Abstract
Objective
Proton-pump inhibitors (PPIs) may disrupt lysosomal acidification, a pathway implicated in age-related macular degeneration (AMD) pathogenesis. We evaluated whether sustained PPI exposure increased the risk of incident AMD in older adults aged ≥60 years.
Methods and analysis
This retrospective observational cohort study included adults aged ≥60 years with sustained PPI exposure, matched 1:1 to unexposed controls on demographics, overall health, healthcare utilisation, cardiometabolic factors. Participants were identified from electronic health records. Sustained PPI exposure was defined as ≥2 years of cumulative use within 3 years of the first prescription. Fine-Gray models treating death as a competing event estimated subdistribution hazard ratios (sHRs) for AMD within 10 years; cause-specific Cox models provided complementary HRs. Analyses included histamine H2-receptor antagonists (H2-antagonists) as a negative exposure control and active comparator.
Results
Incident AMD occurred in 814/49 796 PPI exposed participants (1.64%) and 668/49 796 controls (1.34%) (overall sHR=1.24, 95% CI 1.12 to 1.38; p<0.001; HR=1.25, 95% CI 1.13 to 1.38, p<0.001). Proportional sHRs were non-constant; sHRs were 1.09 (95% CI 0.91 to 1.29) before 2.5 years and 1.33 (95% CI 1.17 to 1.51) thereafter. For non-exudative AMD, 640 PPI exposed (1.29%) and 510 unexposed (1.02%) participants developed disease (overall sHR=1.28, 95% CI 1.14 to 1.44; p<0.001, HR=1.29, 95% CI 1.14 to 1.44, p<0.001). Among those with non-exudative AMD, 190/2503 PPI-exposed participants (7.59%) and 179/2503 controls (7.15%) progressed to exudative AMD (sHR and HR=1.03, 95% CI 0.84 to 1.26; p=0.802 and p=0.801, respectively). No association was observed between H2-antagonists and AMD (sHR=1.02, 95% CI 0.88 to 1.19; p=0.784; HR=1.03, 95% CI 0.88 to 1.20, p=0.729). Results were consistent in the active-comparator analysis (reciprocal overall sHR=1.24, 95% CI 1.04 to 1.48; p=0.016; HR=1.25, 95% CI 1.05 to 1.49, p=0.012).
Conclusion
Sustained PPI exposure was associated with a modestly higher AMD risk, whereas no association was observed with H2-antagonists. The consistency of the active-comparator findings supports the potential specificity of this association and warrants further investigation to determine its clinical implications.
Keywords: Drugs, Epidemiology, Pharmacology, Retina
WHAT IS ALREADY KNOWN ON THIS TOPIC
Proton-pump inhibitors (PPIs) are widely used, often long term in older adults. Experimental evidence suggests PPIs can disrupt lysosomal acidification, a pathway implicated in age-related macular degeneration (AMD) pathogenesis.
WHAT THIS STUDY ADDS
In this new-user, population-based cohort study of adults aged ≥60 years, sustained PPI exposure was associated with a 24% higher risk of incident AMD compared with matched unexposed controls, whereas no association was observed for H2-receptor antagonists.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
These findings support independent replication and reinforce the importance of careful risk–benefit assessment of prolonged PPI therapy in older adults.
Introduction
Age-related macular degeneration (AMD) is one of the leading causes of blindness in people older than 50 years projected to affect 288 million people by 2040.1 Risk factors include increasing age, female sex, white race, smoking, obesity, cardiovascular disease, stroke, hypertension and dyslipidaemia.2 3 Given its impact on functional independence, mental health, and healthcare utilisation, identifying modifiable risk factors for AMD is a major public health priority.4
Dysfunction and deterioration of the retinal pigment epithelium (RPE) has been implicated in AMD pathogenesis.5 RPE cells contain mitochondria, phagosomes and lysosomes. Lysosomes are the main digestive organelle,6 responsible for degrading both extracellular and intracellular material.7 They perform autophagy, in which waste products are encapsulated in an autophagosome and fused with a lysosome for degradation and recycling.7 With ageing, the autophagic and phagocytic abilities of the RPE decline, resulting in accumulation of waste proteins, lipids and lipofuscin within RPE cells.6 8 Lipofuscin accumulation increases lysosomal pH, further impairing degradative capacity and contributing to AMD pathology.9 Impaired autophagy drives RPE degeneration and atrophy.6 8
Vacuolar-type H+-ATPases (V-ATPase) establish lysosomal acidification by transporting protons from the cytoplasm to the lumen.6 Inhibition of this ATP-driven proton pump increases lysosomal pH, inhibits hydrolases and causes lysosomal dysfunction.10 This leads to accumulation of undigested waste products, lipofuscin formation and deposition of undegraded material, contributing to AMD pathology.8 9
Proton-pump inhibitors (PPIs) are prodrugs that typically become activated in the low pH of the parietal cell canalicular membrane of the stomach, but may become activated in other acidic environments, including lysosomes.11 PPIs are effective at inhibiting V-ATPases in vitro and in vivo.12 They are among the most used medications in adult care.13 Approximately 100 million prescriptions are dispensed annually in the USA, and long-term use is frequent in older adults.14 Internationally, more than 35 million prescriptions were issued in the UK in 2022–2023, and 2.87 billion PPI tablets were sold in Germany during 2020–2021.15 16
Given the widespread long-term use of PPIs and the central role of lysosomal dysfunction in AMD, we sought to evaluate whether sustained PPI exposure is associated with incident AMD. Our primary aim was to assess the association between sustained PPI exposure and incident AMD in older adults aged ≥60 years. Secondary objectives were to evaluate associations with incident non-exudative AMD and progression from non-exudative to exudative AMD.
