| |
Proton-Pump Inhibitor Use Is Not Associated With Osteoporosis or Accelerated Bone Mineral Density Loss
|
| |
| |
Gastroenterology March 2010
Laura E. Targownik Corresponding Author Informationemail address, Lisa M. Lix, Stella Leung, William D. Leslie
Section of Gastroenterology, Division of Internal Medicine, University of Manitoba, Winnipeg, Manitoba, Canada
Manitoba Centre for Health Policy, Department of Community Health Sciences, Faculty of Medicine, Winnipeg, Manitoba, Canada
School of Public Health, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
Section of Nuclear Medicine, Department of Radiology, University of Manitoba, Winnipeg, Manitoba, Canada
Section of Endocrinology & Metabolism, Division of Internal Medicine, University of Manitoba, Winnipeg, Manitoba, Canada
Received 17 March 2009; accepted 12 November 2009. published online 19 November 2009.
"Although osteoporosis is a well-recognized risk factor for future fracture, approximately 50% of low-velocity fractures occur in patients without osteoporotic BMD as determined by DXA scanning.35, 36, 37 Using DXA scanning to assess for BMD provides an incomplete picture of bone strength. Bone geometry, trabecular bone composition, and microarchitecture are also important determinants of bone strength and cannot be directly assessed by DXA scanning.38, 39, 40 The vacuolar proton pump, similar in structure to those found in the stomach, can be found in osteoclasts, where they are responsible for bone turnover and modeling by assisting in acidification.41 It has been shown that PPIs are capable of blocking the osteoclast-based vacuolar proton pump, leading to decreased bone turnover"
ABSTRACT
Backgrounds & Aims
Recent studies have shown an association between proton-pump inhibitor use (PPI) and hip fracture. The mechanism by which PPI use promotes the development of hip fracture is uncharacterized. Therefore, we sought to determine whether PPI use is associated with osteoporosis or accelerated bone mineral density (BMD) loss.
Methods
We used the Manitoba Bone Mineral Density Database to determine the relationship between chronic PPI use and osteoporosis on an initial assessment of BMD and on BMD loss between successive assessments of BMD. In the cross-sectional study, cases with osteoporosis at the hip or lumbar vertebrae (T-score ≤-2.5) were matched to 3 controls with normal BMD (T-score ≥-1.0). In the longitudinal analysis, the change in BMD among PPI users and nonusers between successive BMD assessments was assessed. Conditional logistic regression and multivariate linear regression were used to obtain estimates of the association between PPI use and osteoporosis and of the annualized change in BMD associated with PPI use.
Results
PPI use was not associated with having osteoporosis at either the hip (OR, 0.84; 95% CI, 0.55-1.34) or the lumbar spine (OR, 0.79; 95% CI, 0.59-1.06) for PPI use >1500 doses over the previous 5 years. In the longitudinal study no significant decrease was observed in BMD at either site attributable to PPI use.
Conclusions
PPI use does not appear to be associated with either the presence of osteoporosis or accelerated BMD loss. The association between PPI use and hip fracture is probably related to factors independent of osteoporosis.
Abbreviations used in this paper: BMD, bone mineral density, BMI, body mass index, DPIN, Drug Programs Information Network, DXA, dual-energy x-ray absorptiometry, PPI, proton-pump inhibitors, SERM, selective estrogen receptor modulator
Osteoporosis is a common condition throughout the developed world, affecting ≤20% of women and 7% of men >50 years.1 Although uncomplicated osteoporosis is generally asymptomatic, the presence of underlying osteoporosis is a main risk factor for the development of fractures of the hip, proximal femur, spinal vertebrae, and forearm.2, 3 The case-fatality rate for hip fractures can exceed 20%, and all osteoporosis-related fractures can lead to significant long-term disability and decreased quality of life.4
Many risk factors for the development of osteoporosis have been identified, including white race, low body mass index (BMI), sedentary lifestyle, smoking, and female sex.5, 6, 7 There are also a number of medication classes whose use has been linked to higher rates of osteoporosis, including corticosteroids and antiepileptic medication.8, 9 Recently, there has been concern about the effect of proton-pump inhibitors, or PPIs, on bone mineral metabolism. This is primarily based on evidence from epidemiologic analyses showing a significant association between chronic PPI use and the risk of hip fracture.10, 11, 12 However, although there are reasonable physiologic mechanisms through which PPIs may promote the development of osteoporosis, the precise effects of PPIs on bone mineral density (BMD) are currently unknown. Thus, it is unclear whether this reported association between hip fracture and PPI use is due to a medication-induced risk in BMD or whether PPI use is merely a marker of other comorbidities that are associated with frailty. Because PPIs are among the most commonly prescribed drugs, any effect of PPI use on the development of osteoporosis would significantly affect clinical practice. Therefore, we sought to determine whether chronic PPI use is associated with an increased risk of osteoporosis based on bone mineral densitometry testing and whether continued use of a PPI leads to an acceleration in the rate of BMD decline on subsequent testing.
