The association between coffee consumption and esophageal cancer risk: a systematic review and meta-analysis of current observational evidence

Introduction

According to GLOBOCAN in 2022, esophageal cancer ranks as the 11th most common malignancy globally in terms of incidence and the 7th leading cause of cancer-related mortality (1). This disease represents a major public health challenge due to its high morbidity and mortality. Esophageal cancer is primarily classified into two main histological subtypes: esophageal squamous cell carcinoma (ESCC), which is predominant in East Asia, and esophageal adenocarcinoma (EAC), which is more common in Western countries (2). The etiology of esophageal cancer is multifactorial, involving a complex interplay of genetic susceptibility, environmental exposures, and lifestyle factors. Among these, dietary habits have gained increasing attention as potential modifiable risk factors, with growing interest in the role of coffee consumption (3).

As one of the most widely consumed beverages globally, coffee and its abundant bioactive compounds (such as caffeine, polyphenols, and diterpenes) have been extensively investigated for their potential impacts on human health. Epidemiological and experimental studies have linked coffee consumption to reduced risks of several chronic conditions, including type 2 diabetes (4) and certain liver diseases (5).

Regarding esophageal cancer, the relationship with coffee consumption remains inconclusive and continues to be actively debated. An earlier umbrella review of meta-analyses suggested that coffee could be associated with a reduced risk of several cancers, though the evidence for esophageal cancer was less clear (6). This was reflected in specific meta-analyses, such as the one by Zheng et al. [2013], which found no significant association between coffee intake and esophageal cancer risk in observational studies (7). However, the landscape has grown more complex with the advent of novel analytical methods. Recent Mendelian randomization (MR) studies, which aim to minimize confounding, have produced conflicting conclusions. A recent MR study (Xue et al., 2025) found no significant causal relationship between coffee consumption and esophageal cancer risk after accounting for hot beverage intake (8). In contrast, a 2025 MR study by Gao et al. reported a significant inverse association between genetically predicted coffee intake and esophageal cancer risk, underscoring the ongoing debate (9).

Despite the existing meta-analyses, several recent systematic reviews have provided nuanced insights into specific exposures and populations. For example, a 2022 meta-analysis on hot tea drinking (10) and a 2023 review on African populations (11) both highlighted the role of high-temperature beverages in esophageal cancer risk. More recently, Ndebia and Kamsu [2024] conducted a meta-analysis focusing on dietary and culinary practices in the East African corridor, reporting a significant association between hot beverage consumption (predominantly tea) and esophageal cancer risk, but found no significant association for coffee consumption specifically [odds ratio (OR) =1.15, 95% confidence interval (CI): 0.88–1.52] (12). However, these studies were either restricted to specific beverage types (tea), specific geographic regions (Africa), or specific dietary patterns, and none provided a dedicated, updated, and global quantification of the association between coffee consumption specifically and esophageal cancer risk.

Given coffee’s widespread consumption and complex bioactive profile—distinct from tea and other beverages—an updated meta-analysis that systematically incorporates recent evidence, addresses the potential modifying effect of beverage temperature, and explores heterogeneity across populations is urgently warranted. Our study specifically addresses these key research gaps by: (I) providing an updated global synthesis of the association between coffee consumption and esophageal cancer risk; (II) evaluating the potential modifying effect of beverage temperature; (III) conducting comprehensive subgroup analyses by geographic region and histological subtype; and (IV) performing dose-response and meta-regression analyses to explore heterogeneity. We present this article in accordance with the PRISMA reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0327/rc) and it was conducted in accordance with PRISMA 2020 guidelines (13).

Methods

Search strategy

A comprehensive literature search was performed using PubMed, Embase, Web of Science, and the Cochrane Library from database inception through January 2026. The complete, verbatim search strategies for all databases, including Boolean operators, MeSH terms, field codes, and date limits are provided in Appendix 1. This systematic review was prospectively registered in PROSPERO (CRD42024562527) on 27 May 2024. Duplicate records were identified using NoteExpress software and removed after manual verification by two reviewers.

Duplicate management: all retrieved records were exported into NoteExpress reference management software. Duplicate records were first identified automatically using the software’s duplicate detection function (based on title, author, year, and journal). Subsequently, two reviewers (initials) manually reviewed the remaining records to identify and remove any additional duplicates that the automated process may have missed. Disagreements regarding duplicate status were resolved through discussion or by consulting a third reviewer (initials).

After removing duplicates, two investigators independently screened titles and abstracts for relevance, followed by full-text review. Disagreements were resolved by consensus or consultation with a senior investigator.

Study selection

Study selection was conducted in two phases, with two reviewers independently screening titles and abstracts against predefined inclusion and exclusion criteria in the first phase, followed by full-text assessment of potentially eligible studies in the second phase.

