Circulating Levels of Inflammatory and Metabolic Biomarkers and Risk of Esophageal Adenocarcinoma and Barrett Esophagus: Systematic Review and Meta-analysis

The incidence of esophageal adenocarcinoma has rapidly increased in Western populations (1), although the increase seems to have slowed down during recent years in some countries, for example, in the United Kingdom and the Netherlands (2, 3). The prognosis of esophageal adenocarcinoma is poor with an overall 5-year survival <15% to 20% (1, 4). There is a striking male predominance in esophageal adenocarcinoma with the male-to-female incidence ratio of up to 9:1, for which the reasons remain unclear (1, 5). Barrett esophagus, a replacement of the native squamous lining of the esophagus with a specialized columnar epithelium (metaplasia), is the precursor of esophageal adenocarcinoma (1, 6).

Gastroesophageal reflux disease (GERD) and obesity are the main risk factors for esophageal adenocarcinoma and Barrett esophagus (1, 4, 6). GERD can damage the esophageal mucosa and lead to esophagitis, Barrett esophagus, and subsequently esophageal adenocarcinoma. Obesity, particularly central obesity (typical male fat distribution), may promote reflux through increased intragastric pressure and disruption of the gastroesophageal junction and the lower esophageal sphincter (7). Obesity is also a systemic disease that might increase esophageal adenocarcinoma risk through other mechanisms, including chronic inflammation and metabolic alterations (6, 8).

Despite these strong and readily assessable risk factors, it has been difficult to identify individuals with a high absolute risk of esophageal adenocarcinoma enough to advocate endoscopic screening or surveillance. Circulating biomarkers could be a useful addition in this respect. Some studies have investigated associations between circulating levels of inflammatory and metabolic biomarkers and the risk of esophageal adenocarcinoma or Barrett esophagus (9, 10). The findings from the individual studies are not consistent, however, and whether associations differ between the sexes and contribute to the strong male predominance is unclear.

To clarify the role of a range of circulating levels of obesity-related inflammatory and metabolic biomarkers in the development and prediction of esophageal adenocarcinoma and Barrett esophagus, we conducted a systematic review and meta-analysis.

A systematic search for studies published in MEDLINE and EMBASE databases from inception through October 2018 was conducted with no language restriction. The search strategy is presented in detail in Supplementary Table S1. Briefly, we used a combination of keywords for inflammatory and metabolic biomarkers and those for the outcomes, esophageal adenocarcinoma and Barrett esophagus, to identify relevant publications. Candidate biomarkers were those that had been reported to be associated with obesity or cancer risk. The search terms for the biomarkers were predefined after a scoping search and referring to two previous systematic reviews on the topic (9, 11). We also reviewed the reference lists of eligible original articles and the two previous systematic reviews to identify additional studies.

Studies meeting the following criteria were included: (i) cross-sectional, case–control, or cohort studies in humans and published as original articles, (ii) measuring the incidence of esophageal adenocarcinoma or Barrett esophagus (rather than mortality) as an outcome, (iii) examining associations between circulating levels of inflammatory or metabolic biomarkers and risk of esophageal adenocarcinoma or Barrett esophagus, and (iv) containing information necessary to estimate RR compared with a reference group and with a measure of precision [e.g. confidence interval (CI), SE, variance, χ² and degree of freedom, or P value]. In case of multiple reports on the same biomarker from the same study population, only the most recent or informative ones were included. Case–control studies in which biomarker levels were measured at or after onset of the disease outcome were considered cross-sectional, because the temporal relation could not be determined.

