The Prognostic Ability of Staging System in Men with Penile Cancer: An Analysis of SEER Database

Background: Penile cancer is now a rare condition. The low incidence of the disease makes a valid estimation of its prognosis difficult. In this study, we made an attempt and propose a nomogram to develop a prognostic rule that could predict the Cancer-Specific Mortality (CSM) free rates in patients with primary penile squamous cell carcinoma of the penis (PPSCC). Methods: This study included 1304 patients diagnosed with PPSCC between the years 2004 & 2011 and treated with penile tumor excision. Subjects were staged as per Surveillance, Epidemiology & End Results stage (SEER), American Joint Committee on Cancer (AJCC), TNM classification and tumor grade (TG). CSM free rates were determined. Univariate and multivariate Cox regression model was used to test the prediction of the CSM free rate. The predictive rule accuracy was created using the receiver operating characteristic curve. Results: The clinico-pathological profile depicts a mean age of 64.66 ± 14.38 yrs. The most common primary site involved was glans penis (n= 483, 37%) and the disease was most commonly diagnosed at AJCC stage I (n= 670, 51.4%) disease. The cumulative 5-year CSM free rates according to Fine & Gray, & Kaplan-Meier methods were 81.8% and 79.8%, respectively. The predictive accuracy as per SEER stage, AJCC stage, TNM stage alone were 68.8%, 70.3%, 72.3%, respectively. When TG was combined, the predictive accuracy increased to 72.8%, 73.1%, and 75.0%, respectively. TNM stage with TG was most accurate in predicting CSM free rate compared to other


Introduction
Penile cancer is now a rare condition, accounting for less than 1% of cancers diagnosed in men in the United States. Most of the penile cancers are squamous cell carcinoma. Though rare, the disease itself is a challenge years. 2 Lower incidence of the disease makes the valid estimation of prognosis of primary penile squamous cell carcinoma (PPSCC) difficult. To address this issue many investigators Kattan et al, Zini et al, and Thuret et al developed various models to predict Cancer-Specific Mortality (CSM) free rates for PPSCC. [3][4][5] The pathological presence of lymph node metastases is the most powerful predictor of cancer-specific survival in patients with PPSCC. 6 In clinically lymph nodenegative patients, this information is only available in those patients when they undergo inguinal lymph node dissection (ILND). However, due to a higher incidence of postoperative local complications all the patients do not undergo ILND. Evidence suggests though micrometastasis may be present in up to 25% of patient with clinically negative nodes. 7 In this present study, we made an attempt to reconfirm the findings of the previous studies after many years now and propose a nomogram to develop a prognostic rule that could be applied to US men to predict the CSM free rates in patients with PPSCC after penile tumor excision (PTE).

Methods
The Surveillance, epidemiology and end results (SEER) database consisting of 18 SEER registries of the National Cancer Institute (NCI) program covers approximately 34.6 % of the US population and is considered the representative of the United States with regard to demographic composition, as well as of cancer incidence and mortality. In this study, it was used to extract information on subjects diagnosed with PPSCC between the years 2004 and 2011. Patients were identified according to diagnostic codes: the tenth revision of the International Classification of Disease (ICD) for Oncology [C60.0-60.9] and the ICD-O-3 codes for histological subtype (squamous cell carcinoma type; ICD-O-3: 8050-8089). All subjects were staged according to the American Joint Committee on Cancer (AJCC) sixth edition (2002), 2002 TNM system based on the SEER Extent of Disease Classification. Variables analyzed were: Age, race, marital status, tumor stage and grade, TNM stage distribution and SEER registry. The extent was grouped as localized (cancer was limited to the penis), regional (cancer extended outside the penis to lymph nodes in the pelvic area), distant metastasis.
The point of observation was from PTE. Only patients who had PTE were included in the study. Those who didn't undergo surgery or when the surgical procedure was not specified or when surgery type was unknown were excluded. Patients with unknown SEER Stage, AJCC stage, unknown grade, and unknown metastatic status were also excluded. Similarly, the number of patients from Alaska Natives and Rural Georgia were very less and were also excluded ( detailed in Figure 1). Patients whose nodal status was not accessed were considered as N0. The patients in whom T stage and M stage could not be accessed were also less in number, and were excluded. The cause of death was defined according to SEER specific cause of death codes. And for the analysis, death from other cause was considered as censored or as other cause mortality on competing risk analysis. (Figure 1) Descriptive statistics were used to summarize the cohort, including mean ± SD, median and range for continuous variables, or counts and percentages for categorical variables. Kaplan-Meier plots were performed to determine CSM-free survival rates in the study population, and after stratifying by SEER and AJCC stage, TNM classification and TG. Log-rank test was performed to compare the categories within AJCC stage, TNM, and TG and p-value was adjusted as per Bonferroni corrections.

