Lymph Node Metastatic Patterns and Survival Predictors Based on Tumor Size in Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal malignancies. Larger tumor size is widely acknowledged to be associated with increased lymph node (LN) metastatic potential. However, the quantitative relationships between tumor size and LN metastasis or survival remain unclear. This study aims to quantify the objective relationship between tumor size and the prevalence of LN metastases across a spectrum primary tumor size. The Surveillance, Epidemiology, and End Results (SEER) database was used to identify 9958 patients with resected PDAC without distant metastasis. The prevalence of LN metastases, LN ratio (LNR), and N2/N1 ratio were assessed amongst different tumor sizes, and the relationships were displayed by matched curves. In the enrolled cohort, age, tumor site, grade, American Joint Committee on Cancer (AJCC) 8th node staging, tumor size, chemotherapy, and radiotherapy were identified as significant independent predictors for overall survival (OS) and cancer-specific survival (CSS). For tumors within 1–40 mm in size, the prevalence of node-positive disease is closely modelled using a logarithmic formula [0.249 × ln (size) + 0.452] × 100%. The prevalence plateaued between 70% and 80% beyond 40 mm. The mean LNR increased in a stepwise manner as tumor size increased from 1–5 mm (LNR = 0.024) to 41–45 mm (LNR = 0.177); then, beyond 45 mm, it plateaued near 0.170. N2/N1 ratio gradually increased along with tumor size from 1–5 mm (N2/N1 = 0.286) to 41–45 mm (N2/N1 = 1.016), and when tumor size reached to 41–45 mm or more, the ratio stabilized around 1.000. In addition, significant survival prediction by AJCC N staging was observed when tumors ranging between 16 and 45 mm in size. Regional LN involvement demonstrated a logarithmic growth with increasing tumor sizes in patients with resected PDAC . The probability of metastasis in each regional LN for resected PDAC with tumors greater than 40 mm in size was near 17.0% and their overall prevalence of LN metastasis was 70–80%. Among which, 50% of patients had an N2 stage. Such prediction may be a potential and promising tool for guiding lymphadenectomy in PDAC surgery.

Results: In the enrolled cohort, age, tumor site, grade, American Joint Committee on Cancer (AJCC) 8th node staging, tumor size, chemotherapy, and radiotherapy were identified as significant independent predictors for overall survival (OS) and cancer-specific survival (CSS). For tumors within 1-40 mm in size, the prevalence of node-positive disease is closely modelled using a logarithmic formula [0.249 9 ln (size) ? 0.452] 9 100%. The prevalence plateaued between 70% and 80% beyond 40 mm. The mean LNR increased in a stepwise manner as tumor size increased from 1-5 mm (LNR = 0.024) to 41-45 mm (LNR = 0.177); then, beyond 45 mm, it plateaued near 0.170. N2/N1 ratio gradually increased along with tumor size from 1-5 mm (N2/N1 = 0.286) to 41-45 mm (N2/N1 = 1.016), and when tumor size reached to 41-45 mm or more, the ratio stabilized around 1.000. In addition, significant survival prediction by AJCC N staging was observed when tumors ranging between 16 and 45 mm in size. Conclusion: Regional LN involvement demonstrated a logarithmic growth with increasing tumor sizes in patients with resected PDAC . The probability of metastasis in each regional LN for resected PDAC with tumors greater than 40 mm in size was near 17.0% and their overall prevalence of LN metastasis was 70-80%. Among which, 50% of patients had an N2 stage. Such prediction may be a potential and promising tool for guiding lymphadenectomy in PDAC surgery.

