Abstract
Various social determinants of health (SDOH) metrics, also known as area-based social measures, are utilized to evaluate access to cancer care and to explain disparities in outcomes. Little prior work has compared the validity of these various geographic metrics. We reviewed all patients surgically treated for PNETs (2006–2022) at a single comprehensive cancer center. We collected patient demographics including self-reported race (White or Black), billing addresses, tumor characteristics and area-based social measures. We then compared between- and within-race differences to understand accuracy across different geographic levels. One hundred seventy-nine patients were included; 49 (27%) Black, a median age of 60.3 years and 86 (48%) females. At the block group/census tract level, compared to White patients, Black patients lived in neighborhoods with lower educational attainment, lower income, higher rates of uninsurance, higher overall social vulnerability index (SVI), and higher area deprivation index (ADI) (all P < 0.05). These differences, however, were masked when examining county-level area-based social measures. Compared to census block group/tract-level data, for White patients, zip code-level metrics underestimated income and overestimated uninsurance level (P < 0.05). County-level metrics underestimated White patients’ income and education level but overestimated poverty, uninsurance rate and SVI (all P < 0.05). For Black patients, zip code-level metrics overestimated poverty and uninsurance rates (P < 0.05); the only inaccurate county-level metric was overestimation of SVI (P < 0.001). Black patients with PNETs experience more vulnerable area-based social measures, a disparity which may be hidden when analyzing large geographic metrics.
Introduction
Pancreatic neuroendocrine tumors (PNETs) are the second most prevalent pancreatic malignancy, with an estimated 4,032 new cases diagnosed annually and rising (Yao et al. 2008, Lubner et al. 2011, Ren et al. 2019, Wang et al. 2019, Gosain et al. 2020, Wu et al. 2024). Unfortunately, disparities exist in PNET treatment and outcomes, with Black patients suffering worse outcomes than their White peers (DePalo et al. 2019, Xu et al. 2021, Zheng-Pywell et al. 2022, Holland & Rose 2024, Mirza et al. 2024). The explanation for this is likely multifactorial and may include differences in access to care and tumor biology (Herring et al. 2022).
One way to explore the causes (and potential solutions) of disparities between demographic groups is by analysis of exposure to adverse social determinants of health (SDOH). These are area-based environmental and societal factors that influence patient outcomes and are known to disproportionately affect racial minorities (Notterman & Mitchell 2015). They may be collected at a variety of geographic levels (Notterman & Mitchell 2015, Schillinger 2020, Johnson et al. 2021). There is evidence that area-based social measures such as rural residence and single marital status are associated with delayed presentation and poorer PNET outcomes, but this has not been explored as a driving factor for the observed racial disparities in outcome (Cheung 2013, Gosain et al. 2020).
In order to answer this question, one must examine the potential mediating effects of adverse area-based social measures on racial disparities in outcomes. Area-based social measures are often measured using the United States Census in order from smallest to largest geographic region: individual reporting, census block group, census tract, zip code and county (Diez-Roux et al. 2001, Krieger et al. 2002, Fontenot et al. 2018). Large national or state-level registries (i.e., SEER and NCDB) provide area-based social measures primarily at larger geographic area levels (i.e., zip code, county) in order to simplify analyses, maximize anonymity and limit the labor-intensive process of gathering individual-level data. This lack of granularity limits applicability of results to individual patients. There are no generally agreed upon validated measures of the most influential area-based social measures on outcome, and this likely varies by population of study and disease state.
We often assume that individual-level measures will correlate with area-based social measures (Diez-Roux et al. 2001). However, this can lead to dangerous overgeneralizations especially when comparing races (Shavers 2007, Pearson et al. 2023). Little prior work has been done in the general or a cancer population to confirm this or compare the various area-based social measures by geographic area. PNET in particular is a malignancy that affects populations uniquely, given it has no screening criteria and is usually found incidentally (Harrelson et al. 2023). This begs the question of whether previously published conclusions using large-scale registries for patients with PNET are valid or generalizable for different races (Fouad et al. 2017, Xu et al. 2021).
The objective of this study was to compare the accuracy and validity of area-based social measures for Black and White patients utilizing single-institutional data in a racially diverse PNET population. To understand and quantify differences in area-based social measures, we compared between- race differences across different geographic levels. We also compared these area based social measures within races to understand the accuracy of each metric when compared to the gold standard of individual, census block group and tract-level metrics. We hypothesized that smaller geographic levels would be most accurate for both races, with a stepwise decrease in accuracy among larger-level measures. Understanding the characteristics of different area-based social measures will enable future discernment when evaluating their impact on driving racial disparities in outcomes for patients with PNET and other cancer patient populations.
