Association of Coffee Intake With Risk of Gastroesophageal Reflux Disease and Complications: A Systematic Review and Meta-Analysis

Abstract

INTRODUCTION:

Patients with gastroesophageal reflux disease (GERD) are commonly instructed to reduce coffee intake. However, previous studies evaluating the effects of coffee on GERD yielded conflicting results. We aimed to perform a comprehensive systematic review and meta-analysis to assess the association between coffee use and risk of GERD and its complications.

METHODS:

A protocolized search strategy was developed for PubMed, EMBASE, and Web of Science databases in accordance with preferred reporting items for systematic reviews and meta-analyses and meta-analyses of observational studies in epidemiology guidelines. Measured outcomes for GERD were compared between coffee drinkers and nondrinkers. Dichotomous events between unmatched groups were used to calculate pooled proportions with rates estimated using random effects models and effect size. Heterogeneity was assessed with I² statistics and publication bias by funnel plot asymmetry and Egger regression.

RESULTS:

A total of 40 studies encompassing 122,074 patients were included (85,400 coffee drinkers vs 36,674 nondrinkers). GERD was more common among coffee users than nonusers (34.9% [CI: 28.5–41.8] vs 30.7% [CI: 25.2–36.7]; OR: 1.18 [CI: 1.03–1.36; I² = 89.38]). There was no significant association between coffee intake and Barrett's esophagus (22.1% [CI: 12.8–35.4] users vs 17.6% [CI:5.5–43.8] nonusers; OR:1.13 [CI:0.79–1.61; I² = 55.5]). There was no evidence of publication bias based on funnel plot and Egger regression testing (P > 0.05 for all analyses).

DISCUSSION:

Coffee use was associated with a small, statistically significant increased rate of GERD, but not Barrett's. The magnitude of this effect, however, is of unclear clinical significance. The role of routine avoidance/reduction of coffee intake as universal lifestyle modification for GERD needs further evaluation.

INTRODUCTION

Gastroesophageal reflux (GER) symptoms are among the most encountered gastrointestinal complaints in the outpatient setting, affecting more than 30% of the American population weekly (1) and accounting for over 5.5 million ambulatory visits annually (2). Although proton pump inhibitors are the mainstay of medical therapy, dietary and lifestyle changes are often recommended to reduce or eliminate symptoms of GER disease (GERD). In particular, avoidance of potential symptom triggers such as coffee is frequently used as a first-line strategy.

Coffee use in the United States (US) has been steadily increasing, with 62% of Americans drinking coffee daily as of 2020. The average American coffee drinker has about 3 cups of coffee daily (3). Coffee use has been suggested as a risk factor of GERD, and it is a common recommendation to reduce coffee intake to alleviate GER symptoms. Previous studies have shown that infusion of coffee into the stomach using a nasogastric tube may trigger heartburn in patients with GERD (4) and that coffee intake increases acid exposure time (AET) and reflux events on 3-hour ambulatory pH monitoring (5).

However, clinical studies have not consistently demonstrated a relationship between coffee intake and GERD (6,7). A prospective study using data from the Nurses' Health Study found that coffee use independently increased the risk of GER symptoms with a dose-dependent response (8). Substituting water for coffee was also found to reduce GER symptoms. By contrast, epidemiologic studies from Norway and Japan found no significant association between coffee use and GERD, even in subgroups drinking multiple cups daily (9,10). Two previous meta-analyses also found no association between coffee intake and GERD (11,12). However, the most recent meta-analysis included only studies targeting young and adolescent patients, thereby limiting generalizability, and newer data have emerged since the last comprehensive meta-analysis by Kim et al (11). The search strategy by Kim et al (11) may also not have captured all relevant studies internationally and their relatively small sample size may have limited the subgroup analyses that could be performed. Moreover, the potential effect of coffee use on complications of GERD, such as Barrett's esophagus (BE), was not assessed. Our primary aim was to perform a comprehensive meta-analysis to determine if coffee use is associated with GERD. We also aimed to assess the relationship between coffee use and complications of GERD including BE, and to explore the mechanism by which coffee may cause GERD.

METHODS

We performed a systematic review and meta-analysis, reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The protocol was registered with PROSPERO (registration number CRD42023398177; https://www.crd.york.ac.uk/PROSPERO/view/CRD42023398177).

Selection criteria

Observational studies evaluating the association between coffee and GERD in subjects 17 years or older were eligible for inclusion. Studies evaluating coffee use in combination with other beverages (such as tea or soda) or evaluating decaffeinated coffee alone were excluded to ascertain the individual effects of coffee on GERD and to limit heterogeneity. Inclusion of other beverages or decaffeinated coffee may misclassify those exposed vs unexposed. GERD was defined using typical symptoms of heartburn and/or regurgitation, validated GERD questionnaires, and/or identification of erosive esophagitis (EE) on endoscopy. Studies evaluating extraesophageal reflux or that defined GERD using atypical symptoms were excluded.

Search strategy and study selection

We searched MEDLINE, EMBASE, and Web of Science from inception through June 9, 2025. See Supplementary Methods (see Supplementary Digital Content, http://links.lww.com/CTG/B510) for details of the search terms used for each database. We performed a recursive search using the bibliographies of all obtained articles to identify additional studies eligible for inclusion. Full texts and published abstracts were evaluated for eligibility. Results of the search were uploaded through the Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia), where duplicates were removed. Studies were then selected for inclusion through a 2-step process using a priori criteria. The first step was the abstract screen, and the second was the full-text review. These were performed by 2 independent reviewers (M.M. and D.H.). Discrepancies were resolved through consensus and consultation with a third reviewer (R.F.). Ultimately, 43 studies were selected for inclusion in the meta-analysis (Figure 1).

