Survival outcomes for lung neuroendocrine tumors in California differ by sociodemographic factors

Introduction

Neuroendocrine neoplasms (NENs) are a heterogeneous group of cancers that arise from neuroendocrine cells in almost any organ in the body. NENs can produce hormones and are classified histologically as well-differentiated neuroendocrine tumors (NETs) or poorly differentiated neuroendocrine carcinomas (NECs). Well-differentiated NETs of the lung account for less than 3% of all primary lung cancers in adults (Rekhtman 2010). However, the lung is the second most common NET primary site after the gastrointestinal tract (Modlin et al. 2003). Lung NETs have a wide spectrum of clinical behavior and a unique pathologic classification as either typical carcinoids (grade 1) or atypical carcinoids (grade 2) (Travis et al. 2015).

There are few established predictors of mortality for lung NETs beyond disease-specific factors like stage and grade (García-Yuste et al. 2007, Steuer et al. 2015). One recent publication found that sociodemographic factors were associated with survival for patients with lung NENs in the United States, including sex, marital status, insurance, race, and county of residence – with better survival noted for women, married, insured, Hispanic and ‘other’ race, and urban populations (Shah et al. 2021). However, this analysis did not account for treatment factors or socioeconomic status (SES) of cases, which may be important mediators of survival outcomes. In addition, the analysis included both lung NETs and NECs (small cell and large cell) in the same survival models. Given that well-differentiated lung NETs have different biology, genomic profiles, underlying risk factors, and treatment strategies than high-grade pulmonary NECs (Fernandez-Cuesta et al. 2014, Rekhtman 2022), we aimed to examine predictors of survival for well-differentiated lung NETs, specifically.

Using the California Cancer Registry (CCR), we characterized the clinical characteristics and survival outcomes of patients diagnosed with well-differentiated lung NETs in California from 1992 through 2019. In addition, we evaluated the impact of sociodemographic and treatment factors on survival.

Materials and methods

Results

Univariate OS

Kaplan–Meier survival analyses showed statistically significant differences in survival by sociodemographic factors and social determinants of health (Fig. 1). OS was worse for men vs women (Fig. 1A) for non-Hispanic Black cases compared with non-Hispanic White (P = 0.014) or Hispanic (P < 0.001) cases (Fig. 1B), for unmarried cases (Fig. 1C), for cases living in suburban counties (Fig. 1D), and for cases living in the lowest nSES quintile (Fig. 1E). For cases diagnosed after 1995, compared to people with private health insurance, people with Medicare had worse survival (P < 0.001), as did those with Medicaid, military, or other public insurance (P < 0.001; Fig. 1F). We also confirmed associations of stage (Fig. 2A) and histology (Fig. 2B) with OS. We found similar differences in survival by histology for the subset of cases diagnosed after 1997 (data not shown). In contrast, OS did not differ by diagnosis decade (Fig. 2C).

View Full Size

Kaplan–Meier overall survival curves by sociodemographic variables. NH, non-Hispanic, nSES, neighborhood socioeconomic status. *Insurance payer was not required in CCR prior to 1996, so insurance data are presented for the n = 5527 cases diagnosed from 1996 onward.

Citation: Endocrine-Related Cancer 31, 1; 10.1530/ERC-23-0068

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Kaplan–Meier overall survival curves by disease characteristics.

Citation: Endocrine-Related Cancer 31, 1; 10.1530/ERC-23-0068

• Download Figure
• Download figure as PowerPoint slide

Discussion

Lung NET epidemiology has been studied less than other primary sites. Nearly all are sporadic, although rarely they are associated with familial genetic syndromes (Oliveira et al. 2001). The contribution of chronic inflammation, underlying pulmonary conditions like asthma or chronic obstructive pulmonary disease, or environmental factors like smoking are poorly understood. Some patients develop diffuse idiopathic pulmonary neuroendocrine cell hyperplasia as a precursor to lung NETs but not a majority (Samhouri et al. 2023). Both the underlying mechanisms driving NET pathogenesis and clinical outcomes may vary by case.

Here, we report a comprehensive summary of patients diagnosed with lung NETs in California’s population-based cancer registry. More cases were diagnosed in the most recent decade of our study compared with the earliest decade, including more with metastatic disease. Underlying contributors to this finding may include better detection of advanced disease by imaging, improved recognition of NET histologies by pathologists, and potentially rising incidence.

Women were a majority of the lung NET population, consistent with prior studies (Steuer et al. 2015, Shah et al. 2021). The reasons for this sex difference are unclear. Women might be diagnosed more frequently than men as a consequence of higher healthcare-seeking behavior and more contact with the healthcare system (Bertakis et al. 2000). However, lung cancer screening by low-dose computed tomography is equally infrequent in women and men (Jemal & Fedewa 2017), and recent screening recommendations would not explain our findings across all decades. Furthermore, we found no difference by sex in stage at diagnosis, suggesting that increased early-stage detection does not explain why lung NETs occur predominantly in women. A biologic mechanism for the sex difference is possible, perhaps through a role for hormones like estrogen in regulating gene expression and promoting neuroendocrine cell proliferation. There may be unrecognized differences in behavior or risk factors that predispose women to lung NETs. Further study is merited to clarify the mechanism.

