Advertisement

Thiazide diuretics and risk of knee replacement surgery among patients with knee osteoarthritis: a general population-based cohort study

  • J. Wei
    Affiliations
    Health Management Center, Xiangya Hospital, Central South University, Changsha, Hunan, China

    Division of Rheumatology, Allergy, and Immunology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA
    Search for articles by this author
  • T. Neogi
    Affiliations
    Section of Rheumatology, Boston University School of Medicine, Boston, MA, USA
    Search for articles by this author
  • R. Terkeltaub
    Affiliations
    Department of Medicine, University of California at San Diego, San Diego, CA, USA

    VA San Diego Medical Center, San Diego, CA, USA
    Search for articles by this author
  • A.Z. Fenves
    Affiliations
    Division of Nephrology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA
    Search for articles by this author
  • C. Zeng
    Affiliations
    Division of Rheumatology, Allergy, and Immunology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA

    Department of Orthopaedics, Xiangya Hospital, Central South University, Changsha, Hunan, China
    Search for articles by this author
  • D. Misra
    Affiliations
    Section of Rheumatology, Boston University School of Medicine, Boston, MA, USA

    Department of Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA
    Search for articles by this author
  • H.K. Choi
    Affiliations
    Division of Rheumatology, Allergy, and Immunology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA
    Search for articles by this author
  • G. Lei
    Correspondence
    Address correspondence and reprint requests to: G. Lei, Department of Orthopaedics, Xiangya Hospital, Central South University, 87 Xiangya Road, Changsha, Hunan, 410008, China.
    Affiliations
    Department of Orthopaedics, Xiangya Hospital, Central South University, Changsha, Hunan, China

    Hunan Key Laboratory of Joint Degeneration and Injury, Changsha, Hunan, China

    National Clinical Research Center of Geriatric Disorders, Xiangya Hospital, Central South University, Changsha, Hunan, China
    Search for articles by this author
  • Y. Zhang
    Correspondence
    Address correspondence and reprint requests to: Y. Zhang, Division of Rheumatology, Allergy, and Immunology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, 55 Fruit Street, Boston, MA, 02114, USA.
    Affiliations
    Division of Rheumatology, Allergy, and Immunology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA
    Search for articles by this author
Open ArchivePublished:June 07, 2019DOI:https://doi.org/10.1016/j.joca.2019.05.020

      Summary

      Objective

      Thiazide diuretic use is associated with higher bone mineral density (BMD) and possibly lower serum magnesium levels than loop diuretic use, and both high BMD and low serum magnesium have been linked to high prevalent knee osteoarthritis. This study aimed to compare the risk of a clinically relevant endpoint, knee replacement (KR) surgery, among initiators of thiazide and loop diuretics.

      Design

      Among patients aged ≥50 years with a diagnosis of knee osteoarthritis in The Health Improvement Network (THIN) in United Kingdom, we conducted a propensity score-matched cohort study to examine the relation of thiazide diuretic initiation vs loop diuretic initiation to the risk of KR over 5 years.

      Results

      Among thiazide and loop diuretic initiators (n = 3,488 for each group; mean age: 73 years; female ratio: 59%), 359 (28.6/1,000 person-years) and 283 (24.1/1,000 person-years) KRs occurred during the follow-up period, respectively. The hazard ratio (HR) of KR for thiazide diuretic initiation vs loop diuretic initiation was 1.26 (95% confidence interval [CI]: 1.08–1.47). The adherence-adjusted HR of KR for continuous use of thiazide diuretics was 1.44 (95% CI: 1.21–1.72).

      Conclusions

      In this population-based cohort of patients with knee osteoarthritis, thiazide diuretic use was associated with a higher risk of KR than loop diuretic use. This association may potentially be due to thiazide diuretics’ effect on BMD and serum magnesium.

