Investigation of Novel Urinary Biomarkers in Hepatocellular Carcinoma Risk in a Predominantly African American Population: A Case-Control Study

Hepatocellular carcinoma (HCC) is the third leading cause of cancer deaths in the world and the sixth most diagnosed cancer based on 2020 statistics [1]. The majority of HCC occurs in patients with underlying chronic liver disease, which usually produces liver inflammation [2]. Racial disparities in death from HCC in the USA have previously been linked to the risk of HCC-related death in Asian Americans, American Indian/Alaskan natives, and African Americans in comparison to non-Hispanic whites [3]. African Americans have also been more frequently diagnosed with advanced liver disease and have lower survival than whites [4]. The molecular etiology of these differences has not been elucidated.

Activation of inflammatory signaling pathways produces an environment that is rich in various cytokines, prostaglandins, and reactive oxygen species (ROS). The cyclooxygenases (COX-1 and COX-2) are key enzymes responsible for conversion of arachidonic acid to prostaglandins, such as prostaglandin E₂ (PGE₂), prostaglandin D (PGD₂), thromboxane A₂ (TXA₂), and prostaglandin I₂ (PGI₂), also known as prostacyclin. Nonsteroidal anti-inflammatory medications (NSAIDs) inhibit the COXs and decrease the production of their downstream products. NSAID use has previously been associated with reduced risks of colorectal and other digestive tract cancers, with several in vivo studies now reporting decreased incidence and recurrence of HCC associated with use as well [5, 6]. Thus, COX inhibition may play a role in inhibiting carcinogenesis (Figure 1).

Of the prostaglandins, PGE₂ is also likely to be the mediator of the effects of COX-2 in liver carcinogenesis. Overexpression of COX-2 or treatment with PGE₂ has been shown to increase HCC invasion and growth [7]. Furthermore, PGE₂ was described as increasing human HCC cell proliferation. Importantly, PGE₂ levels are increased in HCC tissue and adjacent non-malignant tissues when compared to normal hepatic parenchyma [8]. The half-life of PGE₂ in humans is very short; thus, quantification of the major urinary metabolite of PGE₂ (9,15-dioxo-11α-hydroxy-13,14-dihydro-2,3,4,5-tetranor-prostan-1,20-dioic acid, PGE-M) is an accurate and precise method to assess endogenous PGE₂ production in humans [9].

PGD_2, on the other hand, seems to decrease tumor progression, and higher concentrations of PGD₂ tend to slow down tumor growth and metastasis in animal models [10]. Elevated levels of PGD₂ can decrease the number of ROS and increase apoptosis in non-small cell lung cancer, slow down the growth of gastric adenocarcinoma, and be associated with decreased number of metastasis in melanoma [11‒13]. The data on association between urinary PDG₂ metabolite (PGD-M) and cancers is limited [14, 15].

PGI₂ is another [16] product of arachidonic acid metabolism by COX-1 and COX-2. It counteracts the effects of ROS in the liver [17] and mediates normal hepatic perfusion. Prostacyclin analogs have been investigated as therapeutic targets for delaying hepatic fibrosis formation [18]. It also has antimetastatic activity on liver metastasis of human pancreatic cancer in a mouse model [19]. The best way to estimate systemic biosynthesis of PGI₂ is to measure its urinary metabolite, 2,3-dinor-6-keto-PGF_1α aka PGI-M.

Thromboxane A₂ (TxA₂) is a mediator of platelet activation. TxA₂ increases cancer proliferation, invasion, and angiogenesis in colorectal and bladder cancers [20, 21]. Furthermore, TxA₂ inhibitors suppress formation of metastasis in pulmonary and hepatic tissue in experimental mouse models [22, 23]. TxA₂ is usually produced in response to physiological stress and has been implicated in mediating hepatic inflammation and injury [24]. 11-dehydro-TxB₂ is a major urinary metabolite of TxA₂ and is readily measured in urine. 11dTxB₂ has previously been shown to be increased in patients with cirrhosis [25, 26]. However, the role of TxA₂ or 11dTxB₂ in HCC has not been investigated. In addition, previous reports identified alterations in the ratio of prostacyclin to thromboxane in the development of adverse cardiovascular events in patients on antiplatelet therapy or with breast cancer or preeclampsia [27‒29].