Materials and methods
This study was approved by the University of Alabama at Birmingham Institutional Review Board: #300001622. We completed a retrospective review of all patients surgically treated for sporadic PNETs from 2006 (start of electronic medical record system) to 2022 at a single NCI-designated cancer center in the Southern United States. We collected patient demographics including self-reported race (White or Black; other groups excluded due to low numbers), age, sex, insurance, billing address at time of surgery, distance from treating facility, tumor characteristics (size, grade, surgical margins, functional status, lymph node involvement and presence of metastasis) and area level social measures. White and Black races were studied because they were the largest racial groups treated at our center and nationally. In addition, these groups have the most well- described outcome disparities. We excluded neuroendocrine carcinomas and high-grade (G3) tumors as these aggressive tumors are less representative of the overall PNET population.
Billing addresses were geocoded and matched to area-level social metrics including census block group, census tract, county-level or five-digit zip code-level data collected from the year closest to their date of surgery. Census block groups are defined by the US Census as small collections of ‘blocks’ or neighborhoods spanning the entire United States (https://www.census.gov/programs-surveys/geography/about/glossary.html). Block groups consist of about 60–3,000 individuals and are the smallest geographic census unit that provides publicly available area-based social measures. Census tracts are larger statistical subdivisions of a county defined by population density. With population sizes between approximately 1,200 and 8,000 people, they are relatively constant and only updated at the time of every decennial census. Here, census block group and census tract levels are combined and described as ‘neighborhood level’ measures.
Neighborhood-level area-based social measures served as the gold standard. Individuals served as their own controls by comparing their neighborhood-level data to their data from the county and zip code levels. Then, data were compared at every geographic level between White and Black patients. Individual social metrics included marital status and insurance type.
SDOH metrics were selected using the Office of Disease Prevention and Health Promotion ‘Healthy People 2030’ domain framework (Office of Disease Prevention and Health Promotion 2025) (Table 1) including: i) education access and quality (percent of geographic level with high school (HS) diploma by age 25), ii) health care access and quality (insurance rate of community and individual), iii) neighborhood and built environment (ADI), iv) social and community context (SVI), and v) economic stability (community-level income and percent of area living below the federal poverty level) (Office of Disease Prevention and Health Promotion 2025). ADI is a composite metric published by the University of Wisconsin and organized by census block group or zip code level (Center for Health Disparities Research 2021). ADI is generated from 17 publicly available environmental and demographic data points and reported as a composite percentile score from 1 (least socioeconomically disadvantaged) to 100 (most disadvantaged). The most recent version was released in 2021.
Detailed description of SDOH domains measured.
| Domain measured | Metric and source | Year available | Geographic unit |
|---|---|---|---|
| Economic stability | Median household income (ACS) | 2015, 2018, 2020 | County, zip code, block group |
| Poverty level (% living below the federal poverty level, ACS) | 2015, 2018, 2020 | County, zip code, census tract | |
| Education access and quality | Education attainment (% graduated HS or equivalent among persons ≥25 years, ACS) | 2015, 2018, 2020 | County, zip code, block group |
| Healthcare access and quality | Small area health insurance estimates (% of adults <65 without health insurance, ACS), insurance provider | 2015, 2018, 2020 | Individual, county, zip code, census tract |
| Neighborhood/built environment | ADI | 2015, 2021 | Census block group ONLY |
| Social/community context | SVI, age, sex, race | 2010, 2012, 2014, 2016, 2018, 2020 | Individual, census tract or county level |
ACS, American Community Survey; HS, high school; ADI, area deprivation index; SVI, social vulnerability index.
The SVI was created by the Centers for Disease Control and Prevention (CDC) & the Agency for Toxic Substances and Disease Registry (ATSDR) and uses 16 US census variables to identify communities vulnerable to the negative effects of human-caused or natural disasters (Centers for Disease Control and Prevention 2024). It has been reported at the US Census Tract or County level every 2 years since 2010. Higher SVI percentiles indicate more vulnerability. Both ADI and SVI have been investigated as SDOH metrics for disparities in cancer care (Centers for Disease Control and Prevention 2024, Center for Health Disparities Research 2021, Mora et al. 2021, Singh 2003, Tran et al. 2023).
