Construction and validation of a prediction model for inguinal lymph node metastasis of penile malignancy

Introduction

Penile squamous cell carcinoma is a relatively rare malignancy among male malignancies. According to the Global Cancer Statistics Report 2020, there are more than 30,000 new cases and more than 10,000 deaths of penile cancer annually (1). According to the 2019 China Cancer Report, the incidence of penile cancer is approximately 6.1 per 100,000 (2). The incidence of penile cancer is higher in remote mountainous areas of Africa, South America, and China. The incidence rate of penile cancer in Africa and other places is 30–40 times higher than that in developed countries, such as European countries and the United States. The remote mountainous areas of Western China also have significantly higher incidence rates of penile cancer compared with more developed coastal cities (1-3).

Penile cancer is highly aggressive and can lead to the early stages of lymph node micrometastasis, with inguinal lymph nodes as the most common site of metastasis. It is widely recognized that inguinal lymph node metastases (ILNMs) have a significant impact on patients’ recurrence-free survival and overall survival. Therefore, early or concomitant inguinal lymph node dissection (ILND) may improve patient survival and prognosis (4). However, approximately 70% of patients who undergo ILND develop postoperative complications, such as infection, wound dehiscence, etc. (5). Therefore, it is important to determine whether patients with penile cancer need ILND, thereby avoiding unnecessary surgery and excessive complications.

In previous studies, ILNM has been linked to various clinical signs and pathologies. Jia et al. (6) uncovered that lymphatic filtration (LVI), a higher grade of tumor, advanced stage, perineural invasion, and younger age at diagnosis are risk factors for ILNM in penile cancer. In terms of biochemistry, overexpression of p53, superoxide dismutase 2 (SOD2), Ki-67, and ID1 are associated with the spread of squamous cell carcinoma to the inguinal lymph nodes. In addition to PD-L1 expression, squamous cell carcinoma antigen (SCC-Ag), neutrophil-to-lymphocyte ratio (NLR), and C-reactive protein (CRP) are involved in the lymph node metastasis of penile cancer. Recently, Fankhauser et al., Zekan et al. (7,8), and others have confirmed that maximum lymph node diameter, tumor pathological T stage, degree of differentiation, lymphovascular invasion, perineural infiltration (PNI), and tumor size are risk factors for ILNM. However, there are few prediction models for ILNM in penile cancer based on clinicopathological features, biochemical markers, and tumor markers. This study retrospectively collected the clinicopathological and biochemical indexes of patients with penile cancer in Yunnan Cancer Hospital to construct a new nomogram. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-24-145/rc).

Methods

Materials

We retrospectively collected patients with penile cancer who visited Yunnan Cancer Hospital from January 2017 to December 2021 did not receive antitumor therapy.

Inclusion criteria were as follows: (I) patients with preoperative pathological biopsy or postoperative histopathological report of squamous cell carcinoma; (II) patients with no systemic or organ-specific infection or other diseases one week before surgery that could affect blood routine; (III) patient not receiving radiotherapy and chemotherapy before surgery and had no history of blood transfusion or surgery within 3 months; and 4. all patients underwent partial penile resection or total resection, and ILND at the same time.

Exclusion criteria were as follows: (I) patients with incomplete medical records or unable to obtain required clinical information, such as preoperative imaging, inflammatory factors, related tumor markers, and other indicators; (II) patients with other malignancies; (III) patients unwilling to provide information about their disease.

All patients signed the informed consent form, and the protocol of the study was approved by the Medical Ethics Committee of Yunnan Cancer Hospital before data collection (No. KYLX2023-215, Kunming, Yunnan Province, China, December 2023). The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

In conclusion, this study included patients who underwent partial or total penile excision plus ILND. All patients had complete imaging information, clinicopathological data and biochemical parameter.

Clinical indicators

Patients’ age, smoking history, lymph node size on imaging, clinical lymph node staging, blood routine, tumor markers, biochemical indicators, and results of postoperative pathological assessment, including histopathological type, pathological T staging, tumor size, differentiation, and other data were collected. The pathological T stage was performed according to the 2019 American Joint Committee on Cancer (AJCC) stage. Clinical lymph node staging was recorded at the first visit. Clinical staging was performed according to preoperative imaging [computed tomography (CT)]. According to the postoperative lymph node metastasis, patients were divided into the lymph node metastasis and non-lymph node metastasis control groups.

