Socioeconomic Disparities Along the Cancer Continuum for Hepatocellular Carcinoma: A Systematic Review

Abstract

Background: Social determinants of health critically impact outcomes along the care continuum of patients with hepatocellular carcinoma (HCC). This systematic review summarizes the effect of socioeconomic status (SES) factors on HCC outcomes in the United States. Methods: Electronic databases were queried for the concepts of “liver cancer”, “health disparities”, and “socioeconomic factors” on 1 March 2021. Eligible studies included an individual- or area-level SES measure such as income, education, employment, and insurance and one of the following outcomes across the clinical continuum of HCC care: incidence, screening/surveillance, diagnosis, treatment, survival, and end-of-life. Results: Of 3331 studies screened, a total of 63 studies encompassing 179 separate analyses were included in our narrative synthesis: 13 on incidence, 5 on surveillance, 19 on diagnosis, 79 on treatment, 61 on survival, and 2 on end-of-life. Insurance was the most frequent SES measure represented (50%), followed by mostly area-level income (39%), education (9%), and employment (2%). The included studies were heterogeneous regarding both SES definitions (e.g., individual vs. area-level measures) and outcome reporting. Trends of worse outcomes were generally observed with lower indicators across all SES domains and HCC outcomes, particularly in analyses using national cancer registry data (e.g., SEER and NCDB). Unadjusted racial and ethnic disparities in outcome were attenuated in six out of 23 analyses that adjusted for an SES measure. Conclusions: Our findings highlight the need for social risk screening and interventions early in the HCC care pathway. Future research should focus on HCC surveillance and end-of-life/survivorship, with greater emphasis on examination of modifiable individual-level social determinants.

1. Introduction

Hepatocellular carcinoma (HCC) is the 6th leading cause of cancer-related deaths in the United States (US) [1]. The American Cancer Society estimated 42,240 new cases of HCC diagnosed and estimated 30,090 related deaths in 2025 [2]. While mortality has plateaued, HCC disproportionately affects the US population, with higher cancer incidence and mortality among non-White racial and ethnic groups [3]. Moreover, racial and ethnic disparities across the continuum of cancer care, from stage at diagnosis to treatment utilization to survival, are well-documented [4,5,6].

While there are likely multiple interconnected factors contributing to these racial and ethnic disparities, to date, studies have identified socioeconomic status (SES) and associated social determinants of health (SDOH) as contributing and potentially unifying factors [7,8]. Independent of race and ethnicity, individuals of low SES carry a higher HCC burden, and they comprise one of the few remaining subgroups with rising incidence [9]. While previous systematic reviews on HCC disparities have focused on race and ethnicity as the primary predictor of interest [3,4], there remains a gap in our knowledge on the role of prominent SES-related factors such as income, insurance, and education, including both area-level and individual-level measures. Furthermore, prior reviews more narrowly focus on one aspect of the care continuum for HCC, such as treatment or survival, which leaves other key stages of the continuum under-explored.

To address this knowledge gap, this study uniquely aims to (1) systematically review the literature on the impact of SES on the entire clinical care continuum of HCC in the US and (2) determine if and how the presence of a socioeconomic parameter influences race and ethnicity estimates in models of cancer outcomes.

2. Materials and Methods

2.1. Literature Search

We performed a literature search across three major electronic databases: PubMed, EMBASE, and CINAHL. The search strategy, developed in collaboration with an experienced medical librarian, utilized the following keyword combinations: liver cancer, socioeconomic factors, and health disparities (final search strategy for each database is in Supplemental Table S1). To maintain a focus on current health disparities within the United States, the search was restricted to human studies published in English since 1 January 2000. The protocol was posted on PROSPERO (ID: CRD42021227615) on 21 January 2021, and the search was performed on 1 March 2021. This review adheres to the PRISMA guidelines [10].

2.2. Study Selection

Search results were imported into Covidence, an online tool for managing systematic reviews, which identified and eliminated duplicate records across the three databases. Titles and abstracts of the remaining studies were independently screened by at least two reviewers (VLT, SZ, and AC) for eligibility. Full-text articles of selected studies were then independently reviewed by two reviewers (VLT and AC). All disagreements were resolved through consensus with a third reviewer (KZ).

The inclusion criteria were as follows: (i) studies involving adults (age ≥ 18) within the United States, (ii) analyses of outcomes related to the hepatocellular carcinoma (HCC) care continuum, (iii) evaluation of the impact of socioeconomic status (SES), and (iv) publication in peer-reviewed journals or presentation at scientific conferences as abstracts. Exclusion criteria were (i) studies involving non-HCC tumors (e.g., cholangiocarcinoma, mixed tumors), (ii) sample sizes fewer than 20, (iii) publication dates prior to 2000, (iv) editorials, review articles, theses/dissertations, and case reports, (v) studies that evaluated survival only among those that received a surgical treatment, such as a transplant, (vi) studies in languages other than English, and (vii) those involving non-human data. For studies initially presented as abstracts, data from the most recent full-text publication were used when available, with earlier versions excluded.

2.3. Data Extraction

Data extraction was performed by one of two reviewers (VLT and CW) using a standardized form and validated by the second reviewer; discrepancies were clarified by a third reviewer (KZ). For the primary covariate of interest, we identified any individual-level SES measures (e.g., income, insurance, education, employment) and area-based SES measures (e.g., census tract, block group, ZIP code-linked survey data) that were included in adjusted, multivariable models assessing outcomes across the HCC care continuum. The outcomes of interest included HCC incidence, surveillance, stage at diagnosis, receipt of treatment, survival, and end-of-life/survivorship. For each included study, we also recorded data on the type of manuscript (abstract or full text), number of patients included, study period, specific population or cohort studied (e.g., if narrowed by age, race/ethnicity, clinical setting, etc.), database used, geographic location, and staging criteria. Summary statistics for the SES predictor, including effect size and confidence intervals, were abstracted. Many studies reported multiple outcomes, and analyses of each outcome were recorded separately; similarly, analytic data were abstracted separately for studies that included more than one SES measure.

2.4. Analytic Plan and Quality Assessment

Meta-analyses were not performed due to significant heterogeneity and insufficient comparability in definition and/or categorization of SES variables and outcome measures. Instead, a narrative synthesis was conducted, which involved grouping and summarizing the results according to the SES domain and HCC outcomes measured within the reviewed studies.

To further explore the effect of SES adjustments on differences in outcomes by race and ethnicity, we abstracted data from studies that included both race and ethnicity and SES covariates and reported results of both univariate (unadjusted) and multivariable (adjusted) models. Among this subgroup of studies, we ascertained (i) frequency of change in significance and (ii) direction of change for the race and ethnicity covariate estimate from univariate models after multivariable adjustment including the SES measure.

Study quality was assessed by one reviewer (JO) and validated by a second reviewer (KZ) using the Newcastle–Ottawa Scale (NOS), which evaluates cohort selection, group comparability, and the ascertainment of exposures or outcomes [11]. “Good” quality studies were defined by a NOS score of seven or greater.

