Introduction
Carcinoid tumours of the lung are rare primary lesions accounting for 0.2% (atypical carcinoids, AC) to 2% (typical carcinoids, TC) of resected lung cancers. According to the 2015 WHO classification [
1], they make up the morphologically defined group of neuroendocrine tumour (NETs) based upon cytological traits, mitotic count and necrosis. Lung NETs comprise low- to intermediate-grade forms (TCs and ACs, respectively), clearly segregated from high-grade malignancies, namely large cell NE carcinomas and small cell carcinomas. Carcinoid tumours, however, translate into a broad clinicopathologic spectrum characterized by highly variable outcomes, which are currently best stratified according to the 2015 WHO classification and 8th edition TNM staging [
1]. Surgery is the best treatment modality choice for both TCs and ACs [
2], with TCs usually exhibiting long life expectancy with 5- and 10-year survival rate by far over 90% [
3]. Conversely, ACs exhibit a more aggressive clinical course and a 5-year survival rate ranging from 56 to 87%, which is lower in node positive patients, suggesting that this subgroup could benefit from some adjuvant treatment [
4‐
8].
In pulmonary NETs, whether biopsy samples or surgical specimens, no grading system is currently accepted for further stratifying single tumour entities beyond the pure tumour histological typing. Indeed, the grading system proposed for gastroenteropancreatic (GEP) NETs does not apply to thoracic NETs, nor Ki-67 plays a role in tumour classification. In the 2015 WHO classification, the Ki-67 prognostic role is not established and is considered valuable in biopsy or cytology samples only [
1]. As a matter of fact, several groups have confirmed the significant prognostic value of Ki-67 in lung NETs [
9], with a 4–5% cut-off being the most widely employed for the sake of reproducibility and reliability [
10] compared to mitotic count and necrosis, especially when dealing with biopsy specimens [
11,
12].
However, in multivariate survival analysis, Ki-67 seems to add little value to histological typing [
13,
14] when a 4–5% cut-off is adopted, but results from several studies are still inconclusive. In order to integrate rather than to replace the histological classification with the information stemming from Ki-67 index determination, Rindi and co-workers have proposed an accurate and innovative grading system for pulmonary NETs (G1 to G3) where mitotic count, necrosis and Ki-67 index were jointly combined on the basis of lung-specific cut-off values [
15]. At the histological level, only TCs by morphology turned out G1, while ACs and the LCNEC/SCLC group crossed G1 to G3 and G2 to G3 categories, respectively, confirming the inherent heterogeneity in behaviour of intermediate- to high-grade lung NETs [
15].
In the GEP area, NET-G3 retaining a well-differentiated morphology but showing proliferation rates consistent with NE carcinomas, either in the whole tumour or in focal areas only, have recently been recognized [
16,
17]. These tumours are associated to genetic alterations of well-differentiated tumours, but with a clinical behaviour intermediate between G2 NETs and more conventional NE carcinomas [
18‐
20]. This phenomenon would identify secondary high-grade NE tumours evolving from pre-existing lower grade carcinoids, as recently demonstrated in the thymus [
21]. In the lung, LCNECs with carcinoid-like appearance have been identified [
22,
23], but little is known about the clinical impact of carcinoid tumours, as defined by current diagnostic guidelines, showing high Ki-67 indices.
Based on these premises, in this study, we sought to investigate the prevalence of carcinoid tumours showing brisk proliferation activity and to define whether values and patterns of Ki-67 labelling alone would allow for a clinically meaningful stratification of a large cohort of lung carcinoids, regardless of the histological type. This could be propaedeutic to the identification of low- to intermediate-grade NETs in the lung that may have the potential to evolve to a higher grade disease, thus sustaining the paradigm shift to secondary high-grade NETs even in the lung.
Discussion
In this study, we showed that up to 13% of lung carcinoids display a Ki-67 proliferation index of ≥10% and that 6% of ACs may show Ki-67 values around 25%. Most importantly, a 10% cut-off for Ki-67 labelling index provided a clinically meaningful stratification of carcinoid tumours of the lung even within the subgroup of ACs, thereby providing evidence that proliferation index assessed by Ki-67 would add useful prognostic information to histological subtyping.
It is currently a matter of debate whether the 2015 WHO classification [
1] is adequate enough to encompass the spectrum of lung NETs. This classification has substantially confirmed the four categories already crystallized in the two previous editions: it is solely based on histological parameters [
27], and the threshold of 10 mitoses/2 mm
2 is the criterion used to take pulmonary carcinoids apart from NE carcinomas [
1]. Here, we show that the rare group of carcinoid tumours is biologically heterogeneous, including a subgroup of lesions with a relatively high proliferative potential as highlighted by the Ki-67 index. From a classification standpoint, the even rarer AC category is monolithic (necrosis and/or 2 to 10 mitoses/2mm
2), even if a cut-off of six mitoses has been proposed, but not envisaged, to separate tumours with different life expectation [
28]. The only diagnostic alternative is the category of high-grade neuroendocrine carcinomas, either LCNECs or SCLCs, which is however ruled out upon mitotic count. Interestingly, even LCNECs may show morphologic [
22,
23] and molecular [
23] features rather akin to ACs, supporting the view of a grey-zone in lung NETs where a biological continuum could probably be more effective than categorical criteria for classification purposes. Most importantly, the optimal treatment options for ACs after surgery, particularly in the metastatic setting, are highly debated and variably including close follow-up, systemic therapies different from or comprising platinum or even biological drugs (somatostatin analogues or m-TOR inhibitors).
