Archives of Osteoporosis (2020) 15:81 https://doi.org/10.1007/s11657-020-00746-z ORIGINAL ARTICLE Alendronate use and bone mineral density gains in women with moderate-severe (stages 3B–5) chronic kidney disease: an open cohort multivariable and propensity score analysis from Funen, Denmark M. Sanni Ali 1,2,3 & Martin Ernst 4,5 & Danielle E. Robinson 2 & Fergus Caskey 6 & Nigel K. Arden 7,8 & Yoav Ben-Shlomo 6 & Mads Nybo 9 & Katrine H. Rubin 4 & Andrew Judge 6,7,8 & Cyrus Cooper 7,8 & M. K. Javaid 7,8 & Anne P. Hermann 10 & Daniel Prieto-Alhambra 2,8,11 Received: 21 August 2019 / Accepted: 30 April 2020 # The Author(s) 2020, corrected publication 2021 Abstract Summary Bisphosphonates are contraindicated in moderate-to-severe chronic kidney disease patients. However, they are used to prevent fragility fractures in patients with impaired kidney function, despite a lack of evidence on their effects on bone density in these patients. We demonstrated that Alendronate had a positive effect on bone in these patients. Purpose This study aimed to assess the association between alendronate use and bone mineral density (BMD) change in subjects with moderate-severe chronic kidney disease (CKD). Methods We created a cohort of CKD stage 3B–5 patients by linking all DXA-based measurements in the Funen area, Denmark, to biochemistry, national health registries and filled prescriptions. Exposure was dispensation of alendronate and the outcome was annualized percentage change in BMD at the femoral neck, total hip and lumbar spine. Individuals were followed from first BMD to the latest of subsequent DXA measurements. Alendronate non-users were identified using incidence density sampling and matched groups were created using propensity scores. Linear regression was used to estimate average differences in the annualized BMD. Results Use of alendronate was rare in this group of patients: propensity score matching (PSM) resulted in 71 alendronate users and 142 non-users with stage 3B–5 CKD (as in the 1 year before DXA). Whilst alendronate users gained an average 1.07% femoral neck BMD per year, non-users lost an average of 1.59% per annum. The PSM mean differences in annualized BMD were + 2.65% (1.32%, 3.99%), + 3.01% (1.74%, 4.28%) and + 2.12% (0.98%, 3.25%) at the femoral neck, total hip and spine BMD, respectively, all in favour of alendronate users. M. Sanni Ali and Martin Ernst contributed equally to this work. * Daniel Prieto-Alhambra 6 Population Health Sciences, Bristol Medical School, University of daniel.prietoalhambra@ndorms.ox.ac.uk Bristol, Bristol, UK 7 1 Faculty of Epidemiology and Population Health, Department of Nuffield Department of Orthopaedics, Rheumatology and Non-communicable Disease Epidemiology, London School of Musculoskeletal Sciences, University of Oxford, Oxford, UK Hygiene and Tropical Medicine, London, UK 8 2 MRC Lifecourse Epidemiology Unit, Southampton, UK Center for Statistics in Medicine, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, 9 Department of Clinical Biochemistry and Pharmacology, Odense University of Oxford, Oxford, UK University Hospital, Odense, Denmark 3 Department of Public Health, Saint Paul Hospital Millennium 10 Medical College, Addis Ababa, Ethiopia Department of Endocrinology, Odense University Hospital, 4 Odense, Denmark OPEN, Department of Health, University of Southern Denmark, Odense, Denmark 11 GREMPAL Research Group (Idiap Jordi Gol Primary Care Research 5 Department of Public Health, Clinical Pharmacology and Pharmacy, Institute) and CIBERFes, Universitat Autonoma de Barcelona, University of Southern Denmark, Odense, Denmark Barcelona, Spain 81 Page 2 of 12 Arch Osteoporos (2020) 15:81 Conclusion In a real-world cohort of women with stage 3B–5 CKD, use of alendronate appears associated with a significant improvement of 2–3% per year in the femoral neck, total hip and spine BMD. More data are needed on the anti-fracture effectiveness and safety of bisphosphonate therapy in moderate-severe CKD. Keywords Alendronate . Bone mineral density . Chronic kidney disease . Osteoporosis . Incidence density sampling . Propensity score Introduction The most important clinical goal of osteoporosis treatment is the reduction in fracture risk, due to high morbidity and mortality Chronic kidney disease (CKD) is a multi-factorial disease as- associated with it, especially in elderly people. Whilst there are sociated with a range of metabolic diseases and complications. effective therapies to reduce the risk of osteoporotic fracture, the The prevalence of these complications increases rapidly in use of oral bisphosphonates (i.e. first-line anti-osteoporosis ther- patients with stage 3B, an estimation of glomerular filtration apies) is restricted in patients with CKD for several reasons. First, rate (eGFR) below 45 ml/min/1.73m2 (1). It is also an inde- there are safety concerns related to the risk of bisphosphonates pendent risk factor for osteoporosis. More than one in four worsening kidney function and other adverse events which are people with osteoporosis have moderate or severe CKD (2). already increased in patients with CKD such as severe CKD has been shown to predict not only low bone mass due hypocalcaemia or hyperphosphatemia, or acute kidney injury to accelerated bone loss (3) but also fracture risk, with a dou- (9). Second, given the biological mechanism of how CKD bled risk in patients with stage 3 CKD (eGFR between 30 and weakens bone differs from osteoporosis, it is far from established 59 ml/min/1.73 m2) (4); a 2.5- to 3-fold risk in those with that bisphosphonates will have a similar beneficial effect in im- stage 3B CKD (4); and a four-times higher fracture incidence proving BMD. Importantly, participants in these trials