      Keywords

      Introduction

      High bone mineral density (BMD) and low serum magnesium levels are both associated with a high prevalence of knee osteoarthritis (OA)
      • Felson D.T.
      • Lawrence R.C.
      • Dieppe P.A.
      • Hirsch R.
      • Helmick C.G.
      • Jordan J.M.
      • et al.
      Osteoarthritis: new insights. Part 1: the disease and its risk factors.
      • Hunter D.J.
      • Hart D.
      • Snieder H.
      • Bettica P.
      • Swaminathan R.
      • Spector T.D.
      Evidence of altered bone turnover, vitamin D and calcium regulation with knee osteoarthritis in female twins.
      • Zeng C.
      • Wei J.
      • Li H.
      • Yang T.
      • Zhang F.J.
      • Pan D.
      • et al.
      Relationship between serum magnesium concentration and radiographic knee osteoarthritis.
      . High BMD is also associated with an increased risk of incident knee OA but may protect against knee OA progression
      • Ernst M.E.
      • Moser M.
      Use of diuretics in patients with hypertension.
      . Such a paradoxical phenomenon is not fully understood, though selection bias due to conditioning on an intermediate stage of knee OA has been postulated as one potential explanation. Identifying risk factors of knee OA progression would provide insightful guidance for secondary prevention of this disabling disease; however, owing to methodological (e.g., confounding by indication and index event bias) and logistic (e.g., repeated assessment of BMD or serum magnesium level) challenges, few, if any, studies have examined the effect of changes in either BMD or serum magnesium on the risk of knee OA progression due to exposure to incident extraneous factors (e.g., medication use) occurring after knee OA diagnosis.
      Diuretics are commonly-used medications with an acceptable side–effect profile
      • Ernst M.E.
      • Moser M.
      Use of diuretics in patients with hypertension.
      . There is increasing evidence for a beneficial effect of thiazide diuretic therapy in preserving BMD
      • LaCroix A.Z.
      • Ott S.M.
      • Ichikawa L.
      • Scholes D.
      • Barlow W.E.
      Low-dose hydrochlorothiazide and preservation of bone mineral density in older adults. A randomized, double-blind, placebo-controlled trial.
      . In contrast, loop diuretics have been found to decrease BMD
      • Rejnmark L.
      • Vestergaard P.
      • Heickendorff L.
      • Andreasen F.
      • Mosekilde L.
      Loop diuretics increase bone turnover and decrease BMD in osteopenic postmenopausal women: results from a randomized controlled study with bumetanide.
      . In addition, there is increasing recognition of the potential effect of diuretic use on lowering serum magnesium levels
      • Ernst M.E.
      • Moser M.
      Use of diuretics in patients with hypertension.
      • Kieboom B.C.T.
      • Zietse R.
      • Ikram M.A.
      • Hoorn E.J.
      • Stricker B.H.
      Thiazide but not loop diuretics is associated with hypomagnesaemia in the general population.
      • Nijenhuis T.
      • Vallon V.
      • van der Kemp A.W.
      • Loffing J.
      • Hoenderop J.G.
      • Bindels R.J.
      Enhanced passive Ca2+ reabsorption and reduced Mg2+ channel abundance explains thiazide-induced hypocalciuria and hypomagnesemia.
      ; the effect appears to be greater for thiazide than loop diuretics
      • Agus Z.S.
      Mechanisms and causes of hypomagnesemia.
      • Neogi T.
      • Li S.
      • Peloquin C.
      • Misra D.
      • Zhang Y.
      Effect of bisphosphonates on knee replacement surgery.
      . Of relevance, the Rotterdam Study recently reported that serum magnesium levels were lower among thiazide users but higher among loop diuretic users than nonusers of either medications
      • Agus Z.S.
      Mechanisms and causes of hypomagnesemia.
      .
      Since diuretics are often used for relatively long periods of time, and thiazide and loop diuretics may have different impact on BMD and serum magnesium levels, initiation of these two medications could serve as a potential indicator for changes in BMD and serum magnesium levels
      • Ernst M.E.
      • Moser M.
      Use of diuretics in patients with hypertension.
      • Kieboom B.C.T.
      • Zietse R.
      • Ikram M.A.
      • Hoorn E.J.
      • Stricker B.H.
      Thiazide but not loop diuretics is associated with hypomagnesaemia in the general population.
      • Nijenhuis T.
      • Vallon V.
      • van der Kemp A.W.
      • Loffing J.
      • Hoenderop J.G.
      • Bindels R.J.
      Enhanced passive Ca2+ reabsorption and reduced Mg2+ channel abundance explains thiazide-induced hypocalciuria and hypomagnesemia.
      • Agus Z.S.
      Mechanisms and causes of hypomagnesemia.
      . Assessment of their relation to the risk of end stage knee OA should shed light on our understanding of the role of both BMD and magnesium in knee OA progression. We conducted a sequential propensity score-matched cohort study to compare the risk of the clinically relevant endpoint of knee replacement (KR) surgery among patients with knee OA who initiated thiazide vs those who initiated loop diuretics.

      Methods

      Data source

      We used The Health Improvement Network (THIN), an electronic medical record database from general practitioners in United Kingdom
      • Neogi T.
      • Li S.
      • Peloquin C.
      • Misra D.
      • Zhang Y.
      Effect of bisphosphonates on knee replacement surgery.
      . THIN contains health information on approximately 17 million patients from 770 general practices in the UK, and previous study has shown that THIN is representative of the UK population in terms of patient demographics and the prevalence of common illnesses
      • Lewis J.D.
      • Schinnar R.
      • Bilker W.B.
      • Wang X.
      • Strom B.L.
      Validation studies of the health improvement network (THIN) database for pharmacoepidemiology research.
      . During consultation with patients, health information is recorded by general practitioners using a computerized system. The information includes socio-demographics, anthropometrics, lifestyle factors, details from general practice visits, diagnoses from specialists’ referrals as well as hospital admissions, and results of laboratory tests. The Read classification system is used to code specific diagnoses. Prescription medications are coded based on a drug dictionary in BNF code and ATC code formats from the Multilex classification system
      • Stuart-Buttle C.D.
      • Read J.D.
      • Sanderson H.F.
      • Sutton Y.M.
      A language of health in action: read Codes, classifications and groupings.
      . Scientific Review Committee for the THIN database and the Institutional Review Board at Xiangya Hospital approved this study, with waiver of informed consent.

      Study population

      We identified knee OA based on Read codes (i.e., a coded thesaurus of clinical terms) that have been used in the National Health Service in the UK since 1985. Read codes provide a standard vocabulary for clinicians to record findings and procedures of the patients when they receive health and social care. We identified knee OA based on the following Read codes: N053611 (patellofemoral osteoarthritis), N05z611 (knee osteoarthritis not otherwise specified), N05zL00 (osteoarthritis not otherwise specified, of knee), and N05zM00 (osteoarthritis not otherwise specified, of tibio-fibular joint). We included individuals (≥50 years) who had a diagnosis of knee OA by Read codes between January 2000 and December 2016. Patients with OA coded only as general, without specifying the knee, were ineligible for the current analysis. We also excluded subjects who had KR prior to knee OA diagnosis, or who were deemed unlikely to be a candidate for KR (i.e., history of joint infection or comorbidities with poor prognosis [end-stage renal disease on dialysis, severe pulmonary disease requiring supplemental oxygen, or any cancer])
      • Neogi T.
      • Li S.
      • Peloquin C.
      • Misra D.
      • Zhang Y.
      Effect of bisphosphonates on knee replacement surgery.
      . In addition, we introduced a 3-month exposure lag period to exclude subjects with KR within 3 months after entering the study cohort because KRs that occur right after the initiation of diuretic were likely to have been scheduled before initiation of the medication.