Statistical analysis

Data were collected using EXCEL and analyzed using SPSS24 and R 4.2.1. Continuous and categorical variables were presented as mean ± standard deviation and numbers (percentages), respectively. The Wilcoxon rank-sum test, t-test, and Fisher exact test were used to compare demographics, clinical features, and differences between the ILNM and non-ILNM groups. Data were imputed by multiple imputations (9), which mainly included the imputation of test indicators.

The performance of the nomogram was evaluated using the logistic least absolute shrinkage and selection operator (LASSO) regression to identify demographic and clinical features increasing the risk of ILNM. A predictive nomogram was constructed for ILNM, and risk factors were screened by LASSO regression, Harrell, S harmony index (C index), and calibration chart. We used leave-one-way cross-validation (LOOCV) to assess internal validation. At the same time, the decision curve analysis (DCA) and receiver operating characteristic (ROC) curves were used to evaluate the prediction ability of the model.

Results

Demographic and clinicopathological features

In total, 112 male inpatients were included in this study. Of them, 34.8% (n=39) had ILNMs. There was a significant difference in the diameter of lymph nodes (P=0.001), cN stage (P=0.008), the absolute value of neutrophils (P<0.001), NLR (P=0.002), and SII (P=0.03) between the two groups. There were no differences in other indicators such as smoking history, age, and test indicators (Table 1). Baseline data between the two groups were basically comparable.

Selection of predictors and establishment of predictive models

In this study, 23 variables were included in the LASSO regression for screening (Figure 1). The LASSO regression indicated 8 variables as the significant predictors of ILNM. We chose λ, which had the smallest mean squared error, because it had the smallest binomial bias, and the predictive model constructed based on independent variables was the most accurate. Finally, eight variables were identified, including age, smoking history, lymph node diameter, T stage, fibrinogen, interleukin 4 (IL-4), IL-2, and NLR.

Figure 1 LASSO regression plots for different penalty parameter values (A) and cross-validation plots for penalty terms (B). In total, 23 variables were included in the LASSO regression for variable screening, of which 8 variables were significant predictors of ILNM. LASSO, least absolute shrinkage and selection operator; ILNM, inguinal lymph node metastasis.

Age (β=−0.005, OR =0.995), smoking history (β=−0.006, OR =0.994), and IL-2 (β=−0.0112, OR =0.989) were protective against ILNM. Lymph node diameter (β=0.3117, OR =1.366), T stage (β=0.1254, OR =1.134), fibrinogen (β=0.0377, OR =1.038), IL-4 (β=0.004, OR =1.001), and NLR (β=0.0355, OR =1.034) were risk factors for ILNM.

Predictive models

As shown in Figure 2, significant factors identified by LASSO regression analysis, including age, smoking history, lymph node diameter, T staging, fibrinogen, IL-4, IL-2, and NLR, were used to construct a nomogram and predict the risk of ILNM in penile cancer.

Figure 2 The nomogram was constructed based on the risk factors of ILNM. The value of each variable provides a score on the point scale axis. By adding up the scores obtained for each variable, we got a total score, from which we estimated the probability of ILNM among patients. The red dots indicate the scores for all variables [427] and the probability of ILNM for the first patient (44.6%). *, the variable has a significant role in the model. ILNM, inguinal lymph node metastasis; NLR, neutrophil-to-lymphocyte ratio; IL, interleukin; FIB, fibrinogen.

The correction curve of the nomogram showed that the predicted and actual risks of ILNM were in good agreement (Figure 3A). At the same time, using the ROC curve to visualize the predicted probability (Figure 3B), it was seen that the area under the curve (AUC) value of the predictive model was 0.81, showing that the model is more accurate. In addition, the model was internally cross-validated by the LOOCV method and validated with a centralized AUC of 0.75. When the AUC value was greater than 0.7, we considered the model to have good consistency. At the same time, we used DCA (Figure 4A), which showed that the model was more beneficial than the extreme curve and had better clinical benefits. Then, the clinical impact curve (Figure 4B), revealed that the number of true positives and the risk of ILNM increased with the threshold probability of the model, which provides a guide to prevent overtreatment in clinical settings.