The following information was collected from the eligible studies into an electronic database by one researcher (S. Rabbani) and independently checked by a second researcher (S.-H. Xie): (i) study design and characteristics (first author, year of publication, study setting, follow-up period, number of participants by group, study population or comparison group, and verification of cases), (ii) participants' age and sex, (iii) examined biomarkers, (iv) statistical analysis strategy (statistical model, covariates matched or adjusted for, and any stratified or sensitivity analysis), and (v) main findings. The quality of the included studies was independently assessed by two researchers (S.-H. Xie and E. Ness-Jensen) and discrepancies were resolved by joint review of reports to reach consensus. The study quality was quantitatively scored according to the Newcastle–Ottawa scale for assessing the quality of nonrandomized studies in meta-analyses (12). This scale contains eight items, which are categorized into three domains, that is, selection, comparability, and assessment of exposure (case–control or cross-sectional studies) or outcome (cohort studies) of study participants, and the assessment provides a score ranging from 0 to 9, where higher scores indicate better quality.

Meta-analyses were performed for associations between circulating biomarker levels and the risk of esophageal adenocarcinoma, Barrett esophagus, and the combined outcome esophageal adenocarcinoma or Barrett esophagus (hereafter labeled as EAC/BE), whenever possible. Pooled ORs with 95% CIs were estimated using a random effect model. The following 13 biomarkers, for which data were available in at least two studies, were examined: (i) adiponectin, (ii) leptin, (iii) ghrelin, (iv) glucose, (v) insulin, (vi) insulin-like growth factor 1 (IGF-1), (vii) IGF-binding protein 3 (IGFBP-3), (viii) triglycerides, (ix) C-reactive protein (CRP), (x) IL6, (xi) IL8, (xii) TNFα, and (xiii) soluble TNF receptor 2 (sTNFR-2). HRs and risk ratios reported in cohort studies were used as proxies of OR, which was justified by the low incidence of esophageal adenocarcinoma and Barrett esophagus in the population. For all biomarkers, except for glucose and triglycerides, we transformed effect sizes into a common scale of comparison before conducting the meta-analysis, that is, comparing the highest versus lowest tertiles, under the assumption that the biomarker level is normally distributed and has a log-linear association with the outcome risk (13, 14). The logarithm of OR for the highest versus lowest tertiles was estimated to be 1.27 times that for the top versus bottom halves, 0.86 times the highest versus lowest quartiles, 0.78 times the highest versus lowest quintiles, and 2.18 times that for per SD increase. When studies reported risk estimates with different degrees of statistical adjustment for covariates, we used the fully adjusted estimates. If multiple measures of obesity were employed, we used the risk estimates adjusted for waist-to-hip ratio, over waist circumference, and over body mass index. Four studies compared biomarker levels in patients with Barrett esophagus separately with two control groups, that is, patients with GERD and another control group from the general population or patients undergoing colonoscopy screening (15–18). For these studies, we used the risk estimates calculated in comparison with patients with GERD in the main analyses. For the two most studied biomarkers, that is, adiponectin and leptin, we conducted separate meta-analyses of risk estimates using different comparators, that is, (i) patients with GERD (for esophageal adenocarcinoma or Barrett esophagus) or patients with Barrett esophagus (for esophageal adenocarcinoma) and (ii) general population or patients undergoing colonoscopy or upper endoscopy. We also conducted meta-analysis excluding one study that compared patients with esophageal adenocarcinoma with patients with Barrett esophagus (19), on the combined outcome EAC/BE for adiponectin. We stratified the analyses by sex for adiponectin and leptin, but not for other biomarkers because of the limited number of available studies.

Statistical heterogeneity across studies was assessed by the I² statistic and Cochran Q test. The I² statistics indicate the proportion of the total variance in risk estimates across studies that is due to heterogeneity rather than chance, and thus, the I² lies between 0% and 100%, with larger values indicating more heterogeneity (20). A P < 0.10 in the Q test was considered statistically significant, as conventionally used (21). Publication bias was assessed by visual inspection of funnel plots and the Begg and Egger tests (22, 23).

The software Comprehensive Meta-Analysis version 3 (Biostat, Inc.) was used for all statistical analyses. All statistical tests were two-sided. We interpreted the results in terms of magnitudes of associations and precision of the risk estimates conveyed by 95% CIs, rather than using P values as measures of significance (24, 25).

