Longitudinal Changes in Ki-67 Indices in Small-Intestinal Neuroendocrine Tumours and Their Impact on Survival

Although small intestinal neuroendocrine tumours (SI-NETs) are rare, their incidence is globally rising with recent reports showing a remarkable increase in their incidence [1‒4]. In most cases, albeit malignant they follow a relatively indolent course with heterogeneous behaviour and are associated with prolonged survival, even in the presence of distant metastases [1, 5]. Survival outcomes mainly depend on the tumour’s biological behaviour and stage, as represented by the Ki-67 proliferation index (PI) and the presence of distant metastases at the time of diagnosis mainly to the liver [6].

World Health Organization (WHO) and European Neuroendocrine Tumour Society (ENETS) classification systems have evolved over the last decades and include a grading system based on the Ki-67 PI and mitotic count and a TNM-based staging system, respectively [7, 8]. Although Ki-67 is of paramount importance in disease prognostication, the span of the observed Ki-67 indices is wide and cases with a higher level of Ki-67 may have a substantially different course. Generally, most SI-NETs are well-differentiated tumours that belong to the WHO G1 and G2 grades, exhibiting Ki-67 indices <10% in more than 80% of the cases [9]. Moreover, there is a sparsity of SI-NETs with Ki-67 indices >20%, currently classified as G3 NETs [9, 10]. We have recently proposed in a collaborative multicentre setting a modified site-specific histopathological grading system for SI-NETs applying Ki-67 cut-off values of 5% and 10% capable of predicting differences in patient survival outcomes more accurately [9].

Following improvements in the management and clinical outcomes of NETs in the last decades, often a subset of tumours become more aggressive during the course of the disease, as reflected by an increase in the Ki-67 PIs. In particular, a concept mainly applied to pancreatic neuroendocrine tumours is that of tumour progression from a lower to a higher WHO grade, evident on re-biopsy after disease radiological progression, such as transition from G1 to G2/G3 or G2 to G3 grades without histologically confirmed dedifferentiation. This phenomenon of grade progression is often linked with worse patient prognosis [11, 12].

To date, NET disease evolution with Ki-67 indices’ increment over time and associated tumour grade progression have solely been described and assessed in pancreatic primaries in contemporary literature [11, 13]. Our aim was to evaluate the proportion of SI-NETs with longitudinal changes in Ki-67 indices and grade progression in our dataset and assess the impact of these changes in overall survival (OS), applying the recently proposed site-specific modified histopathological grading system with Ki-67 cut-offs of 5% and 10% for SI-NETs.

We extracted patient data from the SI-NET databases of 5 collaborating European tertiary referral centres: the EKPA-Laiko ENETS Centre of Excellence, Athens, Greece; the ARDEN NET Centre, ENETS Centre of Excellence, University Hospitals Coventry and Warwickshire NHS Trust, Coventry, UK; the Neuroendocrine Tumour Unit, KHP ENETS Centre of Excellence, King’s College Hospital, London, UK; the Department of Surgery, Örebro University Hospital, Örebro, Sweden; and the Department of Endocrinology and Neuroendocrine Neoplasms, ENETS Centre of Excellence, Medical University of Silesia, Katowice, Poland. We only included patients with a definite histopathological diagnosis of well-differentiated SI-NET with available both baseline and ≥1 follow-up tissue specimens obtained later on, outside of the primary diagnostic and/or therapeutic process (i.e., primary locoregional resective surgery or liver biopsy at diagnosis) to obtain with Ki-67 index estimates given in the pathology reports.

The diagnoses were made between June 15, 1993, and March 8, 2021, and the patients were followed until death or May 15, 2021. The study was conducted according to the 1975 Declaration of Helsinki and approved by the pertinent Human Research Ethics Committee of each institution. In the collaborating centres from the UK, the study was officially registered as an audit. Written informed consent for the study was obtained from alive study participants.