      Assessment of exposure and active comparator

      We identified individuals who initiated thiazide or loop diuretics through the following ATC codes: C03AA01 (Bendroflumethiazide), C03AA04 (Chlorothiazide), C03BA04 (Chlortalidone), C03BA12 (Clorexolone), C03AA03 (Hydrochlorothiazide), C03AA02 (Hydroflumethiazide), C03BA11 (Indapamide), C03BA05 (Mefruside), C03AA08 (Methyclothiazide), C03BA08 (Metolazone), C03AA05 (Polythiazide), and C03BA10 (Xipamide) for thiazide diuretics; C03CA02 (Bumetanide), C03CC01 (Etacrynic), C03CA01 (Furosemide), C03CA03 (Piretanide), and C03CA04 (Torasemide) for loop diuretics, and BNF (thiazide diuretics: 2.2.1; loop diuretics: 2.2.2). The “initiation” of thiazide or loop diuretic was defined as the first prescription of thiazide or loop diuretic after the knee OA diagnosis during the study period in THIN database. Subjects with a history of prescription of thiazide or loop diuretic before entering the study were considered prevalent users and excluded from the study. In addition, subjects were required to be continuously enrolled with the general practice for ≥ 1 year in THIN database before the date of first prescription date of either thiazide or loop diuretic (i.e., index date).

      Assessment of outcome

      The outcome was the first (i.e., incident) primary KR that occurred 3 months after initiation of either thiazide or loop diuretics during the follow-up period. Primary KR included both total and partial KR, identified by Read codes. Previous studies have used this approach to identify KRs in THIN and Clinical Practice Research Datalink (a similar database to THIN)
      • Neogi T.
      • Li S.
      • Peloquin C.
      • Misra D.
      • Zhang Y.
      Effect of bisphosphonates on knee replacement surgery.
      • Kuo C.F.
      • Chou I.J.
      • See L.C.
      • Chen J.S.
      • Yu K.H.
      • Luo S.F.
      • et al.
      Urate-lowering treatment and risk of total joint replacement in patients with gout.
      • Yu D.
      • Jordan K.P.
      • Snell K.I.E.
      • Riley R.D.
      • Bedson J.
      • Edwards J.J.
      • et al.
      Development and validation of prediction models to estimate risk of primary total hip and knee replacements using data from the UK: two prospective open cohorts using the UK Clinical Practice Research Datalink.
      • Misra D.
      • Lu N.
      • Felson D.
      • Choi H.K.
      • Seeger J.
      • Einhorn T.
      • et al.
      Does knee replacement surgery for osteoarthritis improve survival? The jury is still out.
      .

      Sequential propensity score-matched cohorts

      We conducted a sequential propensity score-matched cohort study to compare the risk of KR among thiazide diuretic initiators with that among loop diuretic initiators. Propensity score is the probability of treatment assignment (e.g., thiazide initiation) conditional on observed baseline characteristics. Propensity score matching is used to mitigate the effects of confounding by indication, especially in the presence of a large number of covariates, in epidemiological studies
      • Seeger J.D.
      • Williams P.L.
      • Walker A.M.
      An application of propensity score matching using claims data.
      . We divided calendar time into 1-year blocks from 2000 to 2016 (i.e., 17 blocks). Specifically, subjects were allocated into 17 blocks based on their index date, which was based on the date of initiation of either their thiazide or loop diuretic. For example, subjects whose initiation date fell between January 1, 2000 and December 31, 2000 would be allocated into the first (year-2000) time block. Within each time block, we assembled a cohort of thiazide initiators, defined as patients who started thiazide during that time block, and a comparator cohort of matched loop diuretic initiators, who started loop diuretic during the same time block. We conducted propensity score matching within each time block using a greedy matching algorithm
      • Parsons L.
      Reducing bias in a propensity score matched-pair sample using greedy matching techniques.
      , i.e., for each thiazide initiator, an initiator of a loop diuretic with the closest propensity score was selected as a comparator from the same time block. Propensity scores (i.e., predicted probability of thiazide initiation) were estimated using logistic regression separately for each time block. The variables included in the model consisted of sociodemographic factors (age at index date, sex, the Townsend Deprivation Index score
      • Morris R.
      • Carstairs V.
      Which deprivation? A comparison of selected deprivation indexes.
      ), body mass index (BMI), duration of osteoarthritis prior to the index date, lifestyle factors (smoking status, and alcohol use), comorbidities and medication use recorded in THIN at any time prior to the index date, and healthcare utilization during the 1 year before the index date (see Table I). Comorbidities and medication use were assessed from the date subjects entered into THIN until their index date.
      Table IBaseline Characteristics of Propensity Score-matched Knee Osteoarthritis Patients (≥50 years) Initiating Thiazide or Loop Diuretics
      Variable listThiazide diuretics (n = 3,488)Loop diuretics (n = 3,488)Standard difference
      Demographics
       Age, mean (SD), y72.6 (9.0)72.5 (9.6)0.010
       Socioeconomic deprivation index score
      The Socio-Economic Deprivation Index was measured by the Townsend Deprivation Index, which was grouped into quintiles from 1 (least deprived) to 5 (most deprived).
      , mean (SD)
      2.8 (1.3)2.8 (1.3)0.018
       Female (%)59.158.40.015
      OA duration, mean (SD), y7.4 (6.6)7.5 (6.7)0.010
      BMI, mean (SD), kg/m229.9 (6.0)29.8 (6.3)0.010
      Lifestyle factors
       Drinking (%)0.025
       None24.623.9
       Past3.63.2
       Current71.872.8
       Smoking (%)0.006
       None54.654.9
       Past33.333.1
       Current12.012.0
      Comorbidity (%)
       Chronic kidney diseases8.28.30.003
       Congestive heart failure2.42.70.018
       Hypertension59.559.80.007
       Atrial fibrillation7.98.30.014
       Chronic obstructive pulmonary disease6.77.00.011
       Myocardial infarction6.56.5<0.001
       Peripheral vascular disease2.42.70.015
       Angina14.414.50.003
       Diabetes19.019.30.007
       Venous thromboembolism5.15.30.012
       Hyperlipidemia17.217.80.016
       Ischemic heart disease20.120.20.004
       Liver disease2.62.80.012
       Pneumonia or infection8.57.70.026
       Stroke5.85.30.021
       Transient ischemic attack5.95.40.021
       Varicose veins15.215.80.017
       Depression14.614.20.012
       Dementia1.41.50.002
       Fall17.417.90.012
       Fracture11.511.20.010
       Peptic ulcer8.79.30.021
       Osteoporosis10.010.10.004
       Rheumatic arthritis2.22.20.004
      Medication (%)
       ACE inhibitors38.038.40.007
       Beta receptor inhibitors36.036.30.006
       Calcium channel blockers39.939.60.006
       Angiotensin receptor blockers11.411.4<0.001
       Statins42.642.80.005
       Anticoagulants8.79.50.031
       Antidiabetic medicine14.614.90.009
       Insulin3.53.60.005
       Aspirin44.243.20.020
       Glucocorticoids25.725.00.017
       Nitrates17.918.40.013
       NSAIDs86.085.70.008
       Opioids45.444.50.020
       Estrogen11.912.70.025
       Bisphosphonates8.68.40.005
       PPIs50.149.70.007
       H2 blockers27.728.20.010
      Healthcare utilization, mean (SD)
       Hospitalizations
      Frequency during the past 1 year.
      0.4 (0.9)0.4 (0.9)0.010
       General practice visits
      Frequency during the past 1 year.
      7.8 (5.8)7.9 (6.4)0.025
       Specialist referrals
      Frequency during the past 1 year.
      0.6 (1.1)0.6 (1.1)0.003
      BMI, body mass index; n, number; y, years; SD, standard deviation; NSAIDs, non-steroidal anti-inflammatory drugs; ACE, angiotensin converting enzyme; OA, osteoarthritis; PPIs, proton pump inhibitors; H2, histamine-2.
      The Socio-Economic Deprivation Index was measured by the Townsend Deprivation Index, which was grouped into quintiles from 1 (least deprived) to 5 (most deprived).
      Frequency during the past 1 year.

