Risk of fall in patients taking proton pump inhibitors: a meta-analysis

Falls is a worldwide public health concern because of the increase in aging population. Falls have negative impacts on the quality of life and healthcare costs, and lead to early institutionalization, prolonged hospitalization and injury-related deaths.1 Approximately one-third of adults aged ≥65 years old fall each year.2 Falls account for 90% of all fractures.3 The risk factors for falls include older age, frailty, polypharmacy, psychosocial characteristics and other comorbid conditions with history of falls being the paramount risk factor.4 Moreover, the age-related metabolism of medications and drug–drug interactions all contribute to the potential risk of injurious falls. Because of the variety of medical conditions in older adults, polypharmacy, defined as concomitant use of 4–8 drugs, is inevitable in this population group.5 However, not all medications increase the risk of falls; thus, Fall Risk-Increasing Drugs (FRID) is the coined term that refers to the medications well-documented to be associated with amplified fall risk.6 These include but not limited to anti-hypertensives, anti-psychotics, narcotics, sedatives and anti-cholinergics.3 Recent studies have demonstrated that long-term use of proton pump inhibitors (PPIs) increased the risk of falls7–10 which could potentially lead to increased risk of fractures. Therefore, PPI could be possibly added to the FRID list. Our study focused on PPIs which are among the multitude of medications overprescribed in the older population due to high efficacy of acid suppression and generally perceived safety profile.11 It is estimated roughly that one-tenth of ambulatory adults in the United States have been prescribed a PPI in the past 30 days.12 With the overly inappropriate use, numerous studies have raised concerns about long-term safety of PPI use. One of the possible long-term effects includes increasing the risk of osteoporosis-related fractures in the elderly, particularly hip and vertebral fractures.13 However, the results were inconsistent among different population groups.14 Avoiding precarious medications reduces the propensity of falls in older adults.15 This emphasizes the importance of prescribing medications only as indicated and weighing regularly the benefits over the risks of continuing treatments particularly long-term PPI therapy. Because of the significant magnitude of PPI utilization, the authors performed a meta-analysis to determine whether PPI is associated with increased risk of fall. For including in the meta-analysis, studies were required to meet the following criteria (i) prospective randomized controlled trials, cohort and case-control studies investigating the association of proton pump inhibitor use with the risk of fall in adult population, (ii) fall definitions were provided and (iii) proton pump inhibitor use were identified. Review articles and case reports were excluded from this meta-analysis. Two authors (K.L. and K.W.) independently searched and reviewed articles in electronic databases through MEDLINE and EMBASE databases from inception to March 2018 without any language restriction using the strategy indicated in Supplementary Material (Supplementary Material S1). Potentially relevant articles were identified for title and abstract review. Two authors (K.L. and C.T.) independently reviewed and assessed the titles and abstracts of all potential articles from the search strategy. Articles met inclusion criteria were included, and study characteristics were collected in standardized data sheets. Full text was assessed by two independent authors (K.L. and C.T.). Quality of the study was evaluated using Newcastle–Ottawa quality assessment scale. Discrepancies were addressed by principal investigator, and final decision was made by consensus between two authors (K.L. and C.T.). Full-text reviews of the eight selected studies6–10,16,17 were done by two authors (K.L. and C.T.). Characteristics of the study including first author, study location, year of publication, study design, characteristics of study sample, proton pump inhibitor use, fall documentation and confounder adjustment were retrieved. Adjusted odds ratio (OR) reported in the study was extracted. Unreported OR of interested outcome was calculated from the data provided in the study. All included studies were combined into a pooled odd ratio (OR) using a random-effects model, generic inverse variance method of DerSimonian and Laird. We tested for heterogeneity using the Q-statistic (P < 0.10 was considered significant) and I2 test. A value of I2 of 0–25% indicates insignificant heterogeneity, 26–50% low heterogeneity, 51–75% moderate heterogeneity and 76–100% high heterogeneity.18 For assessment of publication bias, we performed funnel plots and calculated Egger’s regression intercept for studies. Analyses were performed with Comprehensive Meta-Analysis 3.3 software