Discussion
Our results indicate that a history of chronic PPI use is not associated with an increased likelihood of having BMD in the osteoporotic range at either the hip or the lumbar spine. Furthermore, increasing intensity of PPI exposure is not associated with an increased risk of having osteoporosis. The lack of an association between PPI use and changes in BMD was further confirmed in the longitudinal assessment. An effect of PPIs on the rate of bone mineral loss was also not detected among subjects who were not using bisphosphonates.
The results of our study appear to be in conflict with earlier published reports that linked PPI use to the development of hip fractures. Currently, 4 published studies have shown an association between chronic PPI use and fractures of the hip,10, 11, 12, 23 and 2 of the studies show greater risk with longer duration or higher intensities of use or both.10, 11 Conversely, Kaye and Jick24 were unable to detect an effect of PPI use on the occurrence of hip fracture in the absence of other risk factors, which was also shown by Corley.23 This raises the suspicion that the detected association between PPIs and hip fracture may be due to the presence of unmeasured confounders that are related to both PPI use and other risk factors for hip fracture.
The most commonly suggested mechanism to explain a possible causal relationship between PPI use and hip fracture is by inhibiting the intestinal absorption of calcium, leading to accelerated bone mineral loss and the eventual development of osteoporosis. However, there is little direct evidence that links the use of PPIs to the development of osteoporosis. In the only other study assessing the association of PPIs and BMD, subjects classified as PPI users did not have a significant acceleration in the rate of bone mineral loss compared with controls not using PPIs.25 However, that study lacked comprehensive data on PPI use, and drug utilization was only assessed at the time of performance of BMD, with use or nonuse assumed for the time between DXA scans. Although there is a generally accepted belief that calcium solubility and its subsequent absorption is facilitated by the presence of gastric acid,26 studies of the effects of hypochlorhydria on calcium absorption have been equivocal.27, 28, 29 PPI use in juvenile rats has been shown to decrease both bone density and peak bone mineral mass,30 although the mechanism by which this occurs is unexplained. Although reductions in BMD have been shown after a gastrectomy,31 this may not necessarily be due merely to the effects of decreased acid secretion. Other consequences of gastrectomy, most notably hypergastrinemia, which results from medically or surgically induced hypochlorhydric states, may induce parathyroid hyperplasia, which may promote bone calcium loss.32, 33 Furthermore, the performance of vagotomy without gastrectomy does not appear to induce accelerated bone mineral loss, further suggesting a limited role of reduced acid secretion in the development of osteoporosis.34
Several possible explanations are available for this apparent discrepancy between this study and those showing an association between PPI use and fracture. First, the published studies that link chronic PPI use to hip fracture have based this association on data collected retrospectively from general health care databases. Although attempts were made to control for possible confounding variables that may be associated with both PPI use and the development of fracture, it remains possible that there are risk factors for which the investigators were unable to properly control and that the association detected between PPI use and hip fracture may be spurious.
Second, it is possible that PPI use may contribute to the development of hip fractures by a mechanism that is independent of its lack of effect on calcium absorption and BMD. Although osteoporosis is a well-recognized risk factor for future fracture, approximately 50% of low-velocity fractures occur in patients without osteoporotic BMD as determined by DXA scanning.35, 36, 37 Using DXA scanning to assess for BMD provides an incomplete picture of bone strength. Bone geometry, trabecular bone composition, and microarchitecture are also important determinants of bone strength and cannot be directly assessed by DXA scanning.38, 39, 40 The vacuolar proton pump, similar in structure to those found in the stomach, can be found in osteoclasts, where they are responsible for bone turnover and modeling by assisting in acidification.41 It has been shown that PPIs are capable of blocking the osteoclast-based vacuolar proton pump, leading to decreased bone turnover. Decreased bone turnover may promote slight increases in BMD but, more importantly, may increase fracture risk by blocking the repair of microfractures and microarchitectural defects. In support of this model, in the rare disease osteopetrosis, in which the vacuolar proton pump is congenitally absent, bone remodeling is inhibited, leading to increases in BMD but recurrent fractures.42
Features of the Manitoba Bone Mineral Density Database may limit the validity of our findings. Although the database contains the results of all DXA testing obtained in Manitoba, patients who obtain BMD testing are not necessarily representative of the community at large. Therefore, systematic differences may exist in comorbidities and medication use among Manitobans who are referred for BMD testing and those who are not tested yet who may still have osteoporosis. Furthermore, even though the controls by definition had normal BMD, they were probably perceived to be at increased risk of having osteoporosis which prompted the referral for testing. However, other studies that have used the Manitoba BMD database have suggested that the distribution of BMD results in the database does not significantly differ from the expected distribution in the population at large.21, 43 Therefore, this minimizes the likelihood of a selection bias significantly influencing our findings. In addition, we used the annualized change in BMD to operationalize bone mineral loss over time, whereas the trends in bone loss associated with medication use may not be linear. However, we were able to detect significant changes in the drug classes most commonly associated with BMD gain (bisphosphonates, estrogens, SERMs) and loss (systemic corticosteroids), suggesting that our use of annualized changes in BMD was appropriate for this analysis and that our findings about the lack of association of bone mineral loss and PPI use are valid.