At diagnosis, patients underwent cross-sectional imaging with either computed tomography (CT) scan or magnetic resonance imaging of the abdomen. Functional imaging with Octreoscan or 68-gallium (⁶⁸Ga) positron emission tomography (PET)-CT was also performed for more accurate disease staging. Fluorodeoxyglucose (FDG)-PET-CT was not routinely performed as part of the primary diagnostic workup. Surveillance protocols with sequential cross-sectional and functional imaging of the abdomen were performed in the disease clinical course at varying intervals to assess progression as per European Neuroendocrine Tumour Society (ENETS) guidelines [6, 14, 15]. Re-biopsies from metastatic sites were acquired due to a suspicion of a change in disease behaviour, based on clinical, biochemical, and/or radiological information. Radiological progression was substantiated following the RECIST 1.1 criteria and re-biopsies were acquired from patients with rapid tumour growth; resistance to therapy of certain lesions, while others had responded; and discrepancies in radiological and functional imaging findings. On the other hand, tissue specimens from primary tumours, when these were not operated upon at initial presentation, were mainly obtained later in the disease course due to locoregional symptoms that led to surgery in a subacute or emergency setting for symptomatic relief. Radiological progression was assessed at the time of re-biopsy according to the Response Evaluation Criteria in Solid Tumours (RECIST 1.1) [16].

Tumour grade was determined in both baseline and follow-up tumour samples from surgical specimens or liver core biopsies according to the Ki-67 PI by a dedicated pathologist at each participating centre [6]. For tumour grading, the 2017 WHO classification system for gastro-entero-pancreatic NETs was used [8] and also a novel SI-NET site-specific modified histopathological grading system with Ki-67 cut-offs of 5% and 10% [9, 17]. For staging, the 8th edition of the American Joint Committee on Cancer (AJCC) was applied [7]. To avoid immortal time bias, baseline for survival estimates was defined as the date of re-biopsy.

The following variables were recorded at baseline: age, gender, Charlson Comorbidity Index, locoregional and carcinoid syndrome symptoms, TNM stage, Octreoscan/⁶⁸Ga-PET positivity, and Ki-67 PI. Furthermore, tissue source, method of tissue acquisition (surgery/biopsy), and the indication for the follow-up tissue sampling were recorded. The Ki-67 PI was specifically expressed as a percentage based on the count of Ki-67-positive cells in 2000 tumour cells in areas of the highest nuclear labelling often observed at the tumour periphery, using the MIB1 antibody [18, 19]. Morphological features of grade progression, such as tumour cell necrosis, and nuclear atypia were also noted. If multiple tissue specimens were obtained during the disease follow-up, the highest available Ki-67 index estimate at re-biopsy was used. To ensure the quality of data reporting, we followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [20].

Nominal data are presented as number of patients and percentages, and scaled data are presented as medians with ranges. Log-rank testing (Mantel-Cox) and multivariable Cox-regression analysis were used to determine whether there was a difference between OS in SI-NET patients stratified by increment of Ki-67 indices over time and/or grade increment. The multivariable model was constructed, after including variables identified as significant at univariable regression analyses. Tests were two-sided, p < 0.05 was considered statistically significant, and the 95% CI was given for OS estimates. All statistical analyses were performed with the SPSS v23.0 software package (IBM SPSS Statistics, Armonk, NY, USA).

We screened a total of 551 SI-NET patients for eligibility and included 45 patients with available baseline and follow-up samples, who fulfilled the study inclusion criteria. The study flow is given in Figure 1. The descriptive statistics of the study cohort at diagnosis are shown in Table 1. Twenty-one of the re-biopsy specimens were from liver metastases; six were from lymph node metastases; nine were from primary tumours operated upon later in the disease clinical course; three from peritoneal tumour deposits; and the remaining six samples from rare metastatic sites, including bone (n = 3), pancreatic (n = 2), and cardiac metastases (n = 1). Mean time to re-biopsy was 29.8 (SD: ±38.2) months, whereas mean time from re-biopsy until death or last follow-up was 52.1 (SD: ±56.1) months.

Primary and follow-up pathology reports were reviewed to confirm the SI-NET diagnosis and assess tumour morphology and Ki-67 index estimates. At diagnosis, median Ki-67 index at SI-NET diagnosis was 2% (range: 0.5–15%). Twenty-five patients had G1 tumours (55.5%), 20 G2 (44.5%), and there were no G3 tumours. When applying 5% and 10% Ki-67 cut-offs according to a modified site-specific histopathological grading system, 31 patients had Ki-67 indices <5% (68.9%), 6 had Ki-67: 5–10% (13.3%), and 8 had Ki-67 ≥10% (17.8%) at diagnosis. Thirty-one patients (68.9%) exhibited radiological progression of locoregional and/or distant metastases according to the RECIST 1.1 criteria at the time of re-biopsy.

