Comparing proton pump inhibitors with histamin-2 receptor blockers regarding the risk of osteoporotic fractures: a nested case-control study of more than 350,000 Korean patients with GERD and peptic ulcer disease

We used the Korean National Health Insurance Service (NHIS) claim data and the biannual regular health examination data. South Korea has a mandatory social health insurance system covering approximately 97% of the entire population. Almost all medical data, including diagnostic codes, procedures, prescription drugs, and personal information, are included in the NHIS database. The NHIS also conducts a standardized national health screening program every 2 years for all insured persons aged over 40 years and every year for employee subscribers aged 20 years or older. National health screening data include anthropometric measurements, laboratory tests, and social history.

Patients with the International Classification of Diseases (ICD)-10 codes for PUD or GERD (K21, K25, K26, K27, and K28) and a prescription history of PPIs or histamine-2 receptor antagonists (H2RAs) from January 1, 2006 to December 31, 2015 were extracted. Subsequently, information on demographic characteristics and lifestyle or health behavior information (age, sex, residence, income level, height, weight, body mass index, smoking, alcohol drinking, exercise, etc.) were extracted.

All personal identification numbers have been encrypted by converting them into scrambled numbers before data processing, to comply with the privacy rules of the health insurance transfer and privacy laws. Therefore, the use of anonymized information exempted this study from the need for written consent. This study was conducted in accordance with the tenets of the Helsinki Declaration. All research procedures and ethical aspects were approved by the Institutional Review Board of the National Evidence-based Healthcare Collaborating Agency (approval no. 17–011).

We included patients with GERD and PUD, who were prescribed PPI or H2RA at least once with this diagnosis. We did not include non-users of gastric acid inhibitor, who have never used PPI or H2RA during the observation period. Comparing PPI users with non-users of a gastric acid inhibitor can lead to an overestimation of PPI risks. Individuals who have never used PPI or H2RA during the observation period are likely to be in relatively good health and may not have peptic ulcer disease or gastritis, reducing calcium absorption and affecting fracture risk. Therefore, we evaluated the risk of osteoporotic fractures associated with PPI use compared to H2RA use to avoid overestimating the PPI effect. The cohort entry date was defined as the date when the first drug (PPI or H2RA) was prescribed. We defined the case group as patients who developed osteoporotic fractures during the follow-up period and the control group as those who did not develop osteoporotic fractures. A nested case-control study was performed to examine the risk of osteoporotic fracture between the PPI use group and the H2RA use group.

A total of 35,520,188 patients were taking either PPI or H2RA for PUD and GERD from January 1, 2006 to December 31, 2015. Only new users who did not have a history of being prescribed with PPI or H2RA within the first year of cohort entry were included. The exclusion criteria were as follows: age < 50 years on cohort entry; medical service use because of cancer (ICD-10 code: C, T), acquired immunodeficiency syndrome (B20–24), or osteoporosis (M80, 81) within 1 year before entering the cohort; or a diagnosis of osteoporotic fracture (spine: S220, 221, 320, 327, M484, 485; proximal humerus: S422, 423; femur: S72; distal radius: S525, 526) within 1 year before cohort entry. Patients with missing values in the matching variables were also excluded. To use the variables included in the health screening data as covariates, those who did not undergo a national health screening were excluded. After the exclusions, 2,388,137 patients remained.

The outcome index (osteoporotic fracture) was defined using the codes requested from the health insurance claims data. First, patients with osteoporosis with pathological fracture were selected (disease classification code: M80). Second, patients who had a diagnosis of osteoporosis (M81, 82) and experienced an osteoporotic fracture after study entry were collected. Osteoporotic fractures were defined as fractures of 4 parts: spine (S220, 221, 320, 327, M484, 485), proximal humerus (S422, 423), femur (S72), and distal radius (S525, 526). Patients in whom osteoporosis was diagnosed before osteoporotic fracture or within 90 days after osteoporotic fracture were defined as the osteoporotic fracture group. The patients who had osteoporotic fractures during the follow-up period (from cohort entry to 2015) was 78,465.

The index date was the time of osteoporotic fracture. When the osteoporotic fracture occurred more than twice during the study period, the time of the first fracture was used in the analysis. Corresponding to the osteoporotic fracture group, the control group was selected, through 1:5 matching according to sex, age group with a 5-year interval, and use of bisphosphonates, among patients who did not develop osteoporotic fractures within the same cohort entry and follow-up period. Finally, we had 59,240 matched cases and 296,200 controls. The process of selecting and excluding patients for the final analysis is shown in Fig. 1.