Figure 1.

PRISMA flow diagram summarizing the search and review strategy. PRISMA, preferred reporting items for systematic reviews and meta-analyses.

Data extraction

Data were independently extracted by 4 reviewers (M.M., D.H., R.F., T.M.). In addition, features related to study design and patient characteristics were further extracted by 2 reviewers (M.M. and D.H.). Study characteristics included the primary author, year of publication, country, sample size, study design, study population, and methods by which the exposure and outcomes were assessed. Patient characteristics included age, sex, and GERD prevalence.

Quality assessment and risk of bias

Quality and risk of bias assessment were independently performed by 2 reviewers (M.M. and D.H.) using The Evidence Project risk of bias tool (13). This tool was used because it can be applied to various study designs. However, this tool does not provide a composite score to stratify studies as low, moderate, or high risk of bias. Therefore, we also used a priori criteria to categorize studies as having low, moderate, or high risk of bias to allow for subsequent sensitivity analyses. These criteria are outlined further in Supplementary Methods (see Supplementary Digital Content, http://links.lww.com/CTG/B510). Discrepancies in bias assessments were resolved through consensus and consultation with a third reviewer (R.F.).

Exposure and outcome measures

The exposure of interest was coffee use. For the primary meta-analyses, coffee use was dichotomized into use vs nonuse. Two subgroup analyses were performed to evaluate the effect of coffee use at specific doses. The first included studies that compared patients who drank at least 1 cup of coffee daily versus those who drank less than 1 cup daily. The second included studies that compared patients who drank at least 1 cup of coffee weekly versus those who drank less than 1 cup weekly.

The primary outcome was GERD, which was defined by any of the following: presence of typical reflux symptoms of heartburn and/or regurgitation, an elevated score on a validated GERD questionnaire, or the presence of EE on endoscopy. The secondary outcome was BE, defined as upward displacement of the z-line by at least 1 cm from the esophagogastric junction with biopsy-proven intestinal metaplasia.

Two subgroup analyses based on different definitions of GERD were performed: (i) reflux symptoms, identified based on the presence of heartburn and/or regurgitation or with an elevated score on a validated GERD questionnaire, and (ii) endoscopic evidence of reflux with EE. Additional subgroup analyses were also conducted based on (i) study geographic region, which included Europe, East Asia, United States, and Middle East, and (ii) study patient population, classified as population-based, hospital-based, student population, and patients referred for esophagogastroduodenoscopy.

Statistical analysis

All statistical analyses were performed using the Comprehensive Meta-Analysis software, version 3 (BioStat, Englewood, NJ). Measured outcomes for GERD were compared among coffee versus noncoffee drinkers. Dichotomous events between unmatched groups were then used to calculate pooled proportions with rates estimated using random effects models and effect size presented as odds ratios (ORs) with corresponding 95% CIs.

We also performed a number of sensitivity analyses: (i) with additional studies that included adolescents (aged 15 years or older); (ii) including additional studies with adolescents and those that defined GERD using ICD codes rather than symptoms or endoscopy; (iii) only analyzing studies that defined GERD based on symptoms at least weekly (rather than lower frequency) in addition to those with EE; (iv) excluding studies felt to disproportionately bias results away from the null; (v) only analyzing studies with low to moderate risk of bias (excluding studies at high-risk); and (vi) only analyzing studies with low risk of bias (excluding moderate to high-risk studies).

Assessment of heterogeneity

Heterogeneity was assessed for the individual meta-analyses using the χ² test and the I² statistic (14). Significant heterogeneity was defined as P < 0.05 using the Cochran Q test or I² > 50%, with values >50% indicating substantial heterogeneity. Further quantification of heterogeneity was categorized based on I² with values of 25%, 50%, and 75% indicating low, moderate, and high amounts of heterogeneity, respectively. If moderate to high heterogeneity is present, subgroup analyses (as noted above) and meta-regression will be performed to determine whether certain study characteristics explain the variance across studies.

Publication bias

A funnel plot was created and visually inspected for asymmetry and quantitatively using Egger regression testing to assess for publication bias (15,16). If evidence of publication bias existed, then the trim and fill method was used to correct for funnel plot asymmetry and provide an adjusted effect (17). The classic fail-safe test was also used to assess the risk of bias across studies.

RESULTS

Study selection

Our search yielded 1,905 studies. After excluding duplicates and title and abstract screening, 119 studies underwent full-text review. Overall, 43 met inclusion criteria and were used for data synthesis (Figure 1).

Study and patient characteristics

Of the 43 studies, 39 were published after 2000, and 22 were published after the last comprehensive meta-analysis. Four of the studies were conducted in the United States0. Otherwise, most of the studies were conducted in Europe (15), East Asia (13), and the Middle East (9). Most of the studies were cross-sectional (33), with the remaining being cohort studies. The median sample size was 1,518 (range 71–43,363), with 50.1% of the total cohort being men. The overall prevalence of GERD was 21.5%. Study characteristics are further outlined in Table 1.

Table 1.