Slides from 45 follow-up samples included herein were available for histopathological re-assessment. With regard to tumour morphology, nuclear atypia or tumour necrosis was found in 7 (15.6%) of available follow-up samples. At re-biopsy, the median change in Ki-67 index (absolute value; follow-up minus time of diagnosis) was 1% (range: −10 to +38%). An increase in Ki-67 occurred in 20 patients (42.6%); in 14 patients, the change in Ki-67 resulted in progression to higher tumour grade following the modified grading system. Only 3 patients exhibited WHO high-grade (G3) progression, whereas 9 patients exhibited grade progression to Ki-67 >10% group following the modified grading system. Among the 9 patients with grade progression to Ki-67 >10%, only 4 cases had Ki-67 <5% at diagnosis. There was no significant difference in the time to re-biopsy between patients with and without radiological progression at re-biopsy (35.6 [SD: ±39.2] vs. 22.9 [SD: +/−29.2] months; p = 0.209). Importantly, there was no evident association of radiological and grade progression at the time of re-biopsy (χ² test p = 0.975).

Within the follow-up time of the study cohort, 15 patients (33.3%) died. The median OS from baseline was 51.4 (95% CI: 43.4–59.5) months. The OS rates were 79% after 2 years and 39% after 5 years. In log-rank OS analysis, patients with an increment in Ki-67 ≥1% had a median OS of 32.9 months versus 80.5 months in patients without (HR = 5.6, 95% CI: 1.42–22.02; p = 0.014; Fig. 2a). There was no OS difference between patients with evident WHO grade progression and those without (median OS: 46 vs. 53.7 months; HR = 2.4, 95% CI: 0.84–6.83; p = 0.101; Fig. 2b). However, when applying a site-specific modified histopathological grading system for SI-NETs with 5% and 10% Ki-67 cut-offs, patients with grade progression had a median OS of 32.9 months versus 53.7 months in those without (HR = 4.61, 95% CI: 1.22–13.54; p = 0.022; Fig. 2c). Finally, the multivariable Cox-regression OS model utilizing the modified histopathological Ki-67 grading system confirmed grade progression as an independent predictor for death (HR = 7.2, 95% CI: 1.58–32.82; p = 0.011) and demonstrated a statistical trend for TNM stage IV (HR = 4.4, 95% CI: 0.94–20.76; p = 0.060; Table 2).

In this study, a large multicentre series of SI-NETs was analysed to detect both longitudinal changes in Ki-67 PIs and tumour grade progression over time and assess the prognostic impact of these parameters, in terms of OS rates. Among the 45 SI-NET patients of the study cohort with available baseline and follow-up tumour samples, approximately 43% had evident increase in the observed Ki-67 indices and 31% displayed progression to a higher tumour grade according to a recently proposed, modified grading system applying Ki-67 cut-offs of 5% and 10% [9]. OS was significantly reduced in the subset of patients in whom the disease evolved toward increased cell proliferation based on sequential tumour biopsies with Ki-67 increment >1%, commonly obtained at radiological disease progression, as compared to those where there was no change in the histopathological indices. In addition, grade progression following the modified grading system to distinguish between different Ki-67 groups was confirmed as an independent predictor with a 7.2-fold increased risk of death in our series.

Importantly, as far as the distribution of Ki-67 indices is concerned, there were no patient cases with Ki-67 index >20% at diagnosis and only 3 cases exhibited grade progression to WHO G3 NETs at re-biopsy. Thus, the G3 NET entity, which was primarily described in pancreatic primaries in the revisited 2017 WHO classification system for gastro-entero-pancreatic NETs, is a relatively rare finding in the setting of the generally low-proliferative SI-NETs [8]. In addition, the term high-grade progression in pancreatic NETs, defined as a transition from WHO G1 to G3 or G2 to G3 groups, does not seem to apply to the majority of SI-NET primaries, possibly due to their relatively more indolent nature and prolonged disease course [11]. Finally, to our knowledge longitudinal changes in Ki-67 in tumour samples in sequential biopsies during the disease follow-up have only been studied in pancreatic NETs [11, 13].