from Biostat, Inc. The 460 relevant articles were identified from MEDLINE and EMBASE, 177 articles were removed because of duplication. As a result, 283 articles were screened. A total of 272 articles were excluded based on title and abstract review. Additionally, two articles were excluded due to lack of outcome of interest and one article did not include PPI in the study. Eight remaining studies (from seven articles) were included in our meta-analysis (Figure 1). PRISMA diagram of the literature search. Our meta-analysis included three case-control studies, four cohort studies, one cross sectional study, with a total of 367 068 patients. The characteristics are demonstrated in Table 1. All studies had moderate to high quality with a Newcastle–Ottawa quality assessment scale of 6 or more. Characteristics of included studies GERD, Gastroesophageal reflux disease; MMSE, Mini–Mental State Examination; GDS, Geriatric Depression Scale; CNS, Central nervous system. Characteristics of included studies GERD, Gastroesophageal reflux disease; MMSE, Mini–Mental State Examination; GDS, Geriatric Depression Scale; CNS, Central nervous system. Eight observational studies with a total of 367 068 patients were enrolled. There was a significant association between the use of PPIs and the risk of fall with the pooled OR of 1.27 (95% CI, 1.07–1.50, I2 = 82%; Figure 2). This association still existed with the pooled OR of 1.23 (95% CI, 1.07–1.50, I2 = 84%) after sensitivity analysis with exclusion of a cross-sectional study17 (Supplementary Material S2). Sensitivity analysis including five studies evaluating the risk of fall injury requiring hospitalization in patients using PPIs was also performed. The pooled OR for fall injury requiring hospitalization in patients using PPIs was 1.61 (95% CI, 1.18–2.20, I2 = 67%) (Supplementary Material S3). Forest plot assessing the association between PPI, H2RA and Fall. Due to high degree of heterogeneity, meta-regression analysis was performed and showed significant positive correlations between risk of fall in patients using PPIs and year of study (slopes = +0.25, P < 0.001) (Figure 3). The data on the risk of fall in patients using H2 receptor antagonists (H2RAs) were limited in three studies. The pooled OR of fall in patients using H2RAs was 0.95 (95% CI, 0.75–1.20, I2 = 11%; Figure 2). Meta-regression. To evaluate publication bias, we examined the contour-enhance funnel plots and calculated Egger’s regression of the included studies that assessed OR of fall (Supplementary Material S2). The Egger’s regression test showed publication bias with P < 0.05 for the association between proton pump inhibitor and risk of fall. If cross-sectional study was excluded,17 the Egger’s regression test showed no publication bias with P = 0.10. As for the association between H2RA and risk of fall, Egger’s regression test showed no publication bias P = 0.35. Our study was the first meta-analysis that established the significant association between the use of PPIs and increased risk of fall. In addition, there was potentially higher risk of fall among PPI users overtime. The meager 1.27 odds ratio of the pooled data should not be underestimated because of indiscriminate prescription of PPIs in all healthcare settings.19 The widespread overuse of PPI has led to superfluous healthcare expenditures and adverse events, thus highlighting the significance of our study.20 Furthermore, there was an increasing overall use of PPI in the last decade with older adults having 3.5 times more likely to be taking PPIs.21 The impact of falls on the elderly population leading to fractures especially of the hip had significant effects on morbidity and mortality. Therefore, this underscores the value of appropriate prescription and de-prescription of PPI. Through meta-regression analysis, our study was also the first to document that there was a significant positive correlation between the risk of fall in patients using PPIs and year of study. This had potential ramifications for the increasing longevity of the aging population and increasing use of PPIs. The authors hypothesized that longer duration of PPI utilization would lead to increased risk of fall. As the population ages, there is an undoubtful chance that the risk of fall would exponentially escalate over time because of longer duration exposure to PPIs. This was corroborated by the consistency of the results among different populations between the various studies. One of the strengths of our study was the representation of the residence of the majority of older adults which was community dwellers rather than nursing homes.22 Seven of the studies included focused on community-dwelling older adults while only one focused on nursing home residents.7 On the other hand, it should be noted that the rate of falls of the institutionalized elderly was twice than that of community dwellers. Currently, the mechanism of how PPIs increases fall is still