The criteria for PPI use were ‘any use’, ‘duration of use’, and ‘regular use’. We defined any use of PPI as being prescribed PPI at least once during the study period. We also defined PPI duration as the cumulative days of use calculated by adding the prescriptions days regardless of continuity. The cumulative days of PPI were categorized as < 30, 30–59, 60–89, 90–179, 180–364, and ≥ 365 days. The recent 1 year prior to osteoporotic fracture was divided into four quarters. We identified the number of quarters with PPI use from one to four. The use of PPI over four quarters was defined as “regular use.”

We considered various covariates associated with the development of osteoporotic fractures due to the use of PPI. These include socioeconomic and demographic factors such as sex, age, residence area, income, and anthropometric and lifestyle factors such as body mass index, drinking, smoking, and exercise. We obtained information on health behaviors through a standardized self- administered questionnaire collected on the day of the national health screenings. According to smoking status, subjects were classified into non-smokers, former smokers, and current smokers. Heavy alcohol drinkers were defined as those who consumed ≥30 g/day of alcohol. Physical activity was classified as follows according to the frequency of ≥20 min of strenuous exercise (none, ≤ four times per week, and ≥ five times per week). The national health screenings also included measurements of body size. Body mass index was calculated by dividing the body weight by the square of the height (kg/m2). Income levels were classified into quintile groups using the NHIS qualification data.

These covariates also consisted of concomitant diseases and concomitant medications that could directly or indirectly affect osteoporotic fracture development. The use of bisphosphonates, glucocorticoids, anticonvulsants, hormone replacement therapy, warfarin, heparin, antacids, selective serotonin reuptake inhibitors, benzodiazepines, and tricyclic antidepressants was considered concomitant medication use. All were defined using the drug code (main ingredient code). Concomitant diseases were defined using the ICD-10 codes and included chronic obstructive pulmonary disease (J44, 45), testicular dysfunction (E29.1), hypothalamic dysfunction (E23.0), hyperthyroidism (E05), hyperparathyroidism (E21), Cushing’s syndrome (E24), hyperprolactinemia (E22.1), vitamin D deficiency (E55.9), idiopathic hypercalciuria (E83.5), diabetes (E11–14 and diabetic agents), anorexia nervosa (F50.0, 50.1), systemic lupus erythematous (M32), hypertension (I10–13, I15, and use of antihypertensive agents), intestinal absorption disorder (K90), inflammatory bowel disease (K50, 51), chronic kidney disease (N18), and secondary amenorrhea (N91.1). Another variable considered as a covariate was the Charlson Comorbidity Index (CCI). Comorbidities and medications used were defined as at least one claim for the relevant drug code and diagnostic code for the period from 1 year before the date of cohort entry to the end of follow-up.

The categorical variables were presented as frequencies and percentages of the patients. For continuous variables, means and standard deviations were presented.

Conditional logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) to investigate the association of PPI use and osteoporotic fracture compared to the exclusive H2RA use. The crude model was not adjusted for covariates, and the full-adjusted model included adjustments for age, sex, body mass index, alcohol drinking, smoking, exercise, concomitant medications, and diseases as covariates. Subgroup analyses were conducted by sex and age. Interaction tests were performed to determine whether the associations differed by sex and age. We also calculated p values for linear trends of osteoporotic fracture risk according to the increase in the cumulative duration and the regular use of PPI. Two-sided P-values < 0.05 were considered statistically significant. All statistical analyses were performed using SAS software (version 9.2; SAS Institute, Cary, NC, USA).

Our results are consistent with those of previous studies suggesting that PPIs are associated with osteoporotic fracture. Yang et al. reported that the use of PPI for > 1 year in patients aged ≥50 years was associated with the risk of hip fracture and that the longer the duration of treatment, the greater the association with fracture [5]. A study in Denmark reported that the use of PPI was associated with all osteoporotic fractures, including hip and vertebral fractures, within 1 year [15]. In a Canadian study, PPI exposure for > 7 years was associated with osteoporotic fracture, and there was no significant association between the use of PPI for ≤6 years and the risk of osteoporotic fracture [31]. They did not adjust the major risk factors for fracture, such as low body mass index, alcohol consumption, and smoking, so their results may differ from our study results. Corley et al. reported that the use of PPI in adults aged > 18 years was associated with an increased risk of hip fracture, and suggested that the risk of osteoporotic fracture was reduced when the drug was discontinued [6]. In a recent study in the United States, the risk of hip fracture in men aged > 45 years has been reported to increase with longer use of PPI [21]. Another study reported that PPI use increases the risk of osteoporotic fracture in elderly people aged > 65 years [20]. It has also been reported that PPI is associated with fractures in hemodialysis patients at high risk of fracture [24].