Characteristics of included studies

Study	Country	Design	Sample size	Population	Age (mean, SD)	Male n (%)	GERD prevalence n (%)	Risk of bias
Secchi et al (18)	Italy	Cross-sectional	701	Patients undergoing EGD at University of Milan	range 20–70	424 (60.5%)	269 (38.4%)	High
Chang et al (19)	China	Cross-sectional	2,044	Patients presenting for annual visit at Taichung Veterans General Hospital	49.7 ± 12.1	1,266 (61.9%)	346 (16.9%)	High
Anderson et al (20)	USA	Cross-sectional	91	Patients on HD and PD at Johns Hopkins Bayview Hospital	53.1 ± 13.9 (PD), 63.3 ± 16.7 (HD)	40 (44%)	27 (29.7%)	High
Locke GR et al (21)	USA	Cross-sectional	1,524	Residents of Olmstead county	50.5 ± 14.2	732 (48%)	304 (19.9%)	Low
Conio et al (22)	Italy	Case-control	600	Patients undergoing EGD at 8 GI departments	58.5 ± 14.1 (BE), 56.7 ± 14.7 (EE), 61.1 ± 15.8 (controls)	367 (61.2%)	321 (53.5%)	High
Diaz-Rubio et al (23)	Spain	Cross-sectional	2,500	National	range 40–79	1,185 (47.4%)	791 (31.6%)	Moderate
Nilsson et al (9)	Norway	Cross-sectional	43,363	National	52 (GERD), 48 (controls)	20,369 (47%)	3,153 (7.3%)	Moderate
Wang et al (24)	China	Cross-sectional	2,789	Residents of Xi'an	range 18–70	1,374 (49.3%)	430 (15.4%)	High
Nocon et al (25)	Germany	Cross-sectional	6,954	National	45 ± 16 (men), 46 ± 16 (women)	3,338 (48%)	2,964 (42.6%)	Moderate
Somi et al (26)	Iran	Cross-sectional	561	Students of Azad University in Tabriz	22.48 ± 1.98	202 (36%)	36 (6.4%)	High
Chen et al (27)	Taiwan	Cross-sectional	482	Patients presenting for annual visit at Taipei Veterans General Hospital	56.9 ± 14.0	280 (58.1%)	58 (12%)	Moderate
Du et al (28)	China	Cross-sectional	2,231	Patients presenting for outpatient visits at 10 hospitals in the Zhejiang Province of East China	median 43, range 18–90	1,263 (56.6%)	701 (31.4%)	Low
Zheng et al (29)	Sweden	Prospective cohort	23,634	Swedish twin registry	women: median 58, range 42–104; men: median 57, range 42–99	10,950 (46.3%)	4,083 (17.3%)	Low
Nasseri-Moghaddam et al (30)	Iran	Cross-sectional	2,057	Residents of Tehran	34.8 ± 13.0	925 (45%)	865 (42.1%)	Low
Kiciński et al (31)	Poland	Case-control	71	Patients hospitalized at Medical University in Lódź	55.9 ± 11.2	45 (63.4%)	36 (50.7%)	High
Nwokediuko et al (32)	Nigeria	Cross-sectional	410	Medical students at the University of Nigeria	25.3 ± 3.5	240 (58.5%)	108 (26.3%)	High
Lukic et al (33)	Croatia	Cross-sectional	180	Patients undergoing EGD at Osijek University Hospital Centre	53.04 ± 14.41	109 (60.6%)	90 (50%)	High
Murao et al (34)	Japan	Cross-sectional	2,853	Patients presenting for annual visit at Japanese Red Cross Society Kumamoto Health Care Center	56.8, range 27–88	1,975 (69.2%)	667 (23.4%)	Low
Pandeya et al (35)	Australia	Cross-sectional	1,580	National	range 30 or older	1,040 (65.8%)	853 (54%)	Low
Martin-de-Argila et al (36)	Spain	Cross-sectional	2,251	Patients using NSAIDs for 15–30 d across 434 primary care centers	56.5 ± 15.52	999 (44.4%)	1,504 (66.8%)	High
Shimamoto et al (10)	Japan	Cross-sectional	8,013	Patients presenting for visit at Kameda Medical Center Makuhari	50.4 ± 8.8	4,670 (58.3%)	2,112 (26.4%)	Low
Ercelep et al (37)	Turkey	Cross-sectional	2,037	Employees of the Cerrahpasa Faculty of Medicine	36.8 ± 9.6 (GERD), 35.9 ± 9.7 (controls)	791 (38.8%)	442 (21.7%)	High
Kubo et al (38)	USA	Case-control	490	Patients of Kaiser Permanente Northern California	range 20–79	334 (68.2%)	308 (62.9%)	Moderate
Park et al (39)	South Korea	Case-Control	2,226	Patients of Hallym University Medical Center	46.3, range 19–87 (cases)	1,380 (62%)	742 (33.3%)	Low
Asl et al (40)	Iran	Case-control	420	Patients of Razi Hospital Gastroenterology clinic in Rasht	45.5 ± 14.5 (cases), 45.4 ± 14.4 (controls)	180 (42.9%)	210 (50%)	High
Chirila et al (41)	Romania	Cross-sectional	184	Residents of Iaşi	49.4 (range 20–79)	78 (42.4%)	57 (31%)	High
Sajja et al (42)	USA	Cross-sectional	2,038	Patients presenting for outpatient visits at 7 primary care clinics	61.6 ± 7.6 (cases), 60.0 ± 9.2 (controls)	1,869 (91.7%)	310 (15.2%)	Low
Chang et al (43)	Taiwan	Cross-sectional	832	Men presenting for checkups at Taichung Veterans General Hospital	51.8 ± 13.0 (cases), 50.2 ± 12.0 (controls)	832 (100%)	180 (21.6%)	High
Eslami O et al (44)	Iran	Case-control	505	Patients referred to GI clinic at Ali-Ibn-AbiTalib Hospital	range 20–60	156 (30.9%)	285 (56.4%)	High
Filiberti et al (45)	Italy	Case-control	1,420	Patients undergoing EGD at 12 endoscopy units	56.2 ± 15.2 (BE), 52.6 ± 14.7 (EE), 53.7 ± 14.1 (controls)	766 (53.9%)	801 (56.4%)	High
Yuan et al (46)	China	Cross-sectional	678	GI outpatients at Third Xiangya Hospital, Central South University	49.01 ± 13.72 (cases), 46.34 ± 13.74 (controls)	348 (51.3%)	402 (59.3%)	High
Martinucci et al (47)	Italy	Cross-sectional	3,012	Students at University of Pisa	median 24, IQR 5	1,136 (37.7%)	792 (26.3%)	High
Yadegarfar et al (48)	Iran	Cross-sectional	1,130	Residents of Qom	39.4 ± 10.0	522 (46.2%)	717 (63.5%)	Moderate
Chen et al (49)	Taiwan	Cross-sectional	3,385	Patients undergoing screening EGD at Kaohsiung Veterans General Hospital	51.29 ± 11.42	1,932 (57.1%)	89 (2.6%)	High
Wei et al (50)	Taiwan	Cross-sectional	1,837	Patients undergoing screening EGD at Taipei Tzu Chi Hospital	51.57 ± 10.21	970 (52.8%)	763 (41.5%)	Low
Yuan et al (51)	China	Cross-sectional	1,518	Outpatients with upper GI complaints at 6 hospitals	48.51 ± 13.22 (cases), 47.45 ± 14.86 (controls)	757 (49.9%)	832 (54.8%)	High
Nadwa et al (52)	Saudi Arabia	Cross-sectional	500	Students at Jouf University in Sakaka City	21.07 ± 2.00	244 (48.8%)	106 (21.2%)	High
Tuerxun et al (53)	China	Cross-sectional	5,080	Residents of Kashgar	43.2 ± 15.4	2,153 (42.4%)	1,187 (23.4%)	High
Youssef et al (54)	Lebanon	Cross-sectional	264	Patients attending dietetic clinics in North and Mount Lebanon	43.02 ± 12.05 (cases), 36.28 ± 13.62 (controls)	133 (50.4%)	132 (50%)	High
Zhilina et al (55)	Russia	Cross-sectional	371	Residents of the Transbaikal territory	53.4 ± 17.47 (cases, Whites), 42.67 ± 11.52 (cases, buryat), 46.2 ± 19.2 (controls, Whites), 37.89 ± 15.54 (controls, buryat)	—	48 (12.9%)	Moderate
Helle et al (56)	Hungary	Cross-sectional	1,690	Blood donors at the Hungarian National Blood Transfusion Service in Szeged and in the settlements of Csongra'd-Csana'd county	39 (range 18–65)	1,004 (59.4%)	559 (33.1%)	Moderate
Wagihet al (57)	Egypt	Cross-sectional	100	Patients undergoing EGD at Al-Azhar University Hospital and Mansoura University Hospital	28.06 ± 5.12 (cases), 28.06 ± 5.12 (control)	52 (52%)	70 (70%)	High
Franzoni et al (58)	Italy	Case-control	449	Patients seeking evaluation of GI complaints in Northeast Italy	47 (range 23–71) (cases), 59 (range 34–75) (controls, cholelithiasis), 48 (range 29–64) (controls, functional dyspepsia)	204 (45.4%)	239 (53.2%)	High