INTRODUCTION
Pancreatic cancer is one of the chief causes of cancer-related deaths worldwide, with a 5-year survival rate of only 10%. The cancer statistics report in 2021 estimated 60,430 new cases and 48,220 cancer-related deaths due to this malignancy [1,2]. For appropriate patients with resectable pancreatic ductal adenocarcinoma (PDAC), oncologic resection of the primary tumor and regional lymph nodes (LNs) remains the standard for potential cure [3,4].
Traditionally in resected PDAC, increasing tumor sizes are generally understood to be a worse prognostic marker, which is directly reflected in current American Joint Committee on Cancer (AJCC) staging guidelines [5,6]. This guideline conforms to the classic model of tumor progression: as the tumor develops, tumor cells obtain additional mutations and finally acquire the capability to metastasize to regional LNs and distant sites [7,8]. So far, only one report in breast cancer has elucidated a non-linear correlation between the prevalence of LN metastases (percentage node-positive) and increasing tumor size [9]. However, there is no quantitative statement about the relationships between tumor size and LN metastases in PDAC.
PDAC with LN metastasis are known to have higher risk of recurrence and shorter survival [3,[10][11][12]. In general, examined LN (ELN) number and LN ratio (LNR) are considered to be fundamental metrics for quality assessment and prognostic stratification in cancer care [10,13]. Given the crucial nature of N staging in PDAC, the College of American Pathologists recommends that a minimum of 12 LNs should be examined at the time of pancreaticoduodenectomy. For accurate nodal staging, current guidelines recommend a minimum number of ELNs (minELN) that ranges between 12 and 17 for a pancreatic head tumor and at least 20 for a body and tail carcinoma [14][15][16]. However, these recommended minELN numbers for accurate staging vary extensively across guidelines, and the optimal examined number to robustly stratify survival has not yet been formulated. Collectively, the undersampling of LNs remains a potential error that may impact the prognostic accuracy of the current staging paradigm. Thus, a quantitative objective standard is needed to assess LN metastatic ability and assist pathologists with an impression of the presence of LN involvement and tumor metastatic essence.
Tumor size is a crucial fundamental indicator of tumor burden and mortality. In breast cancer, an incremental amount of evidence reveals that, rather than the timing of diagnosis within the clinical window, the risk of metastasis is to a great extent determined by intrinsic biology [9]. Marchegiani et al. [17] reported that in PDAC, larger tumors are significantly associated with more positive LNs, higher LNRs, and worse survival. Therefore, it is interesting to explore the relationship between tumor size and the prevalence of LN metastases with high resolution to provide insight into the biological behavior of tumor growth and its impact on LN metastasis.
To quantitatively identify the objective relationship between the prevalence of LN metastases and tumor size across a size range of 1-100 mm, a large population-based cohort of patients with PDAC without distant metastasis undergoing resection from the Surveillance, Epidemiology and End Results (SEER) database was employed. Stratified by tumor sizes, the prevalence of LN metastases, LNR, and N2/N1 ratio were plotted, and survival predictors were analyzed.

Patients and Data Source
All patients analyzed in our study were collected from SEER database portal (https://seer.cancer. gov/). The up-to-date database, SEER*Stat Database: Incidence-SEER 18 Regs Custom Data (with additional treatment fields), Nov 2018 Sub (1975-2016 varying), was used. Access was approved by SEER database for research, and all data were freely available.
Patients with PDAC diagnosed between 2004 and 2016 without distant metastasis were included. In addition, all patients with neoadjuvant treatments, unknown tumor size, examined/metastatic lymph node number, or distant metastatic information were excluded. Next, only patients with surgical resection were included. Given the fundamental nature of T and N staging in PDAC, the College of American Pathologists recommends that a minimum of 12 LNs [18] should be examined at the time of pancreaticoduodenectomy and since tumor size greater than 10 cm is rare, only patients with 12-50 (including 12 and 50) examined LNs, and no more than 10 cm in tumor size were analyzed. Information regarding age, gender, grade, primary site, tumor size, number of examined LNs and positive LNs, chemotherapy, radiotherapy, and survival information was also retrieved from the database.
According to AJCC 8th edition of N staging, no regional positive LN was defined as N0, 1-3 regional positive LNs was defined as N1, and at least 4 positive LNs as N2. The AJCC 8th M1 stage was evaluated according to the following codes: derived AJCC M, 7th ed Overall survival (OS) was defined as the time interval between surgery and death or last follow-up, and the time interval between surgery and cancer-related death or last follow-up was defined as cancer-specific survival (CSS). The LNR was calculated as the number of positive LNs divided by the number of examined LNs. The prevalence of LN metastases was defined as the proportion of cases with at least one positive regional metastatic LN (N1 ? N2) among total included cases (N0 ? N1 ? N2). The protocol of this study was approved by the Ethics Committee of Zhongshan Hospital, Fudan University.