Inclusion criteria: (I) observational studies (cohort or case-control) examining the association between coffee consumption and esophageal cancer risk; (II) studies providing quantitative data on coffee consumption and sufficient raw data to calculate effect sizes [OR/hazard ratio (HR)/risk ratio (RR)] with 95% CIs; (III) human studies published in English.

Exclusion criteria: (I) non-observational studies (e.g., experimental studies, in vitro/in vivo animal studies); (II) reviews, comments, case reports, conference abstracts, and duplicate publications; (III) studies without reporting quantitative data on coffee consumption; (IV) studies without a control group or with insufficient data to calculate effect sizes; (V) non-English publications.

Data extraction

Two reviewers independently extracted data using a standardized form. For the primary analysis, we extracted the maximally adjusted OR or HR reported in each original study to minimize confounding bias from known risk factors such as tobacco smoking and alcohol consumption. The following information was collected: first author, publication year, study design, country, population characteristics, study period, cancer incidence, coffee consumption levels, adjustment factors, and the risk estimates (comparing the highest versus lowest coffee consumption level) with their 95% CIs (Table 1). Where necessary, corresponding authors were contacted to obtain missing data.

Quality assessment and Grading of Recommendations Assessment, Development and Evaluation (GRADE) certainty

Study quality was evaluated independently by two reviewers using the Newcastle-Ottawa Scale (NOS), adapted for the specific research question of coffee consumption and esophageal cancer risk. For cohort studies, quality was assessed based on selection (4 items), comparability (2 items), and outcome (3 items) domains, with a maximum score of 9 points (Table 2). For case-control studies, quality was assessed based on selection (4 items), comparability (2 items), and exposure (3 items) domains, also with a maximum score of 9 points (Table 3). Discrepancies were resolved through discussion. The detailed NOS scores for each included study are presented in Table 4.

A formal GRADE certainty-of-evidence assessment was performed for all outcomes (Table 5). The evidence for each outcome was rated as high, moderate, low, or very low based on five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. The GRADE assessment was conducted separately for the overall analysis, histological subtypes (ESCC and EAC), geographic regions (Asian vs. European/American), and the coffee temperature subgroup. The results of the GRADE assessment are summarized in the “Results” section and incorporated into the “Conclusions” section.

Statistical analyses

All analyses were performed using Stata software (version 17.0). RevMan 5.4 was used for supplementary analyses. A two-tailed P value <0.05 was considered statistically significant.

Effect measures and handling of HRs and ORs

We pooled HRs from cohort studies and ORs from case-control studies separately. Given the conceptual difference between these measures, HRs and ORs are conceptually distinct: the OR approximates the RR only under the rare disease assumption (incidence <10%). As esophageal cancer incidence varies across populations, including some high-risk regions (e.g., Ethiopia, Taiwan), this assumption may not hold for all included studies. Therefore, we performed sensitivity analyses to address this concern: (I) a separate meta-analysis restricted to cohort studies (pooling HRs only); (II) a separate meta-analysis restricted to case-control studies (pooling ORs only); and (III) a comparison of the direction, magnitude, and statistical significance of effects between the two designs. These sensitivity analyses confirmed that the overall pattern of results did not materially differ when cohort and case-control studies were analyzed separately. Accordingly, we present the primary analysis with combined estimates for completeness, with the understanding that pooled HRs and ORs should be interpreted as relative measures of association rather than absolute risk estimates.

For the primary analysis, we extracted the maximally adjusted OR or HR reported in each original study to minimize confounding bias from known risk factors such as tobacco smoking and alcohol consumption.

Handling of the Hashibe study

In one study (Hashibe et al., 2015), only an RR was reported. Due to the extremely low baseline event rate (P₀≈0.07%) in the control group, the conversion from RR to HR resulted in negligible numerical differences according to VanderWeele [2020] (14). However, we acknowledge that the proportional hazards assumption could not be formally verified from the published data. Therefore, this study was excluded from the primary pooled estimate. A sensitivity analysis including this study (with RR-to-HR conversion) is presented separately (see “Results”—“Sensitivity analyses” section). The results including the converted RR (reported as HR) are presented as a secondary sensitivity analysis.

Dose-response analysis

A one-stage dose-response meta-analysis was performed using the generalized least squares method for trend estimation (Greenland & Longnecker, 1992). Studies were included if they reported three or more exposure categories with a true zero reference (e.g., “0 cups/day”). Dose levels were assigned using the midpoint for reported ranges and lower bound ×1.2 for open-ended upper categories, after standardizing all units to cups/day (mL/day ÷ 150; times/week ÷ 7). A restricted cubic spline model with three knots (placed at the 10th, 50th, and 90th percentiles of the dose distribution) was used to test for non-linearity.