Statistical analysis
Descriptive statistics were completed for all patients, including cohort means, medians and standard deviation/interquartile range as appropriate. Normality was assessed with histograms. Data that were normally distributed were described using parametric tests such as t-tests and chi-squared tests. Non-normally distributed data were analyzed using Wilcoxon rank sum tests. Data were analyzed using R and SAS 9.4 software and a two-sided P value of <0.05 was considered significant.
Power analysis was completed for each of the SDOH domains (Dziak et al. 2020). For median household income at the block level: with 130 White patients and a median of $58.1 K (standard error of $23.3 K), and 49 African American patients with a median neighborhood income of $40.1 K (standard error of $21.4 K), we yielded a >90% power to show the difference between the two race groups with a significant level of 0.05. The calculation for the other domains was similar: a power of 88.5% to show the difference in neighborhood poverty between groups, a power of 79.6% to show the difference in neighborhood HS graduation rate, and a power of 81.3% to show a difference in neighborhood uninsurance rates (P < 0.05).
Results
Overall cohort demographics
A total of 179 patients were surgically treated for PNET at our institution, including 49 (27%) patients who self-identified as Black (Table 2). The median age at surgery was 60 (IQR: 51–68) years and 86 (48%) patients were female. Fifty percent of patients had private insurance coverage. At the neighborhood level, the overall cohort’s median HS graduation rate was 87.4% (IQR: 78.9–94.0%), median population living below the federal poverty level was 15.3% (IQR: 9.8–21.4%), median income was $49 K (IQR: $35.3–66.3 K), median SVI was the 50th percentile (IQR: 30th–70th percentile) and median ADI was the 62nd percentile (IQR: 43rd–83rd percentile).
Baseline pancreatic neuroendocrine tumor patient demographics by race.
| Demographic | Overall (n = 179) | African American/Black (n = 49) | White (n = 130) | P-value |
|---|---|---|---|---|
| Age at time of surgery, median (IQR) | 60.3 (50.9–67.7) | 57.2 (47.8–64.6) | 62.3 (52.9–68.6) | 0.009 |
| Gender | ||||
| Female | 93 (52.0%) | 34 (69.4%) | 59 (45.4%) | 0.004 |
| Male | 86 (48.0%) | 15 (30.6%) | 71 (54.6%) | |
| BMI (mean +/−SD) | 31.4 +/−8.2 | 31.7 +/−9.7 | 31.3 +/−7.7 | 0.812 |
| Married | 126 (70.4%) | 25 (51.0%) | 101 (77.7%) | 0.004 |
| Insurance type | ||||
| Private | 93 (50.0%) | 25 (48.1%) | 68 (50.7%) | 0.953 |
| Governmental | 84 (45.2%) | 23 (44.2%) | 61 (45.5%) | |
| Uninsured | 4 (2.2%) | 1 (1.9%) | 3 (2.2%) | |
| Distance to treating facility, median (miles, IQR) | 78.5 (36.5–120.6) | 88.9 (54.2–139.3) | 73.9 (33.2–102.7) | 0.078 |
| Tumor size (cm, mean ± SD) | 3.4 ± 2.8 | 4.1 ± 3.7 | 3.1 ± 2.4 | 0.031 |
| Surgical margin status: involved | 27 (15.3%) | 8 (16.7%) | 19 (14.7%) | 0.9441 |
| Differentiation: well | 126 (71.2%) | 33 (68.8%) | 93 (72.1%) | 0.840 |
| Tumor grade: 1 | 80 (45.2%) | 20 (41.7%) | 60 (46.5%) | 0.8122 |
| Tumor grade: 2 | 57 (32.2%) | 15 (31.3%) | 42 (32.6%) | |
| Tumor grade: unknown | 40 (22.7%) | 13 (27.6%) | 27 (20.9%) | |
| Functional PNET | 24 (13.4%) | 5 (10.2%) | 19 (14.6%) | 0.270 |
| Metastatic disease | 21 (11.7%) | 7 (14.3%) | 14 (10.8%) | 0.332 |
Bold indicates statistical significance, P < 0.05.
Black patients were younger at diagnosis (median 57 years) compared to White patients (median 62 years, P = 0.009) and were more likely to be female (69 versus 45%, P < 0.01). Tumor sizes among Black patients were larger, with a mean tumor size of 4.1 cm compared to 3.1 cm (P = 0.03) among White patients.
Between race neighborhood level differences
Compared to White patients, Black patients with PNETs more frequently lived in disadvantaged neighborhoods with lower educational attainment (median HS graduation 84 versus 88%), lower income (median household $34 K versus $53 K), higher poverty rate (median 21 versus 14%), higher uninsured rate (20 versus 15%) and higher overall SVI (60th vs 50th percentile) (all P < 0.05) (Table 3). State-level ADI was significantly higher among Black patients (60th percentile, IQR: 30–90) compared to White patients (30th percentile, IQR: 20–60) (P < 0.001).