The Southern Community Cohort Study (SCCS) is a large prospective cohort study including over 80,000 study participants and was designed to evaluate health disparities in cancer. About two-thirds of participants are African Americans and about two-thirds live below poverty. SCCS is also one of the few cohort studies in which baseline urine samples have been banked. In this study, we aimed to prospectively evaluate baseline urinary PGE-M, PGI-M, 11dTxB₂, and TN-E in the SCCA cohort and correlate them with risk of incident HCC.

Detailed methods for the Southern Community Cohort Study (SCCS) have been previously published [30]. Briefly, the SCCS was designed to study cancer disparities including racial (black/white), socioeconomic, and rural/urban disparities within 12 southeastern US states [30]. SCCS participants were recruited in-person from community health centers (86%) or from a random sample of adults who responded to a mailed survey (14%). A total of 86,000 participants enrolled in the study between 2002 and 2009. All study participants were aged 40–79 years at the time of enrollment. Study participants completed an extensive questionnaire during enrollment which solicited information on the participants’ behaviors, health, and environment.

During 2004–2009, a subset of participants (n = 23,400) provided an approximately 30 mL spot urine sample at baseline. Ascorbic acid was added to the samples, and they were shipped cold daily to the study laboratory. Urine samples were spun at 2,000 × g for 10 min using a refrigerated centrifuge (at 4°C). The urine was then removed and pipetted into 7 sterile 3.5 mL cryovials. The pellet (including exfoliated cells) was suspended in 1.5 mL TE buffer and transferred to a 3.5 mL cryovial. All samples are stored in −80°C freezers until use.

In this case-control study, 50 incident liver cancer cases diagnosed after enrollment and who had a urine sample were identified. Out of 50 cases, 48 were identified from state cancer registries, with 43 identified as HCC and 5 as “neoplasm-malignant.” An additional 2 cases were identified from the National Death Index and their morphology is unknown. These 50 cases were matched to 50 controls who did not develop liver cancer and were cancer-free at the time of the diagnosis of the case. Matching was based on sex (male/female), racial group (black, white, other/unknown), NSAID use (defined as ever/never use of two or more times/week for 1 month of low-dose aspirin, aspirin, Advil, Motrin, Celebrex, Vioxx, Bextra), age at enrollment (+/− 5 years), and date of urine collection (+/−12 months). An additional 101 cancer-free controls were included in this analysis from a separate nested case-control study of colorectal cancer. Cancer diagnosis was obtained through annual linkage with the 12 state cancer registries and the National Death Index (NDI). Participants were followed for a median of 10 years for cancer incidence. No participants had a cancer diagnosis prior to enrollment.

Urine samples from all cases and controls were analyzed at the same time. Concentrations of 9,15-dioxo-11α-hydroxy-13,14-dihydro-2,3,4,5-tetranor-prostan-1,20-dioic acid (PGE-M), tetranor-PGE₁ (TN-E), 9α-hydroxy-11,15-dioxo-13,14-dihydro-2,3,4,5-tetranor-prostan-1,20-dioic acid (tetranor PGD-M), 11-dehydro-thromboxane B₂ (11dTxB₂), and 2,3-dinor-6-keto-PGF₁α (PGI-M) in urine were measured in the Eicosanoid Core Laboratory at Vanderbilt University Medical Center. Urine (1 mL) was acidified to pH 3 with HCl. [²H₄]-2,3-dinor-6-keto-PGF1a, [²H₄]-11-dehydro-TxB₂, and [²H₁₁]-tetranor-PGE₁ were added, and the sample was treated with methyloxime HCl to convert analytes to the O-methyloxime derivative. The derivatized analytes were extracted using a C-18 Sep-Pak (Waters Corp., Milford, MA, USA) and eluted with ethyl acetate as previously described [31, 32]. A [²H₆]-O-methyloxime PGE-M deuterated internal standard was then added for PGE-M and PGD-M quantification. The sample was dried under a stream of dry nitrogen at 37 °C and then reconstituted in 75 μL mobile phase A (see below) for LC/MS analysis.