Assessment of heterogeneity

Heterogeneity was assessed using the I² statistic and the Chi-squared test. A fixed-effects model was applied if I²<50% and P>0.10; otherwise, a random-effects model was used.

Meta-regression analysis was performed to explore potential sources of heterogeneity. Univariate meta-regression using the restricted maximum likelihood (REML) method was conducted with geographic region as a pre-specified covariate (coded as 0= East Asian, 1= non-East Asian). Meta-regression for study quality (NOS score) was not performed due to a highly unbalanced distribution (only one study scored 6; the remaining 10 scored 7–9); instead, a sensitivity analysis excluding the NOS-6 study was conducted. Population body mass index (BMI) could not be analyzed due to missing data, and adjustment for smoking/alcohol could not be analyzed because all included studies adjusted for both factors. The adjusted R² statistic was used to quantify the proportion of between-study variance explained by geographic region.

Publication bias

Funnel plots were generated for visual inspection. In addition, both Begg’s rank correlation test and Egger’s linear regression test were performed to statistically assess publication bias.

Sensitivity analyses

We conducted the following sensitivity analyses to test the robustness of our findings: (I) leave-one-out analyses to identify influential studies; (II) comparison of fixed-effects vs. random-effects models; (III) analyses excluding studies with extreme or imprecise estimates (very wide CIs); and (IV) analyses restricted to prospective designs (cohort studies). For the temperature subgroup, which showed substantial heterogeneity, additional sensitivity analyses were performed excluding the outlier study (Deybasso, 2021).

Subgroup analyses

The following subgroup analyses were pre-specified a priori: histological type (ESCC vs. EAC) and geographic region (Asian vs. European/American). The subgroup analysis by coffee temperature (based on subjective perception: highest = burning/very hot, lowest = warm/cold) was conducted as an exploratory, post hoc analysis, as temperature data were not consistently reported across all included studies.

Results

Literature search

The systematic search across four databases initially identified 52 articles. Following the removal of duplicates, 50 records remained. After title and abstract screening, 13 records were excluded as they did not report information on the relation between coffee drinking and risk of esophageal cancer incidence, leaving 37 articles for which full texts were sought. Of these, 5 full-text articles were excluded (2 review articles, 2 with data deficiencies, and 1 non-human study). Ultimately, 22 studies met the eligibility criteria and were included in the meta-analysis. The study selection process is detailed in the PRISMA flow diagram (Figure 1).

Figure 1 PRISMA flow diagram of the study selection process.

Study characteristic

The 22 studies that were finally selected originated from various countries, adding a geographical dimension to the meta-analysis (15-36). The countries represented in the studies included Japan, China, UK, Europe, America, Ethiopia, Malawi and Tanzania and so forth. This diversity in geographical distribution ensured a broad representation of various ethnicities and demographics.

The key characteristics of the included studies are summarized in Table 1. For each study, we extracted the following data: first author, publication year, country, study design, study period, number of subjects (cases/controls or cohort size), categories of coffee consumption (lowest and highest), effect estimates (OR or HR with 95% CI), and covariates adjusted for in the multivariate analysis. Studies were categorized by design (case-control or cohort). Furthermore, cases were classified by pathological type—ESCC or EAC—where such data were available.

Quantitative assessment (meta-analysis)

Association between highest versus lowest coffee consumption and esophageal cancer risk

Stratified by study design, both cohort and case-control studies indicated no statistically significant association between coffee consumption and esophageal cancer risk. Cohort studies yielded a pooled HR of 0.87 (95% CI: 0.74–1.03; I²=24.2%, P=0.24; Figure 2), while case-control studies showed a pooled OR of 0.94 (95% CI: 0.81–1.08; I²=66.8%, P=0.001; Figure 3). Funnel plots for the main analysis were visually inspected and appeared largely symmetrical (Figure S1). Begg’s and Egger’s tests showed no evidence of significant publication bias (cohort studies: Begg’s P=0.45, Egger’s P=0.85; case-control studies: Begg’s P=0.10, Egger’s P=0.08; Table S2).

Figure 2 Forest plot for the association between coffee consumption and esophageal cancer risk in cohort studies (highest vs. lowest). CI, confidence interval; HR, hazard ratio.

Figure 3 Forest plot for the association between coffee consumption and esophageal cancer risk in case-control studies (highest vs. lowest). CI, confidence interval; OR, odds ratio.

