Primary-site variation in recorded metastatic patterns and survival associations in extrapulmonary poorly differentiated neuroendocrine carcinoma
Year: 2026
Type: article
Study design: Observational Study
Abstract: Background: Extrapulmonary poorly differentiated neuroendocrine carcinoma (EP-PD-NEC) is aggressive and heterogeneous. Our primary objective was to assess primary-site differences in bone, brain, liver, and lung metastases recorded at diagnosis. Secondary analyses examined prognostic effect modification and the incremental predictive information in these variables. Methods: We analyzed Surveillance, Epidemiology, and End Results data from 2010-2023. Strict EP-PD-NEC was defined by ICD-O-3 codes 8013 and 8041-8045. Bone, brain, liver, and lung metastases recorded at diagnosis were evaluated using modified Poisson regression with individual-level heteroscedasticity-consistent standard errors. Overall survival (OS) was analyzed with multivariable Cox models containing primary-site-by-organ interactions. Associations identified in 2010-2017 were reassessed in 2018-2022. Four prespecified ridge-penalized Cox models were developed and temporally evaluated in distant-stage cases with complete four-organ status. Primary prevalence analyses included patients with known status for the organ analyzed; sensitivity analyses recoded unknown status as negative or used multiple imputation. Results: Among 6,044 eligible patients, 5,820 (96.3%) had complete four-organ information and 2,356 had at least one recorded organ involved. Recorded metastatic prevalence differed across six primary-site groups for every organ (all false-discovery-rate-adjusted P<0.001). Versus genitourinary primaries, pancreatic primaries had higher adjusted prevalence of liver (prevalence ratio [PR] 3.80, 95% confidence interval [CI] 3.39-4.25) and brain metastases (PR 2.86, 95% CI 1.75-4.66). An increasing recorded four-organ count was associated with shorter OS. The global primary-site-by-organ interaction was significant (likelihood-ratio χ²=43.49, 20 df, P = 0.00176), but the full interaction matrix added negligible temporal predictive value (ΔC-index=0.00169). In 1,374 later-period patients, the spectrum model had a C-index of 0.607 (95% CI 0.588-0.624), a 12-month area under the curve of 0.672 (95% CI 0.639-0.706), and a Brier score of 0.196 (95% CI 0.185-0.207). Compared with the four-organ-count model, paired bootstrap differences were 0.0085 (95% CI 0.0035-0.0139) for the C-index, 0.0119 (95% CI 0.0033-0.0231) for 12-month AUC, and -0.0020 (95% CI -0.0038 to -0.0004) for the Brier score. Missing-status analyses produced similar prevalence estimates. Conclusions: Recorded metastatic patterns differed by primary site in EP-PD-NEC. Organ identity added little predictive information beyond the recorded four-organ count, and site-by-organ interactions yielded negligible improvement. Overall discrimination remained modest.
Source: Frontiers in Endocrinology
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $3,801
Funder: National Cancer Institute
OpenAlex ID: W7223174014