Impact of neuroendocrine morphology on cancer outcomes and stage at diagnosis: a UK nationwide cohort study 2013–2015

Background The diagnosis of neuroendocrine neoplasms (NENs) is often delayed. This first UK population-based epidemiological study of NENs compares outcomes with non-NENs to identify any inequalities. Methods Age-standardised incidence rate (ASR), 1-year overall survival, hazard ratios and standardised mortality rates (SMRs) were calculated for all malignant NENs diagnosed 2013–2015 from UK national Public Health records. Comparison with non-NENs assessed 1-year overall survival (1YS) and association between diagnosis at stage IV and morphology. Results A total of 15,222 NENs were identified, with an ASR (2013–2015 combined) of 8.6 per 100,000 (95% CI 8.5–8.7); 4.6 per 100 000 (95% CI, 4.5–4.7) for gastro-entero-pancreatic (GEP) NENs. The 1YS was 75% (95% CI, 73.9–75.4) varying significantly by sex. Site and morphology were prognostic. NENs (predominantly small cell carcinomas) in the oesophagus, bladder, prostate, and female reproductive organs had a poorer outcome and were three times more likely to be diagnosed at stage IV than non-NENs. Conclusion Advanced stage at diagnosis with significantly poorer outcomes of some NENs compared with non-NENs at the same anatomical site, highlight the need for improved access to specialist services and targeted service improvement.

Intelligence & Surveillance Unit and Public Health Agency Queen's University Belfast. These registries had coverage of a 2015 midyear total population of 65,110,000. Socioeconomic status (not available for Scotland) was measured by deprivation quintiles based on: area deprivation for England and N. Ireland, produced by the Ministry of Housing, Communities and Local Government (formerly the Department for Communities and Local Government); and aspects of deprivation experienced in income for Wales, produced by the Welsh Government using the Welsh Index of Multiple Deprivation. English data for non-NENs was also available for comparison.
Tumour classification NENs were defined by the WHO 2015 classification excluding diffuse pulmonary neuroendocrine hyperplasia. Tumours occurring at all anatomical sites between C00 and C80 according to the 10th edition of the WHO International Classification of Disease (ICD-10) codes were included, and morphology codes included 8013 (excluding lung [C34 and C78]), 8041-8045 (excluding lung), 8150-8158, 8240-8247, 8249 and 9091 according to the WHO International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3). All were behaviour code 3 (malignant), 6 (metastatic) or 9 (malignant, uncertain whether primary or metastatic).
Tumours were staged according to the American Joint Committee on Cancer (AJCC) TNM staging for NENs 5 and the European Neuroendocrine Tumor Society (ENETS) recommendations for the staging of GEP-NENs 6 and graded using the ENETS grading system (WHO 2010) for GEP-NENs, [7][8][9] or according to pathological grading of differentiation for other sites. Ki-67 status was not uniformly captured and is not recognised yet in many organ systems. Due to this being a registry study, we were unable to revisit each tissue specimen to acquire it, most grades were therefore based on a morphology of "poorly differentiated" or "well-differentiated" so grade 1 and grade 2 were combined to reduce inaccuracies. Tumour grades were classified: grade 1 and grade 2 combined-NET G1/G2 (functioning tumours including insulinomas, glucagonomas etc. with morphology codes 8150-8153, 8155-8158 and tumours with morphology codes 8240-8242, 8249, and 9091); grade 3-NEC G3 (large cell carcinoma, small cell carcinoma, combined small cell carcinoma, neuroendocrine carcinoma NOS with morphology codes 8013, 8041-8045, and 8246); Mixed Adeno-Neuroendocrine Carcinoma (MANEC) (Now known as 'goblet cell adenocarcinomas') goblet cell carcinoid, mixed adenoneuroendocrine carcinoma, tubular adenocarcinoid tumour with morphology codes 8243-8245); and "other" (pancreatic endocrine tumour malignant, mixed pancreatic endocrine and exocrine tumour malignant, and Merkel cell carcinoma morphology codes 8150, 8154 and 8247, respectively).  10 and the 2013 European standard population. Of the cohort, 15,106 persons were included in survival and mortality analyses, excluding those with a death certificate only registration and including the first tumour in those with multiple tumours, or the tumour with known stage if multiple tumours had the same diagnosis date. 1-year overall survival (1YS) estimates (also for non-NENs) were made using the Kaplan-Meier method with log rank tests. Multivariable survival analysis using Cox (proportional hazards) regression and likelihood ratios were performed with the model adjusting for clinically relevant independent variables sex, age, stage, site, morphology, grade, deprivation and year of diagnosis (all assessed for significant association with overall survival at p < 0.05). Standardised mortality rates, for those diagnosed between 1 January 2013  and 31 December 2015, for all-cause mortality over the disease  duration, adjusting for explanatory variables age, sex and calendar  period, were calculated by means of the life-table approach using ONS death registrations, including persons aged 0-100 years. 11,12 Multiple logistic regression estimated the odds ratios (OR) and 95% confidence intervals (CI) for association of diagnosis at stage IV with morphology (NEN or non-NEN), for all people diagnosed with cancer between 2013 and 2015 in England, adjusting for sex and age. Vital status on 1 January 2017 was acquired from ONS.