Stratified analysis by pathological type: coffee consumption and the risk of ESCC versus adenocarcinoma

In cohort studies, comparing the highest with the lowest coffee consumption yielded pooled HRs of 0.98 (95% CI: 0.74–1.29) for EAC and 0.95 (95% CI: 0.63–1.42) for ESCC (Figure 4), indicating no association. In case-control studies, these analyses are exploratory and critically underpowered (only 3 studies for each histological subtype). The pooled estimates are highly unstable and should not be interpreted as confirming or excluding an association. Therefore, no definitive conclusions regarding the association between coffee consumption and ESCC or EAC risk can be drawn from these case-control subgroup analyses. The pooled OR for EAC was 1.17 (95% CI: 0.65–2.09; 3 studies) and for ESCC was 1.09 (95% CI: 0.54–2.18; 3 studies; Figure 5). The substantial heterogeneity in the ESCC subgroup (I²=77.3%) likely reflects the very limited number of studies.

Figure 4 Forest plot for histological subtype analysis (EAC vs. ESCC) in cohort studies. CI, confidence interval; EAC, esophageal adenocarcinoma; ESCC, esophageal squamous cell carcinoma; HR, hazard ratio.

Figure 5 Forest plot for histological subtype analysis (EAC vs. ESCC) in case-control studies. CI, confidence interval; EAC, esophageal adenocarcinoma; ESCC, esophageal squamous cell carcinoma; OR, odds ratio.

Subgroup analysis by coffee temperature

In an exploratory (post hoc) analysis restricted to case-control studies (n=7), the pooled OR for very hot versus warm/cold coffee was 1.49 (95% CI: 0.97–2.29), with substantial heterogeneity (I²=87.3%, P<0.001; Figure 6). A leave-one-out sensitivity analysis identified the Deybasso [2021] study as a significant outlier. Excluding this study reduced the pooled OR to 1.04 (95% CI: 0.79–1.37) and heterogeneity to moderate levels (I²≈43.5%; Figure S5). Given the considerable heterogeneity (I²=87.3%), the pooled estimate for the temperature subgroup should be interpreted with caution; these results are presented as exploratory and hypothesis-generating.

Figure 6 Forest plot for high-temperature vs. low-temperature coffee consumption and esophageal cancer risk in case-control studies. CI, confidence interval; OR, odds ratio.

Geographic variation in the coffee-cancer association: a subgroup of Asian versus European and American countries

Among cohort studies, the pooled HR was 0.74 (95% CI: 0.54–1.02) for Asian countries and 0.94 (95% CI: 0.78–1.13) for European/American countries (Figure S7). Among case-control studies, the pooled OR was 0.79 (95% CI: 0.62–1.01) for Asian countries and 0.96 (95% CI: 0.79–1.16) for Western countries (Figure S8). No statistically significant association was observed in any region. Meta-regression showed that geographic region was not a significant source of heterogeneity (coefficient =0.315, 95% CI: −0.495 to 1.125, P=0.39). The adjusted R-squared was −12.32%, indicating that geographic region explained none of the between-study variance. Residual heterogeneity remained substantial (I²_res=67.94%).

Dose-response meta-analysis

A total of 4 case-control studies met the inclusion criteria for dose-response meta-analysis. Cohort studies could not be included as none provided a true zero reference category. As shown in Figure 7, there was no evidence of a linear or non-linear association between coffee consumption and esophageal cancer risk across the range of 0 to 6 cups/day. The pooled OR per 1 cup/day increment was 0.99 (95% CI: 0.96–1.03; I²=24%), with no evidence of non-linearity (P=0.58).

Figure 7 Dose-response relationship between coffee consumption (cups/day) and esophageal cancer risk based on 4 case-control studies.

Sensitivity analyses

A series of sensitivity analyses confirmed the robustness of our findings.

Leave-one-out analysis showed no individual study unduly influenced the pooled estimates. For cohort studies, the pooled HR remained stable between 0.82 and 0.88 after excluding each individual study; for case-control studies, the pooled OR remained stable between 0.88 and 1.09 (Figures S2,S3).

Fixed-effects vs. random-effects models yielded nearly identical conclusions. For cohort studies, the fixed-effects model (HR =0.87, 95% CI: 0.74–1.03) and random-effects model (HR =0.88, 95% CI: 0.74–1.04) gave nearly identical results. For case-control studies, the random-effects model (OR =1.02, 95% CI: 0.77–1.35) and fixed-effects model (OR =0.94, 95% CI: 0.81–1.09) both indicated no significant association (Figures S4,S5).

Excluding the Hashibe study (which reported only RR) from the primary analysis did not materially change the pooled HR. The primary analysis (excluding Hashibe, n=7) yielded a pooled HR of 0.87 (95% CI: 0.74–1.03), while a sensitivity analysis including the Hashibe study (n=8) yielded a pooled HR of 0.82 (95% CI: 0.69–0.96). The leave-one-out analysis (Figure S2) confirmed that no individual study, including Hashibe, unduly influenced the pooled estimate.