Differences in area-based social measures by race (n = 179).
| Neighborhood | Zip code | County | |||||||
|---|---|---|---|---|---|---|---|---|---|
| AA/Black | White | P-value | AA/Black | White | P-value | AA/Black | White | P-value | |
| Median household income (in $1,000s) (IQR) | $33.7 ($25.8–$51.2) | $52.8 ($41.0–$71.1) | <0.01 | $34.4 ($30.3–$45.8) | $47.3 ($37.5–$61.6) | <0.01 | $45.0 ($37.7–$45.8) | $45.6 ($39.2–$51.3) | 0.44 |
| Median % HS graduate (IQR) | 84.1 (75.5–92.6) | 87.9 (80.2–95.0) | 0.02 | 82.4 (76.5–88.1) | 85.1 (79.8–91.1) | 0.03 | 86.6 (79.3–88.4) | 85.4 (79.3–88.4) | 0.43 |
| Median % impoverished (IQR) | 21.4 (13.8–30.3) | 13.7 (7.6–18.8) | <0.01 | 25.2 (18.9–32.6) | 15.2 (9.8–20.7) | <0.01 | 18.8 (17.4–23) | 18.8 (15.6–21.3) | 0.59 |
| Median % uninsured (IQR) | 20.2 (13.2–24.3) | 15.2 (11.4–21.4) | 0.04 | 22.1 (17.8–25.2) | 17.2 (13.4–20.4) | <0.01 | 17.1 (15.5–20.7) | 18.6 (17.1–21.5) | 0.09 |
| Median SVI percentile (IQR) | 0.6 (0.5–0.9) | 0.5 (0.2–0.6) | <0.01 | 0.3 (0.1–0.5) | 0.3 (0.1–0.5) | 0.08 | |||
| Median national ADI | 80 (60–93) | 55 (39–77) | <0.01 | ||||||
AA, African American; ADI, area deprivation index; SVI, social vulnerability index. Block group-level measures: household income and graduation rate. Census tract-level measures: SVI, % uninsured, % impoverished. Bold indicates statistical significance, P < 0.05.
These differences, however, were masked when comparing county-level social measures. At the county level, there appeared to be no significant differences in the number of patients from marginalized social groups by race for educational attainment, income, insurance rate or SVI (all P > 0.05).
At the zip code level, there were significant differences in the number of patients from marginalized social groups between races for all metrics (P < 0.05). However, differences were exaggerated between races for poverty and insurance status, with seemingly larger differences between groups than when compared to the differences at the neighborhood level (P < 0.05) (Table 3).
Within race neighborhood level differences
When examining within-group differences among White patients, compared to the neighborhood level, the zip code-level metrics underestimated neighborhood income and insurance rates (P < 0.05) (Table 4). County-level metrics underestimated White patients’ neighborhood median household income, insurance rates and HS graduation rates, but overestimated poverty rates and SVI (all P < 0.05). For Black patients, compared to the neighborhood level, zip code-level metrics overestimated poverty level and rates of the uninsured (P < 0.05). Only SVI was overestimated at the county level (P < 0.001).
Differences in area-based social measures by geographic level (n = 179).
