1. Introduction
Ampullary adenocarcinoma (AC) is a rare cancer type, comprising less than 1% of gastrointestinal malignancies [
1]. Surgical resection is the standard of care curative approach for localized cancers, but approximately 40–50% of patients are ineligible due to medical comorbidities or locally advanced disease [
2,
3,
4]. Patients with localized AC who are able to undergo surgery have a one-year OS of greater than 90% [
5,
6]. The outcomes of patients with unresectable, non-metastatic disease are less clear. Although management typically consists of chemotherapy and/or radiotherapy according to institutional best-practices, no guidelines exist as to the optimal regimen, sequencing of therapies, or radiotherapy dosing. The utility of neoadjuvant and adjuvant therapies has been reported [
3,
5,
6,
7,
8,
9,
10] but evidence for the optimal non-operative treatment of localized AC is lacking. Due to the rare nature of AC, prospective clinical trials are challenging to perform, underscoring the need for retrospective data with large patient numbers to identify treatment patterns and generate hypotheses for future studies. The primary objective of this study was to characterize practice patterns and survival rates for patients with AC unsuitable for surgical resection. A secondary objective was to compare the outcomes of non-operative strategies to potentially identify the optimal treatment strategy for this subset of patients. We hypothesized that patients treated with a combination of chemotherapy (CT) and definitive radiation therapy (DRT) would have superior one-year OS compared to those treated with CT alone.
2. Materials and Methods
2.1. Patient Selection
A retrospective review of the National Cancer Database (NCDB) from 2004 to 2017 was performed to identify patients ≥ 18 years old with non-metastatic ampullary adenocarcinoma. The NCDB is a combined effort of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society and represents >70% of cancer diagnoses in the United States from over 1500 healthcare facilities [
11]. The CoC’s NCDB and the hospitals participating in the CoC NCDB are the source of the deidentified data, and they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions presented in this study. Use of the NCDB is not considered human subject research and approval by our institutional review board was deemed not necessary. Clinical, T, N, and M staging are defined using the criteria from the American Joint Committee on Cancer (AJCC) [
12]. Histologic subtypes are classified according to World Health Organization criteria. A summary of exclusion criteria is provided in
Figure 1.
In brief, all primary AC patients with metastatic disease, unknown clinical stage, non-adenocarcinoma histology, lack of follow-up, non-abdominal RT, unknown receipt of CT or RT, unknown RT dose, and non-definitive surgical intervention (e.g., debulking only) were excluded. Patients who underwent primary surgical intervention were included for portions of the analysis. Demographic information, including age, sex, year of diagnosis, distance to care, and race were obtained. An age ≥80 was defined as “older”. Socioeconomic status was estimated by matching the 2012 American Community Survey data against the patient’s home zip code [
13] and stratified by less than or greater than/equal to the median income (
$48,000). Comorbidity information was evaluated with Charlson/Deyo scoring and was recorded as no comorbidities or 1+ comorbidities. Treatment data included receipt of CT and RT, RT dose, number of RT fractions, and RT method (3D, IMRT, etc.).
2.2. Treatment Groups
We defined six treatment groups: no treatment, palliative radiotherapy (PRT) alone, CT alone, CT with PRT, DRT alone, and CT with DRT. PRT was defined as any course less than 45 Gy that was not otherwise classified as stereotactic body radiotherapy (SBRT). All RT regimens categorized as SBRT or those with a prescription of 45 Gy or more were classified as DRT. There was no time requirement between CT and RT for a patient to be included in the CT + PRT/DRT groups, although this association was explored in later analyses.
2.3. Statistical Methods
Logistic regression was utilized to analyze factors that may have been predictive of receiving different forms of treatment. All variables were extracted from NCDB data. Variables included age, gender (M vs. F), comorbidities (Charlson-Deyo 0 vs. 1+), race (white vs. non-white), income (<$48,000 annual income vs. ≥$48,000 annual income), year diagnosed (2004–2010 vs. 2011–2017), clinical stage (I vs. II/III), T-stage (1–2 vs. 3–4), N-stage (N0 vs. N1), and distance from site of treatment (<10.2 mi vs. ≥10.2 mi, 10.2 mi being the median). Logistic regression was performed using all variables to determine the odds of receiving treatment vs. no treatment, palliative treatment (PRT alone, CT alone, CT + PRT) vs. definitive treatment (DRT alone, CT + DRT), and CT vs. CT + DRT. One-year OS differences between groups were evaluated using Kaplan–Meier curves and compared using the log-rank test. Cox proportional hazards models were used to estimate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for 1-year OS. Univariate analysis included age, sex, race, income, comorbidities, distance to care, year of diagnosis, and stage, as previously described. Variables with p < 0.10 on univariate analysis were included in the multivariable Cox model.