Excluding two imprecise case-control studies (Deybasso 2021, Garidou 1996) changed the pooled OR from 1.02 to 0.88, but the conclusion of no association remained unchanged (Figure S6). The I² decreased from 66.7% to 40.7%. Restricting the analysis to prospective designs (cohort studies only) did not change the conclusion, as already reported in the cohort analysis (HR =0.87, 95% CI: 0.74–1.03).

GRADE certainty of evidence

Based on a formal GRADE assessment, the overall certainty of evidence for coffee consumption and esophageal cancer risk was rated as low, downgraded due to serious inconsistency (substantial heterogeneity in case-control studies) and serious imprecision (CIs crossing the null). The certainty of evidence for the temperature subgroup was rated as very low, downgraded due to very serious inconsistency (I²=87.3%) and serious imprecision.

Discussion

Summary of main findings

This meta-analysis of 22 studies involving over 1.7 million participants systematically evaluated the association between coffee consumption and esophageal cancer risk. Our findings demonstrate no statistically significant association between coffee intake and esophageal cancer risk, with consistent null results across both cohort (HR =0.87, 95% CI: 0.74–1.03) and case-control studies (OR =0.94, 95% CI: 0.81–1.08). The consistency of these findings across study designs, coupled with low heterogeneity in cohort studies and no evidence of publication bias, reinforces the robustness of these null results. However, a formal GRADE assessment rated the overall certainty of this evidence as low, primarily due to heterogeneity in case-control designs and imprecision.

Comparison with previous meta-analyses

Our findings align with earlier meta-analyses that reported no significant association between coffee consumption and esophageal cancer risk (7). However, our study extends previous work by incorporating more recent evidence, conducting dose-response and meta-regression analyses, and formally grading the evidence certainty. Notably, while previous meta-analyses were limited by smaller sample sizes and less comprehensive subgroup assessments, our study benefits from a larger pooled sample and more detailed stratification by pathological type, geographic region, and beverage temperature. These extended analyses allow for a more thorough examination of potential effect modifiers that may have been overlooked in prior research.

Recent systematic reviews have continued to explore this relationship from complementary angles. A 2022 meta-analysis focusing on hot tea drinking reported a significantly increased risk of esophageal cancer associated with high-temperature beverage consumption, highlighting thermal injury as a potential carcinogenic mechanism (10). Although this study focused exclusively on tea, its findings support the biological plausibility of our exploratory temperature analysis. Separately, a 2023 systematic review and meta-analysis investigating environmental and lifestyle risk factors for ESCC in African populations found that consumption of hot food and beverages, including coffee, was associated with a significantly increased risk (pooled OR =1.68) (11). Our updated synthesis confirms the persistent null finding for coffee consumption volume while highlighting the low certainty of the evidence base and suggesting that temperature may warrant further investigation.

Pathological type does not modify the association

A key strength of our analysis was the stratification by esophageal cancer histological subtype. Despite the distinct etiologies of ESCC and EAC—with ESCC being more prevalent in Asia and linked to smoking and alcohol, while EAC predominates in Western populations and is associated with gastroesophageal reflux and obesity—our results showed no significant association between coffee consumption and either subtype. The pooled HRs for ESCC (0.95, 95% CI: 0.63–1.42) and EAC (0.98, 95% CI: 0.74–1.29) in cohort studies, along with similar null findings in case-control studies, suggest that the lack of association is consistent across both histological types. This finding is important because it suggests that the biological pathways specific to each subtype are not differentially influenced by coffee consumption. However, it should be noted that the exploratory case-control subgroup analyses for histological subtypes were underpowered (only 3 studies each) and should be interpreted with caution.

Beverage temperature: an inconclusive and hypothesis-generating finding

The relationship between coffee temperature and esophageal cancer risk deserves particular attention. Previous research has established that thermal injury from repeatedly consuming very hot beverages may damage the esophageal mucosa, potentially initiating carcinogenesis (37). Our exploratory post hoc analysis comparing high versus low temperature coffee consumption yielded a pooled OR of 1.49 (95% CI: 0.97–2.29), based exclusively on case-control studies. While the magnitude of the point estimate—substantially larger than that observed for consumption volume alone—suggests that temperature may be an important consideration, this finding was not statistically significant and was heavily influenced by a single outlier study (Deybasso, 2021). Excluding this study attenuated the estimate to near-null (OR =1.04), and the substantial heterogeneity (I²=87.3%) further precludes a reliable pooled estimate.

