Management of Large Cell Neuroendocrine Carcinoma

Background Large cell neuroendocrine carcinoma (LCNEC) is a rare, aggressive cancer with a dismal prognosis. The majority of cases occur in the lung and the gastrointestinal tract; however, it can occur throughout the body. Recently advances in the understanding of the molecular underpinnings of this disease have paved the way for additional novel promising therapies. This review will discuss the current best evidence for management of LCNEC and new directions in the classification and treatment of this rare disease. Methods We performed a PubMed search for “Large cell neuroendocrine carcinoma” and “High grade neuroendocrine carcinoma.” All titles were screened for relevance to the management of LCNEC. Papers were included based on relevance to the management of LCNEC. Results Papers were included reviewing both pulmonary and extra pulmonary LCNEC. We summarized the data driven best practices for the management of both early and advanced stage LCNEC. We describe emerging therapies with promising potential. Discussion LCNEC are rare and aggressive neoplasms. In advanced disease, the historical regimen of platinum based therapy in combination with etoposide or irinotecan remains among the commonly used first line therapies, however for extra thoracic LCNEC regimens like FOLFOX, FOLFOIRI and CAPTEM can also be used. Further effective and safe treatment options are desperately needed. Recently, new advances including a new understanding of the genetic subcategories of LCNEC and immunotherapy agents may guide further treatments.


INTRODUCTION
Large cell neuroendocrine carcinoma (LCNEC) is a rare subgroup of high grade neuroendocrine cancer that can occur throughout the body (Figure 1). The most common primary site is the lung, however, LCNEC also occurs in the gastrointestinal tract and in other locations including cervix, uterus, kidney, bladder, prostate, pharynx, larynx and many other primary sites ( Figure 2) (1-6). LCNEC is an aggressive, fast growing neuroendocrine carcinoma, similar to small cell lung cancer (SCLC). The incidence of LCNEC is increasing worldwide. Most LCNEC originates in the lungs and pulmonary LCNEC represents 2 to 3.5% of all lung cancers (7). In studies using SEER data, age adjusted incidence of pulmonary LCNEC is 0.3 per 100,000 with a rise by 0.011 people per 100,000 per year from 2004-2015 (8,9). In addition to pulmonary LCNEC, there is evidence that LCNEC of all sites are increasing. A Dutch registry study of 47,800 patients with neuroendocrine tumors examined LCNECs, revealing an increase in incidence of LCNEC of all sites from 0.01 per 100,000 people to 1.8 per 100,000 people in 2010 (10). In this same population 5 year overall survival (OS) of patients with LCNEC was 20% (10,11). This manuscript is intended to review current best evidence of management of this rare and deadly disease as well as several promising new directions for classification and treatment.

METHODS
The papers included in this review were primarily selected based on expert opinion. However, for completeness a PubMed search was performed for "Large cell neuroendocrine carcinoma" and "High grade neuroendocrine carcinoma." 3658 results were obtained. All titles were screened for relevance to the management of LCNEC. Of these 606 were further examined with review of abstract and/or paper. Papers were included based on relevance to the management of LCNEC. Data search was completed on 12/31/2020. Only papers available in English were included. Author VC completed the data review and paper selection. Randomized control trial and large retrospective reviews were prioritized, as were publications using the most recent WHO neuroendocrine grading criteria for pulmonary or digestive system neuroendocrine tumors. Given limited studies available describing pure extra-pulmonary LCNEC, the majority of extra-pulmonary studies presented included both SCNEC and LCNEC. Breakdown of the number of LCNEC patients was included when available. As the GI tract is the most common site of extra-pulmonary disease, studies describing extrapulmonary LCNEC focused on GI primary tumors.

HISTOLOGIC CLASSIFICATION
As with all neuroendocrine tumors correct histopathological classification of LCNEC is critical to diagnosis and treatment planning. The classification of LCNEC has evolved rapidly in the last 30 years. Pulmonary LCNEC was first named in 1991 when Travis et al. proposed that LCNEC represented a distinct category from SCLC (12). The 2004 WHO classification included a category of non-small cell lung cancer called large cell carcinoma which encompassed LCNEC and also several other subtypes of large cell carcinomas including rhabdoid phenotype, basaloid carcinoma, and lymphoepithelioma-like carcinoma (13). During this period, there was debate over  whether LCNEC should be classified as a non-small cell lung cancer (NSCLC) or as a variant of SCLC (14). The 2015 WHO classification of lung cancer (Table 1) created the current distinct classification of pulmonary LCNEC from the more general category of large cell carcinoma. In this classification pulmonary high grade neuroendocrine carcinoma (HGNEC) including LCNEC and SCLC have mitoses >10/2mm 2 and the presence of necrosis (13,15). LCNEC can exist in combination with other types of lung cancer including SCLC, adenocarcinoma, and squamous cell carcinoma. Adenocarcinoma is the most common combined component; in one retrospective study median OS was not significantly different in pure LCNEC versus combined adenocarcinoma and LCNEC (16). 2021 WHO classification for lung neuroendocrine neoplasm will formally introduce the combined category. This will be defined as carcinoma with at least 25% of LCNEC or SCLC component mixed with non-small cell carcinoma component. LCNEC typically do not secrete vasoactive amines like some functional low grade neuroendocrine tumors, however a variety of paraneoplastic conditions have been described with pulmonary LCNEC including ectopic ACTH production (17)(18)(19)(20), cancer associated retinopathy (21,22), limbic encephalitis (23), Lambert Eaton Syndrome (24,25), and syndrome of inappropriate antidiuretic hormone (SIADH) (26).
