Gastro-oesophageal reflux disease is what happens when stomach contents come back up into the oesophagus often enough to cause symptoms or damage. The classic symptoms are heartburn and regurgitation, and they are extremely common: a systematic review of population-based studies put prevalence at 18.1–27.8 percent in North America, 8.8–25.9 percent in Europe and lower in East Asia, using a definition of heartburn and/or regurgitation on at least one day a week [1] [2], with the clinical spectrum characterised in an Olmsted County population study [3]. It is, as the American College of Gastroenterology puts it, arguably the most common disease a gastroenterologist encounters [4].
Three things make it worth a careful review rather than a shrug.
It is mostly a quality-of-life problem, and that is not nothing. For most people reflux is a nuisance that disturbs sleep, meals and exercise. It is not dangerous, and treating it is about restoring normal life rather than preventing anything.
A minority develop real complications — erosive oesophagitis, stricture, and Barrett's oesophagus, in which the oesophageal lining changes to an intestinal type and carries a small risk of progressing to adenocarcinoma.
And the treatment is one of the most-used drug classes on earth. Proton pump inhibitors are among the top ten most widely used drugs in the world [5]. When a drug is taken by that many people for a condition that is usually benign, the safety questions deserve a careful answer rather than either reassurance or alarm.
Start here: what refluxes, and why it matters that it does
The barrier between stomach and oesophagus is the lower oesophageal sphincter, supported by the diaphragm. When it relaxes inappropriately, or when a hiatal hernia separates the two components, acidic gastric contents enter an oesophagus that has none of the stomach's defences against them.
Obesity is the major modifiable risk factor, and its effect runs the whole length of the disease. A systematic review and meta-analysis examined body mass index against GERD symptoms, erosive oesophagitis, Barrett's oesophagus and adenocarcinoma of the oesophagus and gastric cardia [6]. That is the point worth holding onto: obesity is associated not just with symptoms but with each successive step of the complication ladder.
And reflux is a genuine cause of oesophageal adenocarcinoma, not merely a marker. A nationwide population-based case-control study in Sweden interviewed 189 patients with oesophageal adenocarcinoma and 262 with adenocarcinoma of the cardia — 85 percent of all eligible cases in the country over three years — against 820 population controls and 167 patients with oesophageal squamous-cell carcinoma [7]. Including a second cancer type as a comparison group is the design detail that matters: squamous-cell carcinoma of the same organ is not caused by reflux, so it acts as a control for the recall and detection biases that plague interview-based case-control studies.
The same group then tested the mechanism from a different angle, which is what makes the causal case strong rather than merely consistent. If reflux causes adenocarcinoma, then anything that increases reflux should increase the cancer — so they examined medications that relax the lower oesophageal sphincter: nitroglycerin, anticholinergics, beta-adrenergic agonists, aminophyllines and benzodiazepines, in the same Swedish population with the same three comparison groups [8]. An exposure that acts only through the proposed mechanism, tested against the same outcome, is a much harder thing to explain away than a symptom history.
But the absolute risk needs stating in the same breath. Oesophageal adenocarcinoma incidence rose sharply in Western countries [9], and some of that rise has been attributed to overdiagnosis and reclassification rather than to real change [10]. Most people with reflux never develop Barrett's; most people with Barrett's never develop cancer. The number is now known with some precision. A Dutch nationwide histopathology registry followed 42,207 patients with a first diagnosis of Barrett's oesophagus, 8 percent of whom had low-grade dysplasia, and found an annual risk of oesophageal adenocarcinoma of 0.4 percent — with older age (hazard ratio 12 above 75), male sex (2.01) and low-grade dysplasia at baseline (1.91) as independent predictors [11]. Risk factors for progression specifically from low-grade dysplasia have been examined separately [12].
That 0.4 percent a year is the number that justifies the guideline arithmetic. The ACG, acknowledging the low progression risk in non-dysplastic Barrett's, lengthened its recommended surveillance intervals to no more frequently than every three to five years [13] — across a four-year interval, 0.4 percent per year is roughly a 1.6 percent chance of cancer arising between endoscopies, which is the trade the interval is making explicit.
Pillar 1: measurement and diagnosis
Mostly clinical
Typical heartburn and regurgitation in someone without alarm features generally does not need investigation. The usual first step is an empiric trial of acid suppression: if it works, that is both treatment and supporting evidence. The ACG guideline sets out the diagnostic modalities and when each is warranted [4].
Alarm features change that — dysphagia, weight loss, gastrointestinal bleeding, anaemia — and prompt endoscopy, which is also how oesophagitis is graded [14] and how Barrett's is looked for.