BE, Barrett's esophagus; EE, erosive esophagitis; EGD, esophagogastroduodenoscopy; GERD, gastroesophageal reflux disease; GI, gastrointestinal; HD, hemodialysis; NSAID, non-steroidal anti-inflammatory drug; PD, peritoneal dialysis.

Association between coffee use and GERD

For analysis of the primary outcome, 40 studies were included, totaling 122,074 patients with 85,400 coffee drinkers and 36,674 nondrinkers. The definition of coffee drinker vs nondrinker varied between studies, with 12 comparing no versus any use and the remaining comparing weekly or daily use to less than weekly or daily consumption. Most of the studies defined GERD using symptoms (27/40), whereas the remaining used EE (7/40) or a combination of symptoms and EE (6/40). Study-specific exposure and outcome definitions are outlined in Supplementary Table 1 (see Supplementary Digital Content, http://links.lww.com/CTG/B510). The pooled rate of GERD among coffee drinkers was 34.9% (95% confidence interval [CI] 28.5–41.8) and was significantly increased statistically compared to nondrinkers at 30.7% (95% CI 25.2–36.7), with a pooled odds ratio (OR) 1.18 ([95% CI 1.03–1.36]; I² = 89.38) (Figure 2a). These findings persisted in all sensitivity analyses, as outlined in Table 2 and Supplementary Figure 1 (see Supplementary Digital Content, http://links.lww.com/CTG/B473).

Figure 2.

Meta-analysis of the association between coffee use and (a) GERD and (b) Barrett's esophagus. GERD, gastroesophageal reflux disease.

Table 2.