We also used a propensity score matched analysis (PSMA) to further control for confounding variables. This analysis included age, year of treatment, clinical stage, sex, race, comorbidities, income, and distance from care. Patients receiving CT alone were matched with those receiving CT + DRT using a 1:1 nearest available neighbor match without replacement using a caliper size calculated as 20% of the standard deviation of the propensity score [
14]. Propensity score distributions were evaluated by computing the standardized difference of the covariates across the two groups. Following PSMA, OS between treatment groups was estimated using the Kaplan–Meier method and the effect of CRT was evaluated with a Cox proportional hazards model. All statistical tests were based on 2-sided probability with the significance level set at
p < 0.05 using RStudio (RStudio, Inc., Boston, MA, USA) and SAS (SAS Institute Inc., Cary, NC, USA).
4. Discussion
This study found that the majority of patients with ampullary carcinoma diagnosed from 2004 to 2017 who did not undergo surgical resection did not receive CT or RT to the primary tumor. The underlying reasons for this trend are not ascertainable in a large, generalized dataset that does not capture all treatment decisions. Uncaptured factors include surgical fitness and willingness to undergo surgery, chemotherapy, or radiation. However, several influential covariates were available and included in the analysis. Of these covariates, year of diagnosis, clinical stage, lymph node status, sex, and age were all predictive of whether or not a patient would receive some form of CT or RT. The finding that most patients receive no treatment is important in light of the additional finding that patients who received some form of treatment were associated with improved survival compared to those who received none. This benefit is potentially tempered due to a selection bias in this population, in which only the most fit patients would be candidates for CT or RT. However, comorbidity status was included in this analysis and was not found to be predictive of survival.
This study also emphasizes the poor prognosis of patients who are unable to undergo surgical resection and adds to the sparse body of knowledge regarding non-operative management of AC. The ABC trial examined the role of cisplatin-gemcitabine vs. gemcitabine in advanced biliary tract carcinoma and included 20 patients with AC [
15]. A cohort of AC tumors combined with bile duct tumors demonstrated a poor overall response rate to chemotherapy (<20%). No significant difference was found between chemotherapy groups for AC (HR 0.62, 95% CI 0.21–1.82). Rostain et al. studied patients with AC diagnosed from 1976 to 2009; specific non-operative treatment strategies were not described, but 1-year OS was reported to be 26.5% [
3].
Based on our analysis, approximately 20% of patients with non-metastatic AC underwent primary non-operative therapy. As a combined cohort, these patients had worse 1-year OS compared to the surgical patients (41.8% vs. 84.3%,
p < 0.001) (
Figure 4B). These numbers are comparable to a population-based analysis from the Netherlands, which demonstrated 1-year OS for non-metastatic AC as approximately 80% with resection and 40% without resection [
4]. In that cohort, nonsurgical patients receiving chemotherapy and/or radiotherapy comprised <1% of all non-metastatic patients, compared to 5.8% of our cohort.
In terms of survival, patients treated with CT + DRT did not have improved one-year OS compared to those with CT alone. This observation remained consistent after PSMA. The LAP07 trial demonstrated a similar result [
16], in which patients with unresectable pancreatic cancer, an anatomically similar but different prognosis malignancy, were randomized to receive concurrent chemoradiation or chemotherapy alone. There was no difference in median OS between groups, though there was decreased local progression in the chemoradiation group. However, in the present study, a third group was also analyzed, DRT alone. When DRT was compared to CT alone and CT + DRT, there was no difference in 1-year OS. This may suggest that patients with unresectable AC may gain a similar benefit from either CT or DRT alone. Similar outcomes across multiple treatment strategies support the notion that treatment decisions for rare malignancies should be made in a multidisciplinary setting so that management can be personalized for each patient. For example, a patient who is not medically fit for surgery or chemotherapy but is willing to come for daily RT should be given that option with a full understanding that her chances of survival are not worse, based on our findings, because of the inability to receive chemotherapy. In another situation, a patient living far from medical care may not want to pursue radiotherapy, as it could be a large financial and time burden, so chemotherapy alone may be a better option. The relatively similar OS among all three treatment modalities offers multiple options for the clinician to personalize care for the patient without compromising outcomes.
This study has several limitations. As previously mentioned, the NCDB does not include sufficient detail to ascertain why patients receive the prescribed treatment modalities. This was mitigated in part through the utilization of multivariate regression. In addition, we performed PSMA for the two largest groups of patients who received treatment. This analysis confirmed a lack of statistical difference in OS between these two groups. Secondly, the NCDB does not have cause-specific survival data, so our analysis of survival is limited to OS. However, in the setting of an aggressive cancer with a relatively poor prognosis, the difference between CSS and OS is not as distinct as for malignancies with a better prognosis. Third, retrospective data is fraught with various forms of bias that cannot be corrected without a prospective, randomized trial. Due to the rare nature of AC, prospective trials are not likely to accrue well, and we must draw conclusions based on the data that is available. Fourth, although not a primary focus of this study, the extent of surgery (Whipple, ampullectomy, etc.) is not available in the NCDB, which somewhat limits the direct comparison of these data to any specific surgical procedure. Quality of life measures are other important factors not captured in the NCDB and are missing from this study. Lastly, there is significant heterogeneity among these treatments that is not adequately captured. For example, the type, dosing and frequency of CT is not reported in the NCDB, but survival may vary between approaches. As general treatment paradigms become more established, single- or multi-institutional studies will indicate optimal CT and RT approaches.