3. Results

3.1. Study Characteristics and Thematic Design

Title and abstract screening were performed for a total of 3582 articles, with 247 duplicates identified manually or through Covidence. The PRISMA flow diagram is presented in Figure 1. Of the remaining 3331 titles, 3173 of these failed to meet inclusion criteria. Of the 158 full-text studies remaining, 95 were then excluded for a total of 63 studies included in this review. Most studies were excluded at full-text review due to either lack of or insufficient information on an SES predictor. Studies could contribute multiple analyses if there were more than one SES domain or eligible outcome reported. In total, there were 179 analyses: 13 on incidence, 5 on surveillance, 19 on diagnosis, 79 on treatment, 61 on survival, and 2 on survivorship (Figure 2). Among SES domains, 89 analyses reported on insurance, 69 on income, 17 on education, and 3 on employment.

Table 1. Summary characteristics of all included studies (N = 63).

Study Author (Year)	Database Type	Geographic Setting	Study Population	Staging System	Date Range	Sample Size
Abara et al. (2020) [14]	Multi-Institution	Danville, PA; Detroit, MI; Honolulu, HI; Portland, OR	Adults with hepatitis C and cirrhosis	N/A	2011–2016	2933
Adler Jaffe et al. (2020) [15]	SEER	National	All adults with HCC	SEER Historic Stage	2010–2015	14,655
Alawadi et al. (2016) [16]	Texas Cancer Registry	State (Texas)	All adults with HCC	AJCC 6th Edition	2000–2008	5037
Artinyan et al. (2010) [17]	SEER	National	All adults with HCC	Metastatic vs. non-metastatic	1973–2004	14,906
Aru et al. (2016) [18]	Mississippi Cancer Registry	State (Mississippi)	Black and white adults with HCC	AJCC 6th edition	1995–2016	421
Bateni et al. (2020) [19]	California Cancer Registry	State (California)	Adults with HCC	AJCC 6th edition	2005–2017	19,555
Bemanian et al. (2019) [20]	Not specified	Milwaukee/Racine metropolitan areas (Wisconsin)	Adults with chronic liver disease	N/A	N/A	18,143
Beutler et al. (2020) [21]	SEER	National	Adults with resectable HCC	AJCC 6th edition	2007–2015	28,518
Bodek et al. (2018) [22]	NIS	National	Adults hospitalized with HCC	N/A	2007–2014	7244
Cheng et al. (2019) [23]	NCDB	Regional (Mountain West region)	Adults with HCC	AJCC pathologic stage	2004–2015	6500
Chidi et al. (2016) [24]	Pennsylvania Cancer Registry	State (Pennsylvania)	Adults with HCC	SEER summary stage	2006–2011	3576
Danos et al. (2018) [25]	Louisiana Tumor Registry and US Census	State (Louisiana)	Black and white adults age > 35	N/A	2008–2012	1418
Estevez et al. (2017) [26]	Multi-institution	Stanford U, Mt. Sinai NY, Mayo MN	Black and white adults with HCC	BCLC Staging	1991–2016	1156
Flores et al. (2019) [27]	SEER	National	All adults with HCC	SEER Historic Stage A variable	2000–2015	45,789
Ford et al. (2017) [28]	New York State Cancer Registry	City (New York City)	All adults	N/A	2001–2012	8827
Goldberg et al. (2014) [29]	Truven Health Analytics Database	National	Adults with non-cirrhotic chronic HBV	N/A	2006–2009	4576
Ha et al. (2016) [30]	SEER	National	All adults with HCC	SEER Historic Stage	2003–2013	61,594
Harlan et al. (2015) [31]	NCI Patterns of Care Study	National	All adults with HCC	BCLC staging	2007	946
Hoehn et al. (2015) [32]	NCDB	National	All adults with AJCC Stage I/II HCC	AJCC 6th edition	1998–2011	43,859
Hoehn et al. (2017) [33]	NCDB	National	All adults with HCC	AJCC clinical stage	1998–2007	143,692
Hood et al. (2020) [34]	Ohio Cancer Incidence Surveillance System	State (Ohio)	All adults with HCC	N/A	2003–2016	8208
Hyder et al. (2013) [35]	SEER-Medicare	National	All adults with HCC	SEER Historic Staging A	1998–2007	6752
Jan et al. (2012) [36]	Single institution	Tulane University Medical Center (Louisiana)	All adults with HCC	N/A; tumor size, # of tumors, vascular invasion	2003–2011	206
Jones et al. (2019) [37]	Florida Cancer Data System	State (Florida)	All adults with HCC	SEER Stage 2000	2004–2013	10,852
Kangas-Dick et al. (2020) [38]	NIS	National	All adults hospitalized with HCC and HCV	N/A	2005–2013	200,163
Kokabi et al. (2016) [39]	SEER	National	All adults with HCC	N/A	2000–2011	23,464
Kokabi et al. (2017) [40]	SEER	National	All adults with HCC	AJCC 6th edition	2000–2010	9368
Kronenfeld et al. (2020, A) [41]	USSNC—Multi-institution	Metro regions of Texas, Florida, Georgia, New York	All adults with HCC	AJCC 6th edition	2012–2014	1620
Kronenfeld et al. (2020, B) [42]	USSNC—Multi-institution	Metro regions of Texas, Florida, Georgia, New York	Adults with HCC, AJCC Stage I/II	AJCC 6th edition	2012–2014	1087
Kwong et al. (2010) [43]	California Cancer Registry (CCR)	State (California)	Adults with HCC from nine Asian ethnic groups	SEER classification	1988–2007	6068
Lee et al. (2020, A) [5]	USSNC—Multi-institution	Metro regions of Texas, Florida, Georgia, New York	All adults with HCC	AJCC 8th edition	2012–2014	1832
Lee et al. (2020, B) [44]	USSNC—Multi-institution	Metro regions of Texas, Florida, Georgia, New York	All adults with HCC	AJCC 8th edition	2004–2014	31,107
Lee et al. (2020, C) [45]	USSNC—Multi-institution	Metro regions of Texas, Florida, Georgia, New York	Adults with HCC treated with only chemotherapy	AJCC 8th edition	2004–2014	1479
Major et al. (2014) [46]	SEER, NIH-AARP Diet Study	National (CA, FL, LA, NC, NJ, PA, Detroit, Atlanta)	Retired adults	N/A	1995–2006	494,988
Mazumder et al. (2021) [47]	HealthLNK Data Repository	Multi-institution (greater metropolitan Chicago area)	Adults with HCC	N/A	2006–2012	11,277
Mokdad et al. (2018) [48]	Texas Cancer Registry	State (Texas)	All adults with HCC	SEER classification	2001–2012	15,932
Muzaffar et al. (2020) [49]	SEER	National	All adults with HCC	SEER classification	1990–2016	87,047
Peters et al. (2017) [50]	SEER	National	Adults with early-stage HCC (AJCC T1/T2)	AJCC 6th edition	2004–2012	13,694
Pham Hang et al. (2018) [12]	Single institution	Emory Medical Center (Georgia)	Adults with non-metastatic HCC	N/A; focused on non-metastatic disease	2013–2016	156
Rho et al. (2019) [51]	NCDB	National	All adults with HCC	N/A; metastatic vs. non metastatic	2004–2015	160,517
Rich et al. (2019) [52]	Multi-institution	Parkland and UTSW (Texas)	All adults with HCC	BCLC staging	2008–2017	1117
Robbins et al. (2011) [53]	NCDB	National	All adults with HCC	AJCC 5th/6th editions	1998–2007	7707
Robinson et al. (2017) [54]	SEER	National	Adults born 1945–1965 with HCC	SEER Historic Staging	2004–2014	38,045
Robinson et al. (2018) [55]	SEER	National	Adults born 1945–1965 with HCC	SEER Historic Staging	2004–2014	38,045
Sarpel et al. (2016) [56]	Single institution	The Mount Sinai Hospital (New York)	All adults with HCC	N/A; within vs. beyond Milan criteria	2003–2013	742
Sarpel et al. (2018) [57]	Single institution	The Mount Sinai Hospital (New York)	Adults with BCLC-C HCC	BCLC staging	2007–2013	742
Scaglione et al. (2020) [58]	Multi-institution	U Mich, Loyola, Parkland, Ben-Taub	All adults with HCC	N/A; within vs. beyond Milan criteria	2012–2013	379
Sellers et al. (2019) [59]	Single institution	Yale Hospital Cancer Registry (Connecticut)	All adults with HCC	BCLC staging	2005–2016	769
Singal et al. (2015) [60]	Single institution	Parkland Health and Hospital System	All adults with cirrhosis	N/A; intra hepatic vs. extra hepatic disease	2008–2011	904
Sobotka et al. (2019) [61]	NIS	National	All adults hospitalized with HCC	N/A; non metastatic vs. single met vs. multiple mets	2010–2013	62,368
Sodagari et al. (2019) [13]	NIS	National	All adults hospitalized with HCC	N/A	1993–2015	701,368
Turse et al. (2022) [62]	Single institution	Valleywise Health Center (Arizona)	All adults with HCC	N/A; within or outside of Milan criteria	2010–2020	161
Uhlig et al. (2019) [63]	NCDB	National	All adults with HCC	AJCC 7th edition	2004–2015	63,877
Wang et al. (2016) [64]	SEER	National	All adults with HCC	SEER Historic Staging system	1983–1992	26,535
Wang et al. (2018) [65]	SEER	National	All adults with HCC	SEER Historic Staging system	2007–2012	32,388
Wang et al. (2020) [66]	SEER	National	All adults with HCC	AJCC 6th edition	2004–2017	83,237
Wasif et al. (2018) [67]	NCDB	National	Adults with HCC undergoing surgery	N/A	2003–2012	3814
Woodrell et al. (2021) [68]	Single institution	The Mount Sinai Hospital (New York)	All adults with HCC-related hospitalization	N/A	2012–2016	842
Yan et al. (2016) [69]	SEER	National	All adults with HCC	SEER Historic Staging System	2003–2013	61,594
Yang et al. (2019) [70]	NCDB	National	All adults with single HCC tumor < 2 cm in diameter	N/A; single HCC tumor < 2cm	2004–2014	6261
Yu et al. (2010) [71]	Single institution	Columbia University Medical Center (New York)	All adults with HCC	AJCC staging	2002–2008	462
Zaydfudim et al. (2010) [72]	Tennessee Cancer Registry	State (Tennessee)	All adults with HCC	AJCC/TNM staging	2004–2006	680
Zou et al. (2018) [73]	Veteran Affairs (VA) database	National	Veterans with BCLC-C/D HCC	BCLC staging	2004–2011	397