Statistical Analysis
In this study, statistical analysis was performed using SPSS version 21.0 (IBM Corporation, Armonk, NY, USA). Continuous variables were presented as medians with interquartile range (IQR). The correlation between tumor size and the prevalence of LN metastases was assessed using a logarithmic regression. Cox proportional hazards regression models were used for the univariable and multivariable analyses to identify independent risk factors for OS and CSS, but ''unknown'' and ''others'' variables were not included in the analysis. Statistical significance was defined as P \ 0.05.

Clinicopathological Characteristics
A total of 9958 patients with histologically confirmed resected PDAC without distant metastasis were identified after applying the aforementioned inclusion and exclusion criteria (Fig. 1). The median age was 67 (ranging from 17 to 95) years, and female patients accounted for 50.4% of the population. Lesions located in the head of the pancreas were almost ten times more common than those in the body or tail. On the basis of the pathological examinations, approximately 9%, 49%, and 34% of patients with PDAC were confirmed as well, moderately, and poorly or undifferentiated, respectively. The median number of examined and positive LNs was 20.8 (IQR 15.0-25.0) and 3.0 (IQR 0-4.0), respectively. N0 stage was found in 2891 patients while 3985 were staged as N1 and 3082 were staged as N2. In addition, the median tumor diameter of the total cohort was 31.0 mm (IQR 25.0-40.0 mm), and the population of each 5-mm segment is displayed in Table 1. Radiotherapy was administered in 34.1% of patients while 72.3% received chemotherapy.
The median OS of this cohort was 21.0 months (IQR 11.0-43.0 months), and the cumulative 5-year OS rate was 17.9%. The median CSS was 23.0 months (IQR 12.0--49.0 months), and the cumulative 5-year CSS rate was 21.2%.

Correlations Between Tumor Sizes and Prevalence of Lymph Node Metastases
The prevalence of LN metastases in this cohort was 71%. There was a non-linear relationship between increasing tumor sizes and the prevalence of LN metastases across the tumor size spectrum (Fig. 2, red line). Patients with tumors between 1 and 5 mm in size had the lowest prevalence of LN metastases (less than 15%).
Alternatively, the prevalence increased in a stepwise fashion as tumor size increased from 6-10 mm (40%) to 36-40 mm (76%); however, beyond 40 mm, the prevalence of LN metastases plateaued between 70% and 80%. Observing the tracing pattern of the initial part of the curve (1-40 mm), which conformed to non-linear correlation, we used a logarithmic regression to match this non-linear curve. The established curve was highly matched (Fig. S1), and its formula was Correlation coefficient R 2 = 0.991 (Y, prevalence of LN metastases; X, tumor size in centimeters).
From Fig. 2, we could find that the prevalence of LN metastases was limited within 80% no matter the size of tumors. The probability of patients with PDAC with tumors within 1-40 mm in size having positive LN metastasis was well matched to the formulated matched curve. Beyond 40 mm, the probability of LN metastasis was steadily within 70-80%.