We observed that lung NET diagnoses are associated with race and ethnicity along with several social determinants of health. Californians self-identifying as non-Hispanic White were a higher proportion of our lung NET cohort compared to statewide demographics (Humes et al. 2011), while other racial and ethnic groups were under-represented in the lung NET population, particularly Hispanic and Asian/Pacific Islander. Non-Hispanic White cases were less likely to have metastatic disease at diagnosis compared with other racial/ethnic groups. People living in the highest statewide nSES quintile were also over-represented in the lung NET cohort, and they were less likely to have metastatic disease at diagnosis compared with the lowest quintile. Finally, people with private health insurance were less likely to have metastatic disease compared to those with public or no insurance.

The impact of sociodemographic factors on NET diagnoses and stage at diagnosis could be an artifact of increased detection of early-stage cancer in privileged groups and a corresponding underdiagnosis of lung NETs (particularly early-stage) in vulnerable populations, including people of color, people living in economically disadvantaged communities, and people with public or no health insurance. NETs can be challenging to diagnose, even for experienced pathologists. Lung NETs are occasionally misclassified as other malignancies (Kasajima et al. 2022), and this misclassification may be more common at medical centers that serve vulnerable populations. Alternatively, sociodemographic differences could be explained by underlying risk factors and/or health-related behaviors associated with both social determinants of health and the pathogenesis of lung NETs. Importantly, we do not believe a genetic explanation for the racial and ethnic makeup of the lung NET cohort is likely, given that the vast majority of cases are sporadic, and considering the lack of consistent correlation between self-identified race and ethnicity, genetic ancestry, and genomic variants (Bryc et al. 2015).

We also report an in-depth analysis of predictors of OS and LCSS for lung NETs in California. Prior studies demonstrated that age, histology, and stage are associated with survival, with worse OS for older patients, atypical vs typical carcinoid histology, and metastatic compared with localized disease (García-Yuste et al. 2007, Filosso et al. 2015). We confirmed these associations in our multivariable analyses. In addition, we identified several sociodemographic factors also associated with survival. As reported in another SEER study (Shah et al. 2021), we found that sex, marital status, and health insurance were associated with OS. Women had better OS than men, even after adjusting for other prognostic factors including sociodemographic factors, disease characteristics, and first course of treatment. There may be differences in biology and other underlying risk factors by sex; for example, we found that primary tumor size differed between women (median 1.8 cm, IQR 1.2, 2.7) and men (median 2.3 cm, IQR 1.5, 3.5; P < 0.001). Non-Hispanic Black Californians had worse OS than non-Hispanic White Californians in a univariate model, consistent with one prior study of atypical carcinoid tumors (Steuer et al. 2015). However, this survival difference was eliminated after adjustment for other prognostic factors and appeared to be mediated by disease characteristics (in particular, stage at diagnosis) along with other social determinants of health.

We also report a new finding that nSES was strongly predictive of OS in lung NETs, with lower all-cause mortality for cases living in higher nSES quintiles. This association persisted despite adjustment for other sociodemographic, disease, and treatment-related variables, suggesting that the environment in which one lives has implications for longevity after a lung NET diagnosis.

As reported in a prior study, we found a clear survival benefit for married/partnered individuals (Shah et al. 2021). Interestingly, stage at diagnosis did not differ by marital status, so the survival benefit is not explained by earlier detection alone. Health insurance was also independently associated with OS. One prior study reported worse survival for people with insurance compared to the uninsured but did not provide granular information about insurance type (Shah et al. 2021). In our analyses, we found worse OS for people with Medicare or public health insurance compared to those with private insurance only, even after stratifying by age and adjustment for disease and treatment variables.

As many lung NET patients experience prolonged survival, the longitudinal risk of noncancer death is important to consider when assessing outcomes. Our analysis of LCSS revealed both similarities and differences compared with OS associations. Non-Hispanic Black cases had worse LCSS in our partially adjusted model including stage, but this attenuated after adjustment for histology, likely reflecting how non-Hispanic Black cases were more likely to have atypical carcinoid histology (Supplementary Table 1). We also report a novel finding that American Indian cases (a group not previously well studied for NET outcomes) had worse LCSS compared with non-Hispanic White cases. This finding should be replicated in additional studies, as the number of cases in our dataset was small. As with our OS analysis, we found that sex and marital status were strongly associated with LCSS across all models. However, neither nSES nor insurance was associated with LCSS in any multivariate models; these variables may be important predictors of general health outcomes rather than lung NET-specific mortality.