with CKD amongst patients with stage 4 CKD (eGFR between 15 and 29 are likely to be healthier and have less comorbidities compared ml/min/1.73 m2) (5, 6) or in renal replacement therapy (7, 8). with patients with CKD in the real-world setting. This leaves Osteoporosis may develop in patients with a reduced eGFR moderate-severe CKD patients with a very high fracture risk due to age-related decrease in renal function or in patients with effectively untreatable. In addition, the relationship between intrinsic renal disease (1). Patients with CKD due to intrinsic changes in BMD during treatment with oral bisphosphonates renal disease often have additional risk factors for osteoporo- and fracture risk is still less clear since the numbers of patients sis such as glucocorticoid use, hyperparathyroidism, and vita- with moderate or severe CKD recruited for the pivotal trials were min D deficiency (9). Bone mass or bone mineral density low (15–17, 19–21). However, clinical trial participants with (BMD) is a recognized key determinant of future fracture risk higher BMD gains on oral bisphosphonate therapy had a signif- in postmenopausal women (10) and patients with CKD, next icantly lower risk of fracture than participants with no change or a to bone microarchitecture, bone geometry and rate of turnover reduction in BMD (20). Hence, BMD is an accepted surrogate (11). Measurement of BMD at the lumbar spine, femur neck marker in bridging studies of anti-osteoporosis treatments (22) to or total hip by dual-energy X-ray absorptiometry (DEXA) is test different dosing regimens (23) and efficacy in other popula- the standard method for diagnosis of osteoporosis and evalu- tions such as men (24) and in patients taking glucocorticoids ating fracture risk, particularly in non-CKD patients. (25). However, its significance and interpretation in CKD patients The relationship between BMD change and fracture risk in have been controversial due to a combination of CKD- postmenopausal osteoporosis versus bone fragility in CKD induced changes in bone and mineral metabolism (1). may differ given the different aetiopathology of bone Recently, several studies have shown that BMD is a good fragility in CKD (26) with particular reference to potential predictor of fracture risk in CKD stages 3–5; as a result, the adynamic bone disease. The aim of this study was to examine 2017 Kidney Disease Improving Global Outcomes (KDIGO) the association between bisphosphonate use and changes in guideline recommended systematic BMD measurement in BMD over time in subjects with moderate-to-severe CKD CKD patients (12–14). BMD measurement is often assessed (stages 3B, 4 and 5). before anti-fracture treatment decisions are made; this is in part due to the fact that therapeutic trials in osteoporosis usu- ally require a low BMD value as an entry criterion. Hence, Methods medications are often licensed for use in patients below a given BMD threshold. In addition, BMD is also monitored Data source as a surrogate marker to evaluate the response of an individual patient to treatment as well as the efficacy of a treatment in We used the Odense University Hospital Database (OUHD) clinical trials (15–18). which holds DXA-measured BMD data for the whole of the Arch Osteoporos (2020) 15:81 Page 3 of 12 81 Funen region, the second-largest island of Denmark, comprising was thus minimized whilst preserving statistical power, as 36,024 individuals examined between August 1990 and initial non-users could become alendronate users and will be February 2015. Serum creatinine is part of the routine panel of included later as users. blood tests performed at the Odense University Hospital as part of osteoporosis care. Primary care practices and hospitals in this Exposure, outcome and confounders Danish region use the same clinical biochemistry laboratory, making all biochemistry values (including serum creatinine tests) Alendronate use, defined by pharmacy dispensations in the available for this study. We were able to calculate the eGFR for Danish Prescriptions Register, was the main exposure. Users of participants in this cohort using the CKD-EPI formula (27). alendronates were identified using pre-specified lists of ATC Pharmacy drug dispensations of primary and secondary care codes. Like prescriptions, dispensation data do not necessarily prescriptions can be tracked back to 1995 from the Danish reflect treatment duration. Treatment episodes were created using Prescriptions Register and linked using unique national patient a previously validated algorithm that accounts for non-adherence identifiers to OUHD. Pharmacy dispensations are recorded using (or non-compliance) and defines treatment episodes of continu- Anatomic Therapeutic Classification (ATC) codes. ous exposure. Any overlapping prescriptions of the same drug The Open Patient Data Exploratory Network (OPEN) pro- were interpreted as an early collection of a repeat prescription. vides a unique platform for linking clinical, dispensation and Any overlapping days between two prescriptions of the same biochemistry data. It is an approved Statistics Denmark insti- drug were added to the end of the period covered by the two tutional partner for linking to diagnoses and comorbidity re- prescriptions. Any two prescriptions of the same drug were corded using ICD-10 codes in Danish hospital records. The concatenated and considered as one prescription if the gap be- Network has previously been used to link clinical biochemis- tween the end of the first prescription and the start of the second try to fracture outcomes in numerous studies (28–32). Ad hoc prescription was less than 30 days. extraction and linking to renal function from the Odense The primary study outcome was annualized