When the diagnosis is unclear
Ambulatory pH or pH-impedance monitoring measures how much acid actually reaches the oesophagus, and is used when the diagnosis is in doubt or before committing someone to anti-reflux surgery [4]. The impedance component adds the ability to detect non-acid reflux.
The reason this matters is a distinction that is easy to miss: having reflux symptoms is not the same as having pathological reflux. Some people with typical heartburn have entirely normal acid exposure — functional heartburn — and they will not respond to acid suppression however much of it they are given, nor to surgery. Objective testing is what separates them. It is also why PPI response has been studied as a diagnostic test in non-cardiac chest pain, where PPI therapy reduces symptoms and may help identify abnormal oesophageal acid reflux [15].
And it is how the largest practical problem in this disease gets sorted out. Patients who still have symptoms on twice-daily acid suppression are a mixed group, and combined multichannel intraluminal impedance with pH monitoring can separate them, because impedance detects reflux episodes regardless of their pH — distinguishing persisting acid reflux from non-acid reflux, which acid suppression cannot fix because it changes the acidity of the refluxate rather than stopping the reflux [16].
Two conditions should be excluded before settling on reflux: eosinophilic oesophagitis, an immune-mediated oesophageal disease with its own consensus diagnostic criteria and its own treatment [17] [18], and, in the right context, infection or malignancy.
The extra-oesophageal presentations deserve particular scepticism. Chronic cough and chronic laryngitis are frequently attributed to reflux and treated with acid suppression on that assumption. The role of PPIs in suspected reflux-related chronic laryngitis is explicitly described as controversial and has been meta-analysed across eight randomised placebo-controlled trials in 344 patients [19], and reflux treatment for prolonged non-specific cough has been assessed by Cochrane in children and adults [20]. Attributing a symptom to reflux because it is common is not the same as demonstrating that reflux causes it in that patient.
An honest gap: the intended figure for this review was the pharmacodynamic one — the fraction of the day intragastric pH is held above 4, or the oesophageal acid-exposure-time threshold that defines pathological reflux. Neither dataset is recoverable from this substrate, which was assembled by citation crawl during an OpenAlex outage. No pH-holding-time series and no Lyon-consensus threshold came back. The figure below uses the dose-response that is grounded instead, and no numerical threshold for acid exposure time is quoted anywhere in this review.
Centerpiece: what more acid suppression actually buys
A primary-care randomised, double-blind, placebo-controlled trial gives the whole dose-response in one place. Patients with troublesome reflux symptoms had an endoscopy first. Those without oesophagitis (n = 261) were randomised to omeprazole 20 mg, omeprazole 10 mg or placebo for four weeks; those with oesophagitis (n = 277) to 20 mg or 10 mg. Everyone symptom-free at the end was followed for six months off treatment [21].
The results, as reported: complete upper-GI symptom relief during week 4 in 41 percent on 20 mg and 35 percent on 10 mg; sufficient control in 73 percent and 62 percent; endoscopic healing of oesophagitis in 76 percent and 56 percent. Among endoscopy-negative patients on placebo, complete relief was 19 percent and sufficient control 35 percent. After six months off treatment, 90 percent of oesophagitis patients and 75 percent of endoscopy-negative patients had relapsed [21].
The check the arithmetic was never given. The trial reports three endpoints at two doses and never divides them by each other. Doing so shows the two symptom endpoints — complete relief and sufficient control, which are two different thresholds applied to the same experience — scaling with dose by ×1.171 and ×1.177. They agree to within 0.6 percent. Two independent readings of the same symptomatic response, responding to a doubling of dose in precisely the same proportion, which is what you would expect if they are measuring one underlying quantity at two cut-points.
And then the endpoint that comes apart. Endoscopic healing scales by ×1.357 — about 16 percent more dose-responsive than either symptom measure. Doubling the dose buys substantially more mucosal healing than it buys comfort. In absolute terms the extra 10 mg heals 20 more patients per hundred, a number needed to treat of 5, and removes 45 percent of the residual non-healing left by the lower dose.
That divergence is the clinically important part, and it points the same way as the guideline advice. If you titrate acid suppression by how the patient feels, you will systematically under-treat the oesophagus — because symptoms improve at a dose that has not yet done the healing. It is the argument for treating erosive oesophagitis to a defined course and dose rather than to comfort.
The relapse panel is the other half of the honest picture. Within six months of stopping, nine in ten oesophagitis patients were symptomatic again. Acid suppression is suppression, not cure. And note which group did worse: the patients whose mucosa had been visibly damaged and then healed relapsed more often than those who never had visible damage — so healing the oesophagus does not buy durable remission once the drug stops.