| Neighborhood level metric* | Zip code level metric* | P-value | Neighborhood level metric* | County evel metric* | P-value | Zip level | County level | P-value | |
|---|---|---|---|---|---|---|---|---|---|
| African American/Black | |||||||||
| Median household income (in $1,000s) (IQR) | $33.7 ($25.8–$51.2) | $34.4 ($30.3–$45.8) | 0.315 | $33.7 ($25.8–$51.2) | $45.0 ($37.7–$45.8) | 0.237 | 34.4 (30.3–45.8) | 45.0 (37.7–45.8) | <0.001 |
| Median % HS graduate (IQR) | 84.1 (75.5–92.6) | 82.4 (76.5–88.1) | 0.736 | 84.1 (75.5–92.6) | 86.6 (79.3–88.4) | 0.150 | 82.4 (76.5–88.1) | 86.6 (79.3–88.4) | 0.055 |
| Median % impoverished (IQR) | 21.4 (13.8–30.3) | 25.2 (18.9–32.6) | 0.042 | 21.4 (13.8–30.3) | 18.8 (17.4–23) | 0.156 | 25.2 (18.9–32.6) | 18.8 (17.4–23) | <0.001 |
| Median % uninsured (IQR) | 20.2 (13.2–24.3) | 22.1 (17.8–25.2) | 0.043 | 20.2 (13.2–24.3) | 17.1 (15.5–20.7) | 0.558 | 22.1 (17.8–25.2) | 17.1 (15.5–20.7) | <0.001 |
| Median SVI percentile (IQR) | NA | 0.6 (0.5–0.9) | 0.3 (0.1–0.5) | <0.001 | NA | ||||
| White | |||||||||
| Median household income (in $1,000s) (IQR) | $52.8 ($41.0–$71.1) | $47.3 ($37.5–$61.6) | <0.01 | $52.8 ($41.0–$71.1) | $45.6 ($39.2–$51.3) | <0.001 | 47.3 (37.5–61.6) | 45.6 (39.2–51.3) | <0.001 |
| Median % HS graduate (IQR) | 87.9 (80.2–95.0) | 85.1 (79.8–91.1) | 0.061 | 87.9 (80.2–95.0) | 85.4 (79.3–88.4) | 0.008 | 85.1 (79.8–91.1) | 85.4 (79.3–88.4) | 0.103 |
| Median % impoverished (IQR) | 13.7 (7.6–18.8) | 15.2 (9.8–20.7) | 0.139 | 13.7 (7.6–18.8) | 18.8 (15.6–21.3) | <0.001 | 15.2 (9.8–20.7) | 18.8 (15.6–21.3) | <0.001 |
| Median % uninsured (IQR) | 15.2 (11.4–21.4) | 17.2 (13.4–20.4) | <0.01 | 15.2 (11.4–21.4) | 18.6 (17.1–21.5) | <0.001 | 17.2 (13.4–20.4) | 18.6 (17.1–21.5) | <0.01 |
| Median SVI percentile (IQR) | 0.5 (0.2–0.6) | 0.3 (0.1–0.5) | <0.01 | ||||||
SVI, social vulnerability index. Block group-level measures: household income and graduation rate. Census tract-level measures: SVI, % uninsured, % impoverished. Bold indicates statistical significance, P < 0.05.
All between-race differences P < 0.05.
Discussion
Moving the needle on race- and socioeconomic status-based disparities in health outcomes is an ongoing challenge. Race-based differences in outcomes are often the easiest to capture, likely due to convenience; most EMRs and cancer registries report race (Boffa et al. 2017, Doll et al. 2018). However, race is a social construct that only tells part of the story and often a part that is not clearly actionable at the individual patient level (Schillinger 2020). Disparities outcomes research is moving toward capturing the intersection of environmental adverse determinants of health that differentially impact minority populations such as housing discrimination, racism and differential economic access (Ward et al. 2004). In order to study these exposures and develop interventions to address them for patients with PNET, they must be accurately and reproducibly measured (Amini et al. 2024, Casey et al. 2024). The ideal area-based social measures would be gathered at the patient level, but historically this has not been captured in large EMRs or registries. There are currently no guidelines specifying the most valid area measure. As a proxy, researchers often use zip code- or county-level area-based social measures of adversity (Dasari et al. 2017). However, there are little data on the validity of this and in which ways these larger area metrics are inaccurate. Our study shows that these large geographic area-based social measures may be inaccurate in unpredictable ways, especially for different races.
Our results show that, in general, large area (zip code, county) based social measures for White patients treated for PNET lead to inaccurate conclusions such as an apparent lower household income, higher level of uninsured and lower level of education when compared to neighborhood-level metrics. County-level metrics were much less accurate than zip code-level metrics. For Black patients, analyzing larger geographic units at times caused these patients to seem more vulnerable, but overall the results were more accurate than those for White patients when compared to neighborhood-level metrics.
When examining between-group differences, using larger geographic units erased the racial differences in the proportion of vulnerable populations. These results are novel and draw attention to the limited validity of large geographic units, especially when examining racial disparities. Although these large area-level metrics are more convenient to use, they may dilute or erase the clinically meaningful differences in rates of marginalized social groups by race. This may also explain why some prior PNET studies demonstrate racial and socioeconomic disparities (Gosain et al. 2020, Zheng-Pywell et al. 2021), while others do not (Marincola Smith et al. 2020).
This study is also unique in that it only includes a surgical cohort, which may have differential access to care compared to other epidemiological studies of cancer patients. The benefit here is that all patients have a billing address, which allows direct comparison between geographic areas. The findings of this study can inform future outcomes-based research.