Gastrointestinal (GI) LCNEC has also undergone a recent evolution in classification. The key distinction between welldifferentiated neuroendocrine tumors and high grade neuroendocrine carcinoma was initially described in the 2000 WHO guidelines, which classified these tumors as poorly differentiated. The 2010 guidelines changed to include distinct groups including (i) grade 3 neuroendocrine carcinoma which encompassed both LCNEC and small cell neuroendocrine carcinoma (SCNEC), (ii) and all tumors with Ki-67 greater than 20 and mitotic index greater than 20 (27). The 2010 criteria also described a new category of mixed adenoneuroendocrine carcinoma (MANEC) with an adenocarcinoma and neuroendocrine component. The 2010 criteria definition of G3 as tumors with mitotic rate >20 mitosis per 2 mm2 and Ki-67 >20% included both well differentiated neuroendocrine tumors with high proliferation and poorly differentiated HGNEC such as LCNEC and SCNEC. The WHO 2019 GI NET criteria included a new category of well differentiated (G3) neuroendocrine tumors ( Table 2). This update reflects the recognition that well differentiated tumors can be high grade but are distinct from SCNEC and LCNEC. The WHO 2019 grading system also changed the terminology of mixed tumors to mixed neuroendocrine-nonneuroendocrine neoplasm (MiNEN) (29).

CLINICAL CHARACTERISTICS OF PULMONARY AND EXTRA-PULMONARY LCNEC
Pulmonary and extra-pulmonary LCNEC share many similar clinical characteristics. Patients with pulmonary LCNEC were more likely to be white males (9,30). Patients with pulmonary LCNEC are less likely to present with advanced stage disease as compared to SCLC (31). Cough is the most common presenting symptom of pulmonary LCNEC (32). Although pulmonary LCNECs can occur throughout the lungs, they are most likely to occur in the upper lung lobes (33). In large studies of GI  HGNEC, most patients are also men, most patients are in their 60s, and most are white. The majority of patients also presented with stage IV disease. In one large study the most common primary site was the colon, followed by pancreas, and esophagus (34). Other studies showed that the stomach and small intestine were also frequent sites of HGNEC (35). Pulmonary LCNEC is commonly associated with a smoking history (36,37). In limited data available, patients with a history of colorectal HGNEC are slightly more likely to describe a history of smoking but the contrast with nonsmokers is less pronounced than in pulmonary disease. In addition common presenting symptoms include abdominal pain, hematochezia, melena, and altered bowel movements (38). Although data is limited in extra-pulmonary LCNEC and OS is variable depending on primary site (39), in large studies 5 year OS in pulmonary LCNEC was 16.7% (9) and 13.3% (40) in colorectal HGNEC. Brain metastasis are more common in pulmonary LCNEC (12)(13)(14)(15)(16)(17)(18)(19).2% at presentation in large studies) (8, 41) versus 1.6% in GI HGNEC (31).

MOLECULAR CHARACTERISTICS OF LCNEC
The most recent important innovations in LCNEC classification have occurred in genomic analysis of these tumors. Prior reviews have comprehensively examined the molecular characteristics of pulmonary LCNEC (42). Briefly, George et al. performed genomic and transcriptome analysis on pulmonary LCNEC, comparing genomic alterations in LCNEC with adenocarcinomas and squamous cell carcinoma. They found frequent mutations in TP53, RB1, STK11 and KEAP1. They also describe less common mutations in the RAS-pathway including KRAS, NRAS, Harvey rat sarcoma viral oncogene homolog (HRAS). They identified two molecularly defined subgroups. The first "type 1 LCNECs" with STK11/KEAP1 alterations and a neuroendocrine profile similar to SCLC with elevated ASCL1 and DLL3. George et al. also describe "type II LCNEC" which has a genetic resemblance to SCLC with TP53 and RB1 alterations, but with reduced activity of typical neuroendocrine markers and higher NOTCH pathway activation (43). PD-L1 (programmed death-ligand 1) and tumor mutational burden (TMB) are known to be predictors of potential response to immunotherapy (44) and have been evaluated in LCNEC. PD-L1 was evaluated in relation to the LCNEC subgroups revealing that tumor PD-L1 ≥ 1% is expressed in 16% of metastatic LCNEC and PD-L1 expression was not significantly different between the molecular subgroups of LCNEC (45). Many other studies have reported a similar percentage of PD-L1 expression in pulmonary LCNEC ( Table 3). Reported values for PD-L1 expression in tumor cells (TC) range from 9.1% (53) to 74% (51). PD-L1 expression was more common in stromal lymphocytes or local immune cells (IC) and ranged from 37% (48) to 75% (47). Further investigations into molecular subtypes of pulmonary LCNEC will have important treatment implications. Rekhtman et al. explored next generation sequencing in pulmonary LCNEC. This study revealed three clusters, each with similar molecular make up, including 40% of cases with a "SCLClike" with TP53 and RB1 alterations, and 56% of cases described as "NSCLClike" with STK11, KRAS, KEAP1-NFE2L2 alterations which are more typical of adenocarcinoma or squamous cell carcinoma. This study also showed both of these groups had elevated TMB slightly above 10 mutations/Megabase (Mb). Interestingly, in 4% of patients they identify a "carcinoid-like subset" with MEN1 alterations and low mutational burden (55). Another group also examined mutational burden in pulmonary HGNEC including 39 LCNEC and 63 SCLC cases. They found that the number of non-synoumous mutations was significantly increased in patients with IC infiltration and PD-L1 expression on ICs but not correlated to PD-L1 expression on TCs (52). Another study examined TMB in pulmonary HGNEC in 300 patients with LCNEC and 887 with SCLC. They found the median TMB in patients with pulmonary LCNEC was 9.9 mutations/Mb, the same as in the SCLC group (56).
The overlap between LCNEC and atypical carcinoid tumors was also explored in another study. One group used clustering analysis to compare the genomic profiles of 116 pulmonary carcinoids (including 35 atypical carcinoids), 75 LCNEC and 66 SCLC. Of note, 6 atypical carcinoids clustered with LCNEC patients with a similar survival. The authors suggest that these "supra carcinoids" may represent a group of atypical carcinoid tumors that progress to LCNEC over time (57).