and

Table 2. Summary of SES predictors and HCC outcomes of all included studies (N = 63).

Study Author (Year)	SES Predictor(s)	HCC Outcome(s)	No. of Included Analyses
Abara et al. (2020) [14]	Insurance	Surveillance	1
Adler Jaffe et al. (2020) [15]	Insurance	Survival (by stage)	3
Alawadi et al. (2016) [16]	Income (area-level % poverty, income type)	Treatment (locoregional, surgical, 6) Survival (1)	7
Artinyan et al. (2010) [17]	Income (area-level median income)	Survival	1
Aru et al. (2016) [18]	Income (area-level median household income)	Diagnosis (late stage)	1
Bateni et al. (2020) [19]	Income (SES tertiles), Insurance	Survival	2
Bemanian et al. (2019) [20]	Income (neighborhood SES disadvantage index)	Incidence	1
Beutler et al. (2020) [21]	Insurance	Survival	1
Bodek et al. (2018) [22]	Income (census-tract-level income quartile), Insurance	Treatment (surgical)	2
Cheng et al. (2019) [23]	Insurance	Treatment (all treatment, locoregional, systemic, 4), Survival (1)	5
Chidi et al. (2016) [24]	Income (area-level median household income), Insurance	Treatment (surgical)	4
Danos et al. (2018) [25]	Income (neighborhood concentrated disadvantage index, CDI)	Incidence	1
Estevez et al. (2017) [26]	Insurance	Survival	1
Flores et al. (2019) [27]	Income (area-level median income)	Survival	1
Ford et al. (2017) [28]	Income (neighborhood poverty index), Insurance	Incidence	2
Goldberg et al. (2014) [29]	Insurance	Surveillance	2
Ha et al. (2016) [30]	Education (area-level % without HS degree), Income (area-level quartile)	Diagnosis (2), Treatment (all treatment, 2)	4
Harlan et al. (2015) [31]	Income (area-level median income), Insurance	Treatment (all treatment, 3, locoregional, 5, surgical, 4, systemic, 4), Survival (2)	18
Hoehn et al. (2015) [32]	Income (area-level median income), Education (area-level % without HS degree), Insurance	Treatment (surgical, 3), Survival (3)	6
Hoehn et al. (2017) [33]	Income (area-level median income), Education (area-level % without HS degree)	Treatment (surgical, 2, systemic, 1), Survival (2)	5
Hood et al. (2020) [34]	Income (neighborhood disadvantage index)	Treatment (all treatment, 1), survival (1)	2
Hyder et al. (2013) [35]	Income (area-level income quartiles)	Treatment (lociregional, 3, surgical, 3, systemic, 3), Survival (2)	11
Jan et al. (2012) [36]	Insurance	Survival	1
Jones et al. (2019) [37]	Insurance	Survival	1
Kangas-Dick et al. (2020) [38]	Insurance	Treatment (all treatment)	1
Kokabi et al. (2016) [39]	Income (area-level mean family income), Education (area-level % of adults with less than HS or bachelor’s), Employment (area-level unemployment rate), Insurance	Survival (5)	5
Kokabi et al. (2017) [40]	Income (area-level mean family income), Education (area-level % of adults with HS degree), Insurance	Treatment (all treatment, 4)	4
Kronenfeld et al. (2020, A) [41]	Income (area-level % below poverty line), Insurance	Diagnosis (2)	2
Kronenfeld et al. (2020, B) [42]	Income (area-level % poverty)	Survival	2
Kwong et al. (2010) [43]	Income (area-level SES quintile)	Survival	2
Lee et al. (2020, A) [5]	Income (area-level poverty), Education (area-level % adults with HS degree), Insurance	Treatment (all treatment, 3), Survival (5)	8
Lee et al. (2020, B) [44]	Income (area-level mean household income), Insurance	Survival (2)	2
Lee et al. (2020, C) [45]	Insurance	Survival (1)	1
Major et al. (2014) [46]	Income (area-level % families below poverty line, SES deprivation), Education (% without HS diploma), Employment, Insurance	Incidence (6)	6
Mazumder et al. (2021) [47]	Insurance	Survival (1)	1
Mokdad et al. (2018) [48]	Income (area-level poverty index)	Survival (1)	1
Muzaffar et al. (2020) [49]	Insurance	Survival (1)	1
Peters et al. (2017) [50]	Insurance	Treatment (locoregional, 1, surgical, 2), Survival (1)	4
Pham Hang et al. (2018) [12]	Employment (area-level unemployment rate)	Treatment (any treatment, 1)	1
Rho et al. (2019) [51]	Insurance	Diagnosis (2)	2
Rich et al. (2019) [52]	Insurance	Survival (1)	1
Robbins et al. (2011) [53]	Income (mean area-level household income), Education (area-level % adults without HS degree), Insurance	Treatment (surgical, 3), Survival (3)	6
Robinson et al. (2017) [54]	Insurance	Diagnosis (4)	4
Robinson et al. (2018) [55]	Insurance	Diagnosis (2)	2
Sarpel et al. (2016) [56]	Insurance	Treatment (surgical, 1)	1
Sarpel et al. (2018) [57]	Income (area-level SES), Insurance	Treatment (systemic, 2)	2
Scaglione et al. (2020) [58]	Insurance	Diagnosis (1), Treatment (any treatment, 1), Survival (1)	3
Sellers et al. (2019) [59]	Insurance	Survival (1)	1
Singal et al. (2015) [60]	Insurance	Surveillance (2)	2
Sobotka et al. (2019) [61]	Insurance	Diagnosis (2), Treatment (locoregional, 2, surgical, 2)	6
Sodagari et al. (2019) [13]	Insurance	Treatment (locoregional)	1
Turse et al. (2022) [62]	Insurance	Survival	1
Uhlig et al. (2019) [63]	Income (area-level mean household income), Insurance	Survival (2)	2
Wang et al. (2016) [64]	Income (area-level SES)	Survival (2)	2
Wang et al. (2018) [65]	Insurance	Diagnosis (2), Treatment (any treatment, 1, surgical, 1), Survival (2)	6
Wang et al. (2020) [66]	Income (area-level mean household income)	Diagnosis (1), Survival (1)	2
Wasif et al. (2018) [67]	Education (area-level % adults without HS degree), Insurance	Treatment (surgical, 2)	2
Woodrell et al. (2021) [68]	Insurance	End-of-life care (1)	1
Yan et al. (2016) [69]	Income (area-level income quartile)	Survival (1)	1
Yang et al. (2019) [70]	Insurance	Survival (1)	1
Yu et al. (2010) [71]	Insurance	Treatment (surgical, 1), Survival (1)	2
Zaydfudim et al. (2010) [72]	Insurance	Treatment (surgical, 1), Survival (1)	2
Zou et al. (2018) [73]	Insurance	End-of-life care (1)	1