It comes as no surprise that social determinants of health are associated with survival. Given that lung NETs often progress over years, there is ample time for social factors to compound and impact outcomes. Marital status has been associated with survival across a range of malignancies (Aizer et al. 2013) and can be a proxy for socioeconomic resources and other health-related factors including diet, physical activity, preventative care, and more aggressive treatment. Interestingly, the association between marital status and survival was not attenuated by controlling for nSES, treatments, or health insurance. Similarly, SES is a fundamental cause of health. Higher SES individuals can flexibly deploy resources to avoid health risks and mitigate the effects of poor health. One limitation of our study is the absence of an individual-level measure of SES. However, nSES may influence health through distinct mechanisms than individual SES; for example, through access to medical care, quality of education, environmental exposures, crime, crowding, recreational opportunities, and municipal services.

Based on a prior population-based study in Ontario, Canada (Hallet et al. 2015), and a SEER publication in the United States (Shah et al. 2021), we hypothesized that rural residents in California would have worse survival. Instead, univariate models suggested that suburban, not rural, residents have the worst OS, and there were no associations between county of residence and survival in either multivariable OS or LCSS models. We also found heterogeneity by nSES in the associations between rural/urban status and mortality. Californians of low nSES had worse OS in suburban counties compared with urban counties. In contrast, for Californians of high nSES, there were no survival differences by county. The reasons for worse survival for low nSES residents of suburbs may reflect differences in access to healthcare resources and/or differences in chronic medical conditions, health-related behaviors, or other aspects of the built environment. Compared to suburban residents, rural residents may be referred more often to NET centers of expertise, which have been shown to improve NET outcomes (Baeg et al. 2021). In contrast, socioeconomically advantaged people may find their way to NET centers of expertise regardless of where they live.

Beyond sociodemographic and clinicopathologic factors, we also confirmed strong associations between lung NET treatments and survival. Surgery was associated with improved OS and LCSS. This finding is consistent with existing literature; surgical resection is the primary treatment for localized lung NETs and is usually done with curative intent (Raz et al. 2015, Steuer et al. 2015). Chemotherapy and radiation were associated with worse OS and LCSS, while immune therapy was associated with worse OS but not LCSS. These treatment associations may reflect residual confounding due to disease aggressiveness and patient comorbidities. In particular, chemotherapy is generally reserved for lung NETs with the most aggressive biology, so we interpret the association with higher mortality as a reflection of disease severity rather than the role of chemotherapy itself. Conversely, patients who undergo surgery are likely healthier overall with a better prognosis. Finally, we found improved OS and LCSS in the more recent two decades of our study compared with the earliest decade, even though cases diagnosed in the most recent decade were more likely to have metastatic disease. This association was independent of treatment variables. However, because the CCR only captures treatment during the first year after diagnosis and many lung NET patients have extended survival, it is still possible that treatment advances may explain recent improvements in survival.

This study has several strengths. As a population-based registry that captures new cancer diagnoses across California, the CCR is less subject to selection bias compared with cases series from single institutions. California is a large and diverse state that is well-suited to address questions of disparities. Our study also has limitations. Given that we only included cases in California, our conclusions may not be generalizable to the broader United States or global population with lung NETs. Because NETs are challenging to diagnose and classify, there is likely some selection bias for who is diagnosed. The classification of tumor grade for NETs has also changed over time, so there is likely some degree of histologic misclassification in the dataset. Because we relied on topography to determine lung NET-specific cause of death, it is possible we overestimated hazards for non-NET cancers arising in the lung. Furthermore, as an observational study, our results are subject to residual confounding. We attempted to control for many potential confounders, but we were limited to information collected in the CCR. Our analysis could not account for treatments throughout the duration of illness, or the impact of other comorbidities and lifestyle factors such as smoking, disease-related factors like Ki-67/mitotic count, or hormone overproduction/functional status on outcomes. Lead time bias may also impact the observed outcome differences of various groups. Type I error is also possible due to multiple comparisons. However, while chance may have theoretically played a role, the consistency of our findings across multiple models is reassuring, and our results have both biologic plausibility and supportive data in other cancer types.

In conclusion, we report novel findings that sociodemographic factors, along with clinicopathologic and treatment factors, are associated with OS and, to a lesser extent, LCSS in lung NETs. We believe these results will influence future research into the pathogenesis of lung NETs and help identify opportunities for interventions to reduce disparities and improve survival for these increasingly common malignancies.

Supplementary materials

This is linked to the online version of the paper at https://doi.org/10.1530/ERC-23-0068.

Declaration of interest

CKM has served as an advisor to Jazz Pharmaceuticals and received research funding to her institution as local clinical trial principal investigator from RayzeBio. MAG has served as an advisor to AnHeart, AstraZeneca, BMS, Cardinal Health, Genentech/Roche, Genzyme, Gilead, Guardant, iTeos, Sanofi, Summit, and Surface and received research funding to his institution as local clinical trial principal investigator from Amgen, Celgene, JNJ, Merck, Novartis, OncoMed, and Trizell. All other authors declare no conflicts of interest.

Funding

This work was supported by a Mentored Award from the Neuroendocrine Tumor Research Foundation (grant number 654559).