      Statistical analysis

      We compared the baseline characteristics of the two cohorts (i.e., thiazide diuretic initiators and loop diuretic initiators). The follow-up time for each subject began from the index date until the date of KR, death, age of 90, date of disenrollment in THIN, or the end of the fifth year of follow-up (because approximately 90% of subjects’ diuretics prescriptions were for 5 years or less), whichever occurred first. We calculated the cumulative incidence rate of KR to depict the risk of KR for each cohort accounting for the competing risk of death
      • Austin P.C.
      • Lee D.S.
      • Fine J.P.
      Introduction to the analysis of survival data in the presence of competing risks.
      . The absolute rate difference (RD) in KR was estimated between the thiazide cohort and loop diuretic cohort. Since the two compared cohorts were balanced at baseline, we were able to calculate the crude RD between two groups. The following formula was used for RD (95% confidence interval [CI]) calculation: RD = rate (exposed) - rate (non-exposed); SERD=aPTa2+bPTb2, where a and b refer to the number of events in each cohort, PTa and PTb refer to the total person-time accumulated in each cohort, and 95% CI: RD ± 1.96*SERD. We fitted cause-specific Cox proportional hazard models to examine the relation of thiazide vs loop diuretic initiation to the hazard of KR while accounting for the competing risk of death
      • Austin P.C.
      • Lee D.S.
      • Fine J.P.
      Introduction to the analysis of survival data in the presence of competing risks.
      .
      To account for time-varying exposures and confounders, we performed a marginal structural model to estimate the average adherence-adjusted hazard ratio (HR) of KR for continuous use of thiazide vs loop diuretic use
      • Hernan M.A.
      • Brumback B.
      • Robins J.M.
      Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men.
      . Specifically, time-varying exposures and confounders were updated each year. We fit pooled logistic regression models to obtain their predicted values for each person-year remaining off thiazide diuretic prescription and uncensored. We then used a SAS data step to calculate stabilized inverse-probability weights for each person-year from the predicted values of the previous models. Last, we used generalized estimating equations to fit the final weighted pooled logistic model that estimated the causal parameter and its robust standard error. Variables in the calculation of the propensity score were included in these models.
      We performed several sensitivity analyses. First, to further minimize potential confounding by indication, we conducted the analyses by excluding patients with chronic kidney disease. Second, we used asymmetric trimming to exclude patients whose propensity score was <2.5th percentile of the propensity score of the exposure (thiazide) cohort and >97.5th percentile of the propensity score of the comparison cohort (loop diuretics); thus, patients who were treated with the agent most contrary to prediction were excluded from the analyses to minimize potential unmeasured confounders. Third, we performed an analysis among subjects who were enrolled in THIN for at least 1 year without a diagnosis of OA prior to inclusion in the study sample (i.e., incident OA
      • Neogi T.
      • Li S.
      • Peloquin C.
      • Misra D.
      • Zhang Y.
      Effect of bisphosphonates on knee replacement surgery.
      ) to minimize potential misclassification of the duration of OA. Fourth, since the analyses may not fully adjust for potential confounders we performed quantitative sensitivity analyses to assess the minimum unmeasured confounding effect that would need to explain away an association observed in the main analyses conditional on the included covariates
      • VanderWeele T.J.
      • Ding P.
      Sensitivity analysis in observational research: introducing the E-value.
      .
      In addition, we conducted a nested case–control study to assess the dose–response relationship between number of prescriptions of thiazides and risk of KR using risk set sampling
      • Breslow N.E.
      • Day N.E.
      Statistical methods in cancer research. Volume II--The design and analysis of cohort studies.
      . Specifically, for each case of KR, we created a risk set that included up to 10 controls who were alive and free of KR when a KR case occurred and matched by sex, year of entry into study, age of entry into study (within ±1 year), and propensity-score (within a caliper of ±0.1). The number of prescriptions of thiazides was calculated from the date of thiazide initiation to the date the case (i.e., KR) and matched controls (assigned the same date as their matched case) were identified. We divided number of prescriptions of thiazides into three categories: non-use of thiazides, 1–5, and ≥six prescriptions of thiazides. We estimated the relation of each thiazide prescription category to the risk of KR using conditional logistic regression and tested a dose–response relationship by number of thiazide prescriptions.
      All P-values were two-sided. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