provide summary characteristics of all included studies. All studies were retrospective in design and analyzed a cohort from an institutional/single-center (14%), multi-center (16%), state (18%), or national database (52%). The following national databases were frequently used: Surveillance, Epidemiology and End Results (SEER), National Cancer Database (NCDB), National Inpatient Sample (NIS), and the SEER-Medicare linked database (descriptions of each are provided in Supplemental Table S2). Multiple studies also utilized the US Safety Net Collaborative (USSNC), which included data from five academic medical institutions and their associated safety-net hospitals. Study sample sizes ranged from 156 [12] to 701,368 individuals [13].

3.2. Incidence

Four included studies [20,25,28,46], encompassing 13 analyses, studied SES predictors and their impact on HCC incidence (Supplemental Table S3). Eight analyses focused on income [20,25,28,46], one each on education [46] and employment [46], and three on insurance status [28,46]. All eight of the analyses with respect to income utilized census tract-level poverty or disadvantage indices as predictors. Four analyses [28] utilized a state-based cancer registry, while one used a national registry (SEER) [46]. Five of eight demonstrated a significant increase in total HCC incidence with higher or more neighborhood poverty [25,28]. However, two demonstrated no significant difference [46], and one demonstrated lower HCC incidence with higher poverty levels [20]. With respect to education and employment, no significant associations were found with HCC incidence between SEER census-tract percentages of retired adults without a high school diploma and unemployment rates [46]. Three analyses examined insurance, including census-tract-level insurance status (insured versus uninsured), percentage of Medicare beneficiaries, and percentage of uninsured patients, all of which found no difference [28,46]. A mixture of area-level and individual-level SES factors may explain these inconsistent findings.

3.3. Screening and Surveillance

Three studies encompassing five analyses [14,29,60] examined the effects of SES predictors on HCC screening/surveillance (Supplemental Table S3), all with insurance status as the predictor. In a single institution safety-net setting, insured individuals had higher rates of HCC surveillance within the past 3 years compared to uninsured individuals, but no difference in rates of at least annual surveillance [60]. In a multi-institution cohort of adults with hepatitis C cirrhosis, insurance was not associated with surveillance; similarly, there was no difference in surveillance rates by insurance type among non-cirrhotic patients with hepatitis B in the Truven Health Analytics national database [14,29].

3.4. Diagnosis

Ten studies [18,30,41,54,55,58,61,74,75,76], encompassing 19 analyses, studied the effects of SES predictors on HCC diagnosis (Supplemental Table S3). Of these, 3 analyses were related to income, 2 to education, and 14 to insurance. For income, no significant associations between household income and distant stage were observed among adults in Mississippi [18]. In a SEER analysis, census-tract median household income levels of less than USD 40,000, USD 40,000–55,000, and USD 55,000–70,000 were all independently associated with higher odds of distant-stage HCC compared to those with median household income greater than USD 70,000 [74]. Census tracts with >25% of the population below the poverty line had higher odds of HCC diagnosis in an emergency department setting across four large metropolitan areas [41]. For education, a higher rate of advanced HCC and a lower rate of HCC within Milan criteria were seen among zip codes with a higher percentage of non-high school graduates in SEER [30].

Thirteen of 14 analyses on insurance utilized a national data registry (i.e., SEER, NIS, NCDB) [41,51,54,55,61,65], while one, the USSNC, was a multi-institution registry [58]. In the USSNC, the uninsured were more likely to be diagnosed in the emergency department compared to the privately insured [41]. Eight analyses focused on localized HCC (by SEER or BCLC staging) or HCC within Milan criteria at diagnosis [54,55,58,65]: six of which found that uninsured or Medicaid patients were less likely to have localized or within-Milan HCC compared to either Medicare and/or privately insured patients [54,55,65], while two failed to demonstrate significant associations between localized HCC and insurance, one conducted among a baby boomer birth cohort [55] and one within a multi-institution registry [58]. One NIS study reported higher odds of metastatic HCC at the time of presentation among Medicare-insured compared to privately insured patients, though not Medicaid nor “other” insurance types [61].

3.5. Treatment

A total of 25 studies [5,12,13,16,22,23,24,30,31,32,33,34,35,40,50,53,56,58,61,65,67,71,72] encompassing 79 analyses were performed with HCC treatment as the outcome: 17 analyses (22%) with a broad outcome of all treatments (including no treatment) [5,12,30,31,34,40,58,65], 15 (19%) focused on locoregional therapies [13,16,23,31,35,50,61], 37 (46%) on surgical therapies [16,22,24,31,32,33,35,53,56,61,65,67,71,72], and 10 (13%) on systemic therapies [23,31,35,57]. The following subsections summarize studies and associated analyses by treatment type (Supplemental Table S4 for all studies). In general, higher income was associated with higher likelihood of treatment, while insurance was the strongest predictor across all treatment modalities, with private and Medicare-insured more likely to receive treatment than Medicaid or uninsured.