In future studies, we caution cancer researchers to seek and utilize the smallest geographic unit of the most robust metric of vulnerability when studying healthcare disparities. These metrics may be used in analyses as an outcome or covariate, and researchers need to be rigorous with the application of these data. In order to reduce disparities, we have to understand what we are measuring and how accurate it may be. Otherwise, misleading, paradoxical or inaccurate conclusions may be drawn. Guidelines and researchers in the field of health outcomes should understand the limitations of large area geographic measurements of unmet needs and should design future registries and trials that collect small area or individually measured metrics; otherwise, conclusions may be unreliable and unreproducible. The NIH All of Us registry is an excellent example of this by preferentially collecting patient-reported, directly measured EMR and survey data to push forward robust health outcomes research (Denny et al. 2019). Although these are more resource-intensive approaches, the present study demonstrates the importance of this approach in NETs as well as other cancer populations.
There are some limitations in this study. We only included patients who underwent surgical resection for PNETs, which may not be generalizable to patients with unresectable tumors or other malignancies. This is a single-institution study, which may have limited generalizability to other centers. Our institution is also situated in the Southern United States, which is a unique region compared to the rest of the United States. This includes overrepresentation of Black and African American identifying patients (26% in Alabama) and a larger amount of persistently impoverished and rural neighborhoods compared to the rest of the United States (Savage et al. 2013, Davis et al. 2014, Gyawu et al. 2015). This may emphasize how large geographic area metrics are even more unreliable in diverse patient cohorts. In addition, we utilized patient billing addresses at the time of surgery, which may not be the patient’s true home address but is likely an adequate proxy (Adkins-Jackson et al. 2022). Finally, this is a study of an uncommon neoplasm (Fowler et al. 2024). However, this is a useful cohort to examine because PNET diagnosis is not linked to access to screening but overall access to care and surgery, representing a disease that is appropriate to measure baseline representation of vulnerable social groups.
Conclusions
Compared to White patients, Black patients with PNETs are more likely to be from vulnerable social groups, a disparity that may be hidden when analyzing large geographic regional variations. For the most robust conclusions about racial disparities driving outcomes in cancer research, investigators must prioritize area-based social measures for individuals that utilize the smallest geographic region.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the work reported.
Funding
This work was supported by the National Cancer Institute grant: 3U54CA118948-17S1.
Author contribution statement
AG, JBR, UM and SB were responsible for project conception. AG, BH, RG and WC contributed to data collection and analysis. CD conducted the statistical analysis. HC, AG, RG, JBR and SB were involved in manuscript writing and revision.
Acknowledgments
We would like to acknowledge the Social Determinants of Health Core at the University of Alabama at Birmingham for their assistance.
References
Adkins-Jackson PB , Chantarat T , Bailey ZD , et al. 2022 Measuring structural racism: a guide for epidemiologists and other health researchers. Am J Epidemiol 191 539–547. (https://doi.org/10.1093/aje/kwab239)
Amini N , Demyan L , Shah M , et al. 2024 Decreasing utilization of surgical interventions amongst patients with pancreatic neuroendocrine tumor with liver metastases. Am J Surg 227 77–84. (https://doi.org/10.1016/j.amjsurg.2023.09.035)
Boffa DJ , Rosen JE , Mallin K , et al. 2017 Using the national cancer database for outcomes research: a review. JAMA Oncol 3 1722–1728. (https://doi.org/10.1001/jamaoncol.2016.6905)
Casey M , Brown A , Romero-Hernandez F , et al. 2024 National practice patterns in the use of endoscopic ultrasound biopsy for resectable pancreatic neuroendocrine tumors: insights into the role of DOTATATE PET/CT in diagnosis. Am J Surg 235 115779. (https://doi.org/10.1016/j.amjsurg.2024.115779)
Center for Health Disparities Research 2021 Area deprivation index v4.0. Neighborhood Atlas. Madison, WI, USA: University of Wisconsin, School of Medicine and Public Health. (https://www.neighborhoodatlas.medicine.wisc.edu)
Centers for Disease Control and Prevention; Agency for Toxic Substances and Disease Registry 2024 CDC/ATSDR social vulnerability index (SVI). Atlanta, GA, USA: Centers for Disease Control and Prevention. (https://www.atsdr.cdc.gov/place-health/php/svi/index.html)
Cheung R 2013 Racial and social economic factors impact on the cause specific survival of pancreatic cancer: a SEER survey. Asian Pac J Cancer Prev 14 159–163. (https://doi.org/10.7314/apjcp.2013.14.1.159)
Dasari A , Shen C , Halperin D , et al. 2017 Trends in the incidence, prevalence, and survival outcomes in patients with neuroendocrine tumors in the United States. JAMA Oncol 3 1335–1342. (https://doi.org/10.1001/jamaoncol.2017.0589)
Davis SK , Gebreab S , Quarells R , et al. 2014 Social determinants of cardiovascular health among black and white women residing in Stroke Belt and Buckle regions of the South. Ethn Dis 24 133–143.
