Liver Metastases in Newly Diagnosed Pancreatic Ductal Adenocarcinoma: A Population Based Analysis

Background Estimates of the incidence and prognosis of developing liver metastases at the pancreatic ductal adenocarcinoma (PDAC) diagnosis are lacking. Methods In this study, we analyzed the association of liver metastases and the PDAC patients outcome. The risk factors associated with liver metastases in PDAC patients were analyzed using multivariable logistic regression analysis. The overall survival (OS) was estimated using Kaplan-Meier curves and log-rank test. Cox regression was performed to identify factors associated with OS.


Background
In 2017, an estimated 55,400 new pancreatic cancer cases were diagnosed in the United States, and 44,330 deaths from pancreatic cancer were reported [1]. Pancreatic ductal adenocarcinoma (PDAC) accounts for up to 95% of pancreatic cancers. PDAC is predicted to become the second leading cause of cancer-related death in the developed world by 2030 because of its poor overall survival (OS) [2]. PDAC development is associated with few or non-specific symptoms, resulting in a large proportion of patients diagnosed at a late stage with metastatic disease [3]. Primary PDAC located in the body or tail of the pancreas is associated with a poorer prognosis because of its late-stage presentation; however, sociodemographic and clinical predictors of outcome have not been well characterized. The OS of PDAC patients with metastatic disease is considerably poorer than that of patients with localized disease, and the 5-year survival rate for metastatic PDAC is approximately 2% [4].
The sites of PDAC metastases include the lung, brain, and most frequently the liver [5,6,7]. There are currently few epidemiological studies reporting on the rate of metastasis in PDAC [8,9]. Data obtained from multicenter studies have yielded varying results of the association between the primary PDAC location and liver metastases (LM) development [10,11,12]. In these few studies analyzed the features and risk factors of LM from PDAC [13,14], the detailed characteristics and estimates of the prognosis of patients with newly diagnosed PDAC and LM are lacking, possibly because of the limited number of clinical cases reported.
The purpose of this study was to characterize the incidence rate and the associated risk factors of LM at the time of PDAC diagnosis on a population-based level using the Surveillance, Epidemiology, and End Results (SEER) database. In addition, survival estimates were quantified, and clinical and sociodemographic predictors of poor survival were examined in patients with PDAC and LM present at the time of PDAC diagnosis.

SEER data
The SEER Program is a coordinated system of population-based cancer registries that collect incidence and survival data on cases reported from 18  autopsy or death certificate, as well as patients of unknown race/ethnicity were also excluded. In total, 42,834 patients were included in the cohort for incidence analysis.
Patients without LM and patients with missing or incomplete information concerning survival were subsequently removed, leaving 16,207 patients eligible for survival analysis.

Statistical analysis
The patients' demographic and tumor characteristics were summarized using descriptive statistics. Univariable and multivariable logistic regression was used to determine whether variables were associated with the presence of LM at diagnosis. Chi-square test was used to analyze the association between the variables and incidence of presenting LM. Survival curves were obtained using the Kaplan-Meier method, and differences between the curves were assessed using the log-rank test. Multivariable Cox regression was performed to identify covariates associated with decreased OS using the same variables used in the logistic regression model described herein. The hazard ratios (HRs) with corresponding 95% confidence intervals (CIs) were used to show the effect of risk factors on OS.
Observations were censored if patients were alive at the time of the last follow-up. All statistical analyses were performed using SPSS 23.0 (IBM, Armonk, NY, USA). Statistical significance was set at p < 0.05.