The literature search identified 7,641 studies. Among these, 19 studies fulfilled the inclusion criteria for meta-analysis, including 12 cross-sectional (15–18, 26–33), two nested case–control (10, 34), and five cohort studies (19, 35–38). The full selection procedure of eligible studies is shown in Supplementary Fig. S1. Fourteen of the eligible studies were conducted in the United States, two in Europe, one was conducted in Australia, and the remaining one study included participants from both the United States and Europe. Eleven of 12 cross-sectional studies measured circulating levels of biomarkers in patients with Barrett esophagus compared with patients with GERD, general population, or patients who had undergone endoscopy, while esophageal adenocarcinoma was the outcome in the remaining one cross-sectional study. All nested case–control and cohort studies investigated associations between biomarker levels and esophageal adenocarcinoma risk, either in the general population or in cohorts of patients with Barrett esophagus. Characteristics of the included cross-sectional and nested case–control studies are presented in Table 1 and the cohort studies in Table 2.

A detailed study quality assessment is shown in Supplementary Tables S2 and S3. Briefly, most studies (14/19) had overall quality scores ranging from 6 to 8, while two cross-sectional studies scored lower (4 or 5) and the remaining one cross-sectional study and two cohort studies scored higher (9). All included studies, except for the two cross-sectional studies with lower quality scores, performed well in terms of comparability, that is, regarding controlling for confounding from the major risk factors for esophageal adenocarcinoma and Barrett esophagus, that is, age, sex, GERD, obesity, and tobacco smoking.

Seven cross-sectional studies reported associations of circulating adiponectin levels with Barrett esophagus risk, and one nested case–control study and one cohort study reported associations with esophageal adenocarcinoma risk. Meta-analysis showed no associations between the highest versus lowest tertiles of adiponectin levels and risk of Barrett esophagus (pooled OR, 0.90; 95% CI, 0.59–1.37), esophageal adenocarcinoma (OR, 0.87; 95% CI, 0.60–1.25), or EAC/BE (OR, 0.88; 95% CI, 0.67–1.16; Figs. 1 and 2). Adiponectin levels were not associated with risk of EAC/BE in the meta-analysis when excluding the study comparing patients with esophageal adenocarcinoma with patients with Barrett esophagus (OR, 0.91; 95% CI, −0.65–1.27). There was a tendency of a decreased risk of EAC/BE associated with higher adiponectin levels in comparison with the general population or patients undergoing endoscopy (OR, 0.79; 95% CI, 0.55–1.12), but not in comparison with patients with GERD or Barrett esophagus (OR, 0.96; 95% CI, 0.56–1.55; Supplementary Fig. S2). Stratified analysis by sex showed no associations between adiponectin levels and risk of EAC/BE either in men or in women (Fig. 3).

Meta-analysis of seven studies (including one pilot study and one validation study in Australian) found an increased risk of Barrett esophagus associated with higher leptin levels (OR, 1.68; 95% CI, 0.95–2.97, comparing the highest vs. lowest tertiles; Figs. 1 and 2). An increased esophageal adenocarcinoma risk was also reported in one cohort study (OR, 1.53; 95% CI, 0.58–4.05). Meta-analysis of these eight studies generated similar estimates for the combined outcome, EAC/BE (OR, 1.64; 95% CI, 1.01–2.68; Fig. 2). Stratified analysis by type of comparators showed that the association for leptin was restricted in comparison with the general population or patients undergoing colonoscopy screening (OR, 1.90; 95% CI, 1.29–2.81) rather than in comparison with patients with GERD or Barrett esophagus (OR, 0.97; 95% CI, 0.49–1.90; Supplementary Fig. S3). The association between leptin levels and risk of EAC/BE was not substantially stronger in men (OR, 2.00; 95% CI, 0.96–4.16) than in women (OR, 1.57; 95% CI, 0.29–8.42; Fig. 3).

