**2. Materials and Methods**

### *2.1. Study Design and Population*

As stated, we used data from Taiwan's NHIRD; Taiwan implemented National Health Insurance (NHI) in 1995, and the database contains NHI claims data for more than 99% of Taiwan's population and provides a means to explore the risk factors or effects of disease interventions. This study used a retrospective cohort design with NHIRD data from 2000 to 2017. Diseases were diagnosed using the International Classification of Diseases, Ninth [Tenth] Edition, Clinical Modification (ICD-9(10)–CM). This study was approved by the Institutional Review Board of Chung Shan Medical University Hospital (approval number CS2-20036).

### *2.2. Study Population*

Records from 2000 to 2017 were collected from the NHIRD, and the study population included patients with hypertension (ICD-9-CM codes 401-405 and ICD-10-CM codes I10-I15). Patients must have used a hypertension medication within 1 year of diagnosis, including beta-blockers (anatomic therapeutic chemical [ATC] code: C07), CCBs (ATC code: C08), alpha-blockers (ATC code: C02CA), and ACEIs or ARBs (ATC code: C09).

We identified a total of 498,738 patients with hypertension. After accounting for the excluded conditions, we finally included 216,867 patients for data analysis. We divided patients into two groups: hypertension with thiazide use (HT-with-thiazide) and hypertension without thiazide use (HT-without-thiazide). A total of 18,620 patients were placed in the HT-without-thiazide group and 198,247 patients were placed in the HT-withoutthiazide group. The index date was defined as the first day 365 days after the diagnosis of hypertension. This setting is because we defined patients who continuously used thiazide for 1 year as stable users and assumed that these stable users would continue to use thiazide during follow-up.

#### *2.3. Characteristics, Comorbidities, and Outcomes*

We identified baseline demographic characteristics (as reported within 365 days before the index date), such as age and sex, and the comorbidities and medications of each participant to evaluate their health status. Baseline comorbidities included diabetes mellitus, hyperlipidemia, ischemic heart disease, cerebrovascular accident, abnormal renal function, chronic obstructive pulmonary disease (COPD), cancer, and depressive disorders. Baseline medications included beta-blockers, CCBs, ACEIs, ARBs, corticosteroids, non-steroidal anti-inflammatory drugs (NSAIDs), proton-pump inhibitors, and hormonal medications.

The study outcome was defined as the diagnosis of a fracture, differentiated into skull, spine and trunk, upper limb, lower limb, and pathological fractures. All study individuals were followed up from the index date to either the study outcome, the occurrence of fracture due to car accident, death, or the end of the study (31 December 2017).

### *2.4. Statistical Analysis*

Statistical analysis was performed using SAS version 9.4 (SAS Institute, Cary, NC, USA), and a *p*-value of <0.05 was considered statistically significant. HT-with-thiazide patients were matched with HT-without-thiazide patients by age (±1 year) and sex at a 1:4 ratio. To reduce potential confounding bias due to measured factors, 1:1 propensity score matching (PSM) was performed by using the greedy nearest neighbor algorithm and noreplacement matching with a caliper width of 0.01; matched variables included birth year, sex, age (±1 year) at the index date, index year, comorbidities, and medication. The absolute standardized difference (ASD) was used to evaluate the differences in covariates between the two study groups; an ASD value of <0.1 indicated that the item was balanced between the groups.

Categorical data were presented as numbers and percentages, and the differences in categorical variables were compared using the chi-square test. Incidence rates with corresponding confidence intervals (CI) and crude hazard ratios (HRs) were calculated using Poisson regression. After the proportional hazards assumption was tested, a Cox proportional hazards model analysis was performed to estimate the HRs for mortality and 95% CIs. Cumulative fracture probabilities were assessed using Kaplan–Meier analysis, in which statistical significance was determined using the results of a log-rank test.

#### **3. Results**

#### *3.1. Characteristics of the Participants*

The selection flow chart for this study is presented in Figure 1. A total of 18,593 patients were included in the HT-with-thiazide group, and an additional 74,372 patients were sex and age (±1 years) matched at a 1:4 ratio to form the HT-without-thiazide control group (Supplementary Table S1). After 1:1 PSM matching by sex, age, urbanization level of place of residence, income, comorbidities, and medications, 18,483 HT with-thiazide and HT without-thiazide participants were obtained; 58% of the patients were male, and 41% were female. Over 57% were aged 46 to 60 years. The baseline characteristics are presented in Table 1.


**Figure 1.** Study flowchart of patient selection.


**Table 1.** Baseline characteristics among study groups.

ASD: Absolute Standardized Difference; COPD: Chronic Obstructive Pulmonary Disease; CCBs: Calcium Channel Blockers; ACEIs: Angiotensin-Converting Enzyme Inhibitors; ARB: Angiotensin Receptor Blockers; NSAIDs: Non-Steroidal Anti-Inflammatory Drugs; PPIs: Proton Pump Inhibitors.

