Effect of metastatic site on survival in patients with neuroendocrine neoplasms (NENs). An analysis of SEER data from 2010 to 2014

Neuroendocrine neoplasms (NENs) are rare malignancies of the aerodigestive, genitourinary and integumentary systems. Their histologies vary from well-to-moderately differentiated neuroendocrine tumors (NETs) to poorly differentiated neuroendocrine carcinomas (NECs) and their natural history has been described in several publications [1‐3]. Most studies are limited due to the small number of cases, inconsistent follow-up or retrospective nature but it is clear, however, that the incidence of NENs is increasing [3] and that, at least for certain subtypes, survival might be improving [2].

NENs show a spectrum of behaviors and this makes their treatment challenging. Some exhibit an indolent, slow growth pattern, while others parallel the more aggressive, rapidly spreading tumors such as small cell lung cancer (SCLC); in between there are neoplasms of intermediate malignant potential. Research so far has identified stage, site of origin [4] and differentiation [5 6], as well as proliferative indices (Ki-67, mitotic count) as important prognostic factors and multiple scores have been published, trying to predict survival in metastatic disease or recurrence after curative surgery [5‐11]. In general, well- differentiated tumors progress slowly and surveillance may be the best approach in some cases, whereas poorly differentiated neoplasms require urgent aggressive chemotherapy and are associated with markedly shorter survivals [12]. Tumors of small bowel origin tend to have a better prognosis [13] compared to NENs originating in the pancreas. The effect of other factors such as age, race [3], resectability [14], performance status [15] or even marital status [16] has similarly been examined in several publications. Most medical decisions nowadays consider tumor of origin, staging, but also tumor differentiation and mitotic indices (values that have formed the basis of the current grading system [17]).

While it is generally accepted that stage IV (presence of metastasis) portends a poor prognosis for most neoplasms including NENs, there is no consensus on the gravity and importance of metastatic sites, or how they interplay with the primary tumor site when it comes to survival estimates. The National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program is an annually updated population data bank which has been used extensively to monitor long term trends in survival trends for rare and common tumors alike. The November 2017 iteration of the program covers 28% of the US population and includes 10,050,814 cases. A variety of publications have utilized the power of SEER data to draw conclusions about NENs [18, 19]. It has been shown for example that the incidence of certain types of NETs has increased and that the survival of patients has improved over time [2]. This has been partially attributed to treatments such as somatostatin analogues [20] (time to progression prolonged by 8 months), targeted therapies such as everolimus [21] and sunitinib [22] (progression free survival benefit of about 5 months for both), and hopefully pazopanib or peptide receptor radionuclide therapy (PRRT) [23] in the future. With the latest iteration, the SEER database was enriched to include details of general metastatic sites, including lung, liver, bone and brain. This presents a unique opportunity to study the behavior of metastatic neuroendocrine tumors across a range of sites and histologies. We sought to explore the behavior of NENs with regards to the site of origin and metastatic areas and hypothesized that site of metastasis will carry different prognostic significance depending on tumor grade and tissue of origin.

Neuroendocrine neoplasms are rare and heterogeneous, but their incidence is rising, their evolution can span multiple years [1 5] and new treatments have been approved in the past decade. In order to have meaningful discussions about prognosis and properly design clinical trials, clinicians need to better understand the factors affecting patient survival. In this paper we have used the SEER database to examine metastatic only patients and showed that both the site of origin and the metastatic site independently influences their survival. We have further attempted to quantify that and provide a prognostic score for use by the clinician in everyday practice.

The varied behavior of patients based on tumor histology and grade is not new knowledge. Multiple publications have shown that, on average, patients with small bowel neuroendocrine tumors fare better compared to pancreatic neuroendocrine tumors (PanNENs) and that higher grade is associated with decreased survival. Yao et al. [3] examined more than 35,000 SEER NEN cases from 1973 to 2004 and showed inferior survival for patients with pancreatic or lung NENs compared to those of the small intestine, something that we have also shown here. More recent publications have shown slightly improved but similar trends between histologies [24] and better survival. Similar to our findings, poorly differentiated tumors had lower median survivals compared to the well differentiated counterparts, reflecting the virulent biology and the limited advances in the treatment of high grade NENs in the past 20 years. Our study adds to this knowledge with detailed estimates for tumors of varied origins and metastatic sites, as well as with a prognostic score for better stratification of the patients. This score emphasizes the importance of metastatic location and shows that brain and liver involvement carry a higher risk of death regardless of other well-known factors, such as origin of the tumor or histology.

It is important to mention that we relied on a uniform reporting style for all origins, including lung neuroendocrine tumors. This has some inherent problems, as classification for lung NENs is different (typical and atypical carcinoids, small and large cell neuroendocrine categories). The WHO classification that designated the typical and atypical categories was developed for resected primary carcinoids, so its usefulness in the metastatic setting is not clear. Moreover, small cell lung cancer, with its well described histology and extensive treatment options, was removed from our database. Our approach in this paper is similar to other large SEER studies [3 24] but can definitely alter the reported results. The GI NEN literature is moving to a more uniform grading of neoplasms and there is a trend to use proliferation markers to predict outcomes in lung NENs as well [25, 26] but this is far from a settled argument. Another issue is that classification of pancreatic NENs (and most recently other GI NENs) has changed to include a category of “well differentiated, high grade” with a prognosis intermediate between G2 and G3. We have approached this problem by factoring in our model both typical/atypical/pancreatic histology and G1/G2/G3 categories but acknowledge that it can cause problems with interpretation of data.