In conclusion, chronic PPI use does not appear to be associated with an increased risk of osteoporosis as determined by BMD testing, despite evidence linking their use to an increased risk of fractures. At this time, we would recommend that long-term PPI users should not discontinue these medications. However, because PPIs are often used in situations in which they are not indicated or other less powerful medications may be substituted, the ongoing requirement for PPI use in any individual patients should be addressed and should strive to limit their use in clinical situations in which continuous PPI therapy is absolutely indicated. Given the limitations of our analysis, further prospective studies of the effects of PPI use on the bone mineral metabolism and on the development of fracture are required before clinicians and patients can be certain that the chronic use of PPIs is not deleterious to bone health.
Results
In the cross-sectional study, 2193 subjects had evidence of osteoporosis at the hip and were matched to 5527 controls with normal hip measurements. A total of 3596 subjects had BMD measurements consistent with osteoporosis at the lumbar spine and were in turn matched to 10,257 normal controls. A description of patient demographics, BMI, medical comorbidities, and medication use is shown in Table 1.
In univariate analyses, PPI use was associated with a lower risk of osteoporosis at the lumbar spine for all levels of PPI use and was further associated with a lower risk of osteoporosis at the hip for subjects who had been dispensed >1500 standard PPI doses over the previous 5 years. However, after adjusting for potential confounders through conditional multivariable logistic regression, no association was observed between PPI use and osteoporosis at either the hip or the lumbar spine, the results of which are summarized in Table 2. Subanalyses limiting the sample to subjects >60 years of age also did not show a relationship between PPI use and having been diagnosed with osteoporosis on the basis of BMD testing (data not shown).
In the longitudinal study, we identified 2549 subjects who underwent 2 separate BMD assessments whose characteristics are displayed in Table 3. The main interval between assessments of BMD was 2.31 ± 0.5 years. The mean BMD at baseline was 1.010 g/cm2 at the lumbar spine and 0.856 g/cm2 at the total hip. Once again, we determined that the use of PPIs at a rate >0.5 standard daily doses over the time between BMD assessments did not have a statistically significant effect on the rate of BMD decline, either at the lumbar spine (change in BMD, 0.03% ± 0.22%; P > .2) or the total hip (change in BMD, -0.17% ± 0.18%; P > .2). Moreover, the use of PPIs at higher intensities (>1.0 standard daily doses) was also not associated with a significant effect on BMD at either site (change in BMD at lumbar spine, -0.08% ± 0.26%; hip, -0.09% ± 0.21%; P > .2 for both). The use of bisphosphonates, systemic estrogens, and SERMs was associated with a significant annual increase in BMD at both sites (adjusted P < .0001), whereas the use of corticosteroids was associated with a significantly greater decline in BMD at both the lumbar spine and hip in the adjusted analyses (Table 4, Table 5). When users of bisphosphonates were excluded, the use of PPIs at 0.5-1.0 doses/d and exceeding 1.0 doses/d was not associated with a significant effect on adjusted BMD change either at the lumbar spine (-0.20% ± 0.29% for 0.5-1.0 standard daily doses; -0.23% ± 0.35% for doses > 1.0 standard daily dose; P > .20 for both) or the hip (-0.11% ± 0.23% for 0.5-1.0 standard daily doses; -0.26% ± 0.28% for >1.0 standard daily doses; P ≥ .2 for both).
Materials and Methods
Two separate analyses were performed. In the cross-sectional study, we compared subjects with established osteoporosis, as determined by BMD testing, to controls with normal BMD. In the longitudinal study, we analyzed the change in BMD on serial assessments between PPI users and non-PPI users. The methodology used in each of these studies is described below in detail.