Several additional factors warrant caution. First, the exclusive reliance on case-control data raises the possibility of recall bias, where cancer patients may over-report “very hot” consumption compared to healthy controls due to heightened awareness of potential risk factors following their diagnosis, potentially inflating risk estimates. Second, the subjective nature of temperature perception (e.g., “very hot” vs. “warm”) may vary across individuals and cultures, introducing additional misclassification. Third, the lack of prospective cohort studies examining beverage temperature represents a critical gap in the literature. Furthermore, a formal GRADE assessment rated the certainty of evidence for the temperature finding as very low, downgraded due to very serious inconsistency and serious imprecision. Therefore, this finding should be considered hypothesis-generating only, and future large-scale cohort studies with standardized, validated measures of beverage temperature are urgently needed.

Geographic variations: clues for future research

Our geographic subgroup analysis revealed an intriguing pattern: although no statistically significant association was observed in any region, point estimates were consistently lower in Asian populations (cohort HR =0.74; case-control OR =0.79) compared with Western populations (cohort HR =0.94; case-control OR =0.96). Several factors may explain this observed trend. First, genetic polymorphisms in coffee-metabolizing enzymes, such as CYP1A2, vary across populations and may influence the biological effects of coffee consumption. Second, preparation methods differ substantially between regions—from espresso in Europe to filtered coffee in America and instant coffee in parts of Asia—potentially affecting the concentration of bioactive compounds. Third, lifestyle factors that correlate with both geographic region and coffee consumption patterns may act as residual confounders. These observations, while not definitive, generate hypotheses for future investigations. Meta-regression analysis confirmed that geographic region was not a significant source of heterogeneity (P=0.39), suggesting that other factors may contribute to the observed between-study variance.

Strengths and limitations

There are several strengths in this meta-analysis. First, the large sample size (over 1.7 million participants) provides adequate statistical power to detect modest associations. Second, the comprehensive subgroup analyses—by study design, pathological type, geographic region, and coffee temperature—allowed for detailed exploration of potential effect modifiers. Third, the consistent null findings across multiple analytical approaches and the absence of publication bias enhance the robustness of our conclusions. Fourth, the high methodological quality of the included studies, with 95.5% (21/22) scoring ≥7 on the NOS, enhances the credibility of our findings. Fifth, the inclusion of dose-response and meta-regression analyses, along with formal GRADE certainty assessment, represents a methodological advancement over previous meta-analyses.

However, several limitations should be acknowledged. First, substantial heterogeneity was observed in case-control studies (I²=66.8%) and the temperature subgroup (I²=87.3%), which may reflect differences in study design, population characteristics, exposure definitions, and adjustment factors. Under such high heterogeneity, the pooled estimates from random-effects models represent a weighted average of divergent study-specific effects rather than a single well-defined underlying parameter, which limits the interpretability of these findings. Second, the temperature subgroup analysis and some histological subgroup analyses (e.g., case-control EAC and ESCC) were exploratory (post hoc) and underpowered (only 3 studies each), and therefore should be interpreted as hypothesis-generating rather than confirmatory. Third, we were unable to perform a dose-response analysis in cohort studies due to the lack of a true zero reference category (most cohort studies used “≤1 cup/day” rather than “0 cups/day” as the reference). Fourth, most included studies did not distinguish between different types of coffee (caffeinated vs. decaffeinated) nor account for variations in preparation methods (e.g., filtered, boiled, espresso, instant). The concentration of bioactive compounds—such as caffeine, chlorogenic acids, and diterpenes (cafestol and kahweol)—varies substantially across coffee types and brewing techniques, and these compounds may exert opposing biological effects on carcinogenesis. The inability to disentangle these potentially differential effects may have contributed to the observed null findings and heterogeneity across studies. Fifth, despite using maximally adjusted estimates from each original study, the possibility of residual confounding cannot be entirely excluded. Sixth, the GRADE certainty of evidence was rated as low for the overall analysis and very low for the temperature finding, indicating that future research is likely to change these estimates. Seventh, only English-language publications were included, which may introduce language bias. Seventh, a sensitivity analysis restricted to high-quality studies (NOS ≥8) could not be meaningfully performed. While all 8 cohort studies had a NOS score of 9 (making such restriction identical to the main analysis), only 2 case-control studies scored ≥8, which is insufficient for a stable pooled estimate. Additionally, a sensitivity analysis comparing pooled adjusted versus unadjusted estimates could not be performed, as only 4 of the 22 included studies (2 cohort and 2 case-control) reported crude effect estimates. The remaining >80% of studies did not provide unadjusted data, precluding a meaningful comparison.