One study examined driver gene status in 94 LCNECs. In this study EGFR mutants were detected from peripheral blood with an amplification refractory mutation system (ARMS) and anaplastic lymphoma kinase (ALK) rearrangement was detected by immunohistochemistry. They found driver mutations including epidermal growth factor receptor (EGFR) (8.3%), anaplastic lymphoma kinase (ALK) (5.8%) (58). Given the limited number of cases the percentage of patients with ALK-mutated LCNEC that respond to tyrosine kinase inhibitor (TKI) treatment is not known. Limited case reports exist of patients with pure and mixed LCNEC who have ALK rearrangements, and a few patients experienced benefit from treatment with ALK inhibitors (59,60). In another study primary tumor location in 125 case of pulmonary LCNEC was examined. Most patients had peripheral tumors. Central tumors were associated with smoking history. Genomic DNA was used for polymerase chain reaction (PCR) amplification and sequencing of EGFR. RB1 was detected on IHC of tumor sample. Peripheral tumors had a higher incidence of EGFR expression. RB1 protein was more frequently expressed in peripheral tumors. In a multivariate analyses of tumor location, resection status, and EGFR mutational status were associated with improved OS (61). Additional case reports of EGFR mutations in LCNEC have been described, along with response to gefintib (62). These findings suggest a potential role for molecular profiling in LCNEC.
There is less information available about the molecular underpinnings of gastrointestinal LCNEC. A retrospective analysis of pancreatic neuroendocrine carcinoma included 49 patients with SCNEC and 18 patients with LCNEC. Immunohistochemistry was reviewed and pRb proteins were identified by monoclonal antibodies. KRAS codon 12 mutations were evaluated using fluorescence-based DNA analysis, and the PCR. Of the LCNEC patients, 17 were evaluable for genetic analysis and 4 patients, (23.5%), had retained RB and KRAS (wild type) and 14 (82.3%) had RB loss and/or KRAS mutation. Of the 18 patients with LCNEC of the pancreas the response rate to first line platinum-based therapy was 40% compared to 65.2% in the 31 SCNEC patients. They analyzed the combined pancreatic LCNEC and SCNEC cohort and found that RB and KRAS mutations were associated with improved response to platinum-containing therapy (63).
Another study examined next generation sequencing and pathology of 19 poorly differentiated pancreatic neuroendocrine carcinomas with 10 LCNEC and 9 SCNEC. Immunohistochemical labeling was done for proteins including pRb, p53, PAX8 among others. Genomic DNA was extracted from and PCR amplification was performed on genomic DNA or whole-genome amplification for KRAS, TP53, CDKN2A/p16, and RB1. In this study 5 of 10 LCNEC patients expressed PDX1 (pancreatic duodenal homeobox 1), 5 of 10 LCNEC also expressed PAX, a transcription factor expressed in well differentiated pancreatic neuroendocrine tumors. Abnormal TP53 expression was found in 9 of the 10 LCNEC cases and RB1 expression loss in 6 of 10 LCNEC patients. Of note, overexpression of BCL-2 was observed in 9 of 9 of the SCNECs and 5 of the 10 LCNECs (64).
Although testing for microsatellite instability (MSI) is routinely done for colorectal adenocarcinoma, few studies reported testing for MSI in gastrointestinal neuroendocrine carcinoma. One study reviewed 40 cases of gastrointestinal neuroendocrine carcinoma including both SCNEC and LCNEC, with 10 cases of colorectal neuroendocrine carcinoma (5 SCLC and 5 LCNEC). MSI cases were detected using immunohistochemistry for the mismatch repair (MMR) proteins MLH1, PMS2, MSH2, and MSH6. Of the 40 cases included, MSI was found in 3 cases, all were cecal neuroendocrine carcinomas (65). Limited studies have evaluated PD-L1 expression in GI HGNEC (Table 4). In one study of 33 patients with GI HGNEC 29% of patients had PD-L1 expression in TCs. TMB was also measured with a range from 0.57 to 11.75 mutations/Mb, with a median TMB of 5.68 mutations/Mb (67). In gastric NEC (4 LCNEC and 39 SCNEC) 48.8% of cases in the combined cohort had TC expression of PD-L1 (68).
As previously described studies in pulmonary LCENC have suggested that a possible subset of atypical carcinoid tumors progress to LCNEC over time (57). A similar study was done in gastroenteropancreatic tumors which evaluated a next generation sequencing analysis of neuroendocrine neoplasms with unsupervised cluster analysis which revealed 3 histology independent clusters. Each cluster contained similar genetic mutations shared by neuroendocrine tumors (NETs) and HGNEC, including well differentiated NETs with high grade components. The authors suggest that some NETs may progress to secondary LCNEC over time (69). Although these theories are still novel and require further exploration, they highlight the many recent innovations in genetic analysis of these tumors.

DIAGNOSIS
Biopsy must be obtained to confirm diagnosis. Diagnosis is made based on neuroendocrine morphology, mitotic count, presence of necrosis, and proliferation rate as assessed by Ki-67. Confirmation of neuroendocrine differentiation is needed using immuno histochemical markers. Core biopsy samples are preferred as in small tumor samples diagnosis can be difficult in fine needle aspirate due to crushed tumor cells (15). It is to be noted that LCNEC are often diagnosed postoperatively based on surgical specimen. Limited fine needle aspirate tissue especially from lung lesion is often not able to confirm the diagnosis. Despite the preference for larger samples endobronchial ultrasound-guided transbronchial needle aspiration has been shown to be effective in diagnosis of LCNEC (70). For pulmonary and gastrointestinal primary sites cross sectional imaging should be done for staging including CT of the chest, abdomen, and pelvis with intravenous contrast. If a gastrointestinal LCNEC primary is suspected, further imaging with MRI of the abdomen and pelvis with contrast (Eovist preferred) while non-contrast chest CT can be considered. MRI of the brain should also be considered for staging at presentation in pulmonary LCNEC. A small retrospective study of 37 patients with LCNEC compared to 76 patients with SCLC, revealed 16.2% of LCNEC patients presented with brain metastasis compared to 18.5% in SCLC patients (71). FDG-PET/CT is also recommended at baseline for assessment of disease burden. Somatostatin receptor scintigraphy with imaging such as Gallium-68 DOTATATE PET/CT can be considered in select cases, especially if the clinical course is following a relatively indolent course, for potential screening for somatostatin receptor type 2 (SSTR-2) targeted clinical trials. One large retrospective study of gastroenteropancreatic neuroendocrine neoplasms examined SSTR-2 in 163 patients with high grade gastroenteropancreatic neuroendocrine tumors or cancer of unknown primary. Of this group 128 were patients with LCNEC, with 36 patients (22.1%) strongly positive for SSTR-2 on immunohistochemical examination (72). Another study examined SSTR-2 expression retrospectively in 218 pulmonary neuroendocrine tumors, including 60 cases of LCNEC. In this study 20 (33%) of these patients were positive for SSTR-2 on immunohistochemical examination (73).