Description of Databases
Population Health Research Data Repository
The Manitoba Centre for Health Policy maintains the Population Health Research Data Repository,13, 14 a comprehensive collection of population-based health utilization data sets provided by the provincial ministry of health. The Research Data Repository contains data sets that capture information on patient demographics, all outpatient and hospital visits since April 1984, including a primary diagnosis and ≤15 secondary diagnoses (25 diagnoses from April 2004). There is no direct link to patient records or to the results of laboratory testing. The Drug Programs Information Network (DPIN) data set contains the drug name, date dispensed, dose dispensed, and quantity dispensed for all prescription drugs dispensed outside of hospitals and personal care homes since April 1995. The DPIN does not contain information on the use of any medication not directly dispensed by a pharmacist, including over-the-counter medications and vitamins (such as vitamin D and calcium). All pharmacies in Manitoba are required to have a direct link with the network, which allows real-time prescription information to be entered into the data set. All data sets are linked with a de-identified personal health information number, giving researchers the ability to construct a longitudinal medical history for any person registered with the Manitoba Health Services Insurance Plan. The Research Data Repository is well validated and has been extensively used in previous clinical research studies of PPI use and bone disease.15, 16, 17, 18 The Scientific and Ethical Advisory Board of the University of Manitoba's Health Research Ethics Board approved the study, and the Manitoba Health Information Privacy Committee approved data access.
Manitoba Bone Mineral Density Database
The reference method for measuring BMD and determining the presence of osteoporosis before fracture is dual-energy x-ray absorptiometry, or DXA scanning.19, 20 In Manitoba, the results of all clinical DXA scans performed since 1990 are maintained in the Provincial Bone Mineral Density Database, along with patient demographics, other medical comorbidities, and anthropomorphic measures, including weight, height, BMI, and soft tissue composition (percentage of lean and fat). As of December 31, 2007, the database contained >69,000 individual records. We limited analysis to scans performed between 2000 and 2007. During this time all scans were performed on 1 of 3 cross-calibrated instruments (Lunar Prodigy; GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom). Follow-up scans are performed on the same instrument >95% of the time. Bone density measurements are obtained from each of the first 4 lumbar vertebrae, the femoral neck, greater trochanter, and total hip. T-scores are calculated as the number of standard deviations that BMD is above or below the mean for healthy young adults (manufacturer USA white reference values for the lumbar spine and the third National Health and Nutrition Examination Survey white reference values for the hip). The database has been carefully validated and extensively used for clinical research, with completeness and accuracy >99%.21 Densitometers showed stable long-term performance (coefficient of variation <0.5%) and satisfactory in vivo precision (coefficient of variation, 1.7% for L1-4 and 1.1% for the total hip).22
Identifications of Cases and Controls
Cross-sectional study
Two separate sets of cases and controls were created, one for the total hip site and another for the lumbar spine. Hip cases were defined as persons who were found to have osteoporosis at the total hip site by BMD testing, whereas lumbar spine cases were those persons who were determined to have osteoporosis based on the first 4 lumbar vertebrae. Osteoporosis was defined based on a T-score of -2.5 or lower at the site of interest. Each case was matched with ≤3 controls, each of whom were found to have BMD in the normal range (ie, T-score of -1.0 or greater at the site of interest). Cases were also matched on sex and age (within 3 years) on the date of the original BMD test. All cases and controls were required to have ≥5 years of complete medication use information to be included in the analysis. Cases and controls who were using medications specifically for osteoprotection, including bisphosphonates, selective estrogen receptor modulators (SERMs), salmon calcitonin, or parathyroid hormone agonists, were excluded to minimize the effects of channeling bias. Long-term residents of personal care homes and patients receiving chronic dialysis were also excluded because medications used by these patients cannot be tracked with our prescription dispensation database.
Longitudinal study
In the longitudinal analysis, we included all subjects who underwent only 2 BMD assessments between 2001 and 2006 in which these BMD assessments were separated by 1-3 years. Any subjects who had >2 additional assessments of BMD within 3 years of the initial BMD assessments were excluded. Subjects who used osteoprotective medications were included in this analysis. The outcomes of interest were the annualized change in the BMD of the lumbar spine and the annualized change in BMD of the total hip. This outcome was obtained by computing the difference in the bone density measurements, dividing by the number of days between assessments, and then dividing by 365.25.