3.5.1. Any Treatment

Three of six analyses on the effects of income on receiving any treatment found a general trend of higher relative income being associated with a greater likelihood of receipt of treatment [30,34,40]. The largest study of the SEER registry (2003–2013) identified a lower likelihood of receiving therapy in the lowest two area-level income quartiles compared to the highest quartile, as well as all lower quartiles of education level compared to the highest quartile [30]. Areas with <20% of adults with a high school diploma (vs. >20%) were associated with lower odds of receiving HCC therapy, but not when the threshold was increased to <30% [40]. In a single-center study [12], employed patients were more likely to receive treatment. Six analyses on insurance reported more treatment among those with private insurance compared to those uninsured or with public insurance in national registries [23,31,40,58]. No significant differences in HCC treatment receipt by insurance type were noted in the USSNC [5].

3.5.2. Locoregional Therapies

Fifteen analyses from seven studies [13,16,23,31,35,50,61] focused on receipt of locoregional treatment: seven related to income [16,31,35] and eight to insurance [13,23,31,50,61]. In the Texas Cancer Registry, no differences were seen in rates of ablation by area-level poverty (ranging from less than 5% below the poverty line to greater than 20%), but there was higher utilization of ablation among zip codes with higher mean retirement income and lower utilization among zip codes with lower mean social security income [16]. In the NCI Patterns of Care study, which linked more comprehensive data collection on HCC patients in SEER over one year (2007), areas with median incomes of USD 37,000–50,000 had lower odds of receiving embolization compared to areas with median incomes less than USD 37,000, while no difference was seen in higher income groups [31]. In an older population within the SEER-Medicare database, there were no significant associations between income quintiles and the receipt of interventional oncology care or being seen by an interventional radiology specialist [35]. Analyses of insurance status and locoregional therapy within national databases were mixed in findings. In the largest study of 62,368 representative patients in NIS, those with Medicaid were less likely to receive ablation compared to those privately insured [61]. However, a second NIS study demonstrated a higher likelihood of interventional radiology intervention as compared to surgery among patients with public compared to private insurance [13], potentially suggesting lower preference given to curative treatments, though neither NIS study accounted for cancer stage. A SEER analysis restricted to those with early-stage (AJCC T1/T2) HCC found no difference in locoregional therapy receipt between insured and uninsured [50].

3.5.3. Surgical Therapies Including Liver Resection and Transplantation

Seventeen studies [16,22,23,31,32,33,35,50,53,56,61,65,67,71,72,77] comprising 36 analyses described the impact of SES predictors on receipt of surgical resection and/or liver transplantation (LT): 15 analyses on income [16,22,24,31,32,33,35,53], 4 on education [32,33,53,67], and 17 on insurance [22,23,24,31,32,50,56,61,65,67,71,72]. Studies on income showed mixed associations: two studies [16,22] found a significant and direct association between area-level income and receipt of LT, while three did not [16,31,53]. A SEER-Medicare study showed higher income brackets were more likely to receive surgery [35], while a study of the Pennsylvania Cancer Registry failed to identify an association between area-level median household income and receipt of surgery [24]. Both higher education and private insurance were more consistently linked to surgical treatments. Three studies on education demonstrated significantly lower odds of receiving surgery between zip codes with >14% of adults without a HS degree compared to areas with <14% of adults without a HS degree [32,33,67], while an NCDB study found no association between area-level HS graduation rates and receipt of LT [53]. With respect to insurance status, of ten analyses [22,23,24,31,32,50,56,61,65,67,71,72], nine demonstrated an overall increased likelihood of receiving surgical therapy in patients with private insurance compared to uninsured populations [22,23,24,31,32,50,56,61,65,67,71,72]. Medicaid and Medicare insured patients were also more likely to receive surgical intervention compared to uninsured patients in most settings [23], but not when examining receipt of surgery in a high-volume hospital only [67], and less likely compared to privately insured [24]. In SEER, receipt of LT was also lower among Medicaid and uninsured patients compared to those with either Medicare or private insurance [65].

3.5.4. Systemic Therapies

Four studies [23,31,35,57] with ten analyses described the effects of SES predictors on receiving systemic therapies for HCC. Two studies examined access to or receipt of Sorafenib as a clinical outcome, with both higher composite area-level SES status (vs. lower) and median household incomes of >USD 67,000 (vs. <USD 37,000) associated with increased rates of Sorafenib prescription (OR 2.05–2.20) [31,57]. Income was not associated with evaluation by a medical oncologist or receipt of systemic chemotherapy in SEER-Medicare [35]. All insurance types were found to have significantly higher rates of systemic therapy access compared to the uninsured in the Mountain West region within NCDB [23]. Neither Medicaid nor Medicare insurance groups were found to have significant differences in receipt of systemic chemotherapy in two analyses compared to private insurance; however, both had higher rates of systemic treatment compared to uninsured [23].

3.6. Survival

A total of 40 studies [5,15,16,17,19,21,23,26,27,31,32,33,34,35,36,37,39,42,43,44,45,47,48,49,50,52,53,58,59,62,63,64,65,70,71,72,74] comprising 61 analyses examined the effect of SES predictors on survival outcomes (Supplemental Table S5): 24 analyses related to income [5,16,17,19,27,31,32,33,34,35,39,42,43,44,48,53,63,64,69,74], 6 to education [5,32,33,39,53], 1 to employment [39], and 30 to insurance [5,15,19,21,23,26,31,32,36,37,39,44,45,47,49,50,52,53,58,59,62,63,65,70,71,72].

3.6.1. Survival and Income

Of the 24 analyses on income, three specified HCC mortality rates as the outcome of interest [16,17,33]. Two demonstrated lower mortality rates with higher area-level income measures [16,17], while one analysis of the NCDB found no significant associations between mortality over 15 years of follow-up and mean area-level incomes [33]. Twenty-one analyses examined overall survival (OS) [5,19,27,31,32,34,35,39,42,43,44,48,53,63,64,69,74]. Fourteen (67%) demonstrated an association between lower relative income and reduced OS [19,27,34,35,39,42,43,48,63,69,74,78]. For example, this relationship was seen for lower SES tertiles/quintiles in the California Cancer Registry [43], higher neighborhood disadvantage index in the Ohio Cancer Incidence Surveillance System [34], lower individual-level poverty indices using American Hospital Association Survey-Texas Cancer Registry linked data [48], and mean area-level family income <USD 85,000 in SEER [39]. Seven analyses demonstrated no significant associations; all utilized a national-level database and area-level predictors [5,31,32,33,35,44,53].

3.6.2. Survival and Education

Only one of five analyses found an association between higher area-level education and increased survival: in the USSNC, census tracts with a 73–90% rate of HS graduation (vs. >90%) were associated with lower OS [5]. Otherwise, no significant associations between area-level education (i.e., percentage without a HS degree) and OS were noted in four analyses all utilizing national cancer databases [32,33,39,53]. One study on unemployment also found no significant difference in OS in HCC patients from US zip codes with a <13% versus >13% unemployment rate [39].