LC was performed on a 2.0 × 50 mm, 1.7 mm particle Acquity BEH C18 column (Waters Corporation, Milford, MA, USA) using a Waters Acquity I-Class UPLC. Mobile phase A was 95/5/0.01 (v/v/v) water/acetonitrile/acetic acid, and mobile phase B was 10/90/0.01 (v/v/v) water/acetonitrile/acetic acid. Samples were separated by a gradient of 85–5% of mobile phase A over 12 min at a flow rate of 0.375 mL/min prior to delivery to a Waters Xevo TQ-XS triple quadrupole mass spectrometer.

Urinary creatinine levels were measured using a test kit from Enzo Life Sciences. The urinary metabolite levels in each sample are normalized using the urinary creatinine level of the sample and expressed in ng/mg creatinine. The ratio of PGI-M to 11dTXB₂ was calculated to reflect the balance of vasodilator to vasoconstrictor in the environment.

The mean CVs for intra-and inter-assay variability of PGE-M were found to be 4.1% and 6.7%, respectively. The mean CVs for intra-and inter-assay variability of TN-E were 8.1% and 6.1%, respectively.

Differences in potential confounders by case/control status were derived from χ² (categorical) or t tests (continuous). The biomarkers were ranked into tertiles according to the control distribution. The PGI-M to 11dTXB₂ ratio was calculated. In order to avoid division by zero, we imputed the lower limit of detection for PGI-M (0.01) and 11dTXB₂ (0.005) and divided it first by the creatinine value and then by the square root of 2. The ratio was then grouped into four categories (0.001–0.24, 0.25–0.49, 0.5–0.99 and ≥1) to approximate quartiles in controls. Any covariates which significantly differed among the two groups were considered potential confounding factors. The potential confounding factors were age (years), sex (male/female), race (Black, white), body mass index (BMI, <25, 25–29, 30–34, 35+ kg/m²), household income (less than $15 K, $15–49 K, more than $50 K per year), education (less than high school, high school or equivalent, college or more), smoking status (current, former, and never smokers), alcohol use [none, moderate (>0 and ≤1 drink per day for females or >0 and ≤2 drinks per day for males), and heavy drinkers (>1 drink per day for females, and >2 drinks per day for males)], and regular NSAID use (yes, no). BMI was missing for 2 participants, education level was missing for 1 participant, and income was missing for 1 participant. The median or mode was assigned to these participants. Odds ratios (ORs) and 95% confidence intervals (CIs) were derived from multivariate logistic regression models. Tests for linear trend were performed by entering the categorical tertile variables as continuous variables in the model. A sensitivity analysis was performed by using conditional logistic regression for the liver cancer cases and their matched controls. Effect estimates were similar to the full pooled study, although the confidence intervals were wide due to the small sample size. All p values were 2-sided with a significance level of <0.05. All statistical analyses were performed using SAS 9.4.

Among the 50 liver cancer cases, 10 were diagnosed in less than 24 months after enrollment and 7 were diagnosed at or over 10 years after, median time to diagnosis of 6 years (range 0–11; results not shown in table). In comparison to controls, patients with liver cancer were more likely to be male (74 vs. 49%, p = 0.002), be a current smoker (72 vs. 45%, p = 0.001), use more NSAIDs (48 vs. 28%, p = 0.01), and have higher rates of moderate and heavy alcohol consumption (46 vs. 38% and 34 vs. 21%, respectively, p = 0.02) (Table 1).