Materials and methods
Study design
This was a retrospective observational cohort study using electronic health record data from patients who attended Mass General Brigham (MGB) hospitals. These include Massachusetts General Hospital, Brigham and Women’s Hospital, Spaulding Rehabilitation Hospital, Newton-Wellesley Hospital, Faulkner Hospital, North Shore Medical Center, McLean Hospital, Mass Eye and Ear, Wentworth-Douglass Hospital and Dana-Farber Cancer Institute. Summarised methodology is present in online supplemental appendix 1, table 1. This study adhered to the Declaration of Helsinki and Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.17
The Complete Patient Data Science Repository (PDSR), a data repository comprising over 5 billion clinical facts, was used to construct the study cohort. The project was reviewed by the Mass General Brigham Institutional Review Board (MGB IRB; Human Research Affairs), which determined that it did not constitute human subjects research because it exclusively used retrospectively collected, fully deidentified data. Therefore, formal IRB approval and individual informed consent were not required. Patients and the public were not involved in the design, conduct, reporting or dissemination plans of this retrospective cohort study using deidentified electronic health record data. Data were extracted in November 2025. Identifiers defining exposures, outcomes and variables are shown in online supplemental appendix 1, table 2.
Inclusion criteria
Patients initiating a PPI between 1 January 2005 to 1 September 2022 were required to accumulate ≥730 days of PPI supply within a 3-year window (database lock: 1 September 2025). The index date was the date on which 730 cumulative exposure days were reached; participants were restricted to those aged ≥60 years at that time. Unexposed individuals had no prior antacid exposure (PPI or histamine H2 receptor antagonist (H2-antagonist)); their index date was the first encounter at age ≥60 years. Participants were required to have ≥6 months of observable follow-up before the first PPI prescription for exposed and before the index encounter for unexposed. A 30-day landmark (defined as index date +30 days) minimised immortal time bias, excluding individuals with AMD, death, end of observation or (for exposed participants) end of PPI exposure before the landmark.
Matching
We used 1:1 propensity-score matching including sex, race, smoking, myopia, diabetes, hypertension, ischaemic heart disease, cerebrovascular disease, obesity and dyslipidaemia. To address structural differences between sustained PPI exposed participants and the unexposed population, we enforced exact matching on: age, Charlson Comorbidity Index (CCI; indicator of overall health status),18 number of Mass Eye and Ear (MEE) visits and number of inpatient admission days in the 2 years before index and index year. The age component was removed from the CCI to avoid duplicating age, which was matched separately. Matching used nearest-neighbour algorithms without replacement, using a calliper of 0.2 on the logit of the propensity score. Propensity scores were estimated using multivariable logistic regression incorporating the prespecified baseline covariates described above. Covariate balance before and after matching was assessed using standardised mean differences (SMDs) (online supplemental appendix 1, figure 1), with an absolute SMD≤0.10 indicating acceptable balance.
Outcomes
The primary outcome was incident AMD within 10 years of the landmark. Follow-up continued until AMD, death, end of observation, PPI exposure ending (exposed only) or administrative censoring at 10 years. Fine-Gray competing-risk models were used to estimate subdistribution hazard ratios (sHRs) with death treated as a competing event. Kaplan–Meier and cumulative incidence curves were generated, and cause-specific Cox proportional hazards models with robust standard errors were fitted as a robustness check.
Secondary outcomes included incident non-exudative AMD and progression from non-exudative to exudative AMD. For the progression analysis, patients with non-exudative AMD were classified as exposed if they had ≥6 months of PPI exposure in the year preceding diagnosis; unexposed participants had no prior PPI or H2-antagonist exposure. A 180-day landmark was applied to ensure adequate exposure ascertainment and minimise reverse causation. Propensity score matching was performed without exact matching to preserve sample size. The same covariates as in the primary analysis were included, with variables used for exact matching in the primary analysis instead incorporated into the propensity score model.
Sensitivity analyses
For all sensitivity analyses, the same analytic framework as in the primary analysis was used. Residual imbalance was addressed stepwise by applying a tighter calliper of 0.1 and adjusting for any persistently imbalanced covariates in the Fine-Gray and Cox models.
Sex-stratified analyses. Sex-stratified analyses were conducted separately in males and females. Effect modification by sex was evaluated by including a multiplicative interaction term between PPI exposure and sex in the Fine-Gray sHRs model.
Negative control exposure. H2-antagonists (ranitidine, cimetidine, famotidine, nizatidine) chosen because they share similar indications with PPIs but lack a hypothesised biological relationship with AMD.
Active comparator analysis. We compared PPI with H2-antagonist exposed participants using identical exposure definitions, to mitigate confounding by indication, healthy-user bias and differences in healthcare surveillance. To prevent cross-exposure contamination, patients were excluded from the PPI group if they had ever received an H2-antagonist, and vice versa. Because the PPI and H2-antagonist cohorts were more clinically comparable than the antacid-unexposed group, exact matching on all covariates was not required. We performed exact matching on index year to control for secular trends in prescribing patterns, diagnostic practices and AMD detection over time, and on MEE visit category to ensure comparable baseline ophthalmic healthcare utilisation and opportunity for AMD ascertainment. Remaining covariates were incorporated into the propensity score model. We additionally performed a sensitivity analysis restricted to patients with ≥1 MEE visit in the 2 years prior to index (defined as the date 730 days of cumulative exposure was reached) to reduce misclassification from undiagnosed prevalent AMD.
Negative outcome analysis. International Classification of Diseases (ICD) category macular cyst, hole or pseudohole, selected because it affects a similar older population, is detected through the same retina care pathways, and is captured within the same subspecialty coding structure without a known relationship to PPI exposure.
Statistical analyses
Statistical analyses were undertaken using R project V.4.4.0 (2024-04-24), RStudio using the MatchIt, survival and cmprsk packages. All statistical tests were two-sided and CIs were reported at the 95% level. A p value<0.05 was considered statistically significant.