      Results

      In total,200,317 subjects with knee OA met our inclusion criteria. Of them 16,431 initiated a thiazide diuretic and 15,069 initiated a loop diuretic. We excluded 9,334 subjects who were deemed unlikely to be a candidate for KR, 269 subjects who had KR within 3 months after initiation of the diuretic because KRs that occur during this period were likely to have been scheduled before initiation of the medication, and 5,091 subjects who had missing information on BMI, Townsend Deprivation Index Score, smoking status and alcohol drinking. Of the remaining (n = 16,806) 3,488 initiators of thiazide were successfully propensity score-matched to the same number of initiators of loop diuretic (Fig. 1). The baseline characteristics of the two propensity score-matched cohorts are shown in Table I. The mean age was 73 years, and slightly more than 40% were men. The characteristics of the thiazide cohort and its matched comparison loop diuretic cohort were well-balanced, with all standardized differences being < 0.1
      • Franklin J.M.
      • Rassen J.A.
      • Ackermann D.
      • Bartels D.B.
      • Schneeweiss S.
      Metrics for covariate balance in cohort studies of causal effects.
      .
      Fig. 1
      Fig. 1Selection process of included subjects.
      The cumulative incidence of KR was higher in the thiazide cohort than in the loop diuretic cohort (Fig. 2). As shown in Table II, 359 kRs (28.6/1,000 person-years) occurred in the thiazide cohort and 283 (24.1/1,000 person-years) occurred in the loop diuretic cohort over 5 years of follow-up. The RD of incident KR in the thiazide cohort vs that in the loop diuretic cohort was 4.5 (95% CI: 1.08 to 1.47) per 1,000 person-years and the corresponding HR was 1.26 (95% CI: 1.08 to 1.47). The adherence-adjusted HR of KR for continuous use of thiazide was 1.44 (95% CI: 1.21 to 1.72).
      Fig. 2
      Fig. 2Time to Knee Replacement over Five Years for the Propensity Score-Matched Cohorts of Knee Osteoarthritis Patients with Thiazide Diuretic Initiation compared with Initiation of Loop Diuretic, adjusting for Competing Risk of Death.
      Table IIAssociation between thiazide diuretic initiation and incidence of knee replacement surgery comparing with loop diuretic initiation
      Thiazide diureticsLoop diuretics
      Total population
      Thiazide initiation also showed a higher risk of knee replacement compared with initiation of loop diuretics (hazard ratio = 1.26, 95% confidence interval: 1.10 to 1.44) without restricting to 5-year follow-up.
       Subject (n)3,4883,488
       Incident knee replacement (n)359283
       Mean follow-up (year)3.63.4
       Rate (1,000 person-years)28.624.1
       RD (1,000 person-years, 95% CI)4.5 (0.4, 8.5)0.00 (reference)
       HR (95% CI)
      Hazard ratios were adjusted for competing event (death).
      1.26 (1.08, 1.47)1.00 (reference)
      Adherence-adjusted, HR (95% CI)
      Hazard ratios were adjusted for competing event (death).
      1.44 (1.21, 1.72)1.00 (reference)
      Excluding CKD, HR (95% CI)
      Hazard ratios were adjusted for competing event (death).
      1.27 (1.08, 1.49)1.00 (reference)
      PS trimming, HR (95% CI)
      Hazard ratios were adjusted for competing event (death).
      1.30 (1.11, 1.52)1.00 (reference)
      Incident OA, HR (95% CI)
      Hazard ratios were adjusted for competing event (death).
      1.23 (1.02, 1.48)1.00 (reference)
      HR, hazard ratio; n, number; RD, rate difference; PS, propensity score; 95% CI, 95% confidence interval; CKD, chronic kidney disease.
      Thiazide initiation also showed a higher risk of knee replacement compared with initiation of loop diuretics (hazard ratio = 1.26, 95% confidence interval: 1.10 to 1.44) without restricting to 5-year follow-up.
      Hazard ratios were adjusted for competing event (death).
      The results from various sensitivity analyses are presented in Table II. Exclusion of subjects with chronic kidney disease (HR = 1.27, 95% CI: 1.08 to 1.49), extreme propensity scores (HR = 1.30, 95% CI: 1.11 to 1.52), or restricting to subjects with incident knee OA (HR = 1.23, 95% CI: 1.02 to 1.48) did not change the association materially. Furthermore, to completely nullify the observed associations (e.g., HR = 1.23 for the smallest effect estimate), the association of residual confounder(s) with either thiazide or with KR should be ≥ an odds ratio of 1.76. Such a strong residual confounder(s) seems unlikely given that many known confounders have been accounted for in the propensity score-matched design.
      The odds of KR increased with longer duration of thiazide use (Table III). Compared with non-use of thiazide, multivariable-adjusted odds ratios (ORs) of KR were 1.16 (95% CI: 0.90 to 1.48) and 1.28 (95% CI: 1.03 to 1.58) for 1–5 and ≥six prescriptions of thiazide, respectively (P for trend = 0.04).
      Table IIIDose–response association between thiazide diuretic initiation and incidence of knee replacement surgery
      Cases
      Incident knee replacement.
      ControlsOdds Ratio (95% CI)
      Total number5992,446
       Non-use of thiazide (n)2271,0391.00 (reference)
       1-5 prescriptions of thiazide (n)1405601.16 (0.90, 1.48)
       6 or more prescriptions of thiazide (n)2328471.28 (1.03, 1.58)
      95% CI, 95% confidence interval; n, number.
      Incident knee replacement.

      Discussion

      We found that the risk of KR was higher among thiazide initiators than those of loop diuretic initiators, and long-term use of thiazide was associated with even higher risk of KR than loop diuretics. Our findings were independent of the major confounders and remained stable in various sensitivity analyses, suggesting that the observed associations were robust.