unclear. It maybe multi-factorial but the extent of the contributions of each mechanism is unknown. PPIs or H2RAs decreased acid secretion but have inconsequential effects on calcium absorption.23 Furthermore, PPIs have not been shown to increase the prevalence of osteoporosis or alter bone strength.24 Inhibition of proton pumps in osteoclasts in the bone mineral matrix should curb bone resorption.25 This suggests that the increased fall risk from PPI use may be independent of bone metabolism. On the other hand, PPIs can impair vitamin B12 absorption due to increased gastric pH26 while B12 deficiency presents with gait disturbance from altered proprioception and muscle weakness leading to a greater risk of falls and cognitive decline.8,27 Several reports documented the association of PPI and hypomagnesemia28 which accounts for lethargy, cramps and weakness escalating the propensity for falls. PPIs have also been associated with visual disturbances, dizziness, functional decline and decreased performance in mobility tests in older adults.8 Because of the multiple associations between PPI and fracture risk, the authors hypothesized that PPIs increases fall ultimately leads to increased incidence of fractures. There were several limitations that should be accounted in our study. First of all, it was difficult to prove the causality between PPIs and increased risk of fall because majority of the studies in the meta-analysis were observational studies. The challenge of eliminating all confounders that can affect the results was nearly impossible to overcome. This was similar to the other associated adverse events with PPI use such as enteral infection, community-acquired pneumonia, Clostridium difficile diarrhea and chronic renal failure, where there was marked heterogeneity in the population groups studied.29,30 Given that most of the studies in our meta-analysis involved older adults with multitude of comorbidities, medications and psychosocial circumstances, it was difficult to ascertain the degree which of the variables contributed to the fall. Thus, there was conceivable confounding by indication. Second, it was challenging to ascertain the temporal profile from the exposure to PPI or H2RA to the fall because there was no clear documentation when the fall occurred. Although we were able to find a positive correlation between the risk of fall with PPI use and year of study, these observational studies suffered the drawbacks of differential and non-differential misclassification. Third, the dose-response effect was unclear because studies have not specified the dose and duration of the exposure to the said drugs. Evaluating the biologic gradient with exposure intensity was limited in this study. Fourth, the generalizability of our results was also limited because of the populations represented. Only one study referred to Asians with a total of 358 patients.6 The others were mainly in European countries.7,9,10,16,17 However, the authors deemed that ethnicity may not play a role in increasing the risk of fall with PPI use. Two studies included only women.8,17 Furthermore, there was inter-study heterogeneity of the definitions of fall among included studies. However, when analysis was performed limited to only studies evaluating fall injury requiring hospitalization, the risk of fall in patients using PPIs remained significant. Our meta-analysis did not show any increased risks of fall with H2RAs use. Generalizability of these results was constrained due to limited number of studies compared to PPIs. However, these are corroborated by findings of a recent meta-analysis by Kwok et al. involving 1 521 062 patients wherein H2RAs were not associated with increased risk of fractures.31 H2RAs maybe even protective of fractures.32 On the other hand, there are still inconsistencies with this association as reported by Corley et al. pointing out that inhibition of acid secretion increases fracture risk in patients who are already at risk for osteoporosis.33 In summary, our study demonstrated the significant association between the use of PPIs and increased risk of fall. There was a significant positive correlation between the risk of fall in patients using PPIs and year of study. In addition, there was potentially higher risk of fall among PPI users overtime. Our results supported the 2015 Updated Beers Criteria by the American Geriatrics Society (AGS), wherein PPIs were included in the list of Potentially Inappropriate Medications (PIM).34 PPIs potentially could be included in the list of Fall Risk-Increasing Drugs (FRID). Continued reassessment of medications in the older population particularly the indications for long-term PPI use is essential not only to prevent unneeded health care costs but also to avoid unnecessary harm. The authors have no commercial associations that might be a conflict of interest about this article. No funding support for this article. Conflict of interest: None declared.