Continuous variables were summarised using mean and SD when symmetrically distributed and median and IQR when skewed. Categorical variables were summarised using counts and frequencies. For binary categorical variables, the SMD was calculated as the difference in proportions divided by the pooled SD. For categorical variables with more than two levels, an overall SMD was calculated by treating the categories as non-redundant indicator variables and using the Mahalanobis distance between the corresponding proportions.
No imputation was performed. Age and index year were complete for all participants. Binary clinical covariates were coded as present when documented in the electronic health record and absent otherwise. Healthcare-utilisation variables were assigned a value of 0 when no MEE visits or inpatient days were recorded during the specified baseline period.
Multicollinearity among variables included in the propensity-score matching specification was assessed using variance inflation factors and dimension-adjusted generalised variance inflation factors for multilevel categorical variables; all values were below 2.0 (online supplemental appendix 1, table 3). Multicollinearity testing was not applicable to the Cox and Fine-Gray outcome models because PPI exposure was the only predictor included in these models.
The proportional sHRs assumption was assessed by adding an exposure-by-log(follow-up time) interaction; its coefficient and p value tested whether the exposure effect changed over time. When this test indicated non-proportionality, flexible competing-risk regression was used to estimate the exposure coefficient non-parametrically over time. Constancy of this coefficient was evaluated using resampling-based Kolmogorov-Smirnov (KS) and Cramér-von Mises (CvM) tests, which assess the maximum and overall departures from a constant effect, respectively. The resulting coefficient curve and confidence bands were examined to identify an approximate transition in the association. This curve-derived threshold was then used to fit a two-period Fine-Gray model, from which separate sHRs were estimated before and after the transition. An exposure-by-period interaction tested whether the two period-specific sHRs differed. Overall sHRs were retained as summaries across follow-up, while period-specific sHRs characterised associations where nonproportionality was detected.
The proportional-hazards assumption for Cox models was assessed using scaled Schoenfeld residuals. Where these suggested potential non-proportionality, an exposure-by-log(follow-up time) interaction was used as a formal test. When non-proportionality was confirmed, period-specific HRs were estimated using the same approach as the Fine-Gray analysis.
Results
We included 49 796 PPI exposed participants and 49 796 matched controls (online supplemental appendix 2, figure 1). The median age was 66 years (IQR, 63–72) in the PPI versus 66 years (IQR 60–72) in the unexposed group. Females comprised 56% of the cohort, and nearly 90% of participants were White. Among comorbidities, hypertension was the most common (PPI: 61% vs controls: 62%), followed by dyslipidaemia (PPI: 38% vs controls: 35%) (table 1). For the exposed group, the median cumulative duration of PPI exposure during the 3-year assessment window was 3.0 years (IQR, 2.9–3.0). At index, most patients were receiving omeprazole (38 704; 77.7%), followed by pantoprazole (9182; 18.4%) and lansoprazole (1447; 2.9%). By strength, the most common exposures were omeprazole 20 mg (27 570; 55.4%), pantoprazole 40 mg (7274; 14.6%) and omeprazole 40 mg (5617; 11.3%) (online supplemental appendix 2, figure 2). The median time from the landmark to the last recorded PPI prescription end date was 3.6 years (IQR, 1.59–6.31 years).
Table 1. Demographics and baseline characteristics of study participants stratified by PPI exposure.
| Characteristic | Unexposed (n=49 796) | PPI exposed (n=49 796) | SMD |
|---|---|---|---|
| Demographics | |||
| Age at index, median (IQR), years | 66 (60–72) | 66 (63–72) | 0.085* |
| Age group, n (%) | <0.001 | ||
| 60–64 | 20 697 (41.6) | 20 697 (41.6) | – |
| 65–69 | 11 985 (24.1) | 11 985 (24.1) | – |
| 70–74 | 7678 (15.4) | 7678 (15.4) | – |
| ≥75 | 9436 (18.9) | 9436 (18.9) | – |
| Female, n (%) | 27 480 (55.2) | 28 280 (56.8) | 0.032 |
| Index year, n (%) | <0.001 | ||
| <2015 | 12 516 (25.1) | 12 516 (25.1) | – |
| 2015–2019 | 17 112 (34.4) | 17 112 (34.4) | – |
| 2020+ | 20 168 (40.5) | 20 168 (40.5) | – |
| Race, n (%) | |||
| White | 43 231 (86.8) | 43 446 (87.2) | 0.013 |
| Black | 1998 (4.0) | 1968 (4.0) | 0.003 |
| Asian | 1030 (2.1) | 1144 (2.3) | 0.016 |
| Comorbidities | |||
| Hypertension, n (%) | 30 741 (61.7) | 30 396 (61.0) | 0.014 |
| Lipid disorder, n (%) | 17 333 (34.8) | 19 091 (38.3) | 0.073 |
| Ischaemic heart disease, n (%) | 8415 (16.9) | 8825 (17.7) | 0.022 |
| Diabetes, n (%) | 6893 (13.8) | 6340 (12.7) | 0.033 |
| Smoker, n (%) | 5870 (11.8) | 6684 (13.4) | 0.049 |
| Obesity, n (%) | 5653 (11.4) | 7240 (14.5) | 0.095 |
| Cerebrovascular disease, n (%) | 2317 (4.7) | 2359 (4.7) | 0.004 |
| Myopia, n (%) | 414 (0.8) | 597 (1.2) | 0.037 |
| CCI, n (%) | <0.001 | ||
| 0 | 29 686 (59.6) | 29 686 (59.6) | – |
| 1 | 6866 (13.8) | 6866 (13.8) | – |
| 2 | 6311 (12.7) | 6311 (12.7) | – |
| ≥3 | 6933 (13.9) | 6933 (13.9) | – |
| Healthcare utilisation | |||
| MEE visits, n (%) | <0.001 | ||
| 0 | 40 821 (82.0) | 40 821 (82.0) | – |
| 1–3 | 5799 (11.6) | 5799 (11.6) | – |
| ≥4 | 3176 (6.4) | 3176 (6.4) | – |
| Inpatient visit days, n (%) | <0.001 | ||
| 0 | 44 939 (90.2) | 44 939 (90.2) | – |
| 1–3 | 3104 (6.2) | 3104 (6.2) | – |
| ≥4 | 1753 (3.5) | 1753 (3.5) | – |
Age is presented as median (IQR) because of its skewed distribution; the SMD was calculated using group means and pooled SD.
