1195P - Neuroendocrine Tumors (NETs): How Big Data (BD) techniques applied to Electronic Health Records (EHR) can improve our understanding of this patient population

Background

Low incidence and heterogeneity of NETs complicates the study of epidemiological and clinical pathways in this population. There is an unmet need for large datasets and BD tools to fill this knowledge gap. Our study aimed to describe the demographic and clinical characteristics of patients with NETs in Spain through the application of BD techniques to EHR.

Methods

This was a retrospective, observational study in adult patients with NETs managed in 5 Spanish centers between 01/01/2014 to 31/12/2019. An unstructured search of free text from EHRs was performed by artificial intelligence (AI) techniques based on SAVANA’s EHRead®, an innovative Natural Language Processing and machine learning system. Patients were included when NET was first identified (Index Date) and classified as `Incident´ or `Prevalent´ if the Index Date was within the study window or not, respectively.

Results

A total of 2,939,309 patients were screened by EHRead and 1,256 NETs cases were found: median age 60y (95% CI 59 - 62), 52% female, 41% with gastroenteropancreatic origin, and 44% were categorized as incident (Table). More than 70% of patients presented symptoms prior to diagnosis and >10% had metastases at diagnosis (>70% were non-resectable). Median time from symptom onset to first diagnosis was 0.6 years (95% CI 0.4 - 0.73). The most frequently detected comorbidities were hypertension (66%), diabetes (35%), asthma (22%), and depression (17%). A previous diagnosis of cancer different from NETs was reported in 21% of cases. Uncommon clinical events included brain metastases (1%) and familial history of NETs (1.1%). Table: 1195P

Classification of patients with NETs

* Paraganglioma, adrenal gland and medullary thyroid carcinomas and pheochromocytomas§ Within +/-6 months of Index Date

Conclusions

BD Techniques should improve generation of large datasets in low incidence neoplasms and help identify uncommon events. NETs diagnosis is delayed about 6m after symptom onset and comorbidities may influence the choice of treatment.

Clinical trial identification

Editorial acknowledgement

Editorial assistance was provided by Content Ed Net (Madrid, Spain)

Legal entity responsible for the study

IPSEN Pharma.

Funding

IPSEN Pharma.

Disclosure

J. Hernando: Financial Interests, Personal, Advisory Board: Eisai, Ipsen, Novartis, AAA, Angelini, Pfizer, Roche. All other authors have declared no conflicts of interest.