Unlike well-differentiated neuroendocrine tumors, high grade neuroendocrine carcinomas are rarely functional. Some institutions use peripheral blood measurements of chromogranin A (CgA) and neuron specific enolase (NSE) as these have been shown to be elevated in some patients with high grade neuroendocrine carcinoma including SCLC and some LCNEC patients (74). In SCLC elevated NSE is a negative prognostic factor (75). NSE was also examined in a retrospective cohort of poorly differentiated gastroenteropancreatic neuroendocrine tumors, which included both LCNEC and SCLC. In this combined cohort elevated NSE greater than two times the upper limited of normal was associated with a worse overall survival (76).
As described above emerging data suggests a potential role for next generation sequencing in patients with LCNEC of all primary sites. Although there is not yet robust data to suggest testing in all patients, next generation sequencing should be considered.

TREATMENT Stage I-III
Surgery should be considered first line treatment in all patients with early stage pulmonary (I-III) LCNEC. Retrospective data suggests that in pathologic limited stage disease, surgical resection, most commonly lobectomy, was associated with a 5 year OS of 49.2% (77). In a large propensity-matched retrospective study using SEER data in patients with pulmonary LCNEC, surgery was associated with improved overall survival. This study examined 473 patients with pathologic stage IA pulmonary LCNEC and compared outcomes with patients with lung adenocarcinoma and squamous cell cancer. For LCNEC patients median OS was 66.0 months and 5 year OS rates 52.5% (33). In terms of the type of surgery, for pathologic stage I and II pulmonary LCNEC, lobectomy and pneumonectomy have been shown to have improved survival (78). To better evaluate the optimal type of surgery Lufti et al. retrospectively examined patients with pathologic stage I LCNEC and compared 5 year OS in patients who had a sub-lobar resection (wedge or segmentectomy) as compared to patients who had lobectomy. They found sub-lobar resection in early stage LCNEC was associated with a lower 5-year OS rate compared to lobectomy on unadjusted and propensity matched analyses (79).
A retrospective study of 139 patients with pulmonary LCNEC was done, all patients underwent surgery (majority were lobectomies) with curative intent. In this group 5 year OS was 53%, 5 year disease free survival was 39% (80). Although there is less data available in extra pulmonary LCNEC, limited studies of high grade gastroenteropancreatic neuroendocrine carcinoma show improved survival in patients who undergo resection for early clinical stage disease. One retrospective study reviewed 600 high grade high-grade gastroenteropancreatic neuroendocrine tumors with 335 patient who presented with limited clinical stage at diagnosis. Of these patients 89% underwent surgery and median survival for patients undergoing surgery was 153 months vs. 71 months for those not undergoing surgery (81). Another study reviewed outcomes in high grade pancreatic neuroendocrine carcinoma patients undergoing surgery. This group included 28 patients who underwent surgery and 14 patients who underwent resection of the primary tumor in clinically early stage disease with curative intent. All patients recurred, the median time to recurrence or metastasis in this group was 7 months (range 2-14 months). Resection of the primary tumor was an independent prognostic factor of improved survival for patients after occurrence of metastatic disease (82).
Three prospective clinical trials have evaluated adjuvant treatments for pulmonary LCNEC ( Table 5). They include a one arm nonrandomized trial of adjuvant cisplatin and etoposide, compared to a control of historical data from the same institution. In this study 15 men received adjuvant cisplatin and etoposide for 1 or 2 cycles after completing lobectomy with lymph node dissection for LCNEC. This trial revealed a 5 year OS of 88.9% as compare to historical control of 47.4% (83). Another prospective phase II study evaluated 4 cycles of adjuvant irinotecan and cisplatin after complete resection of pulmonary LCNEC and SCLC tumors (95% of patients received a lobectomy). In the 23 patients with LCNEC, OS at 3 years was 86% (84). Recently the results were published of a phase III randomized control trial comparing adjuvant cisplatin and irinotecan versus etoposide plus cisplatin in patients with pathologic stage I-IIIA resected HGNEC. The initial primary endpoint of this study was overall survival; however, the primary endpoint was changed to relapsed free survival (RFS) as there were too few events for analysis of OS after accrual was completed. In total 221 patients were enrolled, of these 74 had pure LCNEC (38 were treated with etoposide and cisplatin and 36 with irinotecan and cisplatin). There was no significant difference in 2 year RFS between the treatment arms, etoposide and cisplatin 65.4% vs irinotecan and cisplatin 69%, P = 0.619. Median RFS was not reached. In a subgroup analysis there was no significance difference among patients with LCNEC or combined LCNEC (85,93).
Several retrospective reviews exist of adjuvant therapy for pulmonary LCNEC. Most studies also suggest benefit of adjuvant In summary most studies suggest that adjuvant chemotherapy improves OS in patients with localized LCNEC. The evidence for this is best in stage IA disease. Platinum based chemotherapy with either irinotecan or etoposide is commonly used. The best evidence for these two agents is the prospective, randomized phase III trial from Kenmotsu et al. (85) which found no significant difference in 3 year RFS between these two regimens.