Based on this evidence, we reasoned that rather than forcing the four-tier histological classification, an approach based on proliferation activity could be beneficial to best depict the clinical variety within carcinoid tumours, thereby integrating a purely histological classification. At present, over 2000 lung NET cases have been published having Ki-67 investigated as part of the morphological and phenotypical description [
9], and most of these studies recognize the usefulness of Ki-67 index reporting. A multicentre study on almost 400 cases proposed a grading system for lung NETs incorporating the Ki-67 index alongside mitotic count and necrosis evaluation [
15].
In our study, we wanted to ascertain whether Ki-67 alone, not engaged in a grading system, would allow a prognostic classification within the category of carcinoid tumours with the specific aim of identifying a niche of carcinoid tumours (defined according to the official parameters) having a proliferation index above the expected values. Our results clearly show a prognostic separation in terms of both overall survival and time to progression when cut-offs of 4 and 10% are adopted. Our analysis also revealed that the tumours with either <4 or 4–9% proliferation indices showed comparable survival rates, thus suggesting the 10% as a more suitable cut-off to identify lung carcinoids with a clinically meaningful brisk proliferation activity.
The adoption of Rindi’s grade (grade 2 versus 1) showed a significant correlation with outcome in univariate but not in multivariate analysis. Although also the Ki-67 thresholds here analysed did not reach statistical significance in multivariate analysis, it is interesting to note that the 10% cut-off was able to tease out a category of tumours harbouring dismal prognosis even within the restricted category of ACs, whereas Rindi’s grade did not hold significant prognostic information in this subset of patients.
Importantly, a heterogeneous Ki-67 proliferation pattern was significantly more common in the category of cases displaying a Ki-67 labelling index ≥10%. This is in line with recent reports suggesting that intra-tumour heterogeneity holds per se a negative prognostic impact, as shown across several distinct diseases [
29].
We documented in our study that a non-negligible number of life-threatening ACs showed quite unexpected high Ki-67 indices, which overlapped with those reported for LCNECs [
1,
9], even if not fulfilling the relevant mitotic count for this tumour category. Although Ki-67 levels and mitoses do not entirely reflect the same cell cycle-associated event, it is known that pancreatic NETs with discrepant Ki-67 and mitotic indices bear different impacts on tumour grading [
30]. Moreover, we speculate that these ACs with high Ki-67 index could harbour a potential of transition towards LCNECs with similar molecular profiles, either synchronously or metachronously, as recently demonstrated in thymus ACs [
21]. Furthermore, a recent study by Simbolo et al. [
31] has also shown that rare lung LCNECs resembling carcinoids harbour
MEN1 mutations—the most frequent alteration in ACs—along with
TERT,
SDHA,
RICTOR,
PIK3CA,
MYCL and
SRC gains are shared with high-grade carcinomas. This is in line with a paradigm shift, even in the lung, that envisages the development of secondary high-grade NETs from pre-existing lower grade NETs, as demonstrated in the thymus [
21]. Regrettably, the numerosity of outlier tumour carcinoids, i.e. either highly proliferating or harbouring genetic features of aggressive disease (
TP53 mutations, hypermutated profile), was too limited in the recent study reported by Simbolo et al. [
31] to draw any informative conclusions. In this scenario, it becomes clear that a thorough genomic characterization focused on a large cohort of highly proliferating carcinoid tumours in strict comparison with LCNECs and SCLCs is eagerly awaited.
In conclusion, although the histological subtyping still represents the cornerstone in lung carcinoid classification, the inter- and intra-tumour heterogeneity of proliferation indices together with different clinical behaviour within the same histological subgroups suggest that the current morphology-based system could be suboptimal for an effective prognostic stratification and therapy planning. Herein, we indicate that a subgroup of ACs showed brisk Ki-67 indices within the spectrum of allowed mitotic count. These patients were associated with specific clinical and pathological characteristics, to some extent regardless of histological subtyping. Preliminary data obtained by our group in thymus ACs on synchronous and metachronous lesions suggest that a progression capability of these lesions may stem form a markedly heterogeneous intra-tumour distribution of Ki-67 labelling indices. The correct identification of these tumours is clinically warranted. A thorough molecular characterization of this specific high proliferative index carcinoid subgroup may uncover the determinants of their biology and/or progression capacity, as well as identify biomarkers of potential therapeutic usefulness.