#### *3.2. Risk of Fracture between HT-with-Thiazide and HT-without Thiazide Group*

As Table 2 indicates, the incidence densities of fractures (per 1000 person-months) were 1.83 (95% CI: 1.76–1.90) and 1.97 (95% CI: 1.94–2.01) in the sex- and age-matched HT-with-thiazide and HT-without-thiazide cohorts, respectively (Supplementary Table S2). After PSM, the values were 1.82 (95% CI: 1.76–1.89) and 1.99 (95% CI: 1.92–2.06) in the HT-with-thiazide and HT-without thiazide cohorts, respectively, and the aHR for the HR-with-thiazide group was 0.93 (95% CI: 0.88–0.98). Kaplan–Meier survival analysis revealed a significantly lower cumulative incidence of fractures in the HT-with-thiazide group (log-rank test; *p* = 0.0012; Figure 2).


**Table 2.** Incidence density of fracture.

\* per 1000 person-months. † Adjusted variables including age, sex, comorbidities and medication.

**Figure 2.** Kaplan–Meier curves of the cumulative proportions of fracture in the use of thiazide and without the use of thiazide.

Multiple cox regression revealed that the aHR of fractures for the HT-with-thiazide group was significantly lower than that of the HT-without-thiazide group. Other significant risk factors for fractures were sex; age; urbanization level; income; the comorbidities of diabetes mellitus, hyperlipidemia, cerebrovascular accident, abnormal renal function, and COPD; and the medications of corticosteroids and NSAIDs (Supplementary Table S3).

In Figure 3, subgroup analysis revealed that the aHR for women and men was 0.98 (95% CI: 0.91–1.05) and 0.86 (95% CI: 0.80–0.93), respectively, and the HR-with-thiazide group exhibited a significantly reduced risk compared to the HR-without-thiazide group. For the age 41 to 50 years group, the aHR was 0.87 (95% CI: 0.76–0.99). For the with-CCBs group, the aHR was 0.92 (95% CI: 0.86–0.90). For the with-ACEI or ARBs group, the aHR was 0.91 (95% CI: 0.86–0.96).

**Figure 3.** Subgroup analysis of adjusted hazard ratio for fractures in patients with thiazide compared with patients without thiazide. \* per 1000 person-months; aHR (adjusted Hazard Ratio) adjusted variables including age, sex, comorbidities and medication; CCBs: Calcium Channel Blockers; ACEIs: Angiotensin-Converting Enzyme Inhibitors; ARB: Angiotensin Receptor Blockers.

#### **4. Discussion**

Hypertension is a common chronic disease among older adults. Several hypotheses accounting for the pathophysiology have been proposed. One of the hypotheses concerns maladaptation to a high-salt diet. Hypertension may be a physiological response intended to excrete excess salt. Natriuresis is a key treatment for hypertension [24]. Diuretics could remove excessive salt to achieve hypertension control. A review concluded that the use of a low-dose thiazide reduced all mortality and morbidity in adult patients with moderate-to-severe primary hypertension [25].

Older people are also more susceptible to osteoporotic fractures and even low-energy injuries due to low bone density and being prone to falls. Hip fracture is also one of the risks associated with higher mortality and morbidity among older adults. The cumulative mortality rate for 12 months was 33%, and the 1-year mortality rate increased significantly by 2% per year [26] Several cost-effective pharmacologic treatments may be used to prevent fractures, such as vitamin D and calcium supplements and bisphosphonates [27]. Denosumab is an effective option for preventing fracture, which reduces bone resorption to achieve bone mass preservation [28].

Our study aimed to determine whether thiazide could be used to reduce fracture risk in patients with hypertension. This has remained controversial in prior research. A meta-analysis determined that the use of thiazide was associated with reduced risk in case-control studies but not in cohort studies. That study concluded that the use of thiazide might not protect against fractures [29]. However, some studies have different conclusions. A recent study using a Swedish database concluded that the use of bendroflumethiazide or hydrochlorothiazide could reduce the risk of hip fractures [30]. That study also noted that the choice of antihypertensive can influence fracture risk. A recent meta-analysis including 22 observational studies concluded the use of thiazide was associated with a lower fracture rate of the hip [31]. A meta-analysis that compared patients with diuresis with and without thiazide determined that thiazide reduced overall fracture risk by 14% and hip fracture risk by 18% [32].

In the present population-based retrospective cohort study, we determined that patients with hypertension and thiazide had a lower overall fracture risk. The adjusted risk ratio for the HT-with-thiazide group was 0.926, indicating thiazide has a protective effect against fracture for patient hypertension.

