY et al. Validation Study of the American Joint Committee on Cancer Eighth Edition Prognostic Stage Compared with the Anatomic Stage in Breast Cancer. Jama Oncol. 2017; 2.

11.Camp RL, Marisa DF, Rimm DL. X-Tile: A New Bio-Informatics Tool for Biomarker Assessment and Outcome-Based Cut-Point Optimization. Clin Cancer Res. 2004; 21: 7252–7259.

12.Balachandran VP, Gonen M, Smith JJ, DeMatteo RP. Nomograms in Oncology: More than Meets the Eye. Lancet Oncol. 2015; 4: e173–180.

13.Shouval R, Labopin M, Gorin NC et al. Individualized Prediction of Leukemia-Free Survival After Autologous Stem Cell Transplantation in Acute Myeloid Leukemia. Cancer-Am Cancer Soc. 2019.

14.Cho CSM, Gonen MP, Shia JM et al. A Novel Prognostic Nomogram is More Accurate than Conventional Staging Systems for Predicting Survival after Resection of Hepatocellular Carcinoma. J Am Coll Surgeons. 2008; 2: 281–291.

15.Wong SL, Kattan MW, McMasters KM, Coit DG. A Nomogram that Predicts the Presence of Sentinel Node Metastasis in Melanoma with Better Discrimination than the American Joint Committee On Cancer Staging System. Ann Surg Oncol. 2005; 4: 282–288.

16.Wu Y, Meyers JP, Shi G et al. A Nomogram for Predicting Survival and Retroperitoneal Lymph Node Dissection Treatment in Patients with Resected Testicular Germ Cell Tumors. J Surg Oncol. 2019; 3: 508–517.

17.Kim BH, Kim K, Chie EK et al. Risk Stratification and Prognostic Nomogram for Post-Recurrence Overall Survival in Patients with Recurrent Extrahepatic Cholangiocarcinoma. Hpb. 2017; 5: 421–428.

18.Kim Y, Park HC, Yoon SM et al. Prognostic Group Stratification and Nomogram for Predicting Overall Survival in Patients Who Received Radiotherapy for Abdominal Lymph Node Metastasis From Hepatocellular Carcinoma: A Multi-Institutional Retrospective Study (KROG 15–02). Oncotarget. 2017; 55: 94450–94461.

19.Epstein JI, Egevad L, Amin MB, Delahunt B, Srigley JR, Humphrey PA. The 2014 International Society of Urological Pathology (ISUP) Consensus Conference on Gleason Grading of Prostatic Carcinoma: Definition of Grading Patterns and Proposal for a New Grading System. Am J Surg Pathol. 2016; 2: 244–252.

20.Sundi D, Wang V, Pierorazio PM et al. Identification of Men with the Highest Risk of Early Disease Recurrence After Radical Prostatectomy. The Prostate. 2014; 6: 628–636.

21.Kim TH, Jeon HG, Jeong BC et al. Development of a New Nomogram to Predict Insignificant Prostate Cancer in Patients Undergoing Radical Prostatectomy. Scand J Urol. 2017; 1: 27–32.

22.Grossklaus DJ, Coffey CS, Shappell SB, Jack GS, Chang SS, Cookson MS. Percent of Cancer in the Biopsy Set Predicts Pathological Findings After Prostatectomy. J Urol. 2002; 5: 2032–2036.

23.Hamada R, Nakashima J, Ohori M et al. Preoperative Predictive Factors and Further Risk Stratification of Biochemical Recurrence in Clinically Localized High-Risk Prostate Cancer. Int J Clin Oncol. 2016; 3: 595–600.

24.Evangelista L, Guttilla A, Zattoni F, Muzzio PC, Zattoni F. Utility of Choline Positron Emission Tomography/Computed Tomography for Lymph Node Involvement Identification in Intermediate- to High-Risk Prostate Cancer: A Systematic Literature Review and Meta-Analysis. Eur Urol. 2013; 6: 1040–1048.

25.Sheng W, Zhang H, Lu Y. Survival Outcomes of Locally Advanced Prostate Cancer in Patients Aged &Lt; 50 Years After Local Therapy in the Contemporary US Population. Int Urol Nephrol. 2018; 8: 1435–1444.

26.Danella JF, DeKernion JB, Smith RB, Steckel J. The Contemporary Incidence of Lymph Node Metastases in Prostate Cancer: Implications for Laparoscopic Lymph Node Dissection. J Urol. 1993; 6: 1488–1491.

27.Amin MB, Greene FL, Edge SB et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing to Build a Bridge From a Population-Based to a More “Personalized” Approach to Cancer Staging. CA Cancer J Clin. 2017; 2: 93–99.

Tables

Table 1. Descriptive characteristics of 14185 locally advanced prostate cancer patients undergoing radical prostatectomy between 2010 and 2016 from the Surveillance Epidemiology and End Results database. Primary 14185 patients were randomly divided into 2 cohorts: training cohort and validation cohort.

 1.Other: American Indian/AK Native, Asian/Pacific Islander

2.Single: Divorced, Separated, Single (never married), Widowed, unmarried

Abbreviation: AJCC: American Joint Committee on Cancer.

Table 2. Univariate and multivariate Cox regression analyses of preoperative prognostic factors influencing cancer-specific survival outcomes in the training cohort

Abbreviation: Ref.: reference, % positive core biopsy: percent of positive cores at biopsy

Table 3. Detailed risk scores of all prognostic factors in the nomogram

Abbreviation: CSS: cancer specific survival