Denny JC , Rutter JL , Goldstein DB , et al. 2019 The “All of Us” research program. N Engl J Med 381 668–676. (https://doi.org/10.1056/nejmsr1809937)
DePalo DK , Lee RM , Lopez-Aguiar AG , et al. 2019 Interaction of race and pathology for neuroendocrine tumors: epidemiology, natural history, or racial disparity? J Surg Oncol 120 919–925. (https://doi.org/10.1002/jso.25662)
Diez-Roux AV , Kiefe CI , Jacobs DR , et al. 2001 Area characteristics and individual-level socioeconomic position indicators in three population-based epidemiologic studies. Ann Epidemiol 11 395–405. (https://doi.org/10.1016/s1047-2797(01)00221-6)
Doll KM , Rademaker A & Sosa JA 2018 Practical guide to surgical data sets: surveillance, epidemiology, and end results (SEER) database. JAMA Surg 153 588–589. (https://doi.org/10.1001/jamasurg.2018.0501)
Dziak JJ , Dierker LC & Abar B 2020 The interpretation of statistical power after the data have been gathered. Curr Psychol 39 870–877. (https://doi.org/10.1007/s12144-018-0018-1)
Fontenot K , Semega J & Kollar M 2018 Income And Poverty In the United States: 2017. In U.S. Census Bureau, Current Population Reports, P60-263. Washington, DC, USA: U.S. Government Printing Office.
Fouad MN , Oates GR , Scarinci IC , et al. 2017 Advancing the science of health disparities through research on the social determinants of health. Am J Prev Med 52 (Supplement 1) S1–S4. (https://doi.org/10.1016/j.amepre.2016.09.032)
Fowler K , Li J & Pommier RF 2024 Hepatic cytoreduction for lung and renal neuroendocrine tumor metastases. Am J Surg 231 41–45. (https://doi.org/10.1016/j.amjsurg.2024.01.036)
Gosain R , Ball S , Rana N , et al. 2020 Geographic and demographic features of neuroendocrine tumors in the United States of America: a population-based study. Cancer 126 792–799. (https://doi.org/10.1002/cncr.32607)
Gyawu R , Quansah JE , Fall S , et al. 2015 Community food environment measures in the Alabama Black Belt: implications for cancer risk reduction. Prev Med Rep 2 689–698. (https://doi.org/10.1016/j.pmedr.2015.08.015)
Harrelson A , Wang R , Stewart A , et al. 2023 Management of neuroendocrine tumor liver metastases. Am J Surg 226 623–630. (https://doi.org/10.1016/j.amjsurg.2023.08.011)
Herring BR , Bonner A , Guenter RE , et al. 2022 Under-representation of racial groups in genomics studies of gastroenteropancreatic neuroendocrine neoplasms. Cancer Res Commun 2 1162–1173. (https://doi.org/10.1158/2767-9764.crc-22-0093)
Holland MM & Rose JB 2024 Touching with your eyes: fluorescent intraoperative imaging in the management of small bowel neuroendocrine tumors. Am J Surg 232 2. (https://doi.org/10.1016/j.amjsurg.2024.01.012)
Johnson AE , Zhu J , Garrard W , et al. 2021 Area deprivation index and cardiac readmissions: evaluating risk-prediction in an electronic health record. J Am Heart Assoc 10 e020466. (https://doi.org/10.1161/jaha.120.020466)
Krieger N , Chen JT , Waterman PD , et al. 2002 Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter? The Public Health Disparities Geocoding Project. Am J Epidemiol 156 471–482. (https://doi.org/10.1093/aje/kwf068)
Lubner SJ , Kunnimalaiyaan M , Holen KD , et al. 2011 A preclinical and clinical study of lithium in low-grade neuroendocrine tumors. Oncologist 16 452–457. (https://doi.org/10.1634/theoncologist.2010-0323)
Marincola Smith P , Baechle J , Tan MC , et al. 2020 Impact of insurance status on survival in gastroenteropancreatic neuroendocrine tumors. Ann Surg Oncol 27 3147–3153. (https://doi.org/10.1245/s10434-020-08359-z)
Mirza MB , Baechle JJ , Marincola Smith P , et al. 2024 Survival disparities in rural versus urban patients with pancreatic neuroendocrine tumor: a multi-institutional study from the US neuroendocrine tumor study group. Am J Surg 233 125–131. (https://doi.org/10.1016/j.amjsurg.2024.03.003)
