Premium
Proton pump inhibitors and risk of all‐cause and cause‐specific mortality: A cohort study
Author(s) -
Brown Jeremy P.,
Tazare John R.,
Williamson Elizabeth,
Mansfield Kathryn E.,
Evans Stephen J.,
Tomlinson Laurie A.,
Bhaskaran Krishnan,
Smeeth Liam,
Wing Kevin,
Douglas Ian J.
Publication year - 2021
Publication title -
british journal of clinical pharmacology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.216
H-Index - 146
eISSN - 1365-2125
pISSN - 0306-5251
DOI - 10.1111/bcp.14728
Subject(s) - hazard ratio , medicine , confidence interval , proportional hazards model , propensity score matching , confounding , cohort , cohort study
Aim To investigate the association between proton pump inhibitors (PPIs) and both all‐cause and cause‐specific mortality. Methods We conducted a cohort study using the UK Clinical Practice Research Datalink GOLD database. We compared 733 885 new users of PPIs to 124 410 new users of H2 receptor antagonists (H2Ras). In a secondary analysis we compared 689 602 PPI new users to 1 361 245 nonusers of acid suppression therapy matched on age, sex and calendar year. Hazard ratios for all‐cause and cause‐specific mortality were estimated using propensity score (PS) weighted Cox models. Results PPI prescription was associated with increased risk of all‐cause mortality, with hazard ratios decreasing considerably by increasing adjustment (unadjusted hazard ratio [HR] 1.65, 95% confidence interval [CI] 1.62‐1.69; PS‐weighted HR 1.38, 95% CI 1.33‐1.44; high‐dimensional PS‐weighted HR 1.31, 95% CI 1.26‐1.37). Short‐term associations were observed with mortality from causes where a causal short‐term association is unexpected (eg, lung cancer mortality: PS‐weighted HR at 6 months 1.77, 95% CI 1.39‐2.25). Adjusted hazard ratios were substantially higher when compared to nonusers (PS‐weighted HR all‐cause mortality 1.96, 95% CI 1.94‐1.99) rather than H2RA users. Conclusions PPI prescription was strongly associated with all‐cause and cause‐specific mortality. However, the change in hazard ratios (a) by increasing adjustment and (b) between comparator groups indicates that residual confounding is likely to explain the association between poor health outcomes and PPI use, and fully accounting for this using observational data may not be possible.