3.6.3. Survival and Insurance

The remaining 30 analyses [5,15,19,21,23,26,31,32,36,37,39,44,45,47,49,50,52,53,58,59,62,63,65,70,71,72] focused on insurance status, 15 of which were conducted in national databases [5,15,21,23,31,32,39,44,45,49,50,53,63,65,70], ten in state and multi-institution registries [5,19,37,44,45,47,52,58,72], and five at single centers [26,36,59,62,71]. Ten SEER analyses had differing categorizations of insurance type and selection of comparator group [15,21,31,39,49,50,65]. Of two studies comparing all insured combined and uninsured, one found lower survival with uninsured patients (HR 1.3, CI 1.27–1.33) amongst all HCC cases [49], while the other, limited to only early-stage patients (AJCC T1/T2), found no difference [50]. Nearly all other analyses comparing to uninsured found significantly higher OS with private insurance [15,21,65], composite private/Medicare [21], and composite public insurance with Medicare/Medicaid [21,39,65]. In the 2007 NCI Patterns of Care cohort, no significant difference was observed by insurance type [31]. Three of five analyses comparing privately to publicly insured individuals demonstrated better survival for those with private insurance [15,65]. Two negative studies only included patients with distant-stage HCC [15,31]. All five studies of NCDB had significant results: higher OS among privately insured and publicly insured as compared to uninsured and higher OS for privately insured as compared to publicly insured [23,32,53,63,70].

Among multi-institution or state-level registries, most identified a survival advantage for private payors compared to either public payors or uninsured [19,37,52,72]. In the USSNC, the uninsured had a survival disadvantage [5,44] except when restricting to only HCC patients that received treatment [79], while no difference was seen comparing public versus private payors in all analyses [5,44,79]. One analysis of several academic medical centers around Chicago, IL, found a survival disadvantage for an “other” insurance cohort (HR 1.14) compared to a Medicare/Medicaid cohort [47]. No survival difference between private and public payors was seen in a multi-institution study of the Midwest region [58] and in the Tennessee Cancer Registry [72]. Conversely, in a combined academic and safety-net cohort, Medicaid insured had higher OS compared to privately insured (HR 0.79, CI 0.73–0.85) with no difference between Medicare and privately insured [52]. Among single-center cohorts, one found higher OS with private insurance (vs. uninsured: HR 0.71, CI 0.53–0.97) [26], and one with higher OS among insured (vs. uninsured: HR 0.47, CI 0.29–0.77) [62], while three comparisons of private to public insurance found no significant difference [36,59,71].

3.7. End-of-Life and Survivorship

Two studies focused on end-of-life care for patients with HCC [68,73] (Supplemental Table S3). A single-center study of inpatients at the Mount Sinai Medical Center in New York found no significant differences in rates of palliative care consultation between those with Medicaid and non-Medicaid insurance [68]. Within a national veteran population, individuals with previously treated advanced HCC and a secondary non-Veteran Affairs insurance plan were less likely to utilize hospice services compared to those with solely Veteran Affairs insurance (OR 0.62, CI 0.39–0.97) [73].

3.8. Impact of SES Adjustment on Racial and Ethnic Disparities

Twenty-three analyses from 15 studies (24% of included studies) provided sufficient data to evaluate the influence of SES variable adjustment on the effect estimate for race and ethnicity, summarized in

Table 3. Influence of adjustment for income on the relationship between race and ethnicity and HCC outcome.

Study	Outcome	SES Variable	Race and Ethnic Variable	UV Results	MV Results	Change in Significance	Change in Direction
Bemanian et al. (2019) [20]	Incidence with liver disease diagnosis	Income	Black	1.459 (p < 0.001)	1.099 (p = 0.435)	Partial	No
			Other	1.392 (p < 0.005)	1.602 (<0.001)
			Multiracial	0.978 (0.975)	0.899 (0.886)
			White	Ref	Ref
Chidi et al. 2016 [24]	Receipt of surgical intervention	Income	White	Ref	Ref	Partial	No
			African Am	0.79 (0.66–0.94)	0.89 (0.73–1.10)
			Hispanic	0.71 (0.48–1.04)	0.72 (0.47–1.09)
			Asian	1.66 (1.20–2.27)	1.48 (1.05–2.11)
			Other/Unknown	1.61 (1.02–2.55)	1.44 (0.87–2.37)
Hyder et al. 2013 [35]	Receipt of surgery	Income	White	Ref	Ref	No
			Black	1.05 (0.93–1.18)	1.00 (0.90–1.10)
			Asian	0.87 (0.81–0.94)	0.90 (0.83–0.96)
			Hispanic	1.05 (0.95–1.15)	0.97 (0.89–1.06)
			Other/Unknown	1.07 (0.77–1.48)	0.94 (0.73–1.23)
Kokabi et al. 2016 [39]	Median overall survival	Mean family income	Caucasian	OS 6.0 (5.8–6.3)	0.95 (0.93–0.97)	No
			African Amer	OS 5.0 (4.6–5.5)
			AAPI	OS 8.0 (7.4–8.7)
			Others	OS 5.4 (4.7–5.9)
Kwong et al. 2010 [43]	Cause-specific survival	Income (SES quintile)	Chinese	0.83 (0.77–0.90)	0.89 (0.82–0.96)	No
			Vietnamese	0.83 (0.76–0.90)	0.86 (0.79–0.94)
			Filipino	0.91 (0.83–0.99)	0.89 (0.81–0.98)
			Korean	0.80 (0.73–0.89)	0.90 (0.82–1.00)
			Japanese	0.89 (0.80–1.00)	0.99 (0.87–1.11)
			Laotian/Hmong	2.08 (1.78–2.44)	1.51 (1.28–1.79)
			Cambodian	1.26 (1.06–1.51)	1.24 (1.03–1.48)
			South Asian	0.72 (0.57–0.92)	0.81 (0.64–1.03)
			Thai	1.17 (0.87–1.57)	1.09 (0.81–1.50)
Kwong et al. 2010 [43]	All-cause overall survival	Income (SES quintile)	Chinese	0.80 (0.75–0.85)	0.85 (0.79–0.91)	No
			Vietnamese	0.81 (0.75–0.87)	0.84 (0.78–0.91)
			Filipino	0.97 (0.90–1.05)	0.94 (0.87–1.02)
			Korean	0.77 (0.71–0.84)	0.86 (0.79–0.94)
			Japanese	0.85 (0.77–0.95)	0.93 (0.83–1.03)
			Laotian/Hmong	1.90 (1.64–2.19)	1.43 (1.23–1.66)
			Cambodian	1.25 (1.07–1.46)	1.23 (1.05–1.44)
			South Asian	0.84 (0.69–1.02)	0.92 (0.76–1.11)
			Thai	1.21 (0.94–1.56)	1.15 (0.89–1.48)
Robbins et al. 2011 [53]	Overall survival	Income	White	Ref	Ref	No
			African Amer	0.64 (0.54–0.76)	0.62 (0.52–0.74)
			Hispanic	0.86 (0.75–0.99)	0.88 (0.76–1.02)
			Asian	0.58 (0.49–0.69)	0.67 (0.56–0.81)
Wang et al. 2016 [78]	Overall survival	Income	Race	1.15 (1.12–1.18)	1.20 (1.17–1.24)	No
Wang et al. 2016 [78]	Overall survival	Income	Race	1.16 (1.12–1.19)	1.20 (1.16–1.24)	No
Wong et al. 2020 [74]	Distant (vs. localized) stage of HCC at time of diagnosis	Income	White	Ref	Ref	No
			African Amer	1.30 (1.23–1.39)	1.32 (1.24–1.40)
			Amer Indian/AK	0.99 (0.82–1.20)	0.99 (0.81–1.23)
			AAPI	0.91 (0.86–0.97)	0.96 (0.90–1.02)
			Hispanic	0.92 (0.87–0.97)	0.93 (0.88–0.98)
Wong et al. 2020 [74]	Overall survival	Income	White	Ref	Ref	No
			African Amer	1.13 (1.10–1.16)	1.07 (1.04–1.10)
			Amer Indian/AK	1.02 (0.94–1.10)	0.98 (0.90–1.07)
			AAPI	0.80 (0.78–0.83)	0.83 (0.81–0.85)
			Hispanic	0.99 (0.96–1.01)	0.96 (0.94–0.98)

Bolded items represent changes in significance between univariate (UV) and multivariate analyses.

and

Table 4. Influence of adjustment for insurance on the relationship between race and ethnicity and HCC outcome.