Most of the studies reported the existence of an association between PPI use and osteoporotic fracture risk. However, it was difficult to make definite conclusions because the target country, target patient group, definition of PPI use, fracture site, and covariates varied across different studies. In many studies, information on medication use was dependent on the patients’ memory recall. Further, most of the studies were Western studies and only a few were reported in Asia. A study in Taiwan indicated that the use of > 29 defined daily doses of PPI was associated with hip fracture development [32]. Asian studies are important because Asian people differ from Westerners in bone mineral density, osteoporotic fracture prevalence [30]. Also, it has been reported that the pharmacodynamics of PPI differ between Asian and Western populations because many Asians have fewer gastric parietal cells and poor metabolism of PPI [29]. Therefore, this study was conducted on almost the entire Korean population, using the exact date of prescription in the claims data and with adjustments for a large number of covariates. In addition, we analyzed the effects of any use, cumulative use, and regular use of PPI on the osteoporotic fracture risk.

There are several mechanisms by which PPIs increase the risk of osteoporotic fractures. First, PPIs reduce gastric acid by inhibiting H⁺/K⁺-ATPase at the last stage of gastric acid production. Calcium solubility plays an important role in calcium absorption, and the acidic environment of the gastrointestinal tract promotes the release of ionized calcium from insoluble calcium salts [33]. PPIs inhibit acid secretion and increase the pH, which slows down the dissolution of calcium salt and, consequently, prevents calcium absorption. The lack of calcium in the bone may increase the risk of osteoporotic fracture in the long term [34]. A second mechanism may be the reduction of vitamin B12 absorption induced by PPI [7]. The acidity of the stomach is important for the absorption of vitamin B12 [35]. It has been reported that inhibition of gastric acid by PPI may inhibit the absorption of vitamin B12 [36]. Vitamin B12 deficiency can cause homocysteinemia, leading to bone loss through increased osteoclastic activity, reduced osteoblast activity, reduced collagen linking, and decreased bone formation [37]. Vitamin B12 is also an essential co-factor for DNA synthesis and cellular energy production [38]. Continuous vitamin B12 deficiency can cause reversible megaloblastic aplastic anemia and demyelinated neuropathy leading to gait and weakness, which can cause neurological dysfunctions such as cognitive and visual impairment [39]. Therefore, vitamin B12 deficiency due to PPI use may be a risk factor for osteoporotic fractures. The third mechanism may be the increase in gastrin levels. Gastric acid is reduced by the use of PPIs, and the gastrin level is increased by the feedback mechanism. Hypergastrinemia due to PPI use has been reported to cause hyperplasia of the parathyroid gland, and increased parathyroid hormone could induce a decrease in bone mineral density [40, 41]. Also, PPIs interfere with the active transport of magnesium and can cause osteoporotic fractures [42].

The first strength of this study is that we included a large number of patients using data from almost all Koreans. This study is also meaningful because it adds to the few Asian studies in the literature to date. Second, our study had a comparatively long observation period from 2006 to 2015. Third, the accuracy of drug use information over a long period is very high because the drug prescriptions were identified from NHIS claims data. Fourth, by using claims data and national health checkup data, various drugs and diseases affecting fracture could be used as covariates. In addition, the database allowed us to use accurate information without recall bias, which is common in retrospective studies.

Our study has some limitations. First, there was a possibility of residual confounding factors despite the attempt to adjust for factors influencing fracture. Due to the lack of data, we were unable to use lifestyle factors as covariates. The sources of residual confounding are also the lack of information regarding the disease severity and symptom severity of GERD and PUD, and the quality of care patients received in general by their healthcare providers. It is possible that healthcare providers using H2RA instead of PPI were more aware of the risk of PPI and were better educated regarding the management of geriatric patients. Second, we were not able to use medical records or imaging test results. Bone mineral density, a diagnostic parameter for osteoporosis, was also not available because this information is not included in the claims data. Third, our study outcome may include some fractures due to other causes, and may not include some osteoporotic fractures. The misclassification bias might not be eliminated.