Associations between ghrelin levels and Barrett esophagus risk were reported only in two studies, providing a pooled OR of 0.93 (95% CI, 0.33–2.63; Fig. 2; Supplementary Fig. S4).

Meta-analysis of one nested case–control study and two cohort studies showed a slightly increased risk of esophageal adenocarcinoma associated with elevated glucose levels (OR, 1.12; 95% CI, 1.03–1.22; Fig. 2; Supplementary Fig. S5).

The pooled estimates of five cross-sectional studies showed that higher insulin levels were associated with an increased risk of Barrett esophagus (OR, 1.47; 95% CI, 1.08–2.00) and EAC/BE (OR, 1.42; 95% CI, 1.05–1.93, comparing the highest vs. lowest tertiles; Fig. 2; Supplementary Fig. S6).

The meta-analysis of two cross-sectional studies and one cohort study showed a possibly decreased risk of EAC/BE associated with higher levels of IGF-1 (OR, 0.60; 95% CI, 0.31–1.16, comparing the highest vs. lowest tertiles), while no such association was found for IGFBP-3 (Fig. 2; Supplementary Fig. S6).

Pooling of three cohort studies showed no association between triglycerides levels and esophageal adenocarcinoma risk (OR, 0.95; 95% CI, 0.88–1.03, comparing the highest vs. lowest categories; Fig. 2; Supplementary Fig. S7).

Meta-analysis of two studies showed an increased risk of esophageal adenocarcinoma associated with higher CRP levels (OR, 2.06; 95% CI, 1.28–3.30, comparing the highest vs. lowest tertiles), while no association between CRP levels and Barrett esophagus risk was found in the only identified study. An increased OR remained for the combined outcome, EAC/BE (OR, 1.43; 95% CI, 1.02–2.01; Fig. 2; Supplementary Fig. S8).

Two studies reported an increased risk of esophageal adenocarcinoma associated with higher levels of IL6 (pooled OR, 1.50; 95% CI, 1.03–2.19, comparing the highest vs. lowest tertiles), while no associations were found for IL8 or TNFα (Fig. 2; Supplementary Figs. S8 and S9). Meta-analysis combining three studies showed an increased risk of EAC/BE associated with higher IL6 levels (OR, 1.58; 95% CI, 1.12–2.22; Fig. 2; Supplementary Fig. S8).

Meta-analysis of one nested case–control study and one cohort study showed that higher prediagnostic sTNFR-2 levels were associated with an increased risk of esophageal adenocarcinoma (OR, 3.16; 95% CI, 1.76–5.65; Fig. 2; Supplementary Fig. S9).

Seven of the included studies also measured serum levels of some biomarkers except for the 13 biomarkers presented above. No meta-analysis was performed for these biomarkers because data were available in only one study for each of these. No associations were found for total cholesterol or high-density lipoprotein cholesterol (35, 38). A cross-sectional study comparing 141 patients with Barrett esophagus with 139 patients undergoing colonoscopy screening reported an increased risk of Barrett esophagus associated with higher levels of IL-12p70 and lower levels of IL10 and IL1β, while no associations were found for IFNγ (27). A cohort study of 397 patients with Barrett esophagus found no associations between plasma levels of sTNFR-1 or F2-isoprostanes and risk of esophageal adenocarcinoma (37). A recent case–control study nested in seven cohorts quantitated 69 circulating inflammation markers (using Luminex-based multiplex assays) in 296 patients with esophageal adenocarcinoma and an equal number of control participants. This study suggested an increased risk of esophageal adenocarcinoma associated with higher levels of soluble IL6 receptor, soluble VEGFR-3, lipocalin-2, resistin, and serum amyloid A, and with lower levels of IL3 and IL17A (10).