Two-thirds of patients with SI-NETs in the present series underwent re-biopsy at a mean follow-up time of 48.8 (SD: ±162.5) months due to radiological progression on cross-sectional imaging. However, no significant association was seen between grade progression and relevant radiological progression according to RECIST 1.1. These findings confirm the notion that SI-NETs are generally more indolent neoplasms, as compared to their pancreatic counterpart. Although time to radiological and/or grade progression is indeed longer than the pertinent figure in pancreatic NETs, the SI-NET patients with longitudinal Ki-67 increments with or without grade progression have a relatively worse prognosis and may need a stricter surveillance protocol and initiation of earlier additional targeted therapies based on the tumour’s biology.

As SI-NETs generally have a prolonged disease course, it may be reasonable to assume that Ki-67 indices might change over time; hence, histological re-evaluation to assess increments in tumour proliferative activity is sometimes performed in cases with rapidly progressive disease, resistance to therapy of certain lesions, and discrepancies in radiological and functional imaging [13]. The sparsity of data regarding longitudinal Ki-67 changes, particularly in SI-NETs, may imply that histological re-evaluation may be meaningful in selected cases. However, the timing and impact of histological re-assessment on SI-NET patients’ clinical management is not clearly delineated in contemporary ENETS guidelines and remains to be further assessed. Furthermore, as Ki-67 PI has an established role in the grading, prognostic classification, and management of SI-NET patients, longitudinal changes in Ki-67 indices at re-biopsy can provide additional information over conventional and functional imaging and alter the intensity of imaging surveillance and overall management in a significant number of SI-NET cases, as compared to biologically stable ones.

Several studies over the last decade have shown that the Ki-67 PI is a crucial histopathologic parameter in predicting the tumour biological behaviour across different NET primaries. However, most of these studies and also the revisited 2017 WHO classification for NETs were based on pancreatic NET histopathology and the underlying molecular discrepancies between neuroendocrine carcinomas, exhibiting TP53 and RB1 mutations and WHO G3 NETs with MEN1 (multiple endocrine neoplasia type 1), DAXX (death-domain associated protein)/ATRX(alpha-thalassemia/mental retardation X-linked) mutations [8, 21, 22]. Importantly, SI-NETs are genetically silent tumours and demonstrate a lesser relationship with mutational driver events. Putative driver mutations in CDKN1 and APC have been identified in SI-NETs; however, these were encountered in only 10% of cases. Therefore, tumorigenesis of SI-NETs seems to depend more on chromosomal alterations, including loss of chromosome 8 and epigenetic events, which converge to hyperactivate the PI3K/mTOR, MAPK, and Wnt pathways [23, 24]. Recently, a modified site-specific histopathological grading system for SI-NETs was proposed applying Ki-67 cut-offs of 5% and 10%, which appears to predict differences in SI-NET patient survival outcomes more accurately, but further translational studies are needed to elucidate the underlying molecular mechanisms and epigenetic alterations in each modified grade SI-NET category [9, 12, 23].

There is indeed an immense need for new, reliable, and more accurate biomarkers, which will be adopted in clinical practice and may guide the management of SI-NETs in the future, given both the proven inefficacy and reduced clinical utility of traditional biomarkers, such as chromogranin A. The expression of the oncofetal protein IMP3, for example, at the nodal site of SI-NETs has been associated with shorter progression-free survival, independently of the Ki-67 index [25]. In addition, the constantly developing field of liquid biopsies is emerging as a promising strategy that may meet the criteria for the management of NETs, compared to tissue biopsies, as it offers data complementary to imaging, without radiation exposure, and can probably reduce the intensive use of radiological examinations. Molecular multigenic analysis of transcriptional derivatives of NETs (NETest) has already been proven superior to chromogranin A, with various clinical applications in the diagnosis and management of NETs [26].

Although accurate determination of Ki-67 PI in SI-NETs is of paramount importance for disease prognostication, SI-NET primaries may be difficult to biopsy owing to their small tumour size and difficult accessibility. In addition, different sites of disease in the same patient may exhibit different Ki-67 PIs and may respond differently to treatment. Dual-tracer PET-CT imaging using SSTR-targeted PET tracers, and FDG, have recently gained widespread acceptance for better assessment of whole-body tumour biology compared to single-site histopathology, guidance of targeted biopsies but also being non-invasive in terms of evaluating inter- and intra-tumoral heterogeneity [27].