Overall, in an unadjusted analysis, the levels of PGE-M were higher in cases versus controls (20.2 vs. 14.7 ng/mg Cr, p = 0.04) as were TN-E levels (7.9 vs. 5.3 ng/mg Cr, p = 0.03), PGI-M levels (0.49 vs. 0.22, p = 0.03), and 11dTxB₂ levels (1.24 vs. 0.40, p = 0.004). A similar pattern was seen using geometric means (see online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000538131). However, after adjustment for potential confounders (Table 2), only 11dTxB₂ was significantly associated with incident cancer (OR = 11.50; 95% CI [2.34–56.47], p = 0.004). This association was present among those diagnosed either less than 5 years after baseline or 5 years or more since baseline (OR = 14.21; 95% CI [1.08–187.36]; p = 0.0005 and OR = 12.02; 95% CI [1.56–92.53]; p = 0.009, respectively).

We then proceeded to analyze PGI-M to 11dTXB₂ ratio. In order to avoid division by zero, for undetectable 11dTXB₂ values, we used the creatinine standardized logarithm of the odds values divided by square root of 2. The ratio was then grouped into four categories (0.001–0.24; 0.25–0.49; 0.5–0.99; and ≥1). PGI-M to 11dTXB₂ ratio of 0.25–0.49 was associated with a fivefold increased risk of liver cancer in comparison to a ratio greater than 1 (Table 3). This association was strengthened after adjustment for the other biomarkers (OR = 5.16; 95% CI [1.44–18.47]; p = 0.001).

In this study, we investigated 5 potential urinary biomarkers of liver cancer, predominantly HCC risk, within the Southern Community Cohort Study, a population of largely black or low-income participants. We found statistically significant associations of 11dTxB₂ and PGI-M/11dTxB₂ with subsequent risk of liver cancer, while PGD-M, PGI-M, PGE-M, and TN-E failed to show a statistically significant association.

This is the first study to report on association of 11dTxB₂ or PGI-M/11dTxB₂ ratio with liver cancer and HCC risk. The association persisted for both cases diagnosed within and longer than 5 years since urine collection. Thromboxane A2 synthase inhibitors have been previously demonstrated to decrease hepatitis C infectivity in ex vivo models. Since viral hepatitis data is not available in our population and only one control subject self-reported hepatitis without further specification, this may represent a surrogate marker of viral hepatitis infection [33]. TxA₂ has prothrombotic properties (platelet aggregation) and vasoconstrictor properties activated as a part of inflammatory response. TxA₂ production increases in the setting of tissue injury and subsequent inflammation, platelet aggregation, and vasoconstriction, thus decreasing the blood supply to the injured site. Platelets have been shown to promote cancer development through angiogenesis and protection of metastatic cells which allows evasion of the immune system in circulation [22, 34]. The observed elevation of 11dTxB₂ in our cohort who went on to develop liver cancer is in line with previous reports of 11dTxB₂ correlation with breast cancer growth and progression [35] as well as with diagnosis of colorectal cancer [36].

As a vasoconstrictor, TxA₂ exists in equilibrium with its physiological antagonist prostacyclin which is a potent vasodilator. We showed that decreased PGI-M/11dTxB₂ ratio, which reflects relative abundance of TxA₂ as compared to prostacyclin, predicts future liver cancer development in the SCCS. Similar findings were reported in breast cancer [37], endometrial cancer [38], and thyroid cancer [39]. This finding likely reflects inflammatory milieu which is associated with tissue injury and possibly carcinogenesis.