Three honest limits. The two symptom percentages are pooled across both endoscopy strata while the placebo percentages come from the endoscopy-negative stratum only, so they are not a like-for-like drug-versus-placebo contrast and the figure does not draw them as one. The placebo response is nonetheless substantial — 19 percent complete relief, 35 percent sufficient control — which is worth remembering when anyone reports uncontrolled improvement on acid suppression. And this is one trial of one drug at two doses; the ×1.36 versus ×1.17 gap is a finding within it, not an established constant of the drug class.
Pillar 2: treatment
Lifestyle first, and one lever dominates
Weight loss is the intervention with the best rationale, because obesity is associated with every rung of the ladder from symptoms to adenocarcinoma [6]. Head-of-bed elevation, avoiding late meals, and identifying individual food triggers are the other standard measures, and the guideline sets out which lifestyle interventions have evidence behind them and which are traditional [4].
Escalating acid suppression
Antacids and alginates neutralise or raft over acid for immediate, short-lived relief. H2-receptor antagonists block one of the three stimulatory pathways to the parietal cell and suit mild or on-demand use.
Proton pump inhibitors are the mainstay. They work by irreversibly inhibiting the gastric H⁺/K⁺-ATPase — the proton pump itself, the final common step of acid secretion — which is why they suppress acid more completely and for longer than anything acting upstream of it. The figure above is what that potency buys. The corollary is a dosing detail patients are rarely told: because the drug binds pumps that are actively secreting, it works best taken before a meal rather than at bedtime or on an empty stomach.
Deprescribing matters precisely because the drugs work. The relapse data above explain why so many people end up on indefinite therapy, and a double-blind trial put numbers on how hard stopping is. Ninety-seven patients on long-term PPIs — mean 48 months of use, 78 percent with GERD, none with a history of ulcer or oesophagitis — were randomised to taper or to a constant dose for three weeks before all stopped. Twenty-seven percent were off PPIs a year later. Tapering did not help (31 percent against 22 percent, not significant). And the split that matters: only 21 percent of the GERD patients were off treatment at a year, against 48 percent of those without GERD [22].
Read that the right way round. Half the people without reflux disease on long-term acid suppression could simply stop — which is an indictment of how the drugs get started and continued, not of the drugs. The guideline addresses the risks and side effects of the main treatments and their implications for management [4] [23].
The long-term safety debate, honestly
PPIs have been associated with a long list of harms in observational data. The two most instructive examples are in this substrate.
Hip fracture. A nested case-control study in the UK General Practice Research Database found an association between long-term PPI therapy and hip fracture, and — importantly — ran the same analysis for H2-receptor antagonists as a comparison [24]. The proposed mechanism cuts both ways, since PPIs may impair calcium absorption through hypochlorhydria but may also reduce bone resorption by inhibiting osteoclastic vacuolar proton pumps.
The gut microbiome. Tag sequencing in 1,815 individuals across three cohorts, 211 of them PPI users, found significantly decreased diversity and changes in 20 percent of bacterial taxa, with multiple oral bacteria over-represented in the faecal microbiome of users [5] — and an independent study reported altered gut microbiota composition in the same direction [25]. This one has a plausible mechanism that does not depend on who takes the drug: less gastric acid means more swallowed organisms survive the stomach. It connects to the observed association with enteric infection, notably Clostridium difficile, and it is the PPI harm with the strongest claim to being real.
Chronic kidney disease has been reported in cohort studies as associated with PPI use, including progression to end-stage disease [26] [27], and other acid-suppressing classes have been examined alongside [28].
And then there is the case that calibrates all of the others. Omeprazole was suspected of blunting clopidogrel's antiplatelet effect through CYP2C19, and a retrospective cohort of 8,205 patients taking clopidogrel after an acute coronary syndrome across 127 Veterans Affairs hospitals appeared to confirm it: death or rehospitalisation for acute coronary syndrome occurred in 20.8 percent of those not prescribed a PPI and 29.8 percent of those who were [29]. That is a nine-percentage-point absolute difference, a 43 percent relative increase, in a large well-conducted observational study.
Then it was randomised. A trial of clopidogrel with or without omeprazole in coronary artery disease was run to settle the question [30] — and the cardiovascular signal did not survive the design change.
That is the single most useful number in this section, and it is not about clopidogrel. A 43 percent apparent excess, in a big cohort, with a plausible pharmacological mechanism already demonstrated in the laboratory, turned out to be confounding. PPI users are older, sicker and more medicated than non-users, and they are prescribed acid suppression because of that. So when you read that PPIs are associated with hip fracture, kidney disease, dementia or pneumonia, the clopidogrel episode is the correct prior: these associations are overwhelmingly observational, confounding by indication is severe and hard to remove, and the one that was tested by randomisation dissolved.