Our results are consistent with the published literature. A study in two major institutions [27] in carcinoid patients with bone and/or liver metastases confirmed that coexistence of bone and liver metastasis was indicative of worse prognosis with a significant difference in overall survival - the small total numbers of patients (691) might have made it more difficult to obtain statistical significance in other histologies. The poor prognosis of brain metastases is not unexpected and is consistent across different tumor types, such as breast [28, 29], lung [30], gastrointestinal [31] and ovarian [32]. It is usually measured in matters of months and may reflect the virulence of the tumor, the severity of CNS dysfunction or the inability of most current therapies to cross the blood brain barrier. In our study, median survival was a mere 7 months and it is worth mentioning that single site brain metastases were more prominent in lung and “other” primary but uncommon in pancreas and small bowel (Appendix 2) and carried the same poor prognosis. Although not analyzed in this study, the median survival of patients with multiple metastatic sites (594 cases identified in our cohort) was expectedly diminished with a median OS of 6 months (95% CI: 6–6.96). The rate of skeletal metastases was relatively low at 7% (compared with 10% in published literature) and conversely, the rate of brain metastasis was high (6%, compared with 0.5% in non SCLC NENs and about 20% in SCLC) [33]. One needs to remember that this is an enriched database, consisting of only M1 patients with single site metastasis and with exclusion of SCLC. In prior iterations of this study, the database included 34,704 unselected patients with any metastatic entry (M0/M1) and there the metastatic sites included brain (2.99%), bone (4.63%), liver (14.82%) and lung (4.10%); still higher, but closer to, actual reported registry data.

The effect of tumor of origin in patients with similar metastases is a poorly understood phenomenon and can be approached by the significantly different genetic makeup of these neoplasias. Genetic syndromes such as multiple endocrine neoplasia type 1 (MEN1), Von Hippel-Lindau (VHL) and neurofibromatosis type 1 (NF1) have been associated with NENs, but they only account for about 10% of observed cases. PanNENs have demonstrated a variety of alterations [34] including inactivation of TSC1/2 and ATRX/DAXX genes and involve alterations in DNA damage repair, chromatin remodeling, mTOR signaling and telomere maintenance [35]. In contrast, whole-exome sequencing on small intestinal NENs (SI-NENs) has shown pretty low mutation rates [36] and it is felt that epigenetic processes such as DNA methylation or histone modifications might be more important in tumor propagation and metastasis. Pulmonary carcinoids share a lower rate of mutations compared to their adenocarcinoma counterparts (including very low rates of TP53 and RB1 gene mutations), but have frequent mutations in chromatin-remodeling genes [37, 38]. High grade NECs on the other hand demonstrate a very aggressive behavior and poor prognosis with higher frequencies of TP53 mutations [39] and RB1 alterations [40], albeit lower than those reported for small cell lung cancer (SCLS). While the above cannot explain, they can definitely suggest different behaviors of various NETs in the metastatic state, in the sense that lower mutation burden/higher reliance on epigenetic processes can be associated with a more indolent behavior.

Our analysis has several weaknesses. It relies on retrospective information collected over a limited period of time. The patient sample was ultimately limited to less than 2200 cases, as we insisted on full sets of data that included grade, origin and single metastatic site. While SEER will document metastasis, only 4 major sites were codified (lung, liver, bone and brain) and there is no data on how that was established. About 13% of patients had no known N status, which, in the setting of metastasis, is of little importance. Moreover, there is concern about accurate reporting of M stage metastasis itself in SEER; some pathologic M0 patients are known to have clinical M1 status. We tried to address this issue by including only patients with a documented site of metastasis; we can assume that most of these have been clinical. The adoption of receptor imaging techniques such as radiolabeled octreotide or Gallium Dotatate PET/CT scans, which can identify occult metastatic disease is not uniform or consistent between practices. This can underestimate the true incidence of the metastatic population in SEER data. Reporting of histology and grade has changed over time, especially for PanNENs and estimation of Ki-67 is notoriously difficult [41], therefore some patients might have been misregistered. We were unable to calculate median survivals for some groups, as the data (especially for small bowel tumors) had not had time to mature. For the majority of patients, we can assume that metastases were not confirmed by biopsy, and we were unable to distinguish between oligometastatic and heavy tumor burden disease. There was no information on grade switch or transformation to more aggressive tumors. Surgery was not included in our analysis, as it was incompletely reported and it is still debatable if optimal cytoreduction plays a role in survival [42]. The data includes no information on initial or subsequent treatments, which can significantly affect survival; for example, everolimus and sunitinib were approved between 2010 and 2014. Our numbers were not large enough to further subcategorize metastatic NENs and allow for multiple comparisons; this will potentially be feasible in subsequent SEER iterations. Finally, our score reflects the results of the analysis and can help guide prognosis estimation but has not been validated in a separate dataset of patients (plans to perform it in an institutional database currently underway). Future studies should incorporate the effect of tumor mutations, especially in low and intermediate grade NENs and allow for comparisons between different treatments.

Site of metastasis plays an important role for survival in metastatic NEN patients and is probably reflective of variable tumor biology, even among NENs of similar origin and grade. It is independent of commonly described prognostic factors and should be considered in survival estimates and design of clinical trials.