Patients
Overall, 15,145 persons were diagnosed with 15,222 NENs between 2013 and 2015, of which 7,640 (50.4%) were female. The average age at diagnosis was 65-69 years old with variation between sites ( Table 1). The median follow-up time was 19.6 months (range 0-49.2 months) and 24.8% (3,766) died within one year of diagnosis.

Incidence
The UK ASR was 8.6 per 100,000 (2013-2015 combined), 8.1 in females and 9.1 in males; 4.6 for GEP-NENs overall, the second most common gastrointestinal tumour. 13 Incidence by other sites are listed in Table 2.
The 2015 UK annual incidence was 8.7 per 100,000 (8.2 [95% CI, 7.8-8.5] in females and 9.2 in males [95% CI, 8.8-9.6]). It increased steadily, from 3.9 cases per 100,000 in 2001 to 7.9 per 100,000 in 2012 in England, at around 0.4 cases per 100,000 per year ( Fig. 1 and Table 2), although it must be noted that these figures are from ICD-O-2 coded data so do not include morphology codes 8013 and 8249.

Survival
The overall 1YS probability for persons diagnosed with NENs was 74% (Table 2), significantly higher in females 78% than in males 71% (p < 0.001 for comparison between sexes). There was a small but significant difference between the least deprived 77% and most deprived 73%. As expected, small cell neuroendocrine carcinoma had the poorest survival probability of all morphologies, 41.4%, whilst NETs had the highest at >90% (inclusive of all sites).
The median survival for oesophageal, prostate and bladder NENs were 5.7 (4.5-7.5), 7.8 (5.8-9.1) and 11.3 (9.9-12.8) months, respectively. Survival for the other sites exceeded 50% at the longest time period so median survival could not be calculated for these.
The 1YS probability for people diagnosed with pulmonary, pancreatic, stomach and small intestine NENs, predominantly well differentiated NETs, was much higher than for non-NENs at these sites. Conversely, the 1YS probability for people diagnosed with oesophageal, prostate, bladder and to a lesser extent female reproductive organ, colon, and breast NENs, predominantly NECs, was much lower than for non-NENs at these sites (Table 3) and these tumours were at least three times more likely to be diagnosed at stage IV. Around 42% of oesophageal NENs were stage IV and grade 3 tumour pathology with a 1YS probability of 22% (95% CI 16.7-28.5); the predominant morphology at this site was small cell neuroendocrine carcinoma (Table 1). Likewise, 75% of bladder and 74% of prostate NENs were small cell carcinomas or combined small cell carcinomas.
Survival decreased more with increasing stage (p < 0.001 for comparison between all stages) and grade (p < 0.001, for comparison between all grades) combined, identifying the combination as potentially a better prognostic indicator than each variable independently (Supplementary Table 2 Multivariable Cox analysis adjusting for predictor variables sex, age, stage, site, morphology, grade, deprivation and year of diagnosis on mortality (Table 2) determined the hazard of death was up to 18% lower for women than men; those aged over 75 years three times that of those aged under 54 years; and those with stage IV NENs more than twice that of those with stage III NENs. Some sites were associated with increased hazard, the hazard in those with gallbladder, anal and oesophageal NENs was similar to that of those with secondary tumours (HR 3.2 95% CI [2.5-4.1]) (Supplementary Table 1). Those who lived in the most deprived areas had up to a 41% increase in hazard of death when compared with the least deprived.

Mortality
SMRs were used to measure survival relative to the general population (Table 2). Deaths occurred nearly four times more frequently in those diagnosed with NENs; varying from nearly three times more in those aged over 75 years to 10 times more in those aged 55-64 years at diagnosis, and with the same frequency as those in the general public for those with stage I to 10 times more in those with stage IV cancer.

