Clinical features of large cell neuroendocrine carcinoma: a population-based overview

Discussion

In this population-based study we confirmed that LCNEC is a rare disease with an average incidence of 0.9% of lung cancer over a 10-year period, while occurrence increased 2.5-fold. The presented results indicate that LCNEC is a highly aggressive form of lung cancer like SCLC, with a poor prognosis in all stages of disease. Nevertheless, there were important differences from SCLC: in stage IV, the prognosis of LCNEC was lower than SCLC, yet similar in selected cases treated with chemotherapy, as well as the clinical presentation, such as the lower proportion of patients with mediastinal lymph node involvement and the currently applied treatment in early disease differed from SCLC.

The total incidence of LCNEC (0.9%) as fraction of all diagnosed lung cancer was lower than that reported by institutional surgical series (1.7–3.0%) [3, 7] but higher than that reported by the population based SEER registry (0.6%). Probably this is caused by differences in analysed time period (i.e. 2003–2007). Overall, lung cancer occurrence increased with 24% in the Netherlands (2003–2012), mainly attributed to the increase in AdC [17]. The increase observed in LCNEC, particularly occurring after 2008, might be explained by growing awareness among pathologists of this relatively new entity, although at that time, no new pathology guideline was published. Additionally, an increased use of immunohistochemical neuroendocrine markers (CD56, synaptophysin and chromogranin-A) in routine diagnostics and an increase in core/needle-biopsy sampling in order to obtain sufficient material for molecular testing (i.e. more tissue) have improved LCNEC diagnosis. In addition, indiscriminate use of markers with low specificity for neuroendocrine differentiation (i.e. CD56 or NSE) could have increased the diagnostic frequency on biopsy tissue specimens. Finally, introduction of the IASLC guideline for diagnosis of lung cancer on biopsies (2011) may have increased awareness of LCNEC diagnosis on biopsies. Currently, LCNEC on small biopsies is referred to as NSCLC, possibly LCNEC when a neuroendocrine morphology and neuroendocrine immunohistochemical staining is confirmed in a biopsy specimen [2, 18].

The clinical characteristics of LCNEC corresponded to that of SqCC and AdC closely in the early stage, whereas LCNEC overlapped with SCLC in metastatic disease. Not only this study but also the SEER registry reported that the stage distribution of LCNEC resembled NSCLC [4]. A possible explanation is that advanced LCNEC was underrepresented (e.g. pathologists recognise LCNEC in surgically resected tissue more easily and may overlook LCNEC when assessing biopsies). However, we also found that LCNEC patients presented less often with stage N2–N3 compared to SCLC, and this matched SqCC and AdC, also in stage IV disease. Therefore, it might well be that early-onset LCNEC does not metastasise as rapidly as SCLC, increasing the chance of diagnosis at an earlier stage, an observation that was recently also seen in genetically engineered mouse models [18]. Nonetheless, whenever LCNEC has metastasised, the metastatic pattern resembles SCLC. Unfortunately, the metastatic pattern could not be confirmed in other studies as numbers of included patients (n=22–86) were too small [3, 7, 19, 20].

Conflicting data are available in the literature with regard to prognosis of LCNEC. The OS of LCNEC in surgical case series was similar to that of SCLC [8, 9, 21], while the SEER registry showed that OS in very early (T1N0M0) LCNEC was comparable to LCC and better than SCLC [4]. Resembling the SEER results, we observed a clear OS difference between LCNEC and SCLC in early-stage disease, but whenever SCLC was surgically treated, OS resembled LCNEC. The prognostic results in stage IV chemotherapy-treated patients were in line with two small cohorts that compared OS of LCNEC and SCLC [10, 11] and overall OS of chemotherapy-treated LCNEC resembled OS of the European phase II trial [12].

At present, there are no guidelines that aid physicians in treating LCNEC but we have shown that over recent years, treatment corresponded to SqCC and AdC more closely than to SCLC. The difference in (adjuvant) chemotherapy treatment between LCNEC and SCLC in stage I–II and IV was considerable. Indeed, if LCNEC is considered equally aggressive as SCLC, one would expect the ratio of chemotherapy-treated patients to be similar. Moreover, in the adjusted multivariate Cox regression analysis, the prognosis of LCNEC was poorer than that of SCLC. After selection of chemotherapy-treated patients, the adjusted multivariate Cox regression showed a nonsignificant difference. This increase in prognosis might be an important sign of possible undertreatment of patients with stage IV LCNEC disease in the overall population, but requires further investigation.

Because of the rarity of LCNEC, the majority of data comes from single-centre, retrospectively diagnosed LCNEC series. In this study, for the first time, we describe the clinical manifestation and treatment of both early- and advanced-disease LCNEC in a large, population-based, histologically diagnosed cohort. Moreover, we were able to comprehensively define the metastatic pattern of LCNEC at diagnosis and to compare this with SCLC, SqCC and AdC. By doing this, we excluded possible interference from treatment, and by excluding patients with previous malignancies, we minimised confounding from other cancers.

The current study has several limitations. Although only histology-selected cases were included, it remains possible that several tumours diagnosed as LCNEC in this registry were incorrectly classified. In clinical trials, up to 25–27% of LCNEC diagnosed on biopsies was reclassified into NSCLC or SCLC after central revision [12, 13]. Furthermore, interobserver studies show poor agreement, underscoring the difficulty in delineating LCNEC from SCLC and NSCLC [22]. However, this population-based study mirrors daily practice. A consequence of the histological selection criteria is the observed relatively high fraction of early-stage NSCLC. Another limitation might be related to the sensitivity of the metastatic pattern analysis as registration of used modalities for diagnostic imaging was not mandatory. Finally, we were not able to transcribe TNM-6 into TNM-7, had no data on smoking status, and were not able to adjust for possible prognostic confounders such as comorbidities, performance score and weight loss, as these variables were not or insufficiently registered in the NCR.

In summary, LCNEC is increasingly encountered, especially in stage IV disease. We have shown that LCNEC in clinical practice is a different entity, resembling SqCC and AdC in early-stage disease but with a comparably poor prognosis and metastatic pattern to SCLC. Nonetheless, possible biological important differences were present, as LCNEC showed less lymphatic N2–N3 pattern than SCLC. In the near future, it is expected that the visibility of LCNEC will increase even more for physicians due to the separate mention of LCNEC in the pulmonary NET chapter of the fourth WHO classification and the recent introduction of criteria for the diagnosis “NSCLC, possible LCNEC” on biopsy specimens [2, 18]. Therefore, collaboratively structured international phase III trials are needed to investigate the role of adjuvant chemotherapy in early-stage disease and optimal disease management of advanced-stage LCNEC. Eventually, this research can lead to establishment of broadly accepted guidelines for the treatment of LCNEC.