Figure 3 Calibration curve (A) and ROC curve (B) of the prediction model. Based on the calibration curve of the nomogram (A), the predicted risk of ILNM was in good agreement with the actual risk. The ROC curve showed that the AUC value of the prediction model was 0.81 (B), indicating that the model had good accuracy. ILNM, inguinal lymph node metastasis; ROC, receiver operating characteristic; AUC, area under the curve; CI, confidence interval.

Figure 4 DCA curve (A) and clinical impact curve (B) of the predictive model. (A) The results showed that the benefit of the model was higher than the extreme value curve, which had a good clinical benefit. The clinical impact curve (B) revealed that as the threshold probability of the model continued to increase, the number of true positive cases and the likelihood of ILNM increased. ILNM, inguinal lymph node metastasis; DCA, decision curve analysis.

Discussion

In patients with penile malignancy, ILNMs significantly reduce overall survival (5). In addition, 6% to 30% of patients with impalpable inguinal lymphadenopathy have lymph node micrometastases, and inguinal lymphadenectomy can prevent lymph node metastasis (5). However, due to the specific location of the groin, the incidence of postoperative complications, such as postoperative wound infection, surgical mouth dehiscence, lymphedema, etc., is as high as 70%, and the recovery time is long, which affects the physical and mental health of patients (5,10). Sentinel lymph node biopsy is an ideal method for diagnosing ILNM, but De Vries et al. (11) found that ILND is necessary for patients with positive lymph node biopsy, and in patients with a negative result, the need for ILND is unclear. In addition, additional models for identifying ILNM were built using clinical and pathological factors, but the models did not show clear clinical benefit (11). Nazzani et al. (12) suggested that bilateral ILND is the gold standard of treatment in node-positive patients and that dynamic lymph node biopsy cannot completely replace ILND (12). Therefore, it is important to establish a predictive model that can predict ILNM.

In previous studies, many researchers have established relevant diagnostic models mostly based on the Surveillance, Epidemiology, and End Results (SEER) database. Most of the indicators included in the modeling were clinicopathological indicators, and few models included biochemical indicators. In most of these studies (13-16), younger age, lymphatic vascular invasion (LVI), high grade, advanced stage (clinical and pathological), increased infiltration, and PNI were associated with ILNM. In addition, p53, Ki-67 protein (Ki-67), diffuse programmed cell death-ligand 1 (PD-L1) expression, and inhibitor of DNA binding 1 (ID1) overexpression were associated with the spread of penile squamous cell carcinoma to inguinal lymph nodes. In terms of laboratory indicators, increased SCC-Ag expression, high NLR, and CRP >20 were associated with elevated ILNM (6). Moreover, several studies have demonstrated that the systemic inflammatory index, lactate dehydrogenase level, and interleukin level are associated with the prognosis of various tumors (17,18). However, we did not find a model that incorporates clinicopathological indicators, biochemical indicators, and tumor markers to predict ILNM in penile cancer. Therefore, in this study, we included 25 clinicopathological factors (age, smoking history, clinicopathological stage, and grade of tumor), biochemical indicators (absolute value of neutrophils, absolute lymphocyte value, platelet, lactate dehydrogenase, fibrinogen, interleukin, etc.), tumor marker SCC-Ag, and other indicators to construct a predictive model We used LASSO regression to screen the variables and found that age (OR =0.995), smoking history (OR =0.994), IL-2 (OR =0.989), lymph node diameter (OR =1.366), T stage (OR =1.134), fibrinogen (OR =1.038), IL-4 (OR =1.001), and NLR (OR =1.034) were closely related to ILNM. Logistic regression was used for model construction. The nomogram was used for model visualization, and the ROC curve and calibration curve were used to evaluate the model, showing that the prediction ability of the model was acceptable. To better understand the clinical benefits of the model, we drew DCA curves and clinical impact curves, both of which showed good clinical benefits. In addition, we cross-validated the model using the LOOCV method and assessed the accuracy of the model. The results showed that the model has good accuracy and clinical applicability. Implementing this model has ushered in a significant enhancement in the quality of clinical decision-making for patients afflicted with penile tumors. The model’s precision in identifying those candidates for ILND has been pivotal in substantially curtailing the number of unwarranted surgical interventions, thus mitigating the risks inherent in overtreatment. This sophisticated approach to patient assessment guarantees that therapeutic interventions are administered only when they are genuinely essential. Furthermore, as these models evolve and refine their capabilities, their integration into clinical practice stands to become an increasingly invaluable asset. This advancement promises to pave the way for more personalized and efficacious treatment strategies tailored to the unique needs of penile cancer patients. Younger age, lymph node diameter, T staging, and NLR were common between our model and previous models, indicating that these factors are important risk factors for ILNM. This similarity also demonstrates that the predictive model constructed in this study has high levels of confidence. An important feature of the tumor microenvironment is persistent inflammation, which is considered to be a “non-healing wound”. Neutrophils, as important components of the tumor microenvironment, are involved in tumorigenesis. Neutrophils secrete or recruit various inflammatory factors, including vascular endothelial growth factor, interleukins, and reactive oxygen and nitrogen species, to reshape the tumor microenvironment, thereby promoting tumor proliferation and migration (19). In this study, IL-2, IL-4, and NLR showed that neutrophils play an important role in the development of penile cancer and lymph node metastasis. In addition, fibrinogen, as an acute phase protein, is associated with the prognosis of several solid tumors, including soft tissue sarcoma and non-metastatic renal cell carcinoma (20). Fibrinogen can bind to or interact with growth factors, including fibroblast growth factor-2 (FGF-2), platelet-derived growth factor (PDGF), and transforming growth factor-β (TGF-β), etc., interact with TGF-β to induce an immunosuppressive environment, and bind to FGF-2 or PDGF to promote cancer cell proliferation, metastasis, and angiogenesis. High levels of fibrinogen also promote tumor cell migration and can induce epithelial-mesenchymal transformation to reduce cytotoxic T cell activity (21). Therefore, in patients with penile cancer, fibrinogen plays an important role in lymph node metastasis (20,21).