Future research directions

Future research should prioritize: (I) large-scale prospective cohort studies with standardized assessment of coffee consumption, including detailed information on coffee type (caffeinated/decaffeinated), preparation method, serving size, and—critically—beverage temperature using validated instruments; (II) adoption of a true zero consumption reference category (“0 cups/day”) to facilitate dose-response meta-analyses; (III) exploration of potential gene-coffee interactions, particularly regarding CYP1A2 polymorphisms; and (IV) investigation of specific coffee bioactive compounds and their mechanisms of action in esophageal carcinogenesis.

Conclusions

This meta-analysis found no significant association between coffee consumption and esophageal cancer risk, with consistent null findings across study designs, geographic regions, and histological subtypes. A formal GRADE assessment rated the overall certainty of this evidence as low. The exploratory analysis on high-temperature coffee consumption was inconclusive, heavily influenced by an outlier, and rated as vey low certainty. Therefore, while these findings do not support coffee as a major modifiable risk factor, high-quality prospective studies with standardized exposure assessment (including beverage temperature and a true zero consumption reference) are needed to confirm these findings.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the PRISMA reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0327/rc

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0327/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0327/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. He S, Xu J, Liu X, et al. Advances and challenges in the treatment of esophageal cancer. Acta Pharm Sin B 2021;11:3379-92. [Crossref] [PubMed]
3. Islami F, Malekshah AF, Kimiagar M, et al. Patterns of food and nutrient consumption in northern Iran, a high-risk area for esophageal cancer. Nutr Cancer 2009;61:475-83. [Crossref] [PubMed]
4. Huxley R, Lee CM, Barzi F, et al. Coffee, decaffeinated coffee, and tea consumption in relation to incident type 2 diabetes mellitus: a systematic review with meta-analysis. Arch Intern Med 2009;169:2053-63. [Crossref] [PubMed]
5. Kennedy OJ, Roderick P, Buchanan R, et al. Coffee, including caffeinated and decaffeinated coffee, and the risk of hepatocellular carcinoma: a systematic review and dose-response meta-analysis. BMJ Open 2017;7:e013739. [Crossref] [PubMed]
6. Poole R, Kennedy OJ, Roderick P, et al. Coffee consumption and health: umbrella review of meta-analyses of multiple health outcomes. BMJ 2017;359:j5024. [Crossref] [PubMed]
7. Zheng JS, Yang J, Fu YQ, et al. Effects of green tea, black tea, and coffee consumption on the risk of esophageal cancer: a systematic review and meta-analysis of observational studies. Nutr Cancer 2013;65:1-16. [Crossref] [PubMed]
8. Xue F, Xue J, Zhao B, et al. The Associations of Tobacco, Alcohol, and Coffee Consumption with Upper and Lower Gastrointestinal Disease Risk: A Mendelian Randomization Study. Gut Liver 2025;19:715-24. [Crossref] [PubMed]
9. Wang X, Zhang W, Lin T, et al. Causal association between modifiable risk factors and esophageal cancer: A Mendelian randomization study. Discov Oncol 2025;16:1329. [Crossref] [PubMed]
10. Zhong Y, Yang C, Wang N, et al. Hot Tea Drinking and the Risk of Esophageal Cancer: A Systematic Review and Meta-Analysis. Nutr Cancer 2022;74:2384-91. [Crossref] [PubMed]
11. Simba H, Kuivaniemi H, Abnet CC, et al. Environmental and life-style risk factors for esophageal squamous cell carcinoma in Africa: a systematic review and meta-analysis. BMC Public Health 2023;23:1782. [Crossref] [PubMed]
12. Ndebia EJ, Kamsu GT. A Comprehensive Meta-Analysis of Dietary and Culinary Practices on Esophageal Cancer Incidence in the East African Corridor. SVU-International Journal of Medical Sciences 2024;7:207-22.
13. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009;6:e1000097. [Crossref] [PubMed]
14. VanderWeele TJ. Optimal approximate conversions of odds ratios and hazard ratios to risk ratios. Biometrics 2020;76:746-52. [Crossref] [PubMed]
15. Hashibe M, Galeone C, Buys SS, et al. Coffee, tea, caffeine intake, and the risk of cancer in the PLCO cohort. Br J Cancer 2015;113:809-16. [Crossref] [PubMed]