Pooled odds ratios for primary, subgroup, and sensitivity analyses

	Odds ratio	95% CI	I2	No. of studies	No. of patients	Coffee drinkers	Coffee nondrinkers
Primary analyses
GERD (primary outcome)	1.18	1.03–1.36	89.38	40	122,074	85,400	36,674
BE (secondary outcome)	1.13	0.79–1.61	55.5	6	6,999	3,515	3,484
Subgroup analyses
GER symptoms	1.09	0.94–1.27	88.1	29	103,176	76,807	26,369
EE	1.24	0.92–1.67	89.68	12	22,023	12,075	9,948
Coffee use defined as consuming 1 cup or more daily	1.03	0.93–1.14	62.49	12	45,377	31,604	13,773
Coffee use defined as consuming 1 cup or more weekly	1.06	0.97–1.15	0	4	52,014	44,362	7,652
Study region (primary outcome)
Europe	1.08	0.96–1.21	67.0	14	80,740	68,942	11,798
East Asia	1.45	1.11–1.91	91.7	12	30,862	11,354	19,508
United States	1.0	0.79–1.27	0	3	2,000	1,453	547
Middle East	0.90	0.44–1.82	94.4	9	6,534	2,257	4,277
Study region (secondary outcome)	0.94	0.48–1.82	62.0	4	1,576	1,377	199
Europe
Study population
Population-based	0.92	0.75–1.12	90.05	14	84,844	67,298	18,546
Hospital-based	1.36	1.05–1.75	90.71	17	27,800	12,338	15,462
Students	1.78	0.97–3.27	88.9	4	4,491	2,756	1,735
Patients referred for EGD	1.08	0.81–1.44	58.3	6	4,310	3,074	1,236
Sensitivity analyses
Additional studies that enrolled adolescents included	1.26	1.13–1.41	83.63	44	128,006	88,078	39,928
Additional studies that enrolled adolescents and used ICD codes to define GERD included	1.25	1.12–1.39	90.37	45	448,244	329,095	119,149
Excluded study that would bias away from null (Youssef et al 2021)	1.15	1.01–1.32	89.11	39	121,810	85,179	36,631
Excluded studies with high risk of bias	1.16	1.01–1.33	85.56	17	96,348	76,765	19,583
Excluded studies with moderate to high risk of bias	1.25	1.01–1.55	90.65	9	40,646	29,406	3,484
GERD restricted to those with heartburn and/or regurgitation at least once weekly and/or EE	1.31	1.09–1.57	90.54	31	55,689	32,082	23,607

BE, Barrett's esophagus; EE, erosive esophagitis; EGD, esophagogastroduodenoscopy; GER, Gastroesophageal reflux; GERD, gastroesophageal reflux disease.

Association between coffee use and Barrett's esophagus

For analysis of the association of coffee drinking with BE, 6 studies were included, totaling 6,999 patients with 3,515 coffee drinkers and 3,484 nondrinkers. The prevalence of BE is higher than the general population as 3 of 6 studies had a case-control design. There was no statistically significant difference in the rate of BE between the 2 groups. The pooled rate of BE was 22.1% among coffee drinkers (95% CI 12.8–35.4) and 17.6% among nondrinkers (95% CI 5.5–43.8), with a pooled OR 1.13 ([95% CI 0.79–1.61]; I² = 55.5) (Figure 2b).

Association between coffee use and GER symptoms

For analysis of the subgroup of studies that defined GERD based on GER symptoms of heartburn and/or regurgitation or using validated questionnaires, 29 studies were included, totaling 103,176 patients with 76,807 coffee drinkers and 26,369 nondrinkers. There was no statistically significant difference in the rate of GER symptoms between coffee drinkers and nondrinkers. The pooled rate of GER symptoms was 34.4% among coffee drinkers (95% CI 26.5–43.2) and 32.2% among nondrinkers (95% CI 24.8–40.6), with OR 1.09 ([95% CI 0.94–1.27]; I² = 88.10) (Figure 3a).

Figure 3.

(a) and (b) Subgroup meta-analyses of the association of coffee use and the presence of (a) reflux symptoms defined by presence of heartburn and/or regurgitation or validated questionnaires and (b) erosive esophagitis. (c) and (d) Subgroup meta-analyses of the association of GERD and the use of (c) 1 or more cup of coffee daily and (d) 1 or more cup of coffee weekly. GERD, gastroesophageal reflux disease.

Association between coffee use and erosive esophagitis

For analysis of the subgroup of studies that required the presence of EE on endoscopy to define GERD, 12 studies were included, totaling 22,023 patients with 12,075 coffee drinkers and 9,948 nondrinkers. Although there was a higher rate of EE among coffee drinkers, the difference between drinkers and nondrinkers was not statistically significant. The pooled rate of EE was 29.8% among coffee drinkers (95% CI 20.4–41.2) and 24.6% among nondrinkers (95% CI 19.3–30.8), with an OR 1.24 ([95% CI 0.92–1.67]; I² = 89.68) (Figure 3b).

Association between coffee use and reflux monitoring

There were 3 studies assessing the relationship between coffee use and objectively measured GER using ambulatory reflux monitoring. All 3 studies used abbreviated, nonstandardized monitoring protocols of 1–3 hours. Two randomized trials (1 comparing regular and decaffeinated coffee; the other comparing coffee with varying caffeine levels, tea, and water) found significant associations between coffee intake and increased AET and total reflux events with 3-hour pH-impedance testing (5,59). A third trial comparing coffee and water intake found no effect, although the protocol of pH-impedance monitoring was limited to 1 hour (60).

Relationship at specific coffee doses

We evaluated subgroups of studies that assessed specific levels of coffee exposure in relation to GERD to determine if there is a consumption threshold at which the risk of GERD increased. The first subgroup evaluated studies that defined coffee drinking as consuming 1 cup or more daily vs less than 1 cup daily. Twelve studies were included in this analysis, totaling 45,377 patients with 31,604 drinkers of ≥1 cup daily and 13,773 with <1 cup daily use. In this subgroup, ≥1 cup daily of coffee use was not significantly associated with increased GERD (OR 1.03, 95% CI 0.93–1.14; I² = 62.49) (Figure 3c).