A meta-analysis found the protective effect on fracture risk is associated with the long duration and continuity of thiazide use [33]. Similar to our study, the protective effect seems more significant after 72 months from Kaplan–Meier curves (Figure 2). Another metaanalysis concluded that the association between the use of thiazide and risk of osteoporotic fracture is not significant. However, this study also concluded that the different general status of a patient might have different levels of benefit [34]. This study gave us an important implication that we should find the specific population that could benefit from thiazide diuretics for preventing fracture. In our study, subgroup analysis was used to determine whether thiazide could protect against fractures in perimenopausal and postmenopausal women.

A retrospective cohort study using the MJ Health Database in Taiwan indicated that the mean age at menopause for women in Taiwan is 50.2 [35]. The adjusted fracture risk ratio was 0.88 in a subgroup of females aged 41–50. (95% C.I. 0.73–1.06) The possible explanation is sample size was not large enough to meet statistical significance. The adjusted fracture risk ratio was 0.87 in a group aged 41–50. (95% C.I. 0.76–0.99) A possible explanation for this is as follows: As people reach their forties and start to lose protection from hormones, thiazide can compensate to improve calcium reabsorption and increase bone density. However, during their fifth decade or older, people completely lose the protection of hormones. Bone mineral loss accelerates and worsens, and thiazide does not provide sufficient protection.

Bone mineral density (BMD) measurements provide a snapshot of bone health, including the presence of osteoporosis and the risk of fractures. Two randomized controlled trials revealed that thiazide achieved significant benefits in BMD in postmenopausal women [36,37]. To sum up, thiazide could produce a positive effect on BMD, but increased BMD does not necessarily indicate reduced fracture risk.

A meta-analysis revealed that the risk of osteoporotic fractures among individuals with hypertension was higher than individuals without hypertension, with an odds ratio of 1.33 [38]. A longitudinal study indicated that hypertension is an independent risk factor for fractures in women but not men [39]. Although our study observed that thiazide had a protective effect on fracture in patients with hypertension, we were unable to determine whether the protective effect originates from the control of blood pressure or thiazide facilitating calcium reabsorption.

The most common indication for thiazide is hypertension; others include edema or ascites secondary to cirrhosis or heart failure. In several studies, the research participants were not specifically individuals with hypertension [21,29,40]. Different from those studies, the participants in our study used thiazide specific for hypertension.

Our study has several limitations. First, the claims records in the NHIRD are mainly used to calculate medical unit services and service costs. Several key indicators recorded in clinical practice are not included in the database, such as the severity, type, and mechanism of patient fractures. We used PSM to eliminate possible confounding factors that were recorded in the database. Second, the retrospective cohort design of this study excluded causal inference and limited the precision of our study. Therefore, we matched the study group with the control group by propensity score to reduce this bias. Third, BMD is a useful indicator of bone health. A previous study concluded that thiazide could improve BMD [41]. However, BMD values are not recorded in the NHIRD. Further cohort studies or studies using hospital electronic medical records systems might help to clarify the relationships between BMD and the use of thiazide.

#### **5. Conclusions**

In conclusion, our study revealed that thiazide use for hypertension can reduce overall fracture risk. When deciding between antihypertensive agents, thiazide may be a favorable choice if the patient is at elevated risk of fragility fractures. Through subgroup analysis, we also determined that women over 40 years of age who used thiazide were at a reduced risk of fractures. This could be an indication for the use of antihypertensive medications in perimenopausal women.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jcm11123304/s1, Table S1: Baseline characteristics among study groups. Table S2: Incidence density of fracture. Table S3: Multiple Cox proportional hazard regression results for fracture.

**Author Contributions:** Conceptualization, C.-H.C., M.-Y.C. and C.-B.Y.; formal analysis, S.-F.Y., P.-L.L. and J.-Y.H.; writing—original draft preparation, C.-H.C., M.-Y.C. and C.-B.Y.; writing—review and editing, C.-H.C., M.-Y.C. and C.-B.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethical Review Board of the Chung Shan Medical University Hospital (CS2-20036) approved our study.

**Informed Consent Statement:** Patient consent was waived by both the National Health Insurance Administration and the Institutional Review Board of Chung Shan Medical University Hospital due to the database-processing nature of the current study.

**Data Availability Statement:** Restrictions apply to the availability of these data. Data were obtained from the National Health Insurance database and are available from the authors with the permission of the National Health Insurance Administration of Taiwan.

**Acknowledgments:** This study was partly based on data from the NHIRD provided by the NHI Administration, Ministry of Health and Welfare, and managed by the Health and Welfare Data Science Center (HWDC) in Taiwan. The interpretation and conclusions contained herein do not represent those of the NHI Administration, Ministry of Health and Welfare, or National Health Research Institutes.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