Our study has several limitations. The multicentre design may account for certain confounding factors owing to differences in treatment strategies and surveillance protocols including different techniques and modalities at varying intervals among the five tertiary centres involved. However, a rigorous study protocol complying to ENETS guidelines and appropriate analytical approach was applied to ensure internal validity and minimize inter-site variability of the present study. Our study is also limited by a lack of central histopathology review that mainly concerns inter-study concordance for Ki-67 assessment of the surgical specimen and liver biopsies across the participating centres. In addition, tumour heterogeneity and possible discordant Ki-67 expression between primary tumour and metastatic site may also have introduced bias in our study analysis [28]. Nevertheless, the data presented are mainly derived from tertiary ENETS Centres of Excellence, where dedicated NET clinicians and expert pathologists with experience in the field of NETs were involved and re-consulted the obtained samples, according to the ENETS recommendations [6, 19]. In addition, as most patients in this series were subjected to re-biopsies due to radiological progression and the remaining for diagnostic purposes or symptomatic relief, a certain selection bias could be introduced in our study; however, grade progression at re-biopsy did not differ significantly between patients with and without radiological progression. Moreover, we had no comprehensive data on functional imaging (68-gallium-PET-CT and FDG-PET-CT) at the time of re-biopsy to make clinical inferences on the association of functional imaging findings with grade progression. Our study is also subjected to referral bias as many of the included cases were clinically challenging and were re-assessed and treated in tertiary referral centres. Finally, in order to eliminate immortal time bias in the present study and also selection bias of cases chosen for re-biopsy, the baseline for OS estimates was set at the time of re-biopsy.

We hereby provide further evidence for the clinical utilization of a novel histopathological grading system with 5 and 10% Ki-67 cut-offs to assess longitudinal changes in patient tumour samples and suggest that evolution into more aggressive disease histopathologically, albeit with preserved tumour differentiation, likely confers worse survival outcomes. As approximately 43% of SI-NET patients undergoing re-biopsy had evident increase in the observed Ki-67 indices, disease evolution could be a common feature of SI-NET biology. Therefore, further studies on molecular pathways of SI-NETs exhibiting grade progression are necessary to shed light on the mechanisms that render a neoplasm more aggressive during the disease course. In the era of personalized NET management, our findings may provide further aid in disease prognostication, with the aims to identify appropriate stage III and IV SI-NET candidates to be selected for tailored surveillance protocols and adjuvant targeted treatments in future clinical and translational studies. Importantly, to effectively manage SI-NET patients, NET specialists must adopt an approach based on multidisciplinary decision-making and the implementation of novel classification systems with precise comprehension of patient and tumour information including longitudinal changes in Ki-67 proliferation rate, as well as information from novel biomarkers and theranostics.

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards and written informed consent was obtained for participation in this study. In particular, the study protocol was reviewed and approved by the Swedish Ethical Review Authority with Drn number: 2021-01277; the Ethics Committee of Medical University of Silesia with Approval No. KNW/0022/KB1/102/II/17/19; and the Scientific Committee of EKPA-Laiko Hospital, National and Kapodistrian University of Athens, with Approval No. 602. In UK collaborating centres, the study was officially registered as an audit in the University Hospitals Coventry and Warwickshire (UHCW) in Coventry with ID number: 828 and the King’s College Hospital in London with ID number: Gastro-020.

The authors declare that they have no conflicts of interest.

The study did not receive any funding.

Kosmas Daskalakis: conceptualization; formal analysis; investigation; methodology; project administration; software; supervision; visualization; writing – original draft; and writing – review and editing. Marina Tsoli and Maria Wedin: data curation; formal analysis; investigation; methodology; and writing – review and editing. Angelika Kogut: data curation and writing – review and editing. Rajaventhan Srirajaskanthan: data curation; investigation; methodology; supervision; and writing – review and editing. Dominique S.V.M. Clement and Georgios Giovos: data curation; investigation; and writing – review and editing. Beata Kos-Kudla: data curation; investigation; supervision; visualization; and writing – review and editing. Martin O. Weickert: data curation; investigation; supervision; and writing – review and editing. Gregory Kaltsas: conceptualization; investigation; methodology; supervision; validation; and writing – review and editing. Martin O. Weickert and Gregory Kaltsas: last authorship.

Martin O. Weickert and Gregory Kaltsas share last authorship.Gregory Kaltsas is an Editorial Board Member for Neuroendocrinology. This is an invited manuscript.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from the corresponding author (K.D.; kosmas.daskalakis@oru.se) upon reasonable request.