The second subgroup evaluated studies that defined coffee drinking as consuming 1 cup or more weekly versus less than 1 cup weekly. Four studies were included, totaling 52,014 patients with 44,362 drinkers of ≥1 cup weekly and 7,652 with <1 cup weekly use. In this subgroup analysis, ≥1 cup weekly of coffee use was not associated with a significant increase in rate of GERD (OR 1.06, 95% CI 0.97–1.15; I² = 0) (Figure 3d).

There were insufficient data to perform metaregression evaluating a dose-response relationship between coffee use and GERD. The doses examined across studies were not standardized (e.g., some evaluated in categories of 2–4 cups/day vs 3–6 cups/day) and the number of studies evaluating varying doses was limited.

Association between coffee use and GERD/BE by study geographic region

For the primary outcome, 4 subgroups were analyzed based on study region: Europe (14 studies with 80,740 patients—68,942 coffee drinkers and 11,798 nondrinkers), East Asia (12 studies with 30,862 patients—11,354 coffee drinkers and 19,508 nondrinkers), US (3 studies with 2,000 patients—1,453 coffee drinkers and 547 nondrinkers), and Middle East (9 studies with 6,534 patients—2,257 coffee drinkers and 4,277 nondrinkers). The significant association between coffee use and GERD persisted in the subgroup of studies from East Asia, with a pooled OR 1.45 ([95% CI 1.11–1.91]; I² = 91.7). The remainder of the regions did not demonstrate a significant association between coffee use and GERD (see Supplementary Figures 2a–2d, Supplementary Digital Content, http://links.lww.com/CTG/B474).

For the secondary outcome, a subgroup of studies from Europe was analyzed, which included 4 studies with 1,576 patients—1,377 coffee drinkers and 199 nondrinkers. Similar to the primary analysis, there was no significant association between coffee use and BE with a pooled OR 0.94 ([95% CI 0.48–1.82]; I² = 62.0) (see Supplementary Figure 2e, Supplementary Digital Content, http://links.lww.com/CTG/B474).

Association between coffee use and GERD by population

Four subgroups were analyzed based on study population: population-based (14 studies with 85,844 patients—67,298 coffee drinkers and 18,546 nondrinkers), hospital-based (17 studies with 27,800 patients—12,338 coffee drinkers and 15,462 nondrinkers), studies using students (4 studies with 4,491 patients—2,756 coffee drinkers and 1,735 nondrinkers), and studies that recruited patients referred for EGD (6 studies with 4,310 patients—3,074 coffee drinkers and 1,236 nondrinkers). The significant association between coffee use and GERD persisted in the subgroup of hospital-based studies with a pooled OR 1.36 ([95% CI 1.05–1.75]; I² = 90.71). The remainder of the subgroups did not demonstrate a significant association between coffee use and GERD (see Supplementary Figure 3, Supplementary Digital Content, http://links.lww.com/CTG/B475).

Quality assessment

The risk of bias assessment using The Evidence Project risk of bias tool is outlined in Supplementary Table 2 (see Supplementary Digital Content, http://links.lww.com/CTG/B510). Using the predefined criteria outlined in the Supplementary Methods (see Supplementary Digital Content, http://links.lww.com/CTG/B510), a little more than half the studies had a high risk of bias (25/43), whereas the remaining were low (10/43) and moderate (8/43) risk. The risk of bias for each study using the predefined criteria is listed in Table 1.

Publication bias and heterogeneity

There was no evidence of publication bias based on funnel plot evaluation and Egger regression testing (P > 0.05) as demonstrated in Figure 4.

Figure 4.

Funnel plot representation of included studies found no significant publication bias.

Subgroup and sensitivity analyses where exposure/outcome definitions were modified and based on study quality, region, and population were performed to account for expected heterogeneity across studies. These analyses highlighted some of the factors contributing to heterogeneity, as I² improved in some of the regional and population subgroups (Table 2). However, there remained significant heterogeneity despite evaluating these subgroups, indicating that other unmeasurable factors are likely contributing to the study differences. Similarly, metaregression evaluating study region, population, study design (cohort vs cross-sectional), and publication year as covariates may explain some of the variance between studies (see Supplementary Figure 4, Supplementary Digital Content, http://links.lww.com/CTG/B477), but the R² = 0.

DISCUSSION

Coffee use has been commonly implicated as a dietary trigger of GERD, with patients often advised to avoid coffee as part of lifestyle management. However, evidence supporting this recommendation has been conflicting. In this meta-analysis including 39 studies and 121,625 patients, we demonstrated that coffee intake is associated with a small, statistically significant increase in risk of GERD. Our findings persisted in all the preplanned sensitivity analyses, such as those including adolescent patients, using expanded definition of GERD, excluding studies with high and moderate risks of bias, and restricting to more frequent/severe GERD, thereby arguing for the validity of our results. Although 2 previous meta-analyses on this topic did not find significant association between coffee intake and GERD (11,12), our analysis encompassed a much larger number of studies and sample size while using a more comprehensive search strategy to capture studies internationally, thus allowing more rigorous analyses.

Although we identified a statistically significant effect on the risk of GERD by coffee intake, the clinical relevance of this association is unclear given the small effect size. In particular, when weighing the marginal risk of GERD with the numerous reported benefits of coffee intake (61), including reduced risk of all-cause mortality (62,63), malignancy (64,65), type 2 diabetes (66), fatty liver (67), and liver fibrosis (68), universal strict avoidance of coffee may not be the most effective or sensible primary treatment strategy for most of the patients with reflux symptoms. Restricting coffee intake may be best reserved for patients who clearly identify coffee as a symptom trigger and report distinct symptom improvement with reduction or cessation of use, consistent with current recommendations in the American Gastroenterological Association guidelines (69). Perhaps our findings may encourage providers to focus on other lifestyle interventions with more robust supporting evidence, such as head-of-bed elevation or right-sided sleeping for nocturnal reflux symptoms (6), in patients where coffee is not an obvious trigger of their symptoms.