BPKMCH
proportionality assumption was tested using Schoenfeld residual graphs and also by the statistical test. An additional 3 Cox regression models were fitted using the same disease stages after adding of TG. Likelihood ratio test was applied to test the additional advantage of TG. These models were used to develop nomograms. The prognostic ability of nomogram was quantified with receiver operating characteristic (ROC) curve derived area under the curve (AUC) estimates. The prognostic ability of the 3 staging systems with and without TG was tested for predicting the 5-year CSM-free rate.
Since a proportion of patients with PPSCC die of other causes, we used univariate and multivariate competing risks regression models, as described by Fine and Gray, to test the significance of the variables in predicting CSM free rates. 8 Competing risks regression models allow us to account for the effect of other cause mortality. All statistical analyses were performed using SEER*Stat (version 8.3.5; National Cancer Institute, Bethesda,Md) and survival curves were generated with the survival function from the R statistical package for Windows version 3.5.2. All statistical tests were done with R statistical package. Statistical significance was set at 0.05.

Results
The clinicopathological profile of all patients of PPSCC who underwent PTE between 2004-2011 is summarized in Table

Figure 2: Cumulative incidence graphs shows CSM-free and other cause mortality-free rates in overall population of 1,304 men (A). Kaplan-Meier survival curve represents CSM-free rate in overall cohort (B). Kaplan-Meier survival curve shows CSM-free rate by localized vs regional vs distant SEER stage (C). Kaplan-Meier survival curve demonstrates CSM-free rate AJCC stages I vs II vs III vs IV (D).
Official Journal of B P Koirala Memorial Cancer Hospital The competing risk regression model was used to adjust the patients for non-penile cancer-related mortality ( Table 3). The competing risk model for predictors of CSM free mortality SEER stage, AJCC stage, TNM stage, and TG stage was observed to be statistically significant. When each stage was combined with TG stage it was still found to be significant except for T4 versus T1.

BPKMCH
The models with or without TG to predict 5-year CSM free mortality for the penile cancer were calibrated using nomograms (Figure 4, A-F). In nomograms based on SEER stage and TG, distant metastasis was the most powerful CSM predictor. Similarly, in nomogram based on AJCC stage and TG, AJCC stage IV PPSCC was the most powerful predictor respectively. However, in each nomogram, high TG was a less influential variable than SEER stage distant and AJCC stage IV. Nodal status N3 also had a powerful effect. However, Nodal status (N1 and N2) and TG II and III had an intermediate effect.
T substage was an even less influential variable. The internal calibration done with nomogram shows a good relationship between predicted and observed rate.

Discussion
This study included 1304 cases from SEER registry that addresses 34.6% of the USA population and it' s one of the largest cancer databases. 9 Here, all penile cancer patients who underwent penile tumor excision were included and analysis was done taking their SEER stage, AJCC stage, TNM stage and combining tumor grade with them. We made an attempt to develop a simple prognostic rule that could be applied to the US men to predict the CSM after penile cancer. This study is done to know if findings in previous studies were consistent even after many years now. [3][4][5] Penile cancer, though more common among elderly men (above 60 years), in small percentage it can also occur

Figure 4: Nomograms predicting CSM-free rate 5 years after primary tumor excision using SEER stage (A), TNM classification (C) and AJCC stage (E) combined with TG. Calibration between predicted (x axis) and observed (y axis) 5-year CSM free rate for SEER stage (B), TNM classification (D) and AJCC stage (F) models.
in younger males. [10][11][12] In this study, we found it be more common in men above 60 years. Men who are unmarried, divorced or separated are more prone to the disease. Moreover, multiple sexual partners or history of sexually transmitted diseases or no use of barrier contraceptives are at risk. 13,14 The disease may occur at any site, however, the glans penis is the most common primary site and well to moderately differentiated squamous cell carcinoma is the most common histological grade. 11

BPKMCH
The current prognostic rule (TNM stage with TG), therefore, more accurately stratify the risk in patients with penile cancer than the use of other models. Such prognostic rule helps us to determine the consideration of adjuvant therapy and also the determination of the frequency of follow up.
As a number of penile cancer patients die of the noncancer-related cause, the Cox regression model may overestimate the true CSM free rates. To avoid such bias, we also used competing risk regression model. CSM free rate according to both models Cox regression and competing risk were calculated.
The CSM free rates according to Kaplan Meir based estimation was strikingly similar to cumulative incidence estimation by Fine and Gray methodology. Both curves superimposed each other. Thus, we relied on the Cox regression model to develop our prediction rule.
Our proposed four variables model TNM stage and TG though may appear more complex than two variable AJCC stage and TG, but this former model had better accuracy than later and this may provide the clinician an excellent ability to predict the CSM free rates.
Even though data collected in the SEER database is considered reliable it has some limitations. It may have variations in data entry or miscoding or incomplete data of various variables. The database didn't include information regarding co-morbidities like phimosis, smoking and other medical problems. Surgical details like margin status, which are important prognostic factors. Lack of central pathology review and accuracy of the available pathological report may inaccurately stage the disease. Likewise, Information regarding disease recurrence, adjuvant therapy, and long-term follow-up data is under addressed. 16,17 The external validity of the nomograms is required for wide acceptance and clinical application of the observation. The ideal validation should be obtained in a prospective series of patients. This SEER database includes patients from USA and for external validation purpose, an additional patient from a second institution, or other countries Asia or South America with similar protocol with respect to surgery type indications need to be included. External validation was not done in this study.

Conclusion
TNM stage with TG and AJCC stage with TG appear to have comparable accuracy to predict the CSM free rate in patients with PPSCC, TNM stage with TG is the most accurate (75%) method to predict the CSM free rates. Addition of TG variable definitely improved the accuracy of these prognostic model.