      Possible explanations

      While the biological mechanisms linking thiazide use to the risk of KR are not fully understood, differential impacts of thiazide vs loop diuretics on changes in both BMD and serum magnesium levels may partly explain these findings. First, randomized controlled trial (RCT) demonstrated that thiazide preserved BMD
      • LaCroix A.Z.
      • Ott S.M.
      • Ichikawa L.
      • Scholes D.
      • Barlow W.E.
      Low-dose hydrochlorothiazide and preservation of bone mineral density in older adults. A randomized, double-blind, placebo-controlled trial.
      , loop diuretics decreased BMD
      • Rejnmark L.
      • Vestergaard P.
      • Heickendorff L.
      • Andreasen F.
      • Mosekilde L.
      Loop diuretics increase bone turnover and decrease BMD in osteopenic postmenopausal women: results from a randomized controlled study with bumetanide.
      . Many studies have shown that high BMD is associated with prevalent and incident knee OA; however, its association with knee OA progression remains controversial. One explanation for such paradoxical phenomena is related to the fact that BMD is a chronic risk factor. Observational studies of the association of BMD with the risk of knee OA progression among subjects with mild-to-moderate knee OA are in effect adjusting for an intermediate stage of disease (i.e., mild-to-moderate knee OA). Such studies are affected by collider bias (index event bias) and are susceptible to potential selection bias.
      Second, previous studies have shown that chronic use of either thiazide or loop diuretics increases magnesium excretion; however, the degree of magnesium depletion from thiazide use appears to be greater than loop diuretic use
      • Ernst M.E.
      • Moser M.
      Use of diuretics in patients with hypertension.
      • Kieboom B.C.T.
      • Zietse R.
      • Ikram M.A.
      • Hoorn E.J.
      • Stricker B.H.
      Thiazide but not loop diuretics is associated with hypomagnesaemia in the general population.
      • Nijenhuis T.
      • Vallon V.
      • van der Kemp A.W.
      • Loffing J.
      • Hoenderop J.G.
      • Bindels R.J.
      Enhanced passive Ca2+ reabsorption and reduced Mg2+ channel abundance explains thiazide-induced hypocalciuria and hypomagnesemia.
      • Agus Z.S.
      Mechanisms and causes of hypomagnesemia.
      . Recently, the Rotterdam Study reported that thiazide use was associated with lower whereas loop diuretic use with higher serum magnesium levels than nonuse, respectively
      • Kieboom B.C.T.
      • Zietse R.
      • Ikram M.A.
      • Hoorn E.J.
      • Stricker B.H.
      Thiazide but not loop diuretics is associated with hypomagnesaemia in the general population.
      . An animal study demonstrated that intra-articular magnesium sulfate attenuated the development of OA
      • Lee C.H.
      • Wen Z.H.
      • Chang Y.C.
      • Huang S.Y.
      • Tang C.C.
      • Chen W.F.
      • et al.
      Intra-articular magnesium sulfate (MgSO4) reduces experimental osteoarthritis and nociception: association with attenuation of N-methyl-D-aspartate (NMDA) receptor subunit 1 phosphorylation and apoptosis in rat chondrocytes.
      , and several cross-sectional clinical studies have found that low levels of serum magnesium were associated with high prevalence of knee OA
      • Hunter D.J.
      • Hart D.
      • Snieder H.
      • Bettica P.
      • Swaminathan R.
      • Spector T.D.
      Evidence of altered bone turnover, vitamin D and calcium regulation with knee osteoarthritis in female twins.
      • Zeng C.
      • Wei J.
      • Li H.
      • Yang T.
      • Zhang F.J.
      • Pan D.
      • et al.
      Relationship between serum magnesium concentration and radiographic knee osteoarthritis.
      . In addition, magnesium is an antagonist of N-methyl-d-aspartate receptors, which plays an important role in nociceptive transmission, modulation and sensitization of pain
      • Herroeder S.
      • Schonherr M.E.
      • De Hert S.G.
      • Hollmann M.W.
      Magnesium--essentials for anesthesiologists.
      . Indeed, results from a meta-analysis demonstrated that systemic administration of magnesium was effective in minimizing postoperative pain
      • De Oliveira Jr., G.S.
      • Castro-Alves L.J.
      • Khan J.H.
      • McCarthy R.J.
      Perioperative systemic magnesium to minimize postoperative pain: a meta-analysis of randomized controlled trials.
      . Thus, thiazide diuretics could increase the risk of KR by decreasing serum magnesium levels.

      Strengths and limitations

      Several characteristics of our study are worth noting. First, BMD or serum magnesium are both chronic factors; thus, they likely occur prior to the occurrence of knee OA. Observational studies of the effects of a prevalent exposure on disease progression are susceptible to potential selection bias. To mitigate this kind of bias, we examined initiation of diuretics after knee OA diagnosis, whereby both BMD and serum magnesium levels may be altered by these medications after knee OA diagnosis. Second, in contrast to observational studies that compared users of anti-osteoporotic drugs or magnesium-supplement with non-users, we assembled two comparative cohorts who initiated different types of diuretics to minimize confounding by indication.
      Third, we postulated that an increased risk of KR among thiazide users may be through its impact on BMD and/or serum magnesium; however, we can't verify these mechanisms owing to lack of BMD or serum magnesium data in THIN. Fourth, although previous studies have used Read codes to define symptomatic knee OA in THIN
      • Gore M.
      • Sadosky A.B.
      • Leslie D.L.
      • Tai K.S.
      • Emery P.
      Therapy switching, augmentation, and discontinuation in patients with osteoarthritis and chronic low back pain.
      • Zeng C.
      • Dubreuil M.
      • LaRochelle M.R.
      • Lu N.
      • Wei J.
      • Choi H.K.
      • et al.
      Association of tramadol with all-cause mortality among patients with osteoarthritis.
      and KR has been generally accepted as a “hard” outcome in cohort studies of knee OA
      • Neogi T.
      • Li S.
      • Peloquin C.
      • Misra D.
      • Zhang Y.
      Effect of bisphosphonates on knee replacement surgery.
      • Yu D.
      • Jordan K.P.
      • Snell K.I.E.
      • Riley R.D.
      • Bedson J.
      • Edwards J.J.
      • et al.
      Development and validation of prediction models to estimate risk of primary total hip and knee replacements using data from the UK: two prospective open cohorts using the UK Clinical Practice Research Datalink.
      • Khan T.
      • Alvand A.
      • Prieto-Alhambra D.
      • Culliford D.J.
      • Judge A.
      • Jackson W.F.
      • et al.
      ACL and meniscal injuries increase the risk of primary total knee replacement for osteoarthritis: a matched case-control study using the Clinical Practice Research Datalink (CPRD).
      • Nielen J.T.
      • de Vries F.
      • Dagnelie P.C.
      • van den Bemt B.J.
      • Emans P.J.
      • Lalmohamed A.
      • et al.
      Use of thiazolidinediones and the risk of elective hip or knee replacement: a population based case-control study.
      , we were unable to confirm a diagnosis of radiographic knee OA and to assess the radiographic progression of knee OA since knee image data were not available in THIN. Nonetheless, 96% of primary KRs are performed for knee OA
      National Joint Registry
      14th Annual Report. Wales, Northern Ireland and the Isle of Man: National Joint Registry for England.
      , though we acknowledge that there is potential for individuals qualifying for KR but not undergoing the procedure due to other factors such as personal preference. In addition, we excluded patients with rheumatoid arthritis (2.2% in thiazide diuretics group and 2.2% in loop diuretics group according to Table I), the result did not change materially (HR = 1.25, 95% CI: 1.07 to 1.47). Fifth, though we took the propensity score-matching method including many comorbidities as covariates to control for potential confounding bias, as in any observational study we cannot rule out residual confounding. Finally, the current study was conducted using UK data. The differences in health care systems between UK and other countries may limit the generalizability of our study findings. Thus, future studies conducted outside UK are warranted to verify the findings.