CCI, Charlson comorbidity index; MEE, Mass Eye and Ear; SMD, standardised mean difference.
PPI exposure versus unexposed matched controls
Over a median follow-up of 3.8 years (IQR, 1.5–7.1 years), 814/49 796 PPI exposed participants (1.64%) developed AMD, versus 668/49 796 controls (1.34%). The incidence rate was 3.70 versus 3.03 per 1000 person-years, respectively (overall sHR=1.24, 95% CI 1.12 to 1.38, p<0.001; HR=1.25, 95% CI 1.13 to 1.38; p<0.001). The proportional sHRs assumption was not met (exposure-by-log-time interaction p<0.001). The coefficient curve (online supplemental appendix 2, figure 3), consistent with the cumulative-incidence curves in figure 1, suggested that the association strengthened after approximately 2.5 years. The sHR was 1.09 (95% CI 0.91 to 1.29; p=0.351) before 2.5 years and 1.33 (95% CI 1.17 to 1.51; p<0.001) thereafter, although the exposure-by-period interaction did not reach statistical significance (p=0.065; online supplemental appendix 2, table 1). The proportional-hazards assumption was satisfied for the cause-specific Cox model (p=0.313; online supplemental appendix 2, table 2).
Figure 1. Cumulative incidence of AMD and death over 10 years after the landmark date among sustained PPI exposed participants and matched unexposed controls. AMD, age-related macular degeneration; PPI, proton-pump inhibitors.
The Kaplan-Meier curves are shown in figure 2A. No interaction by sex was identified (p=0.910). Among females (online supplemental appendix 2, table 3, SMDs <0.1), 484/27 310 PPI exposed participants (1.77%) and 416/27 310 controls (1.52%) developed AMD (overall sHR=1.20, 95% CI 1.05 to 1.37, p=0.006; HR=1.21, 95% CI 1.06 to 1.37, p=0.005). Among males (online supplemental appendix 2, table 4, SMDs <0.1), 287/21 081 PPI exposed participants (1.36%) and 235/21 081 controls (1.12%) developed AMD (overall sHR=1.24, 95% CI 1.04 to 1.47, p=0.016; HR=1.24, 95% CI 1.05 to 1.48, p=0.013).
Figure 2. Kaplan-Meier curves with risk tables for incident AMD in (A) PPI versus unexposed, (B) H2-antagonist versus unexposed and (C) PPI versus H2-antagonist exposed. AMD, age-related macular degeneration; H2-antagonist, histamine-2 receptor antagonist; PPI, proton-pump inhibitor.
There was no significant overall difference in AMD risk across PPI molecules (global p=0.189). Of the 814 AMD events in the PPI cohort, the largest number occurred among patients receiving omeprazole 20 mg (n=444), reflecting the distribution of index exposures (online supplemental appendix 2, table 5).
Non-exudative AMD developed in 640/49 796 PPI exposed participants (1.29%) and 510/49 796 controls (1.02%), corresponding to incidence rates of 2.91 and 2.31 per 1000 person-years, respectively. The overall sHR for non-exudative AMD was 1.28 (95% CI 1.14 to 1.44, p<0.001), and the HR was 1.29 (95% CI 1.14 to 1.44; p<0.001). No interaction by sex was identified (p=0.810). For the non-exudative to exudative AMD analysis (online supplemental appendix 2, table 6, SMDs <0.1), 190/2503 PPI-exposed participants (7.59%) and 179/2503 controls (7.15%) developed exudative AMD. The incidence rate was 27.7 versus 27.1 per 1000 person-years, respectively (table 2; sHR and HR were 1.03, 95% CI 0.84 to 1.26; p=0.802 and p=0.801, respectively). Kaplan-Meier curves for incident non-exudative AMD and for progression to exudative AMD are presented in online supplemental appendix 2, figures 4 and 5.
Table 2. Fine-Gray competing-risk and Cox proportional hazards model results for incident AMD, incident non-exudative AMD and progression from non-exudative to exudative AMD.
| Comparison and outcome | Exposed (events/ total) | Matched controls (events/ total) | Overall sHR (95% CI) |
sHR p value | Overall HR (95% CI) |
HR p value |
|---|---|---|---|---|---|---|
| PPI vs unexposed matched controls | ||||||
| PPI vs unexposed—AMD incidence | 814/49 796 (1.64%) | 668/49 796 (1.34%) | 1.24 (1.12 to 1.38) |
<0.001 | 1.25 (1.13 to 1.38) |
<0.001 |
| PPI vs unexposed—non-exudative AMD | 640/49 796 (1.29%) | 510/49 796 (1.02%) | 1.28 (1.14 to 1.44) |
<0.001 | 1.29 (1.14 to 1.44) |
<0.001 |
| PPI vs unexposed—progression to exudative AMD | 190/2503 (7.59%) | 179/2503 (7.15%) | 1.03 (0.84 to 1.26) |
0.802 | 1.03 (0.84 to 1.26) |
0.801 |
| H2-antagonists vs unexposed matched controls | ||||||
| H2 vs unexposed—AMD incidence | 322/26 427 (1.22%) | 340/26 427 (1.29%) | 1.02 (0.88 to 1.19) |
0.784 | 1.03 (0.88 to 1.20) |
0.729 |
| H2 vs unexposed—non-exudative AMD | 261/26 427 (0.99%) | 272/26 427 (1.03%) | 1.03 (0.87 to 1.23) |
0.694 | 1.04 (0.88 to 1.23) |
0.646 |
| H2 vs unexposed—progression to exudative AMD | 74/1095 (6.76%) | 81/1095 (7.40%) | 0.90 (0.65 to 1.23) |
0.493 | 0.89 (0.65 to 1.21) |
0.457 |
| PPI vs H2-antagonist | ||||||
| PPI vs H2 (reciprocal)—AMD incidence | 274/16 449 (1.67%) | 224/16 449 (1.36%) | 1.24 (1.04 to 1.48) |
0.016 | 1.25 (1.05 to 1.49) |
0.012 |
| PPI vs H2 (reciprocal)—non-exudative AMD | 226/16 449 (1.37%) | 181/16 449 (1.10%) | 1.27 (1.04 to 1.54) |
0.017 | 1.28 (1.05 to 1.55) |
0.013 |
| PPI vs H2 (reciprocal)—progression to exudative AMD | 17/260 (6.54%) | 15/260 (5.77%) | 1.09 (0.55 to 2.16) | 0.819 | 1.11 (0.54 to 2.25) | 0.784 |
HRs and sHRs represent overall estimates across the 10-year follow-up. Proportional hazards and proportional subdistribution hazards assumptions were formally assessed. Where proportionality was violated, period-specific estimates are reported in online supplemental appendix 2, table 1. No exposure×period interaction was significant.