In extra pulmonary LCNEC no prospective studies were identified evaluating adjuvant chemotherapy. Several retrospective studies have been done (Table 6), most include HGNEC with cohorts combining patients with SCNEC and LCNEC. In a review of 119 patients with high grade pancreatic neuroendocrine carcinoma patients who underwent surgical resection there was improved 3 year OS compared to those without resection. The patients who underwent resection in this study also underwent adjuvant chemotherapy, the majority with cisplatin and etoposide or carboplatin and etoposide. These data suggests that patients who could tolerate more than four courses of adjuvant chemotherapy had improved outcomes (82). In another retrospective study of 39 gastric high grade neuroendocrine carcinomas, including 39 SCNECs and 4 LCNECs, all patients had surgical resection and most received postoperative chemotherapy. The largest group (14) of patients received adjuvant fluoropyrimidine-based regimens, including 5fluorouracil, leucovorin, oxaliplatin combination regimen (FOLFOX) and capecitabine plus oxaliplatin. For those that received postoperative chemotherapy the median overall survival was 44 months compared to 15 months in 5 patients who did not receive chemotherapy post-operatively (96). In a series of 126 patients with colon, rectal, and anal high grade neuroendocrine carcinoma, surgical resection and adjuvant chemotherapy in localized disease did not improve overall survivial (94). A recent large study of 759 patients in the National Cancer Database examined outcomes with postoperative chemotherapy after curative resection. In this study 37.7% of patients had primary tumors in the pancreas, 25.7% had gastric, and 36.6% of patients had tumors that originated in the small intestine. This study used inverse probability of treatment weighting (IPTW) to reduce selection bias and compared postoperative chemotherapy with observation. In this study, 28.1% of patients received postoperative chemotherapy after curative resection, and IPTW showed no OS benefit in the overall group. In a subgroup analysis, paradoxically there was improved survival benefit in the observational arm of the small intestinal group as compared the postoperative chemotherapy group (35). Taken in total the data for post-operative chemotherapy after surgery is mixed. One explanation for mixed results is that the very aggressive nature of this disease leads to recurrence of disease despite surgical resection and adjuvant chemotherapy. Further trials, especially prospective, randomized control trials are needed to clarify the true benefit of adjuvant chemotherapy in resected GI HGNEC.

Radiation
In several retrospective studies post-operative radiation after resection for early stage pulmonary LCNEC does not improve survival. In addition when compared with surgery followed by radiation, patients who received surgery combined with chemotherapy had improved survival (97,98).
In patients with early stage pulmonary LCNEC that are not surgical candidates, radiation may be considered as primary therapy. For patients with stage III disease (not surgical candidates) radiotherapy was associated with a significant increase in overall survival (97). One retrospective study of SEER data used propensity-matched analysis to compare stereotactic ablative body radiotherapy (SABR) to conventionally fractionated radiation therapy in patients with early stage pulmonary LCNEC that were not surgical candidates. In this study SABR was associated with improved OS (99).
There is limited data on the role of prophylactic cranial radiation (PCI) in early stage LCNEC. In retrospective studies PCI is used significantly less in LCNEC as compared to SCLC, and in one study, only 4% of patients received PCI (100). Cumulative risk of brain metastases in LCNEC increases over time in patients with advanced disease treated with first line carboplatin and etoposide. In patients with stage III or IV disease at diagnosis, the risk of brain recurrence reaches 58% to 48% at 18 months after diagnosis (71). Another large study found that in 23 patients with metastatic LCNEC without brain metastasis, 6 (23%) went on to develop brain metastases during clinical follow up (41). Another study of pulmonary LCNEC using SEER data noted a rate of brain metastases of 19.2% as compared to 16.7% in SCLC (8). A retrospective study of 72 patients with advanced or metastatic LCNEC and SCLC included 21 patients treated with PCI (17 SCLC and 4 LCNEC). In this small group of LCNEC patients, PCI was associated with higher progression free survival (PFS) and median OS (101). In another retrospective study, 70 patients with pulmonary LCNEC were reviewed. Almost all (94%) of patients underwent surgery as the first line treatment. In this study 20% patients received adjuvant chemotherapy and/or radiotherapy and prophylactic whole brain radiotherapy (WBRT). They found that in patients who did not receive PCI at median follow up of 23.4 months, 25% developed brain metastasis. At 5 years after diagnosis overall survival was 43% (100). Further trials are needed to clarify the role of PCI in this population. If patients with pulmonary LCNEC are found to have brain metastasis, stereotactic radiosurgery (SRS) is associated with improved survival (102).
There is limited data available on the role of tumor directed radiation and PCI in extra pulmonary LCNEC however in retrospective series fewer brain metastasis are seen in extra pulmonary as compared to pulmonary LCNEC. In one retrospective review of high grade neuroendocrine carcinoma of the colon and rectum (combining both LCNEC and SCNEC) only 2 of 126 patients presented with brain metastasis at diagnosis (94). In another large retrospective study of high grade extra pulmonary neuroendocrine carcinoma (including both LCNEC and SCNEC) only 1.6% of patients had brain metastases at presentation (34). Given the low risk of brain metastases PCI is not routinely done in this patient population.

Stage IV
In patients who present with advanced pulmonary LCNEC there have been 2 prospective trials examining first line systemic treatment and several retrospective studies (Table 7).
Niho et al., performed a single arm phase II study of irinotecan and cisplatin as first line therapy in advanced pulmonary LCNEC. Of 44 patients enrolled, response rate in the intention to treat group was 54.5%. Median PFS was 5.9 months and median OS was 15.1 months. Upon central pathology review, 20 of these patients had pure LCENC. The response rate in the pure LCNEC group was 46.7%. Median overall survival was 12.6 months in the pure LCNEC group (14). In a second trial Le Treut et al. performed a prospective, single arm phase II study of advanced stage LCNEC treated with cisplatin and etoposide. This study enrolled 42 patients with advanced LCNEC of the lung. Upon central pathologist review, 29 of the 40 patients with tissue available for review were pure LCNEC. The objective response was 38%, with a 64% disease control rate. Median PFS in the intention to treat population was 5.3 months and the median overall survival was 7.7 months (103). Given the small number of patients included in each of these trials and the fact that the two regimens have never been directly compared in randomized control trials, it is hard to draw any specific conclusions about irinotecan and cisplatin versus cisplatin and etoposide in this patient population.