This approach may be further supported by our observation that coffee intake did not correlate with an increased risk of BE. The pathophysiology of BE involves several independent risk factors in addition to GERD, including genetic predisposition, smoking, and obesity (70). Any elevation in reflux burden from coffee use alone may not be sufficient to increase the risk of BE without other risk factors. It is also possible that any potential deleterious effects of coffee use may be overshadowed by other more significant risk factors or outweighed by possible protective anti-dysplastic/carcinogenic properties. However, although observational studies suggested that coffee may protect against the development of various cancers, benefits against esophageal adenocarcinoma have not been specifically demonstrated (71,72).

We attempted to determine if a threshold of coffee use is needed to induce an increase in reflux. However, no significant associations were found at 1 cup weekly and 1 cup daily thresholds. These results may suggest that higher coffee consumption amounts are needed to induce GERD, or they may reflect the limited power from a smaller number of studies included in these subgroup analyses. Notably, Mehta et al identified a dose-dependent relationship between coffee use and GERD at 1 cup daily increments (range: 1–6 cups daily) utilizing prospectively collected data from the Nurses' Health Study (8), although this has not been replicated in other studies. Although most health benefits reported with coffee were observed at high doses of consumption (2–6+ cups/day) (61,73,74), a clear threshold for safe coffee intake regarding GERD cannot be established based on current aggregate data and require further evaluations.

Through subgroup analyses, we also attempted to gain insights into the mechanism by which coffee may precipitate or aggravate reflux symptoms. Specifically, we compared studies defining GERD by patient-reported symptoms alone and those requiring endoscopic evidence of reflux-related mucosal injury. Increased symptoms in the former may be driven by increased reflux, reflux hypersensitivity, or a functional cause, whereas the latter would more specifically reflect increased reflux burden and mucosal injury as the likely mechanism. Neither of these subgroup analyses revealed significant associations with coffee use, which may indicate a multifactorial mechanism underlying the relationship between coffee and GERD, or limitations in power for these subgroup analyses. However, there was, notably, a larger difference in EE rates between coffee drinkers and nondrinkers (27.19% vs 21.58%) compared with that for reflux symptoms (33.87% vs 31.39%). Moreover, 2 randomized trials requiring ambulatory reflux monitoring have demonstrated increased reflux burden (AET and reflux events) with coffee use, although both utilized a nonstandardized 3-hour pH-impedance test (5,59).

The properties of coffee that may potentially contribute to increased reflux are not fully understood. Some studies have attributed this to caffeine content, whereas others have investigated different methods of roasting. Regarding caffeine content, 2 randomized trials using pH-impedance testing and including both caffeinated and decaffeinated coffee demonstrated conflicting results—1 study showed that coffee led to increased reflux regardless of caffeine content (5), whereas the other found increased reflux with regular but not decaffeinated coffee (59). In the study by Mehta et al, stratification by caffeine status did not impact the association between coffee and GERD (8). Notably, this study also found an association between tea consumption and reflux symptoms, although only with decaffeinated tea. Based on current evidence, whether caffeine is an etiologic factor of reflux remains unclear, although any impact of coffee use on GERD may be due to more than caffeine content alone.

Different roasting practices may affect the levels of chlorogenic acids, N-alkanoyl-5-hydroxytryptamides (C5HT), and caffeine levels, which may in turn affect gastric acid. However, there are limited studies evaluating the potential effect of coffee roasting processes on GERD. A study of healthy volunteers assessing gastric pH levels after infusing dark and medium roast coffee at similar caffeine levels found that dark roast coffee led to smaller decrease in gastric pH and lower levels of chlorogenic acids and C5HT (75). However, whereas different coffee roasts (light vs medium vs dark) are primarily a function of varying roast temperature and duration, there are other factors constituting the roasting process (76). Another randomized crossover trial comparing conduction versus convection roasting found no differences in heartburn or dyspepsia symptoms, with both types of coffee leading to symptoms in most patients (77).Finally, the outer waxy layer of coffee beans, which contains high levels of C5HT, has been suggested as a trigger for gastrointestinal symptoms. Dewaxing coffee beans before roasting has been shown to reduce C5HT levels (78), and a randomized crossover trial comparing dewaxed and standard coffee found a significant reduction in GERD symptoms with dewaxed coffee use (79). Despite these isolated studies, there are currently insufficient data to conclude any potential impact of coffee sourcing, roasting practices, and preparation on GERD.

The pathophysiologic impact of coffee on the esophagus that may increase reflux is unclear, although increased lower esophageal sphincter (LES) relaxation (4,80–83) and gastric hypersecretion (84,85) have both been implicated. One study measured changes in LES pressure in healthy controls and patients with EE after coffee ingestion and found that coffee at varying pH (4.5 and 7) significantly dropped LES pressure when consumed alone or with a meal (80). Another study of healthy volunteers found significant variability through which different coffee preparations led to LES relaxation, increased gastric acid secretion, and elevated gastrin levels, suggesting that differences in esophagogastric response to coffee may be related to how coffee beans are processed. It also highlighted that the mechanism by which coffee precipitates GERD is likely multifactorial, as most of the coffee preparations led to drops in LES pressure (regardless of caffeine content), but caffeinated coffee stimulated more gastric acid secretion than decaffeinated preparations (84).