Patterns Between Tumor Sizes and Lymph Node Metastatic Status
LNR measures the metastatic ability of each examined LN. The global LNR of the total cohort was 0.145 (IQR 0-0.214), and in patients with at least one positive LN metastasis, it was 0.205 (IQR 0.077-0.278). Following the tracing pattern of average LNR across the tumor size spectrum, the average LNR increased stepwise as tumor size rose from 1-5 mm (LNR = 0.024) to 41-45 mm (LNR = 0.177); then, beyond 45 mm, it too plateaued near 0.170 (Fig. 2, blue line). Thus, when tumor size reached 41-45 mm or more, the probability of metastasis in each regional LN was steadily near 17.0%.
As expected, patients with a higher N stage conferred a higher metastatic presence. The number of stage N2 patients/the number of stage N1 patients (N2/N1) represents the distribution of patients with different LN metastatic severity. The tracing pattern of N2/N1 was similar to the curve for the average LNR. N2/N1 ratio gradually increased along with greater tumor size from 1-5 mm (0.286) to 41-45 mm (1.016), and when tumor size reached 41-45 mm or more, the ratio stably wavered around 1.000 except for size group 61-65 mm (Fig. 2, yellow line). Therefore, nearly half of LN-positive patients may have N2 stage disease when tumor size reaches 40 mm.

Prognostic Significance of AJCC 8th N Staging in Different Tumor Sizes
When considering the prognostic significance of AJCC 8th N staging on OS, there was no significant difference in patients with tumor between 1 and 5 mm in size. Additionally, no significant difference was found between stage N2 and N1 when tumor size was from 6 to 15 mm. However, once tumor sizes reached the range of 16-45 mm, the survival curves could be stratified significantly on the basis of N stage. Again, when tumor size ranged between 46 and   (Table 2). Similarly, in the analysis for CSS, significant differences and discriminations by N stage were found in tumor size of 16-45 mm, similar to OS. The other detailed descriptions when tumor size was less than 16 mm or more than 45 mm are displayed in Table 2. Overall, significant stratification and discrimination was only achieved in tumors ranging between 16 and 45 mm in size. Furthermore, other detailed prognostic indicators, such as age, gender, grade, chemotherapy, and radiotherapy in different tumor sizes, were analyzed and displayed in Supplementary Tables 1-3.