AMD, age-related macular degeneration; H2, histamine-2 receptor antagonist; PPI, proton-pump inhibitor; sHR, subdistribution HR.
H2-antagonist exposure versus unexposed matched controls
We matched 26 427 H2-antagonist exposed participants to 26 427 controls (online supplemental appendix 2, table 7, SMDs <0.1). Over a median follow-up of 3.1 years (IQR, 1.4–5.8 years), 322 H2-antagonist exposed participants (1.22%) and 340 controls (1.29%) developed AMD, corresponding to incidence rates of 3.25 and 3.40 per 1000 person-years, respectively. The sHR was 1.02 (95% CI 0.88 to 1.19, p=0.784; figure 2B), and the HR was 1.03 (95% CI 0.88 to 1.20; p=0.729; figure 3). Results were consistent for non-exudative AMD; 261/26 427 (0.99%) H2-antagonist exposed versus 272/26 427 (1.03%) controls (sHR=1.03, 95% CI 0.87 to 1.23, p=0.694). For progression from non-exudative to exudative AMD (online supplemental appendix 2, table 8), 74/1095 (6.76%) H2-antagonist exposed participants and 81/1095 (7.40%) controls developed exudative AMD (sHR=0.90, 95% CI 0.65 to 1.23, p=0.493).
Figure 3. Summary of overall HRs and subdistribution HRs for AMD outcomes across exposure comparisons. Estimates are shown for PPI versus unexposed, H2-antagonist versus unexposed and PPI versus H2-antagonist exposed. Squares represent sHRs, triangles represent cause-specific HRs. AMD, age-related macular degeneration; H2-antagonist, histamine-2 receptor antagonist; PPI, proton-pump inhibitor; sHR, subdistribution HR.
PPI exposure versus H2-antagonist exposure
We matched 16 449 H2-antagonist exposed participants to 16 449 PPI exposed participants (online supplemental appendix 2, table 9; SMDs <0.1). During follow-up, 274 PPI (1.67%) and 224 H2-antagonist exposed participants (1.36%) developed AMD. The incidence rate was 5.22 among PPI versus 4.09 per 1000 person-years among H2-antagonist exposed participants (reciprocal overall sHR=1.24, 95% CI 1.04 to 1.48; p=0.016; reciprocal HR=1.25, 95% CI 1.05 to 1.49; p=0.012). The Kaplan-Meier curves are shown in figure 2C. No interaction by sex was observed (p=0.220). For non-exudative AMD, 226 PPI (1.37%) and 181 H2-antagonist exposed participants (1.10%) developed disease (reciprocal overall sHR=1.27, 95% CI 1.04 to 1.54; p=0.017; online supplemental appendix 2, figure 6). Among 260 individuals with non-exudative AMD in each group (online supplemental appendix 2, table 10, SMDs <0.1), 17 PPI (6.54%) and 15 H2-antagonist exposed participants (5.77%) progressed to exudative AMD (reciprocal sHR=1.09, 95% CI 0.55 to 2.16; p=0.819).
In the MEE-restricted sensitivity analysis, we identified 3719 H2-antagonist exposed participants and 3719 matched PPI exposed participants (online supplemental appendix 2, table 11). During follow-up, 105 PPI (2.82%) and 71 H2-antagonist exposed participants (1.91%) developed AMD. The incidence rate was 8.93 versus 5.97 per 1000 person-years, respectively (reciprocal sHR=1.49, 95% CI 1.10 to 2.01; p=0.010; reciprocal HR=1.49, 95% CI 1.10 to 2.00; p=0.009).
Negative outcome analysis
When using macular cyst, hole or pseudohole as the outcome, event counts were similar between PPI exposed participants (197/49 636) and matched controls (197/49 636). No association was observed (sHR=1.00, 95% CI 0.82 to 1.22, p=0.965; HR=1.01, 95% CI 0.83 to 1.23, p=0.939).
Discussion
In this population-based cohort study, PPI exposure was associated with a 24% higher risk of incident AMD. In absolute terms, AMD developed in 814 PPI exposed participants (1.64%) compared with 668 matched controls (1.34%), corresponding to an absolute risk increase of 0.29 percentage points, or approximately three additional AMD cases per 1000 PPI exposed participants over a median follow-up of 3.8 years. Although this individual-level absolute risk increase was modest, the high prevalence of long-term PPI use suggests potential population-level relevance. An estimated 28 million individuals across 23 countries use PPIs,19 while in the USA, PPI use is reported by 8.6% of adults and approximately 15% of those aged ≥55 years, equivalent to an estimated 15.7 million older adults in 2026.13 20 Thus, the potential association between long-term PPI use and AMD warrants further investigation.