Study	Outcome	SES Variable	Race and Ethnic Variable	UV Results	MV Results	Change in Significance	Change in Direction
Adler Jaffe et al. (2020) [15]	Survival of localized tumors	Insurance	NH White	Ref	Ref	Partial
			NH Black	1.27 (1.16–1.40)	1.11 (1.01–1.22)
			NH A/PI	0.72 (0.64–0.80)	0.79 (0.71–0.89)
			NH AI/AN	0.92 (0.68–1.25)	0.83 (0.60–1.14)
			Hispanic	1.11 (1.03–1.21)	0.94 (0.86–1.02)		Yes
Adler Jaffe et al. (2020) [15]	Survival of regional tumors	Insurance	NH White	Ref	Ref	Partial
			NH Black	1.20 (1.10–1.32)	1.08 (0.98–1.19)		No
			NH A/PI	0.95 (0.85–1.06)	1.01 (0.90–1.12)
			NH AI/AN	0.94 (0.71–1.24)	1.00 (0.75–1.35)
			Hispanic	0.95 (0.87–1.05)	0.91 (0.83–1.00)
Chidi et al. (2016) [24]	Receipt of surgical intervention	Insurance	White	Ref	Ref	No
			African Am	0.97 (0.64–1.47)	1.02 (0.64–1.61)
			Hispanic	0.70 (0.64–1.47)	0.64 (0.27–1.50)
			Asian	2.45 (0.98–6.16)	2.29 (0.90–5.79)
			Other/Unknown	0.44 (0.21–0.93)	0.41 (0.19–0.86)
Jones et al. (2019) [37]	Overall survival	Insurance	Hispanic	Ref	Ref	No
			White	1.07 (1.00–1.14)	1.09 (1.02–1.17)
			Black	1.29 (1.18–1.40)	1.17 (1.07–1.29)
			Asian	0.94 (0.80–1.10)	1.01 (0.85–1.21)
			Unknown	1.02 (0.75–1.39)	1.29 (0.97–1.72)
Kangas-Dick et al. (2020) [38]	Utilization of liver-directed services and procedures	Insurance	White	Ref	Ref	No
			Black	0.62 (0.56–0.69)	0.60 (0.54–0.66)
			Hispanic	0.79 (0.70–0.89)	0.83 (0.74–0.93)
			AAPI	1.31 (1.12–1.52)	1.26 (1.08–1.48)
			Native Amer	0.79 (0.51–1.21)	0.80 (0.51–1.23)
			Other	1.04 (0.84–1.28)	0.98 (0.79–1.22)
Kangas-Dick et al. (2020) [38]	Inpatient mortality	Insurance	White	Ref	Ref	No
			Black	1.18 (1.07–1.30)	1.23 (1.11–1.36)
			Hispanic	0.99 (0.89–1.10)	0.98 (0.88–1.09)
			AAPI	1.06 (0.90–1.24)	1.13 (0.96–1.33)
			Native Amer	1.09 (0.74–1.60)	1.08 (0.72–1.60)
			Other	1.12 (0.91–1.37)	1.19 (0.98–1.45)
Lee et al. (2020) [45]	Overall survival	Insurance	White	Ref	Ref	Partial
			Black	1.20 (1.02–1.41)	0.78 (0.62–0.97)		Yes
			Asian	0.72 (0.53–0.97)	0.65 (0.43–0.99)
			Amer Indian/AK	2.17 (0.30–15.4)	0.17 (0.02–1.29)
			Other	0.79 (0.30–2.12)	0.21 (0.03–1.50
			Unknown	1.33 (0.66–2.68)	2.06 (0.64–6.67)
Mazumder et al. (2021) [47]	All-cause mortality (with just baseline covariates)	Insurance	Black vs. White	1.24 (1.14–1.26)	1.27 (1.14–1.40)	No
Peters et al. (2017) [50]	Receipt of resection (treatment)	Insurance	Caucasian	Ref	Ref	No
			African Amer	1.27 (1.07–1.52)	1.67 (1.13–2.48)
			Amer Indian	0.63 (0.31–1.3)	1.62 (0.46–5.71)
			AAPI	2.82 (2.47–3.22)	2.34 (1.70–3.21)
			Non-Hispanic	Ref	Ref
			Hispanic	0.45 (0.38–0.53)	0.47 (0.30–0.74)
Peters et al. (2017) [50]	Receipt of transplantation (treatment)	Insurance	Caucasian	Ref	Ref	No
			African Amer	0.64 (0.54–0.75)	0.54 (0.36–0.79)
			Amer Indian	0.45 (0.25–0.83)	0.97 (0.35–2.69)
			AAPI	0.76 (0.66–0.88)	0.71 (0.52–0.98)
			Non-Hispanic	Ref	Ref
			Hispanic	0.74 (0.66–0.84)	0.76 (0.57–1.01)
Peters et al. (2017) [50]	Overall survival	Insurance	Caucasian	Ref	Ref	Partial
			African Amer	0.88 (0.76–1.02)	1.24 (0.92–1.66)
			Amer Indian	1.39 (0.95–2.03)	1.35 (0.60–3.03)
			AAPI	1.49 (1.33–1.68)	0.96 (0.74–1.25)		Yes
			Non-Hispanic	Ref	Ref
			Hispanic	0.71 (0.64–0.80)	0.63 (0.48–0.82)
Rich et al. (2019) [52]	Overall survival	Insurance	White	Ref	Ref	No
			Hispanic	1.07 (0.78–1.50)	0.83 (0.74–0.94)
			Black	1.34 (1.11–1.61)	1.12 (1.10–1.14)

Bolded items represent changes in significance between univariate (UV) and multivariate (MV) analyses.

[15,20,24,35,38,39,43,47,50,52,53,74,79,80]. Along the HCC continuum of care, one described the impact on incidence [20], one on the stage at diagnosis [74], six on treatment [24,35,38,50], and 15 on survival [15,37,38,39,43,47,50,52,53,74,78,79]. SES domains included 11 on income (Table 3) [20,24,35,39,43,53,74,78] and 12 on insurance status (Table 4) [15,24,37,38,47,50,52,79]. The racial and ethnic categories represented varied broadly, from comparisons of major racial and ethnic groups to focusing on subgroup ethnicities of the Asian race [43], for example. Among these analyses, six (26%) exhibited a change in significance from unadjusted to adjusted models of the effect estimate for the race and ethnicity variable after SES adjustment [15,20,24,50], while 17 (74%) remained unchanged [14]. Of the six analyses with a change, five associations went from a significant to non-significant association [15,20,24,50], while one remained significant but with a change in direction of effect [79].