DISCUSSION
In this retrospective population-based analysis, 9958 patients with resected PDAC and without distant metastasis were analyzed to investigate the impact of tumor size on LN metastatic status and survival. As expected, tumor size and the AJCC 8th N staging were both independent predictors of OS and CSS for patients with resected PDAC without distant metastasis. Both tumor size and LN status have been integral parts for staging according to the AJCC guidelines. A retrospective study of 6145 patients with PDAC by Park et al. [19] sought to confirm whether the AJCC 7th TNM staging system accurately represented tumor size versus extrapancreatic extension, and demonstrated that tumor size was more determinant of prognosis than extrapancreatic extension. A large number of other studies also confirmed that tumor size was an independent risk factor in patients with PDAC [6,17,20]. As a result, the most recent AJCC 8th TNM staging guidelines, as of January 1, 2018, solely utilizes tumor size in its T stage classification and no longer considers extrapancreatic extension [21]. In addition, our previous study addressed the controversy of the current AJCC staging [22] and showed that a modified stage IIIA (T1-2N2M0) had a significantly more prolonged survival than a modified stage IIIB (T3N2M0, T4NxM0). This is similar to a report that N2 patients had improved survival rates when compared to T4 patients [23]. Thus, tumor size may have a wider effect on survival stratification than LN metastasis, and the predictive power of N assessment may be limited, just like the regional LN metastatic ability may Fig. 2 Relationships depicted by trend charts between tumor sizes and prevalence of lymph node metastases, average lymph node ratio, and N2/N1 ratio  [7] demonstrated that LN involvement significantly interacts with tumor T stage, thus indicating that a relationship may exist between them. Interestingly, our study supports the existence of trends and where a nonlinear relationship exists in tumor sizes from 1 to 40 mm. We found a stable correlation (70-80%) once beyond 40 mm. In fact, the nonlinear variation between 1 and 40 mm was almost perfectly matched by a formulated logarithmic regression equation. However, such a plateaued curve might be partly explained by the fact that the large tumors with less metastatic potential (and/or less aggressive invasion) would be resected selectively. Additionally, we found that the probability of metastasis in each regional LN steadily plateaued near 17%, and the stage N2/N1 ratio was nearly 50% in node-positive patients when tumor size reached 40 mm. Therefore, the general prevalence of LN metastases can be predicted by radiologists via tumor size measurements in the preoperative setting. Surgeons and pathologists may then be more confident about the likelihood of LN metastasis in pathological specimens.
Ma et al. [24] and Kassardjian et al. [25] reported that preoperative contrast-enhanced with resectable PDAC before surgery, a bias of about 5 mm between imaging modalities and gross examinations should be considered. In our study, a tumor size of 40 mm is the potential demarcation point for the prevalence of LN metastases, the probability of metastasis in each regional LN, and the N2/N1 ratio in node-positive patients. Takahashi et al. [26] determined that a tumor size of 20 mm as the ideal cutoff and found that patients with tumors larger than 20 mm were more likely to have positive LNs. This was partially consistent with our findings that the prevalence of LN metastases, the probability of metastasis in each regional LN, and the N2/N1 ratio in node-positive patients for tumor sizes greater than 20 mm. Meanwhile, international validations have proved that the AJCC 8th TNM staging guideline demonstrates a modestly improved prognostic accuracy with more equal distribution among stages [27,28]. This AJCC new staging system formulates a tumor size cutoff of 40 mm for T staging. Cong et al. [29] confirmed that the pT classification in the 8th edition is significantly superior to that in the 7th edition when stratifying patients by OS as compared to N staging. This indicates that a tumor size of 40 mm may play an important role in tumor progression. Petermann et al. [30] tried to illustrate the impact of tumor sizes on LN metastasis in a more limited dataset (n = 114), but showed that in patients with tumor sizes of at most 20 mm, the prevalence of LN metastases is 41%. This is consistent with our large population-based study and formulated equation.
Furthermore, we found that the AJCC 8th N staging may provide a significant survival prediction and risk stratification in patients with tumor sizes from 16 to 45 mm. Researchers reported that very small primary tumors might be correlated with decreased OS among LNpositive patients with PDAC, which raises the possibility that small tumors are more capable of LN metastasis and hence may represent more biologically aggressive cancers [7]. However, in patients with resected PDAC, the prevalence of LN metastases, LNR, and N2/N1 ratio were relatively stable as tumor size moved beyond 40 mm. This could explain the reason why a lack of significance existed in survival stratification. It was reported that the correlation between tumor size and survival was linear for patients with localized tumors, but stochastic in patients with regional and distant stages. Increased tumor size was associated with worse CSS among patients with resected tumors, rather than that in patients with unresected tumors [6].
There were certain limitations in our study. First, this is a retrospective study with limited information about patients' clinicopathological characteristics from the SEER database. Only the resected cases without distant metastases were included in this study, which would increase the selection bias to some extent. The prevalence of LN metastases, LNR, and N2/N1 ratio of large tumors might be underestimated owing to selectively resection of large tumors with low metastatic potential. Thus, a prospective, international, and large-scale cohort containing cases with distant metastases for further validation would be needed. Second, the number of ELNs may be influenced by other confounders not available in the data provided. These may include the patient's immune status, body mass index, surgical standards, tumor biological behaviors, difficulties in separating individual LNs in dissected specimens, and evaluators' expertise.

CONCLUSION
This study is the first to formulate the associations between tumor sizes and prevalence of LN metastases in PDAC using real-world cohorts with generalizable and representative results by robust statistics. The relationship curves suggest that the probability of metastasis in each regional LN for resected PDAC cases with tumors greater than 40 mm in size is near 17.0% and their prevalence of LN metastasis is 70-80%. Among which, 50% of patients may have a higher stage (stage N2). Such prediction may be a potential and promising tool for guiding lymphadenectomy in PDAC surgery.