PPIs have been associated with other age-related neurodegenerative disorders.7 21–26 Several large observational studies have reported increased risks of dementia among PPI exposed participants with effect estimates ranging from approximately 1.20 to 1.44.21 22 24 Evidence suggests a cumulative exposure relationship, with greater dementia risk observed after more than 4.4 years of PPI exposure.23 This parallels our sHR findings, in which the association with AMD appeared stronger approximately 2.5 years after the landmark, by which point exposed participants had accumulated at least approximately 4.5 years of PPI exposure. This temporal pattern may be consistent with an association requiring prolonged exposure, delayed manifestation or both.
Lysosomal dysfunction, including impaired acidification and degradative capacity, has been implicated in Alzheimer’s disease, Parkinson’s disease and AMD.8 27–29 PPIs can elevate lysosomal pH by inhibiting the lysosomal V-ATPase and impair lysosomal function, disrupting both phagocytosis and autophagy.11 30 Elevation of lysosomal pH in RPE cells impairs degradation of photoreceptor outer segments and produces lipofuscin-like waste accumulation, a central feature of early AMD.9 31 Conversely, restoring lysosomal acidity rescues degradative function, underscoring that acidification failure itself is pathogenic.32 PPIs have also produced dose-dependent inhibition of lysosomal enzymes across several cell types, an effect not observed with the H2-antagonist cimetidine.11 Reduced expression of the V-ATPase subunit V1G1 has been associated with α-synuclein accumulation in Parkinson’s disease,33 while lansoprazole has increased β-amyloid levels in experimental cell and mouse models of Alzheimer’s disease.34 Although these findings provide indirect biological plausibility, the relevance of these experimental observations to PPI exposure and AMD in humans remains uncertain. Lysosomal alkalinisation therefore represents a potential mechanism that warrants further investigation rather than a pathway established by the present study.
Another hypothesised pathway through which PPIs may influence neurodegenerative processes is via modulation of the gut microbiome.35–38 PPI exposure is associated with reduced microbial diversity, depletion of short-chain fatty acid-producing commensals, impaired gut-barrier integrity and systemic inflammation; these changes appear less pronounced with H2-antagonists.35 36 Such microbiome shifts favour expansion of proinflammatory genera and reduction of anti-inflammatory species, altering gut-barrier integrity.36 37 These changes elevate circulating cytokines that prime microglia towards a proinflammatory state.37 39 Gut dysbiosis has similarly been documented in AMD,40 with studies across populations demonstrating reduced microbial diversity and enrichment of proinflammatory taxa and pathways.40 41 Greater exposure to broad-spectrum antibiotics, another potent disruptor of the gut microbiome, has also been associated with increased odds of AMD.42 Dysbiosis-driven systemic inflammation and microglial priming could potentially amplify the innate immune dysregulation involved in AMD.43
This study has limitations. First, the retrospective design introduces the possibility of residual confounding. Although SMDs <0.1, considered acceptable by epidemiologic standards,44 small numerical differences in baseline characteristics between cohorts may have influenced the observed associations. Moreover, propensity-score matching can balance only measured covariates. Unavailable or incompletely captured factors, including socioeconomic status, family history of AMD, lifestyle factors and genetic susceptibility, could have either inflated or attenuated the estimated association. Second, differential healthcare utilisation may influence outcome detection. To address this, we performed an active-comparator analysis; however, this analysis may not fully eliminate confounding by indication. Patients prescribed PPIs may have more severe or refractory acid-related disease, different indications for treatment or greater medication and comorbidity burdens than patients prescribed H2-antagonists. Third, exposure misclassification is possible because prescribing and dispensing does not ensure medication adherence. We minimised this by implementing a sustained-use definition, reducing the influence of sporadic or non-adherent use. Additionally, our new-user design, which required at least 6 months without exposure before cohort entry, reduced the chance of inadvertently classifying prevalent users as new initiators. However, given these medications are also available without prescription, exposed patients may have been misclassified as unexposed. Because administration frequency was not available, and PPIs may be taken as multiple lower-strength tablets, a dose–response analysis was not performed. Fourth, residual systematic bias may exist. To evaluate this, we conducted a negative-outcome analysis using macular hole, a condition unrelated to gastric acid suppression but detected through the same ophthalmic care pathways as AMD. The absence of an association between PPI exposure and this negative outcome suggests minimal influence of unmeasured systematic bias. Fifth, outcome misclassification remains possible because AMD was identified using ICD-10 codes, which may be imperfectly coded in routine clinical practice, and the absence of baseline eye examinations may have omitted patients with disease at baseline.
Future research should validate our findings in independent, diverse populations and examine whether associations vary by PPI type, dose and duration of exposure. Prospective studies incorporating detailed medication adherence and ophthalmic imaging are needed to clarify the temporal relationship and determine whether PPI exposure is associated with specific AMD phenotypes. Furthermore, although our current analysis did not demonstrate an association between PPI exposure and progression to exudative AMD, future work should explore this phenotypic transition further to determine whether extended exposure durations or specific genetic risk profiles influence neovascular pathways. Experimental studies should also investigate the potential roles of impaired lysosomal acidification and gut microbiome disruption in mediating this association.
Conclusion
In this population-based retrospective cohort study, sustained PPI exposure was associated with a modestly increased risk of AMD; however, the observational design precludes causal inference. Replication in independent populations and prospective studies is needed to assess generalisability and clarify the temporal relationship, alongside experimental investigation of hypothesised mechanisms such as lysosomal dysfunction and microbiome-mediated effects. Given the widespread long-term use of PPIs, these findings reinforce the importance of reassessing the indication and risk–benefit balance of prolonged therapy in older adults, but do not alone justify changing clinically indicated treatment at this stage.
Supplementary material
Footnotes
Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Provenance and peer review: Not commissioned; externally peer reviewed.
Patient consent for publication: Not applicable.