PGE-M has been shown to be elevated in individuals with gastric cancer, colorectal cancer, as well as in smokers and overweight/obese patients [40‒42]. 2,3,4,5-tetranor-PGE₁ (TN-E) is a secondary metabolite of PGE₂ excreted in urine [43]. Formation of this metabolite is not dependent upon the action of 15-prostaglandin dehydrogenase (15-PGDH). 15-PGDH has been shown to be downregulated in human hepatoma cells with high expression of COX-2. Quantification of both PGE-M and TN-E gives a more comprehensive representation of endogenous PGE₂ production. However, our analysis of two primary urinary metabolites of PGE₂ was not associated with subsequent liver cancer risk. These findings are in concordance with the only previously published study of PGE-M levels and HCC risk in an Asian population, which also failed to show a significant association with HCC [44]. Possible confounding of this lack of correlation may lie in our inability to assess for concurrent liver disease at the time of urine collection. Underlying liver disease may interfere with PGE₂ metabolism, thus resulting in lower systemic levels of PGE₂ and its metabolite, TN-E. Our study also failed to demonstrate an association between PGD-M levels and the risk of liver cancer. There are no previous studies for comparison.

Our study has some limitations. Although it is the first study to explore several of these biomarkers with liver cancer risk, only a small number of cases were included which may have affected our ability to detect an association. Unfortunately, the incidence of cirrhosis and viral hepatitis was not available in our dataset. We acknowledge that this represents a major limitation in methodology of this paper, considering that the presence of cirrhosis and hepatitis B are major factors for liver cancer. Another limitation is that observed relationship could be a result of reverse causation bias, whereas the unrecognized presence of HCC many leads to elevated levels, which may make these biomarkers a useful attribute in diagnosis of HCC. Likewise, the urine sample obtained provides a single, baseline measurement of urinary metabolites and may not reflect the most etiologically relevant time period. However, the relationships were consistent for diagnoses that occurred within 5 years of biospecimen collection and diagnoses that occurred more than 5 years after collection as well as analyses that excluded cases diagnosed within the first 2 years (results not shown). Future studies should evaluate longer time periods. The prospective analysis and well-characterized population are additional strengths of this study which has newly evaluated several of these biomarkers with liver cancer risk.

Despite these limitations, we believe that this study provides grounds for future prospective studies of urinary TxA₂ markers in HCC risk. 11dTxB₂ and PGI-M/11dTxB₂ ratio may also be urinary markers of HCC risk that warrant future study as well as the development of easily accessible methods.

This work has been presented at Digestive Disease Week 2022 as a poster presentation entitled Tu1278: Investigation of Novel Urinary Biomarkers in Hepatocellular Carcinoma Development in Predominantly African American Patient Population, available at https://www.sciencedirect.com/science/article/abs/pii/S0016508522636990?via%3Dihub.

All participants provided written informed consent through the Vanderbilt University Medical Center Institutional Review Board. The IRB number is 010345.

The authors have no conflicts of interest to declare.

The Southern Community Cohort Study is funded by U01CA202979 from the National Cancer Institute (NCI) at the National Institutes of Health (NIH). Data collection for the Southern Community Cohort Study was performed by the Survey and Biospecimen Shared Resource, which is supported in part by the Vanderbilt-Ingram Cancer Center (P30CA68485). Measurements of the biomarkers in this study were supported by a pilot project award from the Vanderbilt Gastrointestinal Specialized Program of Research Excellence (SPOREs), funded by grant P50CA236733 from the National Cancer Institute.

All authors contributed equally to development of the project, data analysis, and manuscript write-up and review; approved the final version of the manuscript; and agree to be accountable for all aspects of the work. Alexandra Shingina, Xijing Han, Lei Fan, Harvey Murff, Robert Coffey, Ginger L. Milne, Qi Dai, and Martha Shrubsole contributed to design of the project and data acquisition, analysis, and interpretation. Alexandra Shingina and Martha Shrubsole drafted the manuscript, and Xijing Han, Lei Fan, Harvey Murff, Robert Coffey, Ginger L. Milne, and Qi Dai critically reviewed it.

Data are available upon request through the Southern Community Cohort Study Data and Biospecimen Use Committee when consistent with the informed consent of study participants. Further inquiries can be directed to the corresponding author. Instructions are available at https://www.southerncommunitystudy.org/.