Measurements of heterogeneity across studies and publication bias for each biomarker are presented in Table 3. The heterogeneity tests suggested moderate to high heterogeneity across studies on adiponectin (I² = 45%; P = 0.070) and leptin (I² = 70%; P = 0.002). No evident publication bias was detected by the funnel plots (Supplementary Fig. S10) or the Begg and Egger tests for these studies.

This study indicates an increased risk of esophageal adenocarcinoma or Barrett esophagus associated with higher circulating levels of some inflammatory and metabolic biomarkers, that is, leptin, glucose, insulin, CRP, IL6, and sTNFR-2. No associations were found for adiponectin, ghrelin, IGF-1, IGFBP-3, triglycerides, IL8, or TNFα.

Among strengths of the study, first is the extensive search strategy adopted to identify all relevant publications covering a wide range of inflammatory and metabolic biomarkers. Second, the harmonization of the reported associations on different scales of comparison into a common form enabled comparison of magnitudes of the associations for most of the studied biomarkers, and also more accurate assessment of heterogeneity and publication bias. There are also limitations. First, most of the included studies were cross-sectional in design, where biomarker levels were measured after the disease onset. Thus, reverse causality could not be ruled out. However, this should not be an issue for the findings for glucose, triglycerides, and sTNFR-2 because they were based on prospective studies. Second, no more than five studies were identified for most biomarkers, except for adiponectin and leptin. Third, substantial heterogeneity across studies was observed, probably due to the combination of different outcomes and types of comparators. We conducted meta-analyses for the three different outcomes, that is, esophageal adenocarcinoma, Barrett esophagus, and the combined outcome EAC/BE, whenever possible. We also stratified the analyses for adiponectin and leptin by type of comparators and sex, but this was not possible for other biomarkers because of the limited number of available studies. Finally, the measurement of biomarker levels was based on a single sample only, even in the prospective studies, making it impossible to assess any influence of longitudinal changes of biomarker levels on their associations with esophageal adenocarcinoma or Barrett esophagus risk.

To the best of our knowledge, this study is the first to systematically summarize evidence for the other examined biomarkers in relation to the risk of esophageal adenocarcinoma or Barrett esophagus, except for the biomarkers leptin, insulin, and adiponectin. The findings of an increased Barrett esophagus risk associated with higher levels of leptin and insulin and no association with adiponectin, are consistent with a previous meta-analysis (9). However, two more recent publications were added in this analysis, and the earlier meta-analysis did not assess the associations of these three biomarkers with esophageal adenocarcinoma risk. Specifically, for the association between leptin levels and Barrett esophagus risk, the newly added study in this updated meta-analysis found a decreased Barrett esophagus risk associated with higher leptin levels (17), which differed from the earlier studies. Such inconsistency might be due to heterogeneity across studies in characteristics of study population (e.g., sex composition), type of comparators, quantitative assay for leptin levels, ascertainment of patients with Barrett esophagus, and adjustment for confounders, or chance. However, the limited number of existing studies precluded exploring potential sources of heterogeneity, for example, using a meta-regression approach. More large-scale studies are needed to examine whether the association between leptin levels and Barrett esophagus risk varies across populations and strata of other factors, including sex, reflux, and obesity.

Leptin is an adipokine secreted by adipose tissue and is involved in the regulation of energy balance, suppressing food intake, and thereby inducing weight loss (39). The circulating levels of leptin positively correlate with the amount of body fat and are also influenced by sex hormones and some other inflammatory cytokines (40). Higher leptin levels have been linked with increased risk of several cancers, including breast, endometrial, colorectal, and prostate cancers. The carcinogenic mechanisms associated with elevated leptin levels may include enhanced cell proliferation, changes in the regulation of certain cell signaling pathways, and promotion of inflammation and angiogenesis (41–43). Our meta-analysis of three prospective studies found that elevated glucose levels were associated with a modestly (12%) increased risk of esophageal adenocarcinoma, which was in line with a recent pooled study of 13 population-based studies showing a 30% increased risk of esophageal adenocarcinoma or esophagogastric junctional adenocarcinoma in patients with diabetes (44). A stronger association between insulin levels and Barrett esophagus risk was indicated, but the evidence was mainly based on cross-sectional studies and the association was attenuated after adjustment for leptin (26).