In summary, younger age, shorter smoking duration, lower IL-2 level, larger lymph node diameter on imaging, later T stage, higher fibrinogen level, higher IL-4 level, and greater NLR are risk factors for ILNM. Smoking is an important risk factor for lymph node metastasis. A younger age at presentation is associated with a greater risk of ILNM.

Conclusions

As a less common but more aggressive tumor, ILNM significantly worsens the prognosis of patients. Age at diagnosis, T stage, lymph node diameter on imaging, fibrinogen, IL-2, IL-4, and NLR can be used as predictors of lymph node metastasis. Nomogram and accuracy assessment of these predictors showed that the model had good predictive ability (AUC =0.81) and clinical benefit. This study was a single-center, small-sized retrospective study, increasing the risk of selection bias and information bias. Only internal cross-validation was performed in this study, and further external and multi-center validation is needed to determine the generalizability of the model. Although several clinicopathological and biochemical indices were included, there may be other important variables that have not been considered and require further study. Furthermore, the development of technologies such as imaging genomics, genomics, proteomics, and artificial intelligence can allow the incorporation of additional characteristic variables to construct a more comprehensive prediction model, thereby enhancing the accuracy and convenience of the model.

Acknowledgments

The authors would like to express their gratitude to EditSprings (https://www.editsprings.cn) for the expert linguistic services provided.

Funding: This study was supported by the National Natural Science Foundation of China (No. 82160511) and National Cancer Center Climbing Fund (No. NCC201925B01).

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tau.amegroups.com/article/view/10.21037/tau-24-145/rc

Data Sharing Statement: Available at https://tau.amegroups.com/article/view/10.21037/tau-24-145/dss

Peer Review File: Available at https://tau.amegroups.com/article/view/10.21037/tau-24-145/prf