femoral neck University Hospital clinical biochemistry database was done BMD percentage change, relative to the previous or baseline as part of this study using a similar approach. In the OUHD, BMD, up to 3 years after alendronate use (or non-use), mea- 30% of individuals examined (over 10,000 subjects) were sured using a DXA. If more than two measurements were recommended osteoporosis treatment, with alendronate as available, the last DXA measurement 2–3 years from the in- the first-line drug. In Denmark, reimbursement for dex date was considered. Any measurements within 1 year alendronate and other alendronates requires a DXA assess- between each other were excluded, as they were unlikely to ment or an X-ray verified spine or hip fracture. The study provide clinically relevant information. Secondary outcomes was approved by the Data Protection Agency (OUH 15/ were similarly defined as annualized BMD percentage chang- 37999) and by Statistics Denmark (ref. 705079). All statistical es, measured at the lumbar spine and total hip. If participants analyses were done on de-identified patient data merged at had serial spine measurements, we used the vertebrae that Statistics Denmark. were still assessable at the last recorded visit to calculate the rate of BMD change. A sensitivity analysis used DXA mea- Study participants and design surements taken 3–5.5 years after the index date, since change in BMD has been shown to plateau after 3 years (33). All patients registered in OUHD aged 40 years or older at the Potential confounders were pre-identified using clinical age of biochemistry testing, with at least one eGFR measure- knowledge and a literature review. Confounders were mea- ment below 45 ml/min/1.73 m2 (CKD stages 3B, 4 and 5), at sured at baseline and included age, sex, BMI, baseline least 2 years of follow-up data available and at least two DXA eGFR, fracture history, comorbidities (Charlson index, renal BMD measurements recorded 2 or more years apart were disease and diabetes), hospital contacts in the previous year eligible for inclusion. Previous users of any anti-osteoporosis and concomitant drug use (number of ATC in the year before medication (except calcium and vitamin D supplements) in index, aromatase inhibitors, antihypertensives, antidepres- the year before eGFR testing and those with more than 1 year sants and antidiabetics). As systemic steroids have an effect between the closest renal measurement (eGFR) and DXA on BMD and are commonly used in certain renal conditions, (BMD measurement) were excluded. The first dispensation we accounted for these drugs in a more granular fashion. We of alendronate (therapy initiation) was considered the index calculated the number of defined daily doses in the previous date for alendronates users. Participants, users of alendronate, year and stratified into quartiles. were risk-set-matched with replacement to non-users on ± 5 years of the year of birth at index date, meaning a non-user Statistical analyses could be matched to several alendronate users. For every alendronate user, up to five non-users at index date were ran- In a primary analysis, propensity scores for alendronate use domly selected from a large pool of non-users. Selection bias were estimated using multivariable logistic regression, 81 Page 4 of 12 Arch Osteoporos (2020) 15:81 including the confounders (Table 1). Interactions between age (ii) A multivariable linear regression model, adjusting for and BMI as well as age and baseline eGFR were added to the the same set of confounders included in the propensity propensity score equation to improve the balance of the base- score model, was fitted to model annualized percentage line characteristics. The calculated propensity scores were BMD change according to alendronate use, with 95% used to match each alendronate user with to up to five com- confidence intervals. parable non-users with a calliper width of at most 0.02 SDs on the logit of the propensity scores. Balance before and after All analyses were conducted using R (34) and STATA 15 matching was checked using the absolute standardized mean (35). Statistical syntax was programmed at Oxford using mock difference (ASMD). An ASMD of 0.10 or less was considered data extracted from the analytical dataset. These programs were acceptable and representative of a good balance. run ‘in situ’ by a statistician at Odense, Denmark. Final results Linear regression modelling was used to estimate average were verified by two of the authors (MSA and DPA). differences (expressed as beta coefficient and 95% confidence intervals) in the annualized BMD change (per cent versus index measure) at the femoral neck (primary outcome), total Results hip and lumbar spine (secondary outcomes). Two post hoc analyses were conducted to deal with the low number of Participants and baseline characteristics alendronate users eligible for inclusion in the final analyses: We identified 36,024 patients in the linked OUHD BMD da- (i) A linear regression model including all eligible tabase who were eligible for the study, including 12,544 alendronate users and non-users and adjusted for the pre- (34.8%) patients who were alendronate users within the study viously calculated propensity scores period. We excluded 233 (1.9%) who were younger than 40 Table 1 Baseline characteristics for eligible and included Eligible Included bisphosphonate users Characteristic Unique patients (N) 270 71 eGFR, mean (SD) 37.2 (7.2) 37.9 (8.0) Dead within 3 years of first bisphosphonate 35 (13.0%) 0 (0.0%) Age, median (IQR) 82 (76–87) 79 (70–83) Gender, female N (%) 268 (99.3%) 71 (100.0%) Index BMDa, mean (SD) Total hip 0.657 (0.123) 0.678 (0.117) Femoral neck 0.551 (0.110) 0.560 (0.106) Spine 0.791 (0.165) 0.782 (0.166) Fractures, N (%) Hip 21 (7.8%) 5 (7.0%) Non-hip other 21 (7.8%) 5 (7.0%) Non-hip osteoporotic 28 (10.4%) 6 (8.5%) Charlson score, N (%) 0 142 (52.6%) 39 (54.9%) 1 64 (23.7%) 18 (25.4%) 2 29 (10.7%) 7 (9.9%) ≥3 35 (13.0%) 7 (9.9%) Concomitant drugs, N (%) Systemic corticosteroids 98 (36.3%) 29 (40.8%) Anticoagulants 163 (60.4%) 35 (49.3%) Statins 106 (39.3%) 30 (42.3%) Antidiabetics 38 (14.1%) 7 (9.9%) Antihypertensives 198 (73.3%) 58 (81.7%) a Up to 1 year before index date BMD, bone mineral density; discont., discontinued; eGFR, estimated glomerular filtration rate; IQR, interquartile range; OP, osteoporotic; SD, standard deviation Arch Osteoporos (2020) 15:81 Page 5 of 12 81 years at the time of therapy initiation, 1323/12,544 (10.5%) between 1 and 3.5 years after their index DXA and were who did not have an eGFR measurement recorded in the pre- eligible for the analysis. vious year and 10,586/12,544 (84.4%) who had an eGFR Baseline characteristics for all 270 eligible alendronate value of 45 or above. Of the remaining 402 eligible users and the final sample of 71 are reported in Table 1. As alendronate users, a third (132/402, 32.8%) had no DXA mea- expected, participants included in the final analyses were surement in the year before therapy initiation. Therefore, 270 younger (79 verses 82 years old on average), healthier alendronate users (264, 98% on alendronate) were eligible for (54.9% versus 52.6% with a Charlson score of 0) and more the study (Fig. 1). Amongst these, 35 (13.0%) died during likely to use systemic corticosteroids (40.8% versus 36.3%), follow-up, 93 (34.4%) discontinued alendronate therapy be- statins (42.3% versus 39.3%) and antihypertensives (81.7% fore having a repeat BMD and 142 (52.6%) did not have a versus 73.3%) than those excluded. repeat BMD between 1 and 3.5 years after the first scan after Table 2 reports baseline characteristics for the included therapy started. We, therefore, included 71 alendronate users participants, stratified by alendronate use. Both groups only in the study. In parallel, 23,713 non-users were initially avail- included women and had an average age of 79–80 years. As able for comparison. Amongst these, 1228 (5.2%) died during expected, alendronate users had a much lower index BMD follow-up. Only 1609 (6.8%) had repeat BMD measures (0.56 versus 0.62 g/cm2 at the femoral neck), were more likely Fig. 1 Construction of treatment episodes based on prescription data. Most of the eligible patients were on alendronate (264/270, 98%) 81 Page 6 of 12 Arch Osteoporos (2020) 15:81 Table 2 Baseline characteristics according to bisphosphonate use Before matching After matching Users Non-users Users Non-users Patients (N) 71 1492a 40 142 eGFR, mean (SD) 37.9 (8.0) 36.1 (7.7) 37.1 (9.1) 38.0 (6.7) Age, median (IQR) 79 (70–83) 80 (76–85) 79 (75–82) 79 (73–84) Gender, N (%) Females 71 (100%) 1492 (100%) 40 (100.0%) 142 (100.0%) Males 0 (0%) 0 (0%) 0 (0%) 0 (0%) Bone mineral density Total hip 0.678 (0.117) 0.754 (0.144) 0.663 (0.119) 0.738 (0.134) Femoral neck 0.560 (0.106) 0.627 (0.121) 0.547 (0.104) 0.617 (0.104) Spine 0.782 (0.166) 0.877 (0.171) 0.806 (0.146) 0.881 (0.176) Fractures, N (%) Hip 5 (7.0%) 28 (1.9%) 5 (12.5%) (n < 5) Non-hip other 5 (7.0%) 99 (6.6%) (n < 5) (n < 5) Non-hip osteoporotic 6 (8.5%) 89 (6.0%) (n < 5) 6 (4.2%) Charlson score, N (%) 0 39 (54.9%) 778 (52.1%) 27 (67.5%) 95 (66.9%) 1 18 (25.4%) 322 (21.6%) 8 (20.0%) 23 (16.2%) 2 7 (9.9%) 213 (14.3%) (n < 5) 9 (6.3%) ≥3 7 (9.9%) 179 (12.0%) (n < 5) 8 (5.6%) Drugs, N (%) Systemic corticosteroids 29 (40.8%) 493 (33.0%) 11 (27.5%) 51 (35.9%) Anticoagulants 35 (49.3%) 818 (54.8%) 18 (45.0%) 61 (43.0%) Statins 30 (42.3%) 522 (35.0%) 17 (42.5%) 51 (35.9%) Antidiabetics 7 (9.9%) 316 (21.2%) 5 (12.5%) 26 (18.3%) Antihypertensives 58 (81.7%) 1.331 (75.8%) 32 (80.0%) 112 (78.9%) Additional diagnoses, N (%) Diabetes (uncomplicated) 5 (7.0%) 171 (11.5%) (n < 5) 16 (11.3%) Diabetes with complications (n < 5) 30 (2.0%) (n < 5) (n < 5) Dementia (n < 5) 40 (2.7%) (n < 5) (n < 5) Cardiovascular disease (n < 5) 17 (1.1%) 0 (0%) (n < 5) Hypertension 16 (22.5%) 328 (22.0%) 7 (17.5%) 26 (18.3%) COPD 8 (11.3%) 143 (9.6%) (n < 5) 14 (9.9%) Heart failure 17 (23.9%) 383 (25.7%) 8 (20.0%) 27 (19.0%) Renal transplant 0 (0.0%) 6 (0.4%) 0 (0%) 0 (0%) Renal dialysis 0 (0.0%) 20 (1.3%) 0 (0%) 0 (0%) a 1492 episodes in 206 unique patients COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; IQR, interquartile range; SD, standard deviation to use systemic corticosteroids (40.8% versus 32.8) and had a Propensity score matching much higher prevalence of the previous fracture (7.0% versus 1.7% for hip fracture history) than non-users. Low bone mass, Before matching, the two groups had little overlap in their prevention of glucocorticoid-induced osteoporosis and sec- propensity scores, with median (interquartile range) of 0.49 ondary fracture prevention are the most likely indicators of (0.17, 0.84) for the alendronate users and below 0.001 (< alendronates in patients with moderate-to-severe CKD. All 0.001, 0.003) for the non-users. After matching, the median of these variables and the confounders listed in Table 1 are (IQR) of the propensity scores were more similar, at 0.19 adjusted for using propensity score and multivariable regres- (0.08, 0.43) in the alendronate users and 0.13 (0.04, 0.43) in sion models. non-users. Arch Osteoporos (2020) 15:81 Page 7 of 12 81 Forty of the 71 alendronate users were matched to 142 non- mean difference (95% CI) in percentage annual change in users. Matching improved balance and reduced differences to femoral neck BMD between alendronate users and non- acceptable limits (ASMD < 0.1) for most of the assessed users favoured the users by + 