Notably, on subgroup analyses by the geographic locations where the studies were conducted, only the pooled findings from East Asian studies remained significant for the association between coffee use and GERD. Findings in the European subgroup reached near statistical significance, whereas the subgroups from the United States and Middle East showed no correlation between coffee and GERD. The latter subgroups contained smaller numbers of studies, with only 3 in the US subgroup in particular. This may explain, in part, the observed nonsignificant pooled findings of these subgroups. It is also possible that the characteristics of the coffee and the manner through which it is prepared or consumed vary by geographic region and culture, which may modulate the risk of GERD. Coffee extraction practices, dairy and nondairy additives, and/or timing of consumption may play a role. Further research into these potential characteristics and consumption practices may help delineate the complex relationship between coffee use and GERD.

Although our study represents the most comprehensive meta-analysis on the topic to date, there are several limitations. First, very few studies focused solely on coffee use and GERD. Rather, most of the studies evaluated a variety of lifestyle factors including coffee. As a result, granular data on baseline characteristics between coffee drinkers and nondrinkers could not be assessed (including concurrent proton pump inhibitors use), and adjusted odds ratios were not always available. Second, there was significant heterogeneity across studies, which we tried to address through multiple subgroup analyses stratifying studies by population, region, risk of bias and how exposure and outcome were defined. Furthermore, we performed meta-regression evaluating study population, design, region, and publication year. However, not all differences could be accounted for in this manner, and there may be other patient characteristics contributing to heterogeneity across studies that cannot be examined further due to a lack of granular data. Third, the bulk of the included studies were cross-sectional or case-control designs, limiting our ability to evaluate causality or linear relationships. Fourth, the definitions for coffee use and GERD were not standardized across the studies. What constitutes a cup of coffee may differ across regions and may explain differences in the results seen across regions. However, the GERD definitions used in the studies meet the standards outlined in current guidelines (6) by using typical symptoms of heartburn and/or regurgitation and/or endoscopic findings of EE, although it is important to note that LA grade A esophagitis, an inconclusive evidence of pathologic reflux per Lyon consensus (86), was included in the EE group in most studies. Finally, data stratifying patients by EE severity or length of BE segments were not universally available.

Despite the limitations, our study had a number of strengths. As noted above, this is the most comprehensive meta-analysis on this topic to date, with a thorough search strategy and large number of studies accounted for internationally. Furthermore, although more than half the studies were at high risk of bias, we conducted multiple sensitivity analyses (including analyses evaluating only studies with moderate and low risk of bias with large sample sizes) in which our overall findings persisted, supporting the validity of the results. Finally, we were able to conduct multiple subgroup analyses that allowed us to explore the mechanisms through which coffee affects GERD and the potential dosing threshold.

In conclusion, this comprehensive systematic review and meta-analysis demonstrates that coffee use is associated with a small, statistically significant increase in the rate of GERD. However, this minor difference is of unclear clinical significance, especially given that no increased risk of BE was observed. Management strategies should be individualized. For patients who identify a clear link between coffee intake and occurrence of reflux symptoms, avoiding or limiting coffee use remains a reasonable, low-risk measure. However, with the reported health benefits associated with coffee use, our findings call into question the universal recommendation to avoid or restrict coffee intake as a routine lifestyle modification for GERD. Perhaps this may encourage providers to focus patient counseling on lifestyle interventions with more robust supporting evidence, like head of bed elevation, or other dietary interventions like alcohol cessation. Future studies should further explore the underlying properties or mechanisms through which coffee may potentially affect GERD, identify any consumption threshold, and risk-stratify patients who may benefit from targeted lifestyle intervention to reduce coffee intake.

CONFLICTS OF INTEREST

Guarantor of article: Walter W. Chan, MD, MPH, FACG.

Specific author contributions: Study concept and design: M.M., T.R.M., W.W.C. Data acquisition: M.M., D.H., R.F., B.H., T.R.M. Paper preparation and statistical analysis: M.M., T.R.M., W.W.C. Critical revisions: M.M., D.H., R.F., B.H., T.R.M., W.W.C. Administrative support and overall study supervision: W.W.C.

Financial support: B.H. is supported by the NIH/NIDDK award T32DK135449.

Potential competing interests: None to report.

Data availability: The data that support the findings of this study are available from the corresponding author upon reasonable request. All authors approve of final submission.

Study Highlights.

WHAT IS KNOWN

• ✓ Coffee use has been suggested as a risk factor of gastroesophageal reflux disease (GERD), although data have been inconsistent.

• ✓ Coffee avoidance is frequently used as a first-line lifestyle strategy for reflux symptoms.

WHAT IS NEW HERE

• ✓ Coffee use is associated with a marginal but statistically significant increase in GERD of unclear clinical significance.

• ✓ No consumption threshold for coffee was identified for the risk of GERD.

• ✓ No significant association between coffee intake and Barrett's esophagus was observed.

Supplementary Material

ct9-17-e00996-s005.pdf ^{(103.8KB, pdf)}

ct9-17-e00996-s006.docx ^{(53.8KB, docx)}

ABBREVIATIONS:

AET

acid exposure time

BE

Barrett's esophagus

C5HT

N-alkanoyl-5-hydroxytryptamides

EE

erosive esophagitis

EGD

esophagogastroduodenoscopy

GER

gastroesophageal reflux

GERD

gastroeosphageal reflux disease

LES

lower esophageal sphincter

OR

odds ratio