      Clinical and research implications

      The present findings may have clinical implications. If replicated and determined to be causal, these findings suggest that thiazide diuretics use may have an unfavorable effect on knee OA progression. In addition, this study may shed light on our understanding of the biological mechanisms linking thiazide use to the risk of knee OA progression. If future studies could collect data on thiazide use, BMD, serum levels of magnesium, as well as changes in knee structures and symptoms, we could assess to what extend the effect of thiazide use on the risk of knee OA progression is mediated via effects on BMD or magnesium levels; such insight could help guide the development of targeted treatment strategies for knee OA prevention and progression.

      Conclusions

      In this population-based cohort of patients with knee OA, thiazide diuretic use was associated with a higher risk of KR than loop diuretic use. Such an association may potentially be due to thiazide diuretics’ effect on BMD and serum magnesium.

      Contributions

      JW, TN, RT, AZF, CZ, DM, HKC, GL and YZ made substantial contributions to the conception and design of the study. JW conducted the data cleaning, and data analysis. All authors contributed to the interpretation of results. JW and CZ wrote the first draft. All authors contributed to the revision of the manuscript. YZ and GL has full access to the data and takes responsibility for the content and guarantees the integrity and accuracy of the work undertaken. All authors have read, provided critical feedback on intellectual content and approved the final manuscript.

      Conflict of interest

      None.

      Funding

      This work was supported by the National Natural Science Foundation of China, China (grant number 81772413, 81702207, 81702206), National Institutes of Health, United States (grant number R21 AR47785), Department of Veterans Affairs Merit Review Grant, United States (grant number I01BX001660 [RT]), and National Institutes of Health, United States (grant number K24 AR070892).

      Role of the funder/sponsor

      The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; praparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

      Ethical approval

      The Institutional Review Board approved this study, with waiver of informed consent.

      Scientific approval

      This study was approved by the THIN Scientific Review Committee (18THIN073).

      Transparency

      The lead author affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.

      Acknowledgement

      Everyone who contributed significantly to the work has been listed.

      Appendix A. Supplementary data

      The following is the supplementary data to this article:

      Reference

        • Felson D.T.
        • Lawrence R.C.
        • Dieppe P.A.
        • Hirsch R.
        • Helmick C.G.
        • Jordan J.M.
        • et al.
        Osteoarthritis: new insights. Part 1: the disease and its risk factors.
        Ann Intern Med. 2000; 133: 635-646
        • Hunter D.J.
        • Hart D.
        • Snieder H.
        • Bettica P.
        • Swaminathan R.
        • Spector T.D.
        Evidence of altered bone turnover, vitamin D and calcium regulation with knee osteoarthritis in female twins.
        Rheumatology. 2003; 42: 1311-1316
        • Zeng C.
        • Wei J.
        • Li H.
        • Yang T.
        • Zhang F.J.
        • Pan D.
        • et al.
        Relationship between serum magnesium concentration and radiographic knee osteoarthritis.
        J Rheumatol. 2015; 42: 1231-1236
        • Ernst M.E.
        • Moser M.
        Use of diuretics in patients with hypertension.
        N Engl J Med. 2009; 361: 2153-2164
        • LaCroix A.Z.
        • Ott S.M.
        • Ichikawa L.
        • Scholes D.
        • Barlow W.E.
        Low-dose hydrochlorothiazide and preservation of bone mineral density in older adults. A randomized, double-blind, placebo-controlled trial.
        Ann Intern Med. 2000; 133: 516-526
        • Rejnmark L.
        • Vestergaard P.
        • Heickendorff L.
        • Andreasen F.
        • Mosekilde L.
        Loop diuretics increase bone turnover and decrease BMD in osteopenic postmenopausal women: results from a randomized controlled study with bumetanide.
        J Bone Miner Res. 2006; 21: 163-170
        • Kieboom B.C.T.
        • Zietse R.
        • Ikram M.A.
        • Hoorn E.J.
        • Stricker B.H.
        Thiazide but not loop diuretics is associated with hypomagnesaemia in the general population.
        Pharmacoepidemiol Drug Saf. 2018; 27: 1166-1173
        • Nijenhuis T.
        • Vallon V.
        • van der Kemp A.W.
        • Loffing J.
        • Hoenderop J.G.
        • Bindels R.J.
        Enhanced passive Ca2+ reabsorption and reduced Mg2+ channel abundance explains thiazide-induced hypocalciuria and hypomagnesemia.
        J Clin Investig. 2005; 115: 1651-1658
        • Agus Z.S.
        Mechanisms and causes of hypomagnesemia.
        Curr Opin Nephrol Hypertens. 2016; 25: 301-307
        • Neogi T.
        • Li S.
        • Peloquin C.
        • Misra D.
        • Zhang Y.
        Effect of bisphosphonates on knee replacement surgery.
        Ann Rheum Dis. 2018; 77: 92-97
        • Lewis J.D.
        • Schinnar R.
        • Bilker W.B.
        • Wang X.
        • Strom B.L.
        Validation studies of the health improvement network (THIN) database for pharmacoepidemiology research.
        Pharmacoepidemiol Drug Saf. 2007; 16: 393-401
        • Stuart-Buttle C.D.
        • Read J.D.
        • Sanderson H.F.
        • Sutton Y.M.
        A language of health in action: read Codes, classifications and groupings.
        Proc AMIA Annu Fall Symp. 1996; : 75-79
        • Kuo C.F.
        • Chou I.J.
        • See L.C.
        • Chen J.S.
        • Yu K.H.
        • Luo S.F.
        • et al.
        Urate-lowering treatment and risk of total joint replacement in patients with gout.
        Rheumatology. 2018; 57: 2129-2139
        • Yu D.
        • Jordan K.P.
        • Snell K.I.E.
        • Riley R.D.
        • Bedson J.
        • Edwards J.J.
        • et al.
        Development and validation of prediction models to estimate risk of primary total hip and knee replacements using data from the UK: two prospective open cohorts using the UK Clinical Practice Research Datalink.
        Ann Rheum Dis. 2019; 78: 91-99
        • Misra D.
        • Lu N.
        • Felson D.
        • Choi H.K.
        • Seeger J.
        • Einhorn T.
        • et al.
        Does knee replacement surgery for osteoarthritis improve survival? The jury is still out.
        Ann Rheum Dis. 2017; 76: 140-146
        • Seeger J.D.
        • Williams P.L.
        • Walker A.M.
        An application of propensity score matching using claims data.
        Pharmacoepidemiol Drug Saf. 2005; 14: 465-476
        • Parsons L.
        Reducing bias in a propensity score matched-pair sample using greedy matching techniques.
        in: Proceedings of the Twenty-Sixth Annual SAS Users Group International Conference 2001. SAS Institute Inc., 2001
        • Morris R.
        • Carstairs V.
        Which deprivation? A comparison of selected deprivation indexes.
        J Public Health Med. 1991; 13: 318-326
        • Austin P.C.
        • Lee D.S.
        • Fine J.P.
        Introduction to the analysis of survival data in the presence of competing risks.
        Circulation. 2016; 133: 601-609
        • Hernan M.A.
        • Brumback B.
        • Robins J.M.
        Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men.
        Epidemiology. 2000; 11: 561-570
        • VanderWeele T.J.
        • Ding P.
        Sensitivity analysis in observational research: introducing the E-value.
        Ann Intern Med. 2017; 167: 268-274
        • Breslow N.E.
        • Day N.E.
        Statistical methods in cancer research. Volume II--The design and analysis of cohort studies.
        IARC Sci Publ. 1987; : 178-229
        • Franklin J.M.
        • Rassen J.A.
        • Ackermann D.
        • Bartels D.B.
        • Schneeweiss S.
        Metrics for covariate balance in cohort studies of causal effects.
        Stat Med. 2014; 33: 1685-1699
        • Lee C.H.
        • Wen Z.H.
        • Chang Y.C.
        • Huang S.Y.
        • Tang C.C.
        • Chen W.F.
        • et al.
        Intra-articular magnesium sulfate (MgSO4) reduces experimental osteoarthritis and nociception: association with attenuation of N-methyl-D-aspartate (NMDA) receptor subunit 1 phosphorylation and apoptosis in rat chondrocytes.
        Osteoarthritis Cartilage. 2009; 17: 1485-1493
        • Herroeder S.
        • Schonherr M.E.
        • De Hert S.G.
        • Hollmann M.W.
        Magnesium--essentials for anesthesiologists.
        Anesthesiology. 2011; 114: 971-993
        • De Oliveira Jr., G.S.
        • Castro-Alves L.J.
        • Khan J.H.
        • McCarthy R.J.
        Perioperative systemic magnesium to minimize postoperative pain: a meta-analysis of randomized controlled trials.
        Anesthesiology. 2013; 119: 178-190
        • Gore M.
        • Sadosky A.B.
        • Leslie D.L.
        • Tai K.S.
        • Emery P.
        Therapy switching, augmentation, and discontinuation in patients with osteoarthritis and chronic low back pain.
        Pain Pract. 2012; 12: 457-468
        • Zeng C.
        • Dubreuil M.
        • LaRochelle M.R.
        • Lu N.
        • Wei J.
        • Choi H.K.
        • et al.
        Association of tramadol with all-cause mortality among patients with osteoarthritis.
        J Am Med Assoc. 2019; 321: 969-982
        • Khan T.
        • Alvand A.
        • Prieto-Alhambra D.
        • Culliford D.J.
        • Judge A.
        • Jackson W.F.
        • et al.
        ACL and meniscal injuries increase the risk of primary total knee replacement for osteoarthritis: a matched case-control study using the Clinical Practice Research Datalink (CPRD).
        Br J Sports Med. 13 January 2018; https://doi.org/10.1136/bjsports-2017-097762
        • Nielen J.T.
        • de Vries F.
        • Dagnelie P.C.
        • van den Bemt B.J.
        • Emans P.J.
        • Lalmohamed A.
        • et al.
        Use of thiazolidinediones and the risk of elective hip or knee replacement: a population based case-control study.
        Br J Clin Pharmacol. 2016; 81: 370-378
        • National Joint Registry
        14th Annual Report. Wales, Northern Ireland and the Isle of Man: National Joint Registry for England.
        2017