The associations between higher levels of CRP, IL6, and sTNFR-2 and risk of esophageal adenocarcinoma or Barrett esophagus support a role of systemic inflammation in the development of esophageal adenocarcinoma, which may be a mechanism underlying the associations of obesity and tobacco smoking with esophageal adenocarcinoma. A mediation analysis indicated that sTNFR-2 accounted for 33% of the association between central obesity (measured by waist circumference) and esophageal adenocarcinoma risk (10). Taken together, the available evidence suggests that systemic inflammation and metabolic disorders may be pathways in the etiology of esophageal adenocarcinoma.

The striking male predominance in esophageal adenocarcinoma seems not to be explained by the two major risk factors, that is, GERD and general obesity, given the similar exposure prevalence and strengths of associations of these factors with esophageal adenocarcinoma risk between the sexes (1, 5). Abdominal obesity, however, which is more common in men than in women, may contribute to the sex difference in esophageal adenocarcinoma. Our findings in this study did not support stronger associations with adiponectin or leptin levels in men than in women, but large prospective investigations including more participants of both sexes are needed to clarify this question. Sex hormonal factors may play a role in the etiology of esophageal adenocarcinoma and Barrett esophagus (45, 46), but the existing evidence is limited (1, 5). If the role of sex hormones in the etiology of esophageal adenocarcinoma is confirmed, an underlying mechanism may be the influence of sex hormones on inflammation, for example, the anti- or proinflammatory effects of certain sex hormones depending on the biological microenvironment, which may subsequently lead to altered cancer risk.

Despite decades of efforts to develop the treatment, the overall prognosis in esophageal adenocarcinoma remains poor, mainly because most patients are diagnosed at an advanced tumor stage (1, 4). Earlier tumor detection, particularly among individuals at high absolute risk, has the potential to reduce the mortality from this cancer. A few risk stratification models have been developed for esophageal adenocarcinoma and Barrett esophagus, showing promising performance (47–51). These models mainly combine clinical and lifestyle risk factors, which are easily captured through questionnaires or medical records, while genetic biomarkers have thus far not improved the identification of high-risk individuals (50). Whether inclusion of a panel of inflammatory and metabolic biomarkers increases the accuracy of risk stratification models needs to be evaluated. However, because most of the identified biomarkers in this study showed modest associations with esophageal adenocarcinoma or Barrett esophagus risk and are also associated with the major risk factors for esophageal adenocarcinoma and Barrett esophagus, the addition of these biomarkers may improve the model performance, but only to a limited extent. Interestingly, this study found a more than 3-fold increased risk of esophageal adenocarcinoma or Barrett esophagus associated with higher prediagnostic sTNFR-2 levels, which particularly warrants further investigations. Nevertheless, any use of circulating biomarkers to identify high-risk individuals who may benefit from screening or surveillance should be scientifically justified before they may be tested in routine clinical practice.

In summary, this systematic review and meta-analysis suggests an increased risk of esophageal adenocarcinoma or Barrett esophagus associated with higher circulating levels of some inflammatory and metabolic biomarkers, that is, leptin, glucose, insulin, CRP, IL6, and sTNFR-2. The available studies were too few to examine sex-specific associations for these biomarkers. More prospective studies are required to identify biomarkers that can help select individuals at high absolute risk of esophageal adenocarcinoma for targeted prevention and early detection.

No potential conflicts of interest were disclosed.

The funding bodies had no role in the study design; the collection, analysis, and interpretation of data; or the writing of the article and the decision to submit it for publication.

This study was supported by the Swedish Research Council (grant no. 340-2013-5478, to J. Lagergren) and the Swedish Cancer Society (grant no. 190043, to S.-H. Xie; grant no. 180684, to J. Lagergren).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.