2.65% (1.32%, 3.99%) per year participant characteristics included in the model (Fig. 2). of alendronate exposure. Average femoral neck BMD change However, relevant differences remained for key confounders, per month for alendronate non-users and users is shown in including baseline (index) femoral neck BMD (0.55 versus Fig. 3a. 0.62 gr/cm2), use of systemic glucocorticoids (27.5% versus In the analyses including the full cohort, alendronate non- 35.9%) and previous hip fracture history (12.5% versus users lost on average 1.67% of their baseline femoral neck 3.5%). Charlson comorbidity index, polypharmacy, number BMD per year, whereas alendronate users gained an average of hospital visits, history of key comorbid conditions (compli- 0.63% per year on treatment. The results of linear regression cated diabetes, hypertension, chronic heart failure and COPD) modelling supported the findings from the propensity- and previous use of antiepileptics or antidiabetic therapies also matched analysis, with a multivariable-adjusted beta coeffi- still had unacceptable imbalances (ASMD > 0.1) after cient (mean difference between groups) of + 2.14% (1.22%, matching. Matching excluded more than 91% of the potential- 3.05%) and a propensity-score-adjusted beta of + 2.15% ly eligible alendronate non-users. (0.97%, 3.34%) per year. In the analysis of secondary outcomes, BMD at the Annualized bone mineral density changes over time spine increased amongst alendronate users by 3.98% per year for the full cohort and by 3.36% in the propensity- In the propensity-matched cohorts (pre-specified primary matched cohort. Alendronate non-users in the full cohort analysis), alendronate non-users lost on average 1.59% of increased their BMD by 0.65% per year and in the matched their baseline femoral neck BMD per year, whereas cohort by 0.34% per year. The mean (95% CI) difference alendronate users increased their BMD by 1.07% per year of in percentage BMD change was again in favour of therapy. The average percentage BMD changes in the alendronate users, with betas of + 3.01% (1.74%, 4.28%) propensity-score-matched cohort are detailed in Table 3. In a per year for the propensity-score-matched analysis, + linear regression model, after propensity score matching, the 2.14% (1.22%, 3.05%) for the multivariable analysis and Fig. 2 Covariate balance before and after matching usiung ASMD 81 Page 8 of 12 Arch Osteoporos (2020) 15:81 Table 3 Results of the analysis of bone mineral density (BMD) Analysis Non-user User Mean difference (95% CI) changes in participants exposed and unexposed to N BMD change N BMD change bisphosphonates Femoral neck PS-matched 142 − 1.59 40 1.07 2.65 (1.32, 3.99) Multivariable 1492 − 1.67 71 0.63 2.14 (1.22, 3.05) PS-adjusted 1492 − 1.67 71 0.63 2.15 (0.97, 3.34) Spine PS-matched 142 0.34 40 3.36 3.01 (1.74, 4.28) Multivariable 1492 0.65 71 3.98 2.14 (1.22, 3.05) PS-adjusted 1492 0.65 71 3.98 2.87 (1.62, 4.12) Total hip PS-matched 142 − 1.16 40 0.95 2.12 (0.98, 3.25) Multivariable 1492 − 2.14 71 0.82 2.29 (1.46, 3.11) PS-adjusted 1492 − 2.14 71 0.82 1.91 (0.82, 3.00) PS, propensity score + 2.87% (1.62%, 4.12%) for the propensity-score-adjusted patients with stage 3B–5 CKD being prescribed alendronates, analysis. Average spine BMD change per month for reflecting the safety concerns and lack of evidence on the alendronate non-users and users is shown in Fig. 3b. efficacy that led to this study. Total hip BMD improved by an average of 0.82% per year The primary propensity-matched cohort analysis identified for the full cohort of alendronate users and by 0.95% for the a significant, clinically relevant effect, with a + 1.1% femoral matched cohort. Alendronate non-users in the full cohort lost neck BMD gain per year in alendronate users, compared with on average − 2.14% BMD per year and those in the a − 1.6% bone loss at the same site per year in matched non- propensity-matched cohort lost − 1.16% per year. The mean users. This is equivalent to a + 2.7% difference in BMD difference was + 2.12 (0.98%, 3.25%) per year in favour of change per year in favour of alendronate users. The full cohort alendronate users in the propensity-score-matched model, + multivariable and propensity adjustment analyses found con- 2.29% (1.46%, 3.11%) for the multivariable model and + sistent results, with a + 2.1% and + 2.65% difference per year 1.91% (0.82%, 3.00%) for the propensity-adjusted model. favouring alendronate use, respectively. Consistent results Average hip bone mineral density (BMD) change per month were found for the secondary outcomes. Lumbar spine and for alendronate non-users and users is shown in Fig. 3c. total hip BMD improved by 3.4% and 1.0% per year in The pre-specified sensitivity analysis using DXA measures alendronate users, compared with that by 0.3% and − 1.2% taken between 3 and 5.5 years after index date could not be (bone loss) per year in the propensity-score-matched non- performed in the matched analysis as only 1 alendronate user users, respectively. This equated to a significant + 3.0% per and 12 non-users had valid follow-up data in this time year average difference in the lumbar spine and a + 2.1% window. difference in the total hip, both favouring alendronate users. Multivariable and propensity-adjusted models found similar results. Discussion Limitations and strengths We used region-wide population-based data from Funen, Denmark. Despite the size of the source dataset, only 71 fe- This cohort analysis has limitations. The eligibility criteria male participants were identified with stage 3B–5 CKD who excluded most of the records in the database including male were given alendronates, continued the treatment for at least 1 patients and users of other oral bisphosphonates. Although