Determination of Drug Use
Cross-sectional study
We searched within DPIN to calculate the total cumulative dosage of PPIs over the 5 years before the date of BMD testing by finding all PPI prescriptions and determining the number of standard doses dispensed. PPI standard doses were defined as omeprazole, rabeprazole, esomeprazole (20 mg), lansoprazole (30 mg), and pantoprazole (40 mg). Patients were classified into one of the following mutually exclusive categories: no PPI use, low PPI use (<750 PPI doses over 5 years), moderate PPI use (750-1500 PPI doses over 5 years), and high PPI use (≥1500 PPI doses over 5 years). To date, PPIs have not been available over the counter in Manitoba; thus, DPIN is able to fully capture PPI dispensations. We also calculated the cumulative doses of other commonly used classes of medications thought to potentially directly affect bone mineral metabolism, including systemic estrogens, antiandrogens, systemic corticosteroids, selective serotonin receptor inhibitors, and antiepileptics. We defined subjects as having significant cumulative use of any of these classes of medications if the total amount of drugs dispensed over the previous 5 years exceeded 365 doses.
Longitudinal study
A subject was considered to be a PPI user for this analysis if the average daily dose dispensed between the 2 BMD assessments exceeded 0.5, suggesting that the subject had used PPIs on ≥50% of the days between BMDs. We also defined high-intensity PPI use if the mean standard daily dose exceeded 1.0 between consecutive BMD assessments. Subjects were also assumed to be users of any other drug of interest if the aforementioned ratio exceeded 0.5 standard daily doses. A subject was assumed to be a corticosteroid user if the total dose of corticosteroids (expressed as prednisone equivalence) exceeded 5 mg/d between the 2 BMD assessments. A subject was considered a nonuser of a medication if he or she was not dispensed any drug from a particular class between the 2 assessments of BMD. Subjects whose average daily drug dose was <0.5 but >0 were considered to be nonusers.
Determination of Other Confounders
Cross-sectional study
Subjects were classified as having a specific medical comorbidity if they had ≥2 outpatient contacts or 1 inpatient contact for a specific medical diagnosis in the 5 years preceding performance of the bone mineral densitometry testing. The comorbidities analyzed included cardiovascular disease, hypertension, diabetes, thyroid disease, chronic renal disease, cirrhosis, inflammatory bowel disease, celiac disease, alcohol and drug abuse, depression, and schizophrenia. We also tracked whether patients had an osteoporosis-related fracture (hip, spine, distal forearm) in the 5 years before BMD testing. Overall comorbidity was also operationalized through enumerating the number of Johns Hopkins Aggregate Diagnosis Groupings with which the patient was diagnosed within the year before BMD testing. Other included comorbidities were BMI categorized as underweight (BMI <20 kg/m2), normal (BMI = 20-27.9 kg/m2), overweight (BMI = 28-35.9 kg/m2), or obese (BMI ≥35 kg/m2), and annual income quintile (obtained from neighborhood level census data on median household income) as an indicator of socioeconomic status.
Longitudinal study
Comorbidities were defined as in the cross-sectional study, with the date of the second BMD test used as the index date. In addition to BMI, we also used change in BMI between BMD measurement occasions, age, and sex as covariates.
Data Analysis
Cross-sectional study
Chi-square analyses were conducted to test for differences between cases and controls on each of the demographic and clinical characteristics at the date of the BMD measurement. Odds ratios for the risk of the different intensities of PPI therapies were calculated with the use of conditional multivariable logistic regression, controlling for all previously mentioned covariates. A P value of <.05 was considered to be statistically significant. Separate subanalyses were also performed, limiting the cases and controls to subjects >60 years of age to detect whether older age altered the relationship between PPIs and osteoporosis. All analyses were conducted with SAS Version 9.1 software (SAS Institute, Cary, NC).
Longitudinal study
Chi-square testing was performed to compare PPI users and nonusers among the longitudinal cohort. We then performed logistic multivariable regression analysis to control for the effects of the aforementioned covariates, with a P value of .05 considered statistically significant. Subanalyses were also performed in which users of bisphosphonate medications were excluded to determine whether any deleterious effects of PPI use were being masked by the bone-preserving effects of bisphosphonates.
Sample Size Calculations
For the cross-sectional analyses, the determination that PPI use is associated with a ≥20% increase in the risk of osteoporosis required 1208 cases and 3624 controls, assuming a 10% rate of PPI use in controls, an α of 0.05 and a ß of 0.20. In the longitudinal study, we required 309 subjects to show an annual change in BMD of ≥0.5% per year, assuming a 10% prevalence of PPI use, 15 confounding variables, an α of 0.05 and a ß of 0.20.
| |
| |
| |
|
|
|