Ethics approval: Not applicable.
Data availability free text: The data used in this study are derived from the Mass General Brigham Patient Data Science Repository (PDSR), a secure clinical research environment that prohibits the export or public sharing of individual-level data. Therefore, the raw data underlying this analysis cannot be shared. Deidentified aggregate summary data supporting the findings of this study are provided in the supplemental material. Analytic code used for this study is available from the corresponding author upon reasonable request.
Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting or dissemination plans of this research.
Data availability statement
No data are available.
References
- 1.Wong WL, Su X, Li X, et al. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta-analysis. Lancet Glob Health. 2014;2:e106–16. doi: 10.1016/S2214-109X(13)70145-1. [DOI] [PubMed] [Google Scholar]
- 2.Age-Related Eye Disease Study Research Group Risk factors associated with age-related macular degeneration. A case-control study in the age-related eye disease study: Age-Related Eye Disease Study Report Number 3. Ophthalmology. 2000;107:2224–32. doi: 10.1016/S0161-6420(00)00409-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mitchell P, Liew G, Gopinath B, et al. Age-related macular degeneration. The Lancet. 2018;392:1147–59. doi: 10.1016/S0140-6736(18)31550-2. [DOI] [PubMed] [Google Scholar]
- 4.Taylor DJ, Hobby AE, Binns AM, et al. How does age-related macular degeneration affect real-world visual ability and quality of life? A systematic review. BMJ Open. 2016;6:e011504. doi: 10.1136/bmjopen-2016-011504. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Dieguez HH, Romeo HE, Alaimo A, et al. Oxidative stress damage circumscribed to the central temporal retinal pigment epithelium in early experimental non-exudative age-related macular degeneration. Free Radic Biol Med. 2019;131:72–80. doi: 10.1016/j.freeradbiomed.2018.11.035. [DOI] [PubMed] [Google Scholar]
- 6.Sinha D, Valapala M, Shang P, et al. Lysosomes: Regulators of autophagy in the retinal pigmented epithelium. Exp Eye Res. 2016;144:46–53. doi: 10.1016/j.exer.2015.08.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Koh J-Y, Kim HN, Hwang JJ, et al. Lysosomal dysfunction in proteinopathic neurodegenerative disorders: possible therapeutic roles of cAMP and zinc. Mol Brain. 2019;12:18. doi: 10.1186/s13041-019-0439-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Golestaneh N, Chu Y, Xiao Y-Y, et al. Dysfunctional autophagy in RPE, a contributing factor in age-related macular degeneration. Cell Death Dis. 2017;8:e2537. doi: 10.1038/cddis.2016.453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Holz FG, Schütt F, Kopitz J, et al. Inhibition of lysosomal degradative functions in RPE cells by a retinoid component of lipofuscin. Invest Ophthalmol Vis Sci. 1999;40:737–43. [PubMed] [Google Scholar]
- 10.Valapala M, Wilson C, Hose S, et al. Lysosomal-mediated waste clearance in retinal pigment epithelial cells is regulated by CRYBA1/βA3/A1-crystallin via V-ATPase-MTORC1 signaling. Autophagy. 2014;10:480–96. doi: 10.4161/auto.27292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Liu W, Baker SS, Trinidad J, et al. Inhibition of lysosomal enzyme activities by proton pump inhibitors. J Gastroenterol. 2013;48:1343–52. doi: 10.1007/s00535-013-0774-5. [DOI] [PubMed] [Google Scholar]
- 12.Spugnini EP, Citro G, Fais S. Proton pump inhibitors as anti vacuolar-ATPases drugs: a novel anticancer strategy. J Exp Clin Cancer Res. 2010;29:44. doi: 10.1186/1756-9966-29-44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Bergin E, Zylberberg HM, Lebwohl B, et al. Trends in use of proton pump inhibitors among adults in the United States from 1999 to 2018. Pharmacoepidemiol Drug Saf. 2023;32:1406–10. doi: 10.1002/pds.5676. [DOI] [PubMed] [Google Scholar]
- 14.Villars JA, Anderson TS, Yabes JG, et al. Proton Pump Inhibitor Use Exceeding the U.S. Food and Drug Administration Approved Treatment Duration for Patients With Peptic Ulcer Disease: A Retrospective Cohort Study. Pharmacoepidemiol Drug Saf. 2025;34:e70152. doi: 10.1002/pds.70152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Plehhova K, Wray J, Aluko P, et al. Prescribing practices for proton pump inhibitors among primary care physicians in England: an evaluation. BJGP Open. 2025;9:BJGPO.2024.0059. doi: 10.3399/BJGPO.2024.0059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Andrawes M, Andrawes W, Das A, et al. Proton Pump Inhibitors (PPIs)—An Evidence-Based Review of Indications, Efficacy, Harms, and Deprescribing. Medicina (B Aires) 2025;61:1569. doi: 10.3390/medicina61091569. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol. 2008;61:344–9. doi: 10.1016/j.jclinepi.2007.11.008. [DOI] [PubMed] [Google Scholar]
- 18.Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–83. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 19.Shanika LGT, Reynolds A, Pattison S, et al. Proton pump inhibitor use: systematic review of global trends and practices. Eur J Clin Pharmacol. 2023;79:1159–72. doi: 10.1007/s00228-023-03534-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.U.S. Bureau of Labor Statistics Population level - 55 yrs. & over [LNU00024230], retrieved from FRED, Federal Reserve Bank of St. Louis. 2026. [22-Apr-2026]. https://fred.stlouisfed.org/series/LNU00024230 Available. Accessed.