year and had at least two recorded BMD measurements. From this was not surprising, as bisphosphonates are to be used around 200 eligible participants not using alendronates, only with caution in patients with stage 3 CKD and are 1609 records of BMD measurements at different times were contraindicated in stages 4 and 5, it is possible that the available. Propensity score matching led to the exclusion of generalizability of the findings might have been almost half of the users and more than 90% of the non-users. compromised. The need for repeat DXA testing to obtain Although matching helped to reduce confounding, imbalance BMD change over time, the outcome of interest in this remained when alendronate users and matched non-users study, excluded more than half of the potentially eligible were compared, including in key confounders such as baseline bisphosphonate users. These patients may have differed BMD, fracture history and use of systemic steroids. The small from the included participants, as BMD monitoring during sample might demonstrate the low likelihood of Danish treatment is not routine practice in actual care conditions. Arch Osteoporos (2020) 15:81 Page 9 of 12 81 Fig. 3 Average bone mineral density (BMD) change per month for bisphosphonate users (left) and non-users (right) in the femoral neck (a), hip (b) and spine (c) Current evidence suggests that routine monitoring of BMD However, the multivariable and propensity adjustment models within 3 years of starting bisphosphonates in that include the full cohort gave reassuringly similar estimates postmenopausal women is clinically unhelpful1 (36) and to the propensity-matched analysis. Only women were includ- may contribute to a lack of follow-up densitometry in an un- ed in these analyses, possibly related to the high rate of exclu- expectedly large proportion of those treated. Trials of sion. The findings are thus limited to the female population. bisphosphonates have shown a significant attenuation of This analysis also has some strengths. The dataset used has BMD gain after the first 12 months of therapy (15, 37). The high granularity and quality and comprises information from a overlap between alendronate users and non-users was small, number of linked data sources. It is a unique data source for excluding almost half of the alendronate users and over 90% research as it includes comorbidity, hospital contacts, bio- of the non-users when propensity score matching was applied. chemistry, drug use, BMD testing and related questionnaires 81 Page 10 of 12 Arch Osteoporos (2020) 15:81 (38). The dataset also includes most of the population of a + 2.7%, + 3% and + 2.1% BMD change per year in the whole region of Denmark, thus maximizing the external va- femoral neck, lumbar spine and total hip, respectively, all lidity of our findings. in favour of alendronate users. These findings were not The observed differences in femoral neck BMD in modified by age, previous fracture history or baseline re- alendronate user versus non-user participants over 1 year were nal function. similar to those observed in trials of postmenopausal women (15). However, our alendronate users had only half the in- Funding information This project was funded by the NIHR HTA (pro- ject number or 14/36/02) and supported by the NIHR Biomedical crease in BMD seen in the trial’s treatment arm after 1 year Research Centre, Oxford. of therapy, where the increase in femoral neck BMD was 2% with alendronate (15). Although not statistically significant, Compliance with ethical standards The study was approved the total hip is certainly the most relevant site of the upper by the Data Protection Agency (OUH 15/37999) and by Statistics extremity of the femur in the context of a follow-up with or Denmark (ref. 705079). All statistical analyses were done on de- without treatment. It is difficult to make generalization given identified patient data merged at Statistics Denmark. the small size of our study in the findings. The difference at the spine in our alendronate users was also smaller than seen in Conflicts of interest NK Arden reports personal fees from Flexion, Freshfields, Janssen, Merck and Regeneron. C Cooper reports lecture fees the trial, where there was a 4.2% increase after 1 year. It is and honoraria from the Alliance for Better Bone Health Amgen, Eli Lilly, difficult to compare our unexposed CKD participants with GSK, Medtronic, Merck, Novartis, Pfizer, Roche, Servier, Takeda and those in the placebo arm of the trial as the trial participants UCB. A Judge is a subpanel member of the NIHR PGfAR, has received who received the placebo also received calcium and vitamin D consultancy fees from Freshfields Bruckhaus Deringer and has held ad- visory board positions (which involved receipt of fees) at Anthera supplements. Pharmaceuticals, INC. MK Javaid reports research support and speaker Early BMD changes during therapy are difficult to interpret fees from Amgen. D Prieto-Alhambra reports research grants from UCB, due to regression to the mean (39) with most patients who lose Amgen and Les Laboratoires Servier and (not personal) speaker and BMD in the first year of alendronate therapy going on to have consultancy fees from UCB and Amgen. D Prieto-Alhambra is a member of the NIHR HTA CET panel since November 2017. DE Robinson, MS significant increases in BMD in the second year. Our attempt Ali, KRubin, M Ernst, P Hermann, M Nybo, Y Ben-Shlomo and F to address this issue in a sensitivity analysis was inhibited due Caskey have nothing to disclose. The views expressed are those of the to a lack of power. The relationship between BMD change authors and not necessarily those of the NHS, NIHR or Department of and fracture risk may differ in