Retrospective studies have also sought to explore these questions. One retrospective from the Netherlands examined 207 patients with LCNEC who were treated with 3 different regimens, a platinum (carboplatin, cisplatin) plus either gemcitabine, docetaxel, paclitaxel or vinorelbine (called by the authors NSCLC-t), pemetrexed (called NSCLC-pt), or etoposide (called SCLC-t). They found improved median OS in the NSCLC-t group of 8.4 months, compared to 5.9 months in the NSCLC-pt group and 6.7 months in the SCLC-t group. Specifically, the biggest advantage to the NSCLC-t treatment program was in treatment of a platinum combined with gemcitabine (104). The same group examined 148 patients with LCNEC who had next generation sequencing performed and correlated molecular markers with overall survival and PFS. They also stratified response for NSCLC chemotherapy including platinum and gemcitabine or taxanes versus platinum and etoposide combinations. They concluded that patients with LCNEC that carry a wild type RB1 gene have improved outcomes with platinum and gemcitabine or taxanes (NSCLC regimens) versus platinum and episode based treatment (105). This study was limited by its retrospective nature and small patient population. Other groups have sought to stratify LCNEC by novel genomic subtypes described above (43). In a retrospective study of 63 patients with advanced LCNEC patients were stratified by tumor mutation into SCLC-like and NSCLC-like LCNEC based on genomic features from tumor DNA or circulating free DNA. They found treatment with platinum and etoposide was associated with superior response and survival in SCLC-like LCNEC. However, treatment with gemcitabine, taxane and platinum was not associated with improved survival in NSCLC-like LCNEC (106). Additional studies are needed to better evaluate how molecular subtypes of LCNEC influence outcomes.
No prospective clinical trials for second and third line treatments in metastatic pulmonary LCNEC were identified (Table 8). Two retrospective studies examined amrubicin monotherapy as a second line treatment for pulmonary LCNEC. The first study included 18 patients with LCNEC or "high-grade non-small-cell neuroendocrine carcinoma" all of NSCLC-t group with median OS of 8.5 months vs median OS of 5.9 months in NSCLC-t group and 6.7 months in SCLC-t group.
• "NSCLC-pt" Platinum based therapy with gemcitabine, docetaxel, paclitaxel or vinorelbline, most platinum/ gemcitabine (N=60) • "NSCLC-t" Platinum and pemetrexed (N=20) • "SCLC-t" Platinum and etoposide (N=64) Derks et al. whom had received platinum based first line therapy. The overall response rate to single agent amrubicin was 11.1%. with median PFS of 4.0 and median OS of 9.1 months (114). Another retrospective study also examined amrubicin in advanced pulmonary LCNEC after progression on platinum-based chemotherapy. This study include 18 patients; objective response rate (ORR) was 28%, median PFS was 3.1 months and OS was 5.1 months (115).
In advanced extra-pulmonary LCNEC two prospective studies have evaluated treatment in GI HGNEC (Table 9). Li et al. completed a prospective, single arm phase II study of irinotecan and cisplatin followed by octreotide LAR maintenance. This study included 8 LCNEC patients. In total ORR was 45% and median OS was 12.8 months (107). A recent two-stage phase II study was done to evaluate the efficacy and safety of a first line capecitabine, oxaliplatin, irinotecan, and bevacizumab (CAPOXIRI-BEV) combination followed by pazopanib plus capecitabine maintenance therapy in patients with advanced GI HGNEC. In the 22 patients enrolled, ORR was 47.7% and overall disease control rate was 78.9%. Median PFS was 13 months and median OS was 29 months. Subcategories of LCNEC versus SCLC were not reported (108). Currently a randomized phase II trial (the SENECA trial) of folinic acid, 5fluorouracil and irinotecan (FOLFIRI) or capecitabine plus temozolomide (CAPTEM) as 2 nd line therapy in advanced extra-pulmonary HGNEC is being planned to better understand this topic (117).
There are several retrospective studies examining efficacy of first line chemotherapy in metastatic GI HGNEC. The efficacy of platinum-based chemotherapy was retrospectively reviewed in 20 patients with GI HGNEC. Of the 20 patients in this review, 7 had LCNEC, and all received both FDG PET/CT scan and Gallium 68 DOTATATE PET/CT scans. Of included patients, 80% had positive FDG PET/CT scans and 35% had positive Gallium 68 DOTATATE PET/CT scans. In this study 75% of patients received cisplatin and etoposide as first line treatment and 25% received carboplatin and etoposide. The ORR was 68% and median OS was 13.5 months. Interestingly, 2 patients received peptide receptor radionuclide therapy (PRRT) as second line treatment (113). A review of the national cancer database of patients with HGNECs of the colon and rectum was performed. 1208 cases were identified, with 46.7% of patients presenting with metastatic disease. By multivariable analysis, resection, chemotherapy, and rectal primary site were associated with better OS. Median OS in this study was 9.0 months (40). There is limited retrospective literature for second line treatment in advanced GI HGNEC (Table 10) retrospective studies with 582 patients. When all studies were taken together median response rate was 18%. The ORR was 0% for single agent first line therapies everolimus, temozolomide, and topotecan. The ORR was 50% for amrubicin. In the whole cohort median PFS was 2.5 months, median OS was 7.64 months (121). Amrubicin was also retrospectively studied in another study as a monotherapy for patients with platinum refractory disease. In this study, 10 patients with progression on platinum based therapy for metastatic GI HGNEC were included who received second line amrubicin monotherapy. In this very small study 2 patients had a partial response for an ORR of 20% (118).