- 21.Pourhadi N, Janbek J, Jensen‐Dahm C, et al. Proton pump inhibitors and dementia: A nationwide population‐based study. Alzheimer’s Dementia. 2024;20:837–45. doi: 10.1002/alz.13477. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Gomm W, von Holt K, Thomé F, et al. Association of Proton Pump Inhibitors With Risk of Dementia. JAMA Neurol. 2016;73:410. doi: 10.1001/jamaneurol.2015.4791. [DOI] [PubMed] [Google Scholar]
- 23.Northuis CA, Bell EJ, Lutsey PL, et al. Cumulative Use of Proton Pump Inhibitors and Risk of Dementia. Neurology. 2023;101 doi: 10.1212/WNL.0000000000207747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Zhang P, Li Z, Chen P, et al. Regular proton pump inhibitor use and incident dementia: population-based cohort study. BMC Med. 2022;20:271. doi: 10.1186/s12916-022-02478-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Lai S-W, Liao K-F, Lin C-L, et al. Association between Parkinson’s disease and proton pump inhibitors therapy in older people. Biomedicine (Taipei) 2020;10:1–4. doi: 10.37796/2211-8039.1048. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kim J-H, Oh J-K, Kim Y-H, et al. Association between Proton Pump Inhibitor Use and Parkinson’s Disease in a Korean Population. Pharmaceuticals (Basel) 2022;15:327. doi: 10.3390/ph15030327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Notomi S, Ishihara K, Efstathiou NE, et al. Genetic LAMP2 deficiency accelerates the age-associated formation of basal laminar deposits in the retina. Proc Natl Acad Sci USA. 2019;116:23724–34. doi: 10.1073/pnas.1906643116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Dehay B, Bové J, Rodríguez-Muela N, et al. Pathogenic lysosomal depletion in Parkinson’s disease. J Neurosci. 2010;30:12535–44. doi: 10.1523/JNEUROSCI.1920-10.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lee J-H, Nixon RA. Autolysosomal acidification failure as a primary driver of Alzheimer disease pathogenesis. Autophagy. 2022;18:2763–4. doi: 10.1080/15548627.2022.2110729. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Bergmann M, Schütt F, Holz FG, et al. Inhibition of the ATP-driven proton pump in RPE lysosomes by the major lipofuscin fluorophore A2-E may contribute to the pathogenesis of age-related macular degeneration. FASEB J. 2004;18:562–4. doi: 10.1096/fj.03-0289fje. [DOI] [PubMed] [Google Scholar]
- 31.Peters S, Reinthal E, Blitgen-Heinecke P, et al. Inhibition of lysosomal degradation in retinal pigment epithelium cells induces exocytosis of phagocytic residual material at the basolateral plasma membrane. Ophthalmic Res. 2006;38:83–8. doi: 10.1159/000090268. [DOI] [PubMed] [Google Scholar]
- 32.Liu J, Lu W, Reigada D, et al. Restoration of lysosomal pH in RPE cells from cultured human and ABCA4(-/-) mice: pharmacologic approaches and functional recovery. Invest Ophthalmol Vis Sci. 2008;49:772–80. doi: 10.1167/iovs.07-0675. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Li Y, Wang Y, Kou L, et al. Plasma exosomes impair microglial degradation of α‐synuclein through V‐ATPase subunit V1G1. CNS Neurosci Ther. 2024;30 doi: 10.1111/cns.14738. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Badiola N, Alcalde V, Pujol A, et al. The proton-pump inhibitor lansoprazole enhances amyloid beta production. PLoS One. 2013;8:e58837. doi: 10.1371/journal.pone.0058837. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Zhu J, Sun C, Li M, et al. Compared to histamine-2 receptor antagonist, proton pump inhibitor induces stronger oral-to-gut microbial transmission and gut microbiome alterations: a randomised controlled trial. Gut. 2024;73:1087–97. doi: 10.1136/gutjnl-2023-330168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Imhann F, Bonder MJ, Vich Vila A, et al. Proton pump inhibitors affect the gut microbiome. Gut. 2016;65:740–8. doi: 10.1136/gutjnl-2015-310376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Cattaneo A, Cattane N, Galluzzi S, et al. Association of brain amyloidosis with pro-inflammatory gut bacterial taxa and peripheral inflammation markers in cognitively impaired elderly. Neurobiol Aging. 2017;49:60–8. doi: 10.1016/j.neurobiolaging.2016.08.019. [DOI] [PubMed] [Google Scholar]
- 38.Kwon D, Zhang K, Paul KC, et al. Diet and the gut microbiome in patients with Parkinson’s disease. NPJ Parkinsons Dis. 2024;10:89. doi: 10.1038/s41531-024-00681-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Kim M-S, Kim Y, Choi H, et al. Transfer of a healthy microbiota reduces amyloid and tau pathology in an Alzheimer’s disease animal model. Gut. 2020;69:283–94. doi: 10.1136/gutjnl-2018-317431. [DOI] [PubMed] [Google Scholar]
- 40.Zhang Y, Wang T, Wan Z, et al. Alterations of the intestinal microbiota in age-related macular degeneration. Front Microbiol. 2023;14 doi: 10.3389/fmicb.2023.1069325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Sun J, Xu J, Yang B, et al. Effect of Clostridium butyricum against Microglia-Mediated Neuroinflammation in Alzheimer’s Disease via Regulating Gut Microbiota and Metabolites Butyrate. Mol Nutr Food Res. 2020;64:e1900636. doi: 10.1002/mnfr.201900636. [DOI] [PubMed] [Google Scholar]
- 42.Moir J, Hyman M, Wang J, et al. The Association of Antibiotic Use and the Odds of a New-Onset ICD Code Diagnosis of Age-Related Macular Degeneration: A Large National Case-Control Study. Invest Ophthalmol Vis Sci. 2023;64:14. doi: 10.1167/iovs.64.12.14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Xiao J, Zhang JY, Luo W, et al. The Emerging Role of Gut Microbiota in Age-Related Macular Degeneration. Am J Pathol. 2023;193:1627–37. doi: 10.1016/j.ajpath.2023.04.006. [DOI] [PubMed] [Google Scholar]
- 44.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28:3083–107. doi: 10.1002/sim.3697. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
No data are available.