postmenopausal osteoporosis Health. and bone fragility in CKD as they have different Open Access This article is licensed under a Creative Commons aetiopathology, with the potential involvement of adynamic Attribution 4.0 International License, which permits use, sharing, adap- bone disease in bone fragility in CKD (26). The disconnect tation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, pro- between treatment-related BMD increases and fracture risk is vide a link to the Creative Commons licence, and indicate if changes were exemplified in trials of fluoride therapy, where increased made. The images or other third party material in this article are included BMD was associated with a paradoxical increased risk of in the article's Creative Commons licence, unless indicated otherwise in a fracture (40). The artefactual effects of vascular calcification credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by on spinal BMD further complicate the interpretation of BMD statutory regulation or exceeds the permitted use, you will need to obtain in patients with CKD (41). permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Conclusions We demonstrated that alendronate therapy had a positive References effect on BMD in female patients with stage 3B–5 CKD. BMD is the best surrogate available for bone strength in 1. Anandarajah S, Tai T, de Lusignan S, Stevens P, O'Donoghue D, clinical practice, and RCTs using BMD as a primary Walker M, Hilton S (2005) The validity of searching routinely outcome have been accepted by regulators as bridge collected general practice computer data to identify patients with chronic kidney disease (CKD): a manual review of 500 medical studies to extend the indication of alendronate therapies records. Nephrol Dial Transplant 20(10):2089–2096 for men (25) and for the prevention or treatment of 2. Najar MS, Mir MM, Muzamil MM et al (2017) Prevalence of steroid-induced osteoporosis (38). We found that BMD osteoporosis in patients with chronic kidney disease (stages 3–5) improved more (by 2% to 3% per year, depending on in comparison with age-and sex-matched controls: a study from skeletal site) in alendronate users than non-users. In fact, Kashmir valley tertiary care center. Saudi J Kidney Dis Transpl 28(3):538–544 BMD decreased over time when participants were off 3. Nickolas TL, Stein EM, Dworakowski E, Nishiyama KK, alendronate therapy, by about 1% per year in the femoral Komandah-Kosseh M, Zhang CA, McMahon DJ, Liu XS, neck and total hip. The result had an average difference of Boutroy S, Cremers S, Shane E (2013) Rapid cortical bone loss Arch Osteoporos (2020) 15:81 Page 11 of 12 81 in patients with chronic kidney disease. J Bone Miner Res 28(8): 19. McClung MR, Geusens P, Miller PD et al (2001) Effect of 1811–1820 risedronate on the risk of hip fracture in elderly women. N Engl J 4. Ensrud KE, Lui LY, Taylor BC, Ishani A, Shlipak MG, Stone KL, Med 344(5):333–340 Cauley JA, Jamal SA, Antoniucci DM, Cummings SR, 20. Hochberg MC, Ross PD, Black D, Cummings SR, Genant HK, Osteoporotic Fractures Research Group (2007) Renal function Nevitt MC, Barrett-Connor E, Musliner T, Thompson D, for the and risk of hip and vertebral fractures in older women. Arch Gen Fracture Intervention Trial Research Group (1999) Larger increases Intern Med 167(2):133–139 in bone mineral density during alendronate therapy are associated 5. Dooley AC, Weiss NS, Kestenbaum B (2008) Increased risk of hip with a lower risk of new vertebral fractures in women with post- fracture among men with CKD. Am J Kidney Dis 51(1):38–44 menopausal osteoporosis. Arthritis Rheum 42(6):1246–1254 6. Maravic M, Ostertag A, Urena P, Cohen-Solal M (2016) Dementia 21. Ott MS (2012) Bisphosphonate safety and efficacy in chronic kid- is a major risk factor for hip fractures in patients with chronic kid- ney disease. Kidney Int 82(8):833–835 ney disease. Osteoporos Int 27(4):1665–1669 22. Bouxsein ML, Delmas PD (2008) Considerations for development 7. Alem AM, Sherrard DJ, Gillen DL, Weiss NS, Beresford SA, of surrogate endpoints for anti-fracture efficacy of new treatments Heckbert SR, Wong C, Stehman-Breen C (2000) Increased risk in osteoporosis: a perspective. J Bone Miner Res 23(8):1155–1167 of hip fracture among patients with end-stage renal disease. 23. Schnitzer T, Bone H, Crepaldi GF et (2000) Therapeutic equivalence Kidney Int 58(1):396–399 of alendronate 70 mg once weekly and alendronate 10 mg daily in the 8. Hansen D, Olesen JB, Gislason GH, Abrahamsen B, Hommel K treatment of osteoporosis. Aging Clin Exp Res 12(1): 1-12 (2016) Risk of fracture in adults on renal replacement therapy: a 24. Orwoll E, Ettinger M, Weiss S, Miller P, Kendler D, Graham J, Danish national cohort study. Nephrol Dial Transplant 31(10): Adami S, Weber K, Lorenc R, Pietschmann P, Vandormael K, 1654–1662 Lombardi A (2000) Alendronate for the treatment of osteoporosis 9. Thomas R, Kanso A, Sedor JR (2008) Chronic kidney disease and in men. N Engl J Med 343(9):604–610 its complications. Prim Care 35(2):329–344 25. Saag KG, Emkey R, Schnitzer TJ et al (1998) Alendronate for the 10. Fujiwara S, Kasagi F, Masunari F (2003) Fracture prediction from prevention and treatment of glucocorticoid-induced osteoporosis. N bone mineral density in Japanese men and women. J Bone Miner Engl J Med 339(5):92–299 Res 18(8):1547–1553 26. Iwasaki Y, Kazama JJ, Fukagawa M (2017) Molecular abnormali- 11. Jiang C, Giger ML, Kwak SM, Chinander MR, Martell JM, Favus ties underlying bone fragility in chronic kidney disease. Biomed MJ (2000) Normalized BMD as a predictor of bone strength. Acad Res Int 2017:1–11. https://doi.org/10.1155/2017/3485785 Radiol 7(1):33–39 27. Levey AS, Stevens LA, Schmid CH, Zhang Y(L), Castro AF III, 12. Iimori