Patient characteristics
The cohort consisted of 42,834 patients with primary PDAC at a specific tumor site in the head, body, or tail of the pancreas. Table 1 shows the baseline characteristics of the patients including sex, race/ethnicity, age at diagnosis, insurance status, marital status, income, primary PDAC location, TNM stage, metastases to liver, and primary PDAC tumor size. Of these patients, 21,302 (49.10%) had M1 stage, and 16,207 (37.84%) had developed LM at the time of diagnosis. LM was present in the majority (77.07%) of PDAC patients with distant metastases. A slightly greater number of patients were men, and most diagnoses were among Non-Hispanic Whites (NHW). The most prevalent age at diagnosis in the cohort was 60-79 years, and most patients were insured. More than half of the patients (54.03%) were married at the time of diagnosis. The proportion of primary PDAC location was in the head (65.02%), body (17.25%), and tail (17.73%) of the pancreas, respectively. The detailed characteristics regarding TNM stage are presented in Table 1. of the pancreas had a significantly higher LM incidence than did the cohort with tumors in the head (28.57%) of the pancreas (p < 0.01) (Fig. 1C). There was a significantly higher incidence rate of LM in the cohort with primary PDAC tumor size larger than 4 cm (44.45%) or unknown primary tumor size (50.18%) than primary tumor size 2.1-4 cm (24.99%) or tumor size < 2 cm (23.86%) (p < 0.01) (Fig. 1D).
Univariate analysis (   Figure 2 shows the OS estimates ( Fig. 2A), the OS estimates stratified by primary tumor location (Fig. 2B), and the OS estimates stratified by extent of identified LM at diagnosis (Fig. 2C). The median survival did not differ significantly according to tumor location among the three primary PDAC-with-LM groups (Fig. 2D).

Prognostic factors for OS in PDAC patients with LM
On multivariable Cox regression analysis (

Discussion
The poor survival of patients with PDAC is largely due to the absence of methods for early diagnosis, and the development of metastases [15]. In cases of aggressive PDAC, metastasis is frequently present at the time of primary diagnosis, occurring in the peritoneal cavity, pleura, lung, and bone, although the most common site of metastases is the liver [16]. Improving our knowledge of the incidence and survival characteristics of PDAC patients with LM, as well as the relative risk and prognostic factors, is critical for the diagnosis and treatment planning for PDAC patients. However, few epidemiological studies have been conducted on the features of metastases in PDAC to date [8,9]. The present US-population based study provides an unbiased and detailed analysis of the incidence of LM among patients with newly diagnosed PDAC and the subsequent survival of such patients.
To the best of our knowledge, the present study is currently the largest report on such issues to use population-based data sets. We found that the risk factors for LM occurrence at PDAC diagnosis were male gender, NHB (vs. NHW) race/ethnicity, younger age, primary PDAC located at the tail or body (vs. head) of the pancreas, and larger tumor size. Among the entire cohort, male (vs. female) patients, NHB (vs. NHW) race/ethnicity, uninsured (vs. insured) status, single (vs. married) status, lower (vs. higher) income, larger (vs. smaller) tumor size and primary PDAC located at the tail or body (vs. head) of the pancreas were associated with a significantly greater HR of worse OS; however, the association of NHB race/ethnicity with poorer OS or the association of single status with poorer OS was not observed in the cohort of PDAC patients with LM. The HR of OS was not associated with poorer T stage or N stage at diagnosis as defined by the 7th edition of AJCC Cancer Staging system in all PDAC patients cohort or PDAC patients with LM sub-cohort. These data are not in agreement with the conclusions reported by other studies regarding general OS trends [17].
The newly published French clinical practice guidelines for the diagnosis, treatment, and follow-up of pancreatic cancer indicated that clinical presentation depends mainly on the primary PDAC location and the stage at diagnosis [18]. Fewer than 10% of patients diagnosed with PDAC are potentially curable through resection, and patients presenting with metastatic disease at diagnosis are not suitable for surgical resection [19].
Multidetector Computed Tomography (MDCT) is normally performed using thin sections and intravenous contrast injection after the diagnosis of PDAC. It is used for the evaluation of tumor resectablility and metastases. Another sensitive and specific alternative imaging modality is diffusion-weighted (DW) magnetic resonance imaging (-MRI), which allows for the characterization of barely visible isoattenuating PDAC and small or indeterminate LM on MDCT. Performing liver DW-MRI before laparotomy is useful for excluding infra-centimetric LM and by subsequently avoiding futile surgery [20].

Conclusions
The present study provides insight into the epidemiology of LM in patients with newly