That does not make them false — the microbiome finding has a direct mechanism and independent replication, and it deserves more weight than the rest. It means the appropriate response is neither reassurance nor alarm but the same thing good practice would require anyway: the lowest effective dose, for a defined reason, reviewed periodically. Expert best-practice guidance takes the same line [23].
Potassium-competitive acid blockers are a newer class that inhibit the same pump reversibly and competitively with potassium, giving faster and more sustained acid suppression than PPIs. This substrate returned no vonoprazan trial, so this review states no efficacy figures for the class.
Surgery and endoscopic options
Fundoplication wraps the gastric fundus around the lower oesophagus to reconstruct the barrier mechanically. It is an option for selected patients — typically those with objectively confirmed reflux who respond to acid suppression but do not want lifelong medication, or who have volume regurgitation, which acid suppression cannot fix because it reduces the acidity of the refluxate without reducing the reflux [16]. Outcomes have been compared against omeprazole maintenance at twelve years [31], and long-term results have been reported specifically in patients with Barrett's oesophagus [32] — where the important caveat is that adenocarcinoma still occurs after antireflux surgery, so an operation is not a reason to stop surveillance [33]. Comparative effectiveness of medical, surgical and endoscopic management is covered in the guideline [4].
The requirement for objective confirmation before surgery is where the functional-heartburn distinction becomes consequential: operating on someone with normal acid exposure will not help them.
Barrett's oesophagus
Screening is targeted rather than universal: the ACG restricts routine screening to men with reflux symptoms and multiple other risk factors [13], with British Society of Gastroenterology guidance covering the same ground [34]. Non-dysplastic Barrett's is surveilled no more often than every three to five years; neither biomarker panels nor advanced imaging beyond high-definition endoscopy is recommended; and endoscopic ablative therapy is recommended for high-grade dysplasia and for T1a adenocarcinoma [13]. Endoscopic resection techniques have their own European guidance [35], curative endoscopic resection of early Barrett's adenocarcinoma has been reported [36], and scheduled surveillance after curative resection controls secondary cancer [37]. Detection can be improved with chromoendoscopy and directed biopsy techniques [38] [39], and molecular markers of progression such as aberrant p53 and aneuploidy have been studied [40] [41] — though the ACG does not currently recommend biomarker panels or advanced imaging beyond high-definition endoscopy [13]. Whether acid-suppressive medication itself reduces adenocarcinoma risk in Barrett's has been examined directly [42].
That is a real success — a cancer precursor that can be found and destroyed through an endoscope.
Pillar 3: what is unresolved
Separating true reflux from functional heartburn. The people who do not respond to acid suppression are a mixed group, and distinguishing those with persisting acid exposure from those whose oesophagus is hypersensitive but chemically normal determines whether the answer is more suppression, surgery, or neither. Objective testing exists [4]; using it consistently is the gap.
Whom to screen for Barrett's, and how often. The guideline has already moved once toward less frequent surveillance on evidence of low progression risk [13]. Whether non-endoscopic screening devices can widen detection without generating overdiagnosis — the phenomenon already implicated in the apparent rise in adenocarcinoma incidence [10] — is open.
PPI safety, settled properly. Almost every association rests on observational data [24] [26] [27] [28]. The clopidogrel question was resolved by randomising it, and the answer differed from the cohort study [29] [30]; most of the others have not been randomised and probably cannot be. Expert best-practice guidance exists in the meantime [23].
Deprescribing at scale. Given how many people are on these drugs [5] and how reliably symptoms return on stopping [21], the practical question is which patients can step down safely and how to support them through the rebound.
Dig deeper in lmmol
Obesity is the central chain here, and the link is stronger than for most of its associations: body mass tracks not only reflux symptoms but erosive oesophagitis, Barrett's oesophagus and adenocarcinoma [6], so weight is the one lever that acts on every rung at once. The Barrett's-to-cancer thread connects to the oncology reviews, where the same logic of a detectable precursor and a targeted-surveillance programme appears — colorectal cancer is the closest analogue, since both remove a precursor lesion endoscopically, and pancreatic-cancer is the instructive opposite, a cancer with no accessible precursor and no screening programme. Inflammatory bowel disease and celiac disease sit alongside as the other common gastrointestinal conditions where endoscopy plus biopsy decides management, and celiac disease shares with GERD the trap that treating before testing destroys the evidence. Obstructive sleep apnea shares both the obesity risk factor and the nocturnal symptom burden. The full collection is at health.