One study retrospectively examined 5-fluorouracil and oxaliplatin (FOLFOX) as second line treatment for high grade neuroendocrine carcinoma. The study included 20 patients, 12 LCNEC, 7 SCNEC and 1 unknown primary; 12 of the 20 patients had tumors of gastroenteropancreatic origin. In this study, 12 patients received FOLFOX as second line treatment and 8 as third line. Median PFS was 3.5 months, median OS was 9.9 months, ORR was 29% (119). In GI HGNEC after progression on platinum with etoposide or irinotecan, promising other agents include 5-fluorouracil based regimens and amrubicin. There will likely also be a role for immunotherapy in this population moving forward, however additional studies are needed.

FUTURE DIRECTIONS
Although atezolizumab plus carboplatin and etoposide has been established as the first line therapy in SCLC (122), the role of immunotherapy has not been clearly established in LCNEC. Most of the available data for the use of immunotherapy agents in this population are from small retrospective case series and case reports. Levra et al. reported an impressive ORR of 60% in a case series of 10 patients with advanced pulmonary LCNEC treated with single agent nivolumab or pembrolizumb as a second line therapy. In this study median PFS was 57 weeks (123). Other case reports have shown response to nivolumab in patients with pulmonary LCNEC (124,125). A recent larger retrospective study reviewed 37 consecutive patients with advanced pulmonary LCNEC. The patients were divided into two groups N=23 patients treated with immunotherapy and N=14 patients not treated with immunotherapy. In the group treated with immunotherapy ORR was 33% and median PFS 4.2 months, median OS with immunotherapy was 11.8 months. These outcomes were very similar to response to immunotherapy in NSCLC (126). The Dual Anti-CTLA-4 and Anti-PD-1 Blockade in Rare Tumors (DART) trial is a prospective, open-label, multicenter phase II clinical trial of ipilimumab plus nivolumab across multiple rare tumor cohorts. Patients were eligible if they had advanced nonpancreatic neuroendocrine tumors with progression following at least one line of standard systemic therapy. The neuroendocrine cohort included 18 patients with HGNEC with gastrointestinal or lung primaries. The breakdown of histologic subgroups, including the specific number of patients with LCNEC was not reported. For the group of all HGNEC patients ORR was 44% (127). Although these studies series do suggest potential responses to immunotherapy in LCNEC, larger studies are needed to both determine benefit of immunotherapy as a monotherapy or in combination with chemotherapy and to establish which patients benefit most from immunotherapy treatment.
Another important area for study is biomarkers for response to immunotherapy in LCNEC. As reviewed above, PD-L1 is an important marker of response to immunotherapy in other cancer types, the role in LCNEC is not yet defined. A recent retrospective study reviewed 13 patients with pulmonary LCNEC treated with either pembrolizuamb or nivolumab. In this study the ORR for anti-PD-1 therapy was 39% and OS was 13.7 months. Perhaps most interestingly among the 13 patients treated with immunotherapy 10 patients had PD-L1 negative tumors. The ORR among patients with negative PD-L1 expression was 40%. High density of CD8 positive tumor infiltrating lymphocytes (TILs) (≥38/mm2) was associated with improved ORR. Response to immunotherapy was associated with TP53 mutation co-occurring with either PIK3CA or RB1 (128). Further larger clinical trials are needed to evaluate whether PD-L1 status will be a useful marker for LCNEC response to biotherapy or if molecular characteristics of these tumors can help guide treatment to anti-PD-1 therapies. Circulating tumor cells (CTCs) and cell free DNA (cfDNA) are used as liquid biopsy biomarkers in solid tumors to monitor and assess response in recurrent or metastatic disease and may have a role in identifying patients who could respond to immunotherapy in LCNEC (129). One recent study evaluated blood T cell repertoire alterations in stage IV pulmonary LCNEC patients enrolled in a trial of carboplatin and paclitaxel. Patients with pulmonary LCNEC had significant T-cell repertoire alterations (p < 0.001) compared to age-matched smokers. Increased T cell repertoire alterations were associated with better treatment response and longer survival (130). Assessment of T cells in the peripheral blood may also play a future role in prognostication and treatment selection of LCNEC patients. As stated above median TMB in pulmonary LCNEC has been reported to be 9.9 mutations/Mb (56) and 5.68 mutations/Mb (67) in GI HGNEC. Use of TMB as marker of response to combined checkpoint inhibition was evaluated in SCLC in the CheckMate 032 trial. This randomized, multicenter, open-label, phase 1/2 trial of nivolumab ± ipilimumab evaluated nivolumab monotherapy and nivolumab plus ipilimumab in advanced SCLC after progression on prior therapy. Although not statistically significant the results suggested a potential survival benefit from nivolumab plus ipilimumab versus nivolumab monotherapy for patients with a high TMB (131). Further studies are needed to better determine prevalence and clinical utility of elevated TMB in LCNEC.
The KEYNOTE-158 study is perhaps most notable for evaluating response to pembrolizumab in patients with elevated tumor mutation burden (TMB) which led to the FDA approval of pembrolizumab for this indication. In this study elevated TMB was defined as least 10 mutations per megabase. Patients with elevated TMB above this cutoff had improved response to pembrolizumab monotherapy (132). It is also important to note that a definition of TMB high as 10 mutations per megabase may not be the most meaningful measure in all tumor types. The TMB cutoffs that is associated with improved overall survival with immunotherapy varies between cancer subtypes. One study evaluated cut off of TMB based on the top 20% of normalized mutation burden for each cancer type. For example in renal cell carcinoma cutoff was found to be 5.9 while in colorectal cancer the cutoff was 52.2 (133). The cutoff for elevated TMB with benefit with immunotherapy in LCNEC is not known.
A retrospective cohort study of 2589 patients examined rare solid tumors including neuroendocrine tumors for elevated TMB (using the same cutoff of KEYNOTE-158 of 10 mutations per Mb). They reviewed the same rare tumor types as in KEYNOTE-158 and found that prevalence of high TMB was highest in SCLC in 122 of 305 patients (40%) and in NETs in 48 of 164 tumors (29.3%) (134). They do not provide breakdown of the NET patients by WHO grading criteria, however a note does mention that of 164 pts with NETs 90 had "lung large cell NET" and half of these patients had elevated TMB. One might speculate that a portion of these patients in fact had LCNEC rather than carcinoid or atypical carcinoid tumors.