Incidence
The incidence of NENs appears to be rising in this and other international studies. 14,15 This rise may be real, or may be an artefact of the use of diagnostic imaging 15 with improved sensitivity, and increased clinical vigilance resulting in incidental detection of asymptomatic lesions. 14 Also, in this study the upward trend in incidence between 2013, the beginning of ICD-O-3 coding in the UK and 2015 is less pronounced than that seen previously. We need future studies using a wider timeframe to determine whether the incidence of NENs is still rising or beginning to plateau.
In the UK, the incidence of NENs in the rectum ranked after lung, small intestine, appendix, pancreas, stomach and colon and caecum combined, in comparison with the most recent international study, in the USA, 15 where rectum was the 3rd most common after lung and small intestine, with appendix last. In the USA, colorectal screening starts at 50 years old 16 compared with 55 years in the UK and could be improving detection; appendiceal NETs only became reportable in the USA from January 2015 (SEER   personal communication), explaining the comparatively lower incidence.
Survival and mortality Multivariable analysis identified sex, age, site, stage, cell morphology and deprivation to be independently associated with mortality. Survival probabilities for NENs in the UK had similar trends to previous international studies with male sex, increasing age, stage, and grade and decreasing socioeconomic status associated with a poorer outcome. 13,15 The much reduced 1YS and mortality in metastatic tumours (stage IV) when compared with localised tumours (stage I-III) was particularly notable and in-line  with a study of "all cancers" diagnosed in the UK, 17 warranting further measures to ensure that the early diagnosis of NENs is a priority. People diagnosed with oesophageal, bladder, prostate, and female reproductive NENs had predominantly small cell carcinoma morphology, significantly poorer outcomes consistent with the results of previous smaller studies and case series, [18][19][20][21] and were more likely to be diagnosed at stage IV. Some morphologies exhibited poorer survival probabilities than expected. This was the first large series study of MANEC reported and demonstrated a survival probability of 84% in comparison with the 95% previously reported. 22 These findings highlight the need for the development of sub-specialist services to match the clinical need in the groups which have poorer outcomes.

Limitations
Tumours with uncertain behaviour (behaviour code 1) are not consistently captured by UK cancer registries because they have historically been deemed "benign" and were not included. This means that many small and indolent NETs of the stomach, rectum, appendix and pancreas may not have been included in historical data. There is an under representation of type 1 gastric NETs and stage I and II rectal NETs-that all have excellent 5-and 10-year survivals. This under representation may be significantly skewing the survival statistics presented in this study. Also, the high incidence of G3 NECs in this study, particularly those in the lung, are not representative of previous findings. In the lung, G3 NEC is used when there has been insufficient information to classify the tumour as either small cell carcinoma or large cell neuroendocrine carcinoma. Thus, the high incidence of G3 NEC in the lung is likely to represent small cell carcinoma, large cell neuroendocrine carcinomas and mixed tumours. This misclassification of G3 NECs could also be skewing the survival analysis. This is a caveat for ongoing classification recommendations for the purposes of analysis and audit. In the future, many of these tumours will be more consistently captured with improvements in WHO terminology. The completeness of stage was only 61% for this cohort but is improving year on year. ICD-O-3 coding was implemented in 95% of the UK in 2013 (the exception Wales in 2016), prior to this morphology codes 8013 and 8249 were not available, many of these tumours coded carcinoma or adenocarcinoma NOS (not otherwise specified). Therefore, our incidence counts are underestimates, and survival analysis limited to only 1-year.
We calculated that the missing tumours from the Welsh cohort would equate to approximately 62 from common sites. Also, an audit of the English data using data collected by the NET Centre of Excellence at King's College Hospital found 14.6% of the tumours captured were not captured by NCRAS, many behaviour 1. This would suggest the true incidence of NENs could be as high as 10 per 100,000 per year.
The capture of NENs irrespective of behaviour code, ICD-O-3 coding in Wales from 2016 and improvements in data completeness will resolve the afore mentioned issues in the future.
We were unable to calculate disease-specific survival and mortality due to death registrations by ICD-10 coding. However, recently it has become possible to link morphology to death registrations which will allow this in future studies.

CONCLUSION
This is the first population-based epidemiological study of NENs diagnosed in the UK, the results of which will be of value in future service planning. We determined that outcomes varied greatly between sites and morphologies. Comparison of NENs with non-NENs at the same site also found disparities. These differing outcomes between patient groups have a direct relationship with the healthcare resource required to manage them; and highlight the need for NENs to be seen at Centres of Excellence. Earlier diagnosis is necessary, and the identification of more diagnostic markers for NENs, particularly for extrapulmonary small cell carcinomas, required.
Future studies will allow 3-and 5-year survival estimates, and the availability of treatment data will allow more in-depth studies, and we are developing ways of collecting more detailed imaging, biochemistry, screening and symptom data which ultimately will be linked to the current cancer registry data. Pathologists have been instructed to include small benign pancreatic NETs as malignant as per WHO terminology and detail Ki67 on all specimen. Ki67 has not been routinely captured by the UK cancer registries to date, however this will change in mid-2020 with the implementation of version 9 of the Cancer Outcome and Services Data set (COSD), the national standard for reporting cancer in the NHS in England, in which Ki67 will be a required field and so captured by the English registry. Although the UK cancer registries do not currently hold complete data on MEN-1 diagnosis, the recent addition of genetic data means that this may also be available for future studies.