S, Mori Y, Akita W et al (2011) Diagnostic usefulness of Feldman HI, Kusek JW, Eggers P, van Lente F, Greene T, Coresh J, bone mineral density and biochemical markers of bone turnover in for the CKD-EPI (Chronic Kidney Disease Epidemiology predicting fracture in CKD stage 5D patients—a single-center co- Collaboration) (2009) A new equation to estimate glomerular filtra- hort study. Nephrol Dial Transplant 27(1):345–351 tion rate. Ann Intern Med 150(9):604–612 13. Yenchek RH, Ix JH, Shlipak MG, Bauer DC, Rianon NJ, 28. Abrahamsen B, Jørgensen HL, Laulund AL et al (2014) Low serum Kritchevsky SB, Harris TB, Newman AB, Cauley JA, Fried LF, thyrotropin level and duration of suppression as a predictor of major for the Health, Aging, and Body Composition Study (2012) Bone osteoporotic fractures - the open thyro register cohort. J Bone Miner mineral density and fracture risk in older individuals with CKD. Res 29(9):2040–2050 Clin J Am Soc Nephrol 7(7):1130–1136 29. Abrahamsen B, Jørgensen HL, Laulund AS, Nybo M, Bauer DC, 14. West SL, Lok CE, Langsetmo L, Cheung AM, Szabo E, Pearce D, Brix TH, Hegedüs L (2015) The excess risk of major osteoporotic Fusaro M, Wald R, Weinstein J, Jamal SA (2015) Bone mineral fractures in hypothyroidism is driven by cumulative hyperthyroid density predicts fractures in chronic kidney disease. J Bone Miner as opposed to hypothyroid time: an observational register-based Res 30(5):913–919 time-resolved cohort analysis. J Bone Miner Res 30(5):898–905 15. Black DM, Cummings SR, Karpf DB, Cauley JA, Thompson DE, 30. Rubin KH, Glintborg D, Nybo M, Andersen M, Abrahamsen B Nevitt MC, Bauer DC, Genant HK, Haskell WL, Marcus R, Ott (2016) Fracture risk is decreased in women with polycystic ovary SM, Torner JC, Quandt SA, Reiss TF, Ensrud KE (1996) syndrome: a register-based and population-based cohort study. J Randomised trial of effect of alendronate on risk of fracture in Bone Miner Res 31(4):709–717 women with existing vertebral fractures. Lancet 348(9041):1535– 31. Praetorius K, Madsen CM, Abrahamsen B, Jørgensen HL, 1541 Lauritzen JB, Laulund ASB (2016) Low levels of hemoglobin at 16. Cummings SR, Black DM, Thompson DE, Applegate WB, Barrett- admission are associated with increased 30-day mortality in pa- Connor E, Musliner TA, Palermo L, Prineas R, Rubin SM, Scott tients with hip f racture. Geriatr Orthop Surg Rehabil 7(3):115–120 JC, Vogt T, Wallace R, Yates AJ, LaCroix A (1998) Effect of 32. Händel MN, Frederiksen P, Cohen A, Cooper C, Heitmann BL, alendronate on risk of fracture in women with low bone density Abrahamsen B (2017) Neonatal vitamin d status from archived but without vertebral fractures: results from the fracture intervention dried blood spots and future risk of fractures in childhood: results trial. JAMA 280(24):2077–2082 from the D-tect study, a population-based case-cohort study. Am J 17. McClung M, Clemmesen B, Daifotis A, Gilchrist NL, Eisman J, Clin Nutr 106(1):155–161 Weinstein RS, Fuleihan G el-H, Reda C, Yates AJ, Ravn P (1998) 33. Black DM, Schwartz AV, Ensrud KE, Cauley JA, Levis S, Quandt Alendronate prevents postmenopausal bone loss in women without SA, Satterfield S, Wallace RB, Bauer DC, Palermo L, Wehren LE, osteoporosis: a double-blind, randomized, controlled trial. Ann Lombardi A, Santora AC, Cummings SR, FLEX Research Group Intern Med 128(4):253–261 (2006) Effects of continuing or stopping alendronate after 5 years of 18. Siris E, Adler R, Bilezikian J et al (2014) The clinical diagnosis of treatment: the fracture intervention trial long-term extension (flex): osteoporosis: a position statement from the national bone health a randomized trial. JAMA 296(24):2927–2938 alliance working group. Osteoporos Int 25(5):1439–1443 81 Page 12 of 12 Arch Osteoporos (2020) 15:81 34. R Core Team (2017) R: a language and environment for statistical 39. Cummings SR, Palermo L, Browner W, Marcus R, Wallace R, computing, R Foundation for Statistical Computing, Vienna, Pearson J, Blackwell T, Eckert S, Black D (2000) Monitoring os- Austria. URL https://www.R-project.org/ teoporosis therapy with bone densitometry: misleading changes and 35. StataCorp (2017) Stata Statistical Software: Release 15 College regression to the mean. JAMA 283(10):1318–1132 Station, StataCorp LP, TX 40. Riggs BL, Hodgson SF, O’fallon WM et al (1990) Effect of fluoride 36. Prieto-Alhambra D, Lalmohamed A, Abrahamsen B, Arden NK, de treatment on the fracture rate in postmenopausal women with oste- Boer A, Vestergaard P, de Vries F (2014) Oral bisphosphonate use oporosis. N Engl J Med 322(12):802–809 and total knee/hip implant survival: validation of results in an ex- 41. Kinsella S, Murphy K, Breen M, O’Neill S, McLaughlin P, Coyle J, ternal population-based cohort. Arthritis Rheum 66(11):3233–3240 Bogue C, O’Neill F, Moore N, McGarrigle AM, Molloy MG, 37. Harris ST, Watts NB, Genant HK, McKeever C, Hangartner T, Maher MM, Eustace JA (2015) Comparison of single CT scan Keller M, Chesnut CH 3rd, Brown J, Eriksen EF, Hoseyni MS, assessment of bone mineral density, vascular calcification and fat Axelrod DW, Miller PD (1999) Effects of risedronate treatment mass with standard clinical measurements in renal transplant sub- on vertebral and non-vertebral fractures in women with postmeno- jects: the ABC heart study. BMC Nephrol 16(1):188 pausal osteoporosis: a randomized controlled trial. JAMA 282(14): 1344–1352 Publisher’s note Springer Nature remains neutral with regard to jurisdic- 38. Rubin KH, Abrahamsen B, Hermann AP, Bech M, Gram J, Brixen tional claims in published maps and institutional affiliations. K (2011) Prevalence of risk factors for fractures and use of DXA scanning in Danish women: a regional population-based study. Osteoporos Int 22(5):1401–1409