16. Brown LM, Swanson CA, Gridley G, et al. Adenocarcinoma of the esophagus: role of obesity and diet. J Natl Cancer Inst 1995;87:104-9. [Crossref] [PubMed]
17. Castellsagué X, Muñoz N, De Stefani E, et al. Influence of mate drinking, hot beverages and diet on esophageal cancer risk in South America. Int J Cancer 2000;88:658-64. [Crossref] [PubMed]
18. Chen YK, Lee CH, Wu IC, et al. Food intake and the occurrence of squamous cell carcinoma in different sections of the esophagus in Taiwanese men. Nutrition 2009;25:753-61. [Crossref] [PubMed]
19. Deybasso HA, Roba KT, Nega B, et al. Dietary and Environmental Determinants of Oesophageal Cancer in Arsi Zone, Oromia, Central Ethiopia: A Case-Control Study. Cancer Manag Res 2021;13:2071-82. [Crossref] [PubMed]
20. Garidou A, Tzonou A, Lipworth L, et al. Life-style factors and medical conditions in relation to esophageal cancer by histologic type in a low-risk population. Int J Cancer 1996;68:295-9. [Crossref] [PubMed]
21. Inoue M, Tajima K, Hirose K, et al. Tea and coffee consumption and the risk of digestive tract cancers: data from a comparative case-referent study in Japan. Cancer Causes Control 1998;9:209-16. [Crossref] [PubMed]
22. La Vecchia C, Ferraroni M, Negri E, et al. Coffee consumption and digestive tract cancers. Cancer Res 1989;49:1049-51.
23. Lagiou P, Talamini R, Samoli E, et al. Diet and upper-aerodigestive tract cancer in Europe: the ARCAGE study. Int J Cancer 2009;124:2671-6. [Crossref] [PubMed]
24. Lukic M, Nilsson LM, Skeie G, et al. Coffee consumption and risk of rare cancers in Scandinavian countries. Eur J Epidemiol 2018;33:287-302. [Crossref] [PubMed]
25. Masukume G, Mmbaga BT, Dzamalala CP, et al. A very-hot food and beverage thermal exposure index and esophageal cancer risk in Malawi and Tanzania: findings from the ESCCAPE case-control studies. Br J Cancer 2022;127:1106-15. [Crossref] [PubMed]
26. Naganuma T, Kuriyama S, Kakizaki M, et al. Coffee consumption and the risk of oral, pharyngeal, and esophageal cancers in Japan: the Miyagi Cohort Study. Am J Epidemiol 2008;168:1425-32. [Crossref] [PubMed]
27. Nakayama I, Goto A, Yamaji T, et al. Coffee Consumption and Risk of Esophageal Squamous Cell Carcinoma: A Population-Based Prospective Cohort Study in Japan. J Nutr 2025;155:2560-7. [Crossref] [PubMed]
28. Oze I, Matsuo K, Kawakita D, et al. Coffee and green tea consumption is associated with upper aerodigestive tract cancer in Japan. Int J Cancer 2014;135:391-400. [Crossref] [PubMed]
29. Ren JS, Freedman ND, Kamangar F, et al. Tea, coffee, carbonated soft drinks and upper gastrointestinal tract cancer risk in a large United States prospective cohort study. Eur J Cancer 2010;46:1873-81. [Crossref] [PubMed]
30. Sharp L, Chilvers CE, Cheng KK, et al. Risk factors for squamous cell carcinoma of the oesophagus in women: a case-control study. Br J Cancer 2001;85:1667-70. [Crossref] [PubMed]
31. Tavani A, Bertuzzi M, Talamini R, et al. Coffee and tea intake and risk of oral, pharyngeal and esophageal cancer. Oral Oncol 2003;39:695-700. [Crossref] [PubMed]
32. Terry P, Lagergren J, Wolk A, et al. Reflux-inducing dietary factors and risk of adenocarcinoma of the esophagus and gastric cardia. Nutr Cancer 2000;38:186-91. [Crossref] [PubMed]
33. Tran KT, Coleman HG, McMenamin ÚC, et al. Coffee consumption by type and risk of digestive cancer: a large prospective cohort study. Br J Cancer 2019;120:1059-66. [Crossref] [PubMed]
34. Tverdal A, Hjellvik V, Selmer R. Coffee intake and oral-oesophageal cancer: follow-up of 389,624 Norwegian men and women 40-45 years. Br J Cancer 2011;105:157-61. [Crossref] [PubMed]
35. Yuan L, Shen P, Zheng S, et al. Analysis of living habit risk factors for esophageal cancer in central China: A bi-center case-control study. Front Oncol 2023;13:1077598. [Crossref] [PubMed]
36. Zamora-Ros R, Luján-Barroso L, Bueno-de-Mesquita HB, et al. Tea and coffee consumption and risk of esophageal cancer: the European prospective investigation into cancer and nutrition study. Int J Cancer 2014;135:1470-9. [Crossref] [PubMed]
37. Zhang X, Zheng X, Gao R, et al. Role of diet in the risks of esophageal adenocarcinoma and squamous cell carcinoma: an updated umbrella review. Eur J Nutr 2024;63:1413-24. [Crossref] [PubMed]