Recently several promising future therapy directions have been reviewed in SCLC including Poly [ADP-ribose] polymerase (PARP) inhibitors in combination with chemotherapy and targeting of histone-lysine N-methyltransferase EZH2 (135). PARP inhibitors may also have a role in LCNEC as LCNEC has been found to have elevated PARP1 in review of IHC of LCNECs. However, strong PARP1 intensity was more frequent in SCLC, in 87.5% of samples versus in 67.6% of LCNEC samples (136). Similarly EZH2 expression is also elevated in LCNEC and agents targeting EZH2 may be useful in LCNEC (137). Another promising target for future immunotherapy research is delta like protein 3 (DLL3) a Notch family receptor involved in the achaete scute complex-like 1 (ASCL1) pathway. In murine models, overexpression of DLL3 was associated with cell proliferation, tumor growth and reduced apoptosis (138). A retrospective study of 94 patients with metastatic pulmonary LCNEC evaluated DLL3 expression. DLL3 expression was considered positive if ≥ 1% of tumor cells showed cytoplasmic/ membranous or dot-like immunostaining. They found positive DLL3 in 70/93 (74%) samples (139). DLL3 is also expressed in GI HGNEC. One study reviewed DLL3 expression in GI HGNEC and found strong expression of DLL3 in patient tissue samples with levels comparable to SCLC (140).
DLL3 expression is a potential therapeutic target for DLL3 targeting agent. One such agent is Rovalpituzumab-tesirine (Rova-T) a DLL3 targeting antibody drug conjugate (139,141,142). Rova-T was evaluated in a phase 1 study and then a phase 2 study which confirmed antitumor activity but was notable for a high percentage of a patients with adverse events (141,143). A new formulation was recently developed called SC-002 which uses a different linker leading to a more uniform drug antibody ratio, which was thought to improve safety. A phase I study of this new formation was done and included 35 patients all with DLL3high (≥75%) expressing SCLC or LCNEC. In this study 66% of patients experienced serious adverse events, of which 37% were considered related to the drug. However, 14% of patients achieved partial response (144). Further research is needed to evaluate whether DLL3 -targeted antibody-drug conjugates can be used in this patient population. In addition to antibody-drug conjugates, bispecific T cell engagers and chimeric antigen receptor T cell therapy targeting DLL3 are also in development (145).
Aurora kinase inhibitors are another potential target in SCLC which are also promising in LCNEC. A recent study examined subgroups of LCNEC including subtypes defined by TTF-1 and c-Myc, which have been previously described as clinically relevant subgroups in SCLC. They also examined DLL3 in this patient population. IHC of lobectomy and biopsy specimens was reviewed for 27 patients with pulmonary LCNEC. Positivity for c-MYC was defined as cases where ≥40% tumors cells reacted with any intensity; positivity for TTF-1 expression was defined cases where >10% of the tumor cells reacted with any intensity. The authors identified 10 cases with only TTF-1 and 15 cases with only c-MYC present, 2 cases were positive for both. DLL3 was found in 15 patients and was associated with TTF-1 expression. They conclude that LCNEC, similar to SCLC can be further divided into subgroups by TTF-1 and c-MYC expression (146). This has important treatment implications as c-MYC expression is associated with aurora kinase expression and may indicate responsiveness to aurora kinase inhibitors (147). Further studies should include biomarker correlates to help identify subgroups of LCNEC that may best respond to specific treatments.
Somatostatin targeting agents including somatostatin analogues and Peptide Receptor Radionuclide Therapy (PRRT) are effective in well-differentiated neuroendocrine tumors. Research is ongoing as to determine the efficacy in HGNEC including LCNEC. Case reports describe promising response to PRRT in patients with pulmonary LCNEC with positive somatostatin receptor scintigraphy (148). Immunohistochemical analysis of SSTR-2 expression in pulmonary neuroendocrine tumors was reviewed retrospectively in a series of 218 "clinically aggressive cases" including 60 LCNEC. They found that SSTR-2 was expressed in 33% of LCNEC and 38% of SCLC (73). Another study reviewed immunohistochemical analysis of SSTR-2 expression of GI HGNEC including 142 patients with LCNEC. In the LCNEC group, 22.1% of these patients had strongly positive expression of SSTR-2 (72). An analysis of somatostatin receptor scintigraphy in pulmonary LCNEC was done in 26 patients with technetium-99m ethylene diamine-diacetic acid/hydrazinonicotinyl-Tyr3octreotide (Tc-TOC). This study found that 100% of primary lesions showed uptake of Tc-TOC while 84% of metastases were visualized (as compared to conventional imaging) (149). Another study retrospectively examined 26 patients with primary pulmonary LCNEC and compared them to 40 patients with adenocarcinoma. In this study, 50% of patients had high expression of somatostatin receptors (150). Another retrospective study examined 18 consecutive patients with resected pulmonary LCNEC who underwent preoperative indium In-111 pentetreotide scintigraphy. Of this group, 10 patients (55.5%) had positive OctreoScan scans. In this study all patients with TNM pathologic stage higher than Ib received post-operative radiotherapy. Those with positive SRS avidity received post-operative octreotide LAR; of these patients 9 were still alive and disease free at the time the study was published (151). Further studies are needed to better understand the role of SSTR-2 directed therapy in LCNEC.

CONCLUSION
LCNEC are rare and aggressive tumors, and current evidence suggests surgery for local disease with platinum and etoposide or platinum and irinotecan based adjuvant chemotherapy is the most effective treatment. In advanced disease, the historical regimen of platinum based therapy in combination with etoposide or irinotecan remains among the commonly used first line therapies, however for extra thoracic LCNEC regimens like FOLFOX, FOLFOIRI and CAPTEM can also be used. Further effective and safe treatment options are desperately needed.

AUTHOR CONTRIBUTIONS
These authors have contributed equally to this work, VC, SA, LA. This author is senior author, AC. All authors contributed to the article and approved the submitted version.