3.9. Quality Assessment of Included Studies

All included studies scored between seven and nine stars according to the Newcastle–Ottawa Scale (NOS) and were therefore assessed as good-quality studies (Supplemental Table S6). The consistent quality of included papers was due to the majority being conducted using large population-based cancer or clinical registries specifically for patients with HCC, implying a high degree of representativeness, secure and reliable recordkeeping, and adequate length of follow-up.

4. Discussion

In this systematic review, we synthesized the peer-reviewed literature on socioeconomic factors and association with outcomes along the clinical continuum of care for HCC. Notably, the distribution of available studies along the continuum differed widely: HCC treatment and survival accounted for 43% and 34% of the 179 included multivariate analyses, while diagnosis, incidence, screening/surveillance, and end-of-life care all had much lower representation at ≤10%. Similarly, insurance status was the dominant SES factor examined (50%), followed by income (39%), and less commonly, education (9%) and employment (2%). Our results highlight several important themes and identify gaps in the literature for future investigation.

While previous systematic reviews have focused on race and ethnicity and HCC outcomes [3,81,82], our review importantly focuses on socioeconomic factors, many of which, unlike race and ethnicity, are potentially modifiable. Across published studies in this review, patients with poorer socioeconomic capital across domains trended towards worse clinical outcomes across the continuum of HCC-related care, whether residing in a lower-income census tract, having lower educational attainment, or being under-/uninsured. Data demonstrating worse outcomes with lower SES indicators was most robust for clinical outcomes of stage at diagnosis, receipt of treatment, and survival (see Supplemental Table S5), mostly based on high-quality population-based and national data. Of note, studies of institutional cohorts were more likely to produce non-significant findings, which suggests that among those with access to the same or similar healthcare systems, socioeconomic disparities become attenuated. Our understanding of how social factors and SDOH lead to HCC disparities has advanced dramatically over the past decade, as evidence accumulates around mechanisms such as nativity [83], health literacy [84], and structural racism [85], to name a few. Targeting specific modifiable SDOH among disparity populations may be an efficient way to improve outcomes.

Our review highlights a dearth of literature on the impact of SES on opposite ends of the continuum: HCC screening/surveillance and end-of-life/survivorship care. HCC surveillance is critical to the early detection of HCC, which impacts all downstream outcomes and healthcare delivery costs. Included studies described increased adherence to surveillance among privately insured individuals compared to public and uninsured groups, not surprising, given the higher barriers experienced by safety-net and lower-income groups due to structural factors such as cost, transportation, and scheduling difficulties [86]. However, other social determinants such as income, educational status, and employment have yet to be explicitly explored, though we would expect similar observations of lower surveillance with more social barriers. Provider-level barriers are also important to consider, such as lack of guidance on HCC screening from the USPSTF, which impacts knowledge of primary care providers [87,88], as well as initial recognition of cirrhosis, which is the rate-limiting step in approximately 18% of those diagnosed with HCC without prior screening [89]. We propose investment in patient navigators and technology-aided patient and provider reminder systems as examples of potential low-cost interventions to improve surveillance rates.

On the other end, palliative care and survivorship are increasing areas of interest, especially with advances in systemic treatments and the rising proportion of HCC diagnoses at advanced age. Early palliative care consultation with HCC patients improves symptom management, advanced care planning (including decision-making and advanced directives), and care coordination, which in turn reduces suffering and healthcare utilization, while improving quality of life and even prolonging survival [90]. The few studies on end-of-life care in this review demonstrate that socioeconomic disparities also extend to palliative care and hospice access for marginalized communities, and greater emphasis on strategies to combat this is needed. Existing proposals include reclassifying hospice care as an essential benefit in state-required Medicaid programs, expanding Medicare’s long-term care role to include hospice services, and facilitating hospice and palliative care within rural and federally qualified healthcare centers [91].

Not surprisingly, insurance coverage is a major determinant of outcomes across the continuum of HCC care: individuals with private insurance largely had better outcomes compared to those with publicly funded insurance plans, and those with public insurance also had better outcomes compared to those with no insurance. The reasons for this are manifold, including but not limited to (a) insurance dictates where a patient can receive care, which can affect the quality of healthcare delivery; (b) treatment access differs by insurance status (i.e., transplant is not uniformly accessible for those with public or no insurance); (c) insurance barriers result in lack of or delays in healthcare utilization; and (d) insurance is a proxy for financial well-being, and financial toxicity is well-known to impact receipt of care. Studies have shown that the creation and expansion of the Affordable Care Act (ACA) were associated with increased detection of pre-existing medical conditions, most noticeable among community health centers in underserved zip codes [92]. Increased rates of HCC screening have resulted directly from Medicaid expansion or access to a health insurance plan [42,93]. The importance of health insurance as a modifiable risk factor impacting patient outcomes cannot be understated. Ensuring our most socioeconomically disadvantaged patients with cirrhosis are enrolled in a health insurance plan early in their clinical course would positively impact prompt access to providers, modification of lifestyle and clinical risk factors for cancer development, higher adherence to surveillance, and early uptake of appropriate treatments.

The intersectionality of race, ethnicity, and socioeconomic status is critical to consider. Six out of 23 (including two out of 11 income and four out of 12 insurance) effect estimates for race and ethnicity among studies with available data in this review were impacted by SES variable adjustment (Table 3 and Table 4), indicative that race likely exerts an independent influence on cancer outcomes but also underscoring the potential for SES as a mediating or confounding factor. While we cannot conclude based on our methodology that SES itself was the driving force behind changes in significance for racial and ethnic associations given the inclusion of other clinical and tumor-related characteristics in adjusted models, we do strongly advocate for the inclusion of SES variables in the evaluation of racial and ethnic disparities in HCC as we move forward in understanding root causes of such disparities.

Our review is the first to systematically and comprehensively characterize the impact of SES/SDOH on the HCC care cascade, serving as a guide to the breadth of knowledge on this topic while stressing remaining gaps. However, there are a few notable limitations. First, the timing of our search limited studies to those older than 2021, and more recently published data were not captured. A separate grey literature search was also not conducted. Second, although we intended to conduct meta-analyses, the heterogeneity of study predictors, outcomes, and populations limited this to a narrative synthesis. Third, it is important to note that most SES factors outside of insurance were linked to cancer databases at the area level (i.e., census-tract aggregate measures obtained from national survey data) among included studies. This was largely due to the unavailability of individual-level data in national and institutional databases and is problematic for many reasons, such as the potential for misclassification of SES [94], ecological bias with aggregated data, and the use of measures that may not lead to actionable changes. Fourth, we also need to acknowledge the potential for publication bias toward studies from high-resource databases. As screening for social risk drivers is increasingly incorporated into medical visits as mandated by the Centers for Medicare and Medicaid Services in 2024, there is an immense research opportunity to more precisely and accurately pinpoint the social determinants, such as transportation, food, and housing insecurity, that individually and collectively impact HCC outcomes in clinical cohorts, while simultaneously developing system-level interventions to connect vulnerable patients to available resources.

5. Conclusions

In conclusion, this systematic review demonstrates that lower SES—reflected by poverty, limited education, and inadequate insurance—consistently predicts worse HCC outcomes. Particularly, health insurance access and coverage is a significant and modifiable factor across the care continuum. Future research should target underrepresented stages of the care continuum, such as surveillance and survivorship, and emphasize modifiable individual-level social determinants of health.