Mora J , Krepline AN , Aldakkak M , et al. 2021 Adjuvant therapy rates and overall survival in patients with localized pancreatic cancer from high Area Deprivation Index neighborhoods. Am J Surg 222 10–17. (https://doi.org/10.1016/j.amjsurg.2020.12.001)
Notterman DA & Mitchell C 2015 Epigenetics and understanding the impact of social determinants of health. Pediatr Clin North Am 62 1227–1240. (https://doi.org/10.1016/j.pcl.2015.05.012)
Office of Disease Prevention and Health Promotion 2025 Healthy People 2030: social determinants of health. Washington, DC, USA: U.S. Department of Health and Human Services. (https://health.gov/healthypeople)
Pearson J , Jacobson C , Ugochukwu N , et al. 2023 Geospatial analysis of patients’ social determinants of health for health systems science and disparity research. Int Anesthesiol Clin 61 49–62. (https://doi.org/10.1097/aia.0000000000000389)
Ren B , Rose JB , Liu Y , et al. 2019 Heterogeneity of vascular endothelial cells, de novo arteriogenesis and therapeutic implications in pancreatic neuroendocrine tumors. J Clin Med 8 1980. (https://doi.org/10.3390/jcm8111980)
Savage RM , Dillon JM , Hammel JC , et al. 2013 The Alabama Coalition for a Healthier Black belt: a proof of concept project. Community Ment Health J 49 79–85. (https://doi.org/10.1007/s10597-012-9488-z)
Schillinger D 2020 The intersections between social determinants of health, health literacy, and health disparities. Stud Health Technol Inform 269 22–41. (https://doi.org/10.3233/SHTI200020)
Shavers VL 2007 Measurement of socioeconomic status in health disparities research. J Natl Med Assoc 99 1013–1023.
Singh GK 2003 Area deprivation and widening inequalities in US mortality, 1969-1998. Am J Public Health 93 1137–1143. (https://doi.org/10.2105/ajph.93.7.1137)
Tran T , Rousseau MA , Farris DP , et al. 2023 The social vulnerability index as a risk stratification tool for health disparity research in cancer patients: a scoping review. Cancer Causes Control 34 407–420. (https://doi.org/10.1007/s10552-023-01683-1)
Wang R , Zheng-Pywell R , Chen HA , et al. 2019 Management of gastrointestinal neuroendocrine tumors. Clin Med Insights Endocrinol Diabetes 12 1179551419884058. (https://doi.org/10.1177/1179551419884058)
Ward E , Jemal A , Cokkinides V , et al. 2004 Cancer disparities by race/ethnicity and socioeconomic status. CA Cancer J Clin 54 78–93. (https://doi.org/10.3322/canjclin.54.2.78)
Wu C , Song Z , Balachandra S , et al. 2024 Charting the course: insights into neuroendocrine tumor dynamics in the United States. Ann Surg. (https://doi.org/10.1097/SLA.0000000000006331)
Xu Z , Wang L , Dai S , et al. 2021 Epidemiologic trends of and factors associated with overall survival for patients with gastroenteropancreatic neuroendocrine tumors in the United States. JAMA Netw Open 4 e2124750. (https://doi.org/10.1001/jamanetworkopen.2021.24750)
Yao JC , Hassan M , Phan A , et al. 2008 One hundred years after “carcinoid”: epidemiology of and prognostic factors for neuroendocrine tumors in 35,825 cases in the United States. J Clin Oncol 26 3063–3072. (https://doi.org/10.1200/jco.2007.15.4377)
Zheng-Pywell R , Fang A , AlKashash A , et al. 2021 Prognostic impact of tumor size on pancreatic neuroendocrine tumor recurrence may have racial variance. Pancreas 50 347–352. (https://doi.org/10.1097/mpa.0000000000001776)
Zheng-Pywell R , Lopez-Aguiar A , Fields RC , et al. 2022 Are we undertreating black patients with nonfunctional pancreatic neuroendocrine tumors? Critical analysis of current surveillance guidelines by race. J Am Coll Surg 234 599–606. (https://doi.org/10.1097/xcs.0000000000000105)