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tau.amegroups.com/article/view/10.21037/tau-24-145/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the Medical Ethics Committee of Yunnan Cancer Hospital (No. KYLX2023-215, Kunming, Yunnan Province, China, December 2023) and informed consent was taken from all the patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Douglawi A, Masterson TA. Penile cancer epidemiology and risk factors: a contemporary review. Curr Opin Urol 2019;29:145-9. [Crossref] [PubMed]
3. Spiess PE, Dhillon J, Baumgarten AS, et al. Pathophysiological basis of human papillomavirus in penile cancer: Key to prevention and delivery of more effective therapies. CA Cancer J Clin 2016;66:481-95. [Crossref] [PubMed]
4. Joshi SS, Handorf E, Strauss D, et al. Treatment Trends and Outcomes for Patients With Lymph Node-Positive Cancer of the Penis. JAMA Oncol 2018;4:643-9. [Crossref] [PubMed]
5. Shao Y, Tu X, Liu Y, et al. Predict Lymph Node Metastasis in Penile Cancer Using Clinicopathological Factors and Nomograms. Cancer Manag Res 2021;13:7429-37. [Crossref] [PubMed]
6. Jia Y, Zhao H, Hao Y, et al. Analysis of the related risk factors of inguinal lymph node metastasis in patients with penile cancer: A cross-sectional study. Int Braz J Urol 2022;48:303-13. [Crossref] [PubMed]
7. Fankhauser CD, de Vries HM, Roussel E, et al. Lymphovascular and perineural invasion are risk factors for inguinal lymph node metastases in men with T1G2 penile cancer. J Cancer Res Clin Oncol 2022;148:2231-4. [Crossref] [PubMed]
8. Zekan DS, Dahman A, Hajiran AJ, et al. Prognostic predictors of lymph node metastasis in penile cancer: a systematic review. Int Braz J Urol 2021;47:943-56. [Crossref] [PubMed]
9. Austin PC, White IR, Lee DS, et al. Missing Data in Clinical Research: A Tutorial on Multiple Imputation. Can J Cardiol 2021;37:1322-31. [Crossref] [PubMed]
10. O'Brien JS, Teh J, Chen K, et al. Dynamic Sentinel Lymph Node Biopsy for Penile Cancer: Accuracy is in the Technique. Urology 2022;164:e308. [Crossref] [PubMed]
11. de Vries HM, Lee HJ, Lam W, et al. Clinicopathological predictors of finding additional inguinal lymph node metastases in penile cancer patients after positive dynamic sentinel node biopsy: a European multicentre evaluation. BJU Int 2022;130:126-32. [Crossref] [PubMed]
12. Nazzani S, Catanzaro M, Biasoni D, et al. Bilateral inguinal lymph-node dissection vs. unilateral inguinal lymph-node dissection and dynamic sentinel node biopsy in clinical N1 squamous cell carcinoma of the penis. Urol Oncol 2023;41:210.e1-8. [Crossref] [PubMed]
13. Maciel CVM, Machado RD, Morini MA, et al. External validation of nomogram to predict inguinal lymph node metastasis in patients with penile cancer and clinically negative lymph nodes. Int Braz J Urol 2019;45:671-8. [Crossref] [PubMed]
14. Zhou X, Zhong Y, Song L, et al. Nomograms to predict the presence and extent of inguinal lymph node metastasis in penile cancer patients with clinically positive lymph nodes. Transl Androl Urol 2020;9:621-8. [Crossref] [PubMed]
15. Nascimento ADMTD, Pinho JD, Júnior AALT, et al. Angiolymphatic invasion and absence of koilocytosis predict lymph node metastasis in penile cancer patients and might justify prophylactic lymphadenectomy. Medicine (Baltimore) 2020;99:e19128. [Crossref] [PubMed]
16. Peak TC, Russell GB, Dutta R, et al. A National Cancer Database-based nomogram to predict lymph node metastasis in penile cancer. BJU Int 2019;123:1005-10. [Crossref] [PubMed]
17. Huang Y, Gao Y, Wu Y, et al. Prognostic value of systemic immune-inflammation index in patients with urologic cancers: a meta-analysis. Cancer Cell Int 2020;20:499. [Crossref] [PubMed]
18. Janicic A, Petrovic M, Zekovic M, et al. Prognostic Significance of Systemic Inflammation Markers in Testicular and Penile Cancer: A Narrative Review of Current Literature. Life (Basel) 2023;13:600. [Crossref] [PubMed]
19. Liu S, Wu W, Du Y, et al. The evolution and heterogeneity of neutrophils in cancers: origins, subsets, functions, orchestrations and clinical applications. Mol Cancer 2023;22:148. [Crossref] [PubMed]
20. Peschek LS, Hobusch GM, Funovics PT, et al. High fibrinogen levels are associated with poor survival in patients with liposarcoma. Sci Rep 2023;13:8608. [Crossref] [PubMed]
21. Xu R, Yang T, Yan B, et al. Pretreatment fibrinogen levels are associated with survival outcome in patients with cancer using immunotherapy as a second-line treatment. Oncol Lett 2023;25:269. [Crossref] [PubMed]