Selective Serotonin Reuptake Inhibitor Use and Hip Fracture Risk Among Patients on Hemodialysis

Original Investigation

Selective Serotonin Reuptake Inhibitor Use and Hip Fracture Risk Among Patients on Hemodialysis Chandan Vangala, Jingbo Niu, Maria E. Montez-Rath, Jingyin Yan, Sankar D. Navaneethan, and Wolfgang C. Winkelmayer Rationale & Objective: Use of selective serotonin reuptake inhibitors (SSRIs) has been associated with hip fracture risk in the general population. This study examined this relationship among patients with kidney failure treated by hemodialysis, a unique high-risk subpopulation, within which the impact of SSRIs on hip fracture risk remains unexplored. Study Design: Case-control study. Settings & Participants: Eligible cases of hip fracture among maintenance hemodialysis patients between January 1, 2009, and September 30, 2015, were identified using the US Renal Data System. Each case was matched on index date with 10 eligible controls. To be eligible, study participants needed to have more than 1 year of Medicare Parts A and B coverage and more than 3 years of Part D coverage. For a separate examination of new short-term SSRI exposure, we selected cases and controls with more than 18 months of Part D coverage and no prior antidepressant use for 1 year. Exposure: During the 3-year Part D coverage period, use of SSRIs characterized as any (≥1 prescription filled), low, moderate, and high use (<20%, 20%-<80%, and ≥80% of days covered by filled prescriptions, respectively). Outcome: We selected cases using International Classification of Diseases, Ninth Revision codes

W

820.xx and 821.xx. In addition to 1 of these codes tied to a hospitalization, we required a corresponding surgical procedural code within 7 days of diagnosis. Analytic Approach: Conditional logistic regression to estimate unadjusted and multivariableadjusted ORs and 95% CIs. Results: We identified 4,912 cases and 49,120 controls. SSRI use was associated with increased hip fracture risk (adjusted OR, 1.25; 95% CI, 1.17-1.35). Risk for fracture was estimated for any, low, moderate, and high SSRI use: adjusted conditional ORs were 1.25 (95% CI, 1.17-1.35), 1.20 (95% CI, 1.08-1.32), 1.31 (95% CI, 1.18-1.43), and 1.26 (95% CI, 1.12-1.41), respectively. The association between hip fracture events and SSRI use was also seen in the examination of new short-term use (adjusted OR, 1.43; 95% CI, 1.23-1.67).

Correspondence to C. Vangala (vangala@bcm. edu) Am J Kidney Dis. XX(XX):110. Published online Month X, XXXX. doi: 10.1053/ j.ajkd.2019.07.015 Published by Elsevier Inc. on behalf of the National Kidney Foundation, Inc. This is a US Government Work. There are no restrictions on its use.

Limitations: Biomarkers of mineral bone disorder were not captured and accounted for in this analysis. Conclusions: We demonstrated an association between increased hip fracture risk and both long- and new short-term SSRI use. The stronger association with new short-term use may suggest an acute mechanism potentially related to falls.

hile both debilitating and life-threatening in the general population, hip fracture events among patients with kidney failure treated by hemodialysis (HD) are particularly impactful. Maintenance HD patients have longer hospitalizations and increased short- and long-term mortality.1,2 These patients have 4 times higher rates of hip fracture compared with the general population.3,4 Mineral and bone disorders, dysautonomia, acidosis, cachexia, and inflammation are often cited as substantial risk factors. Additionally, patients treated by maintenance dialysis also have more comorbid conditions that also contribute to fracture risk. Despite increased medication burden to manage these numerous comorbid conditions, limited understanding of the risk contributed by commonly prescribed medications exists. Antidepressants are frequently prescribed to manage anxiety and depression associated with advanced organ failure. Patients with kidney failure treated by kidney replacement therapy (KRT) are no exception, with 31.7% of Medicare Part D enrollees having received a prescription AJKD Vol XX | Iss XX | Month 2019

Complete author and article information provided before references.

for a psychotherapeutic agent in 2015.5 The bidirectional relationship between chronic medical illness and depression creates a cycle that is more difficult to break than the depression that is witnessed among patients without chronic comorbid conditions. A trial that randomly assigned patients with chronic kidney disease stage 3 or worse (but not receiving KRT) to either sertraline or placebo did not demonstrate a difference in depressive symptoms.6 Although antidepressants, particularly selective serotonin reuptake inhibitors (SSRIs), are a substantial component of therapy, their utility in patients treated by maintenance dialysis remains uncertain.7 A recently completed trial of 120 maintenance dialysis patients reported a modestly greater decrease in depressive symptoms with sertraline compared to cognitive behavioral therapy.8 Long-term use without further assessment of efficacy may subject patients to hazardous side effects, including those that affect hip fracture risk. Within the general population, antidepressants have been linked to increased fracture risk,9,10 and more 1

Original Investigation specifically, SSRIs have been associated with hip fractures.11,12 Effects on both bone metabolism13,14 and falls15,16 have been touted as potential mechanisms of increased risk. However, the bone pathology associated with kidney disease is both diverse and distinct from osteoporosis. Thus, the impact of SSRIs on hip fracture risk in maintenance dialysis patients remains unexplored, and this investigation aims to identify any potential associations that may exist between short- and longer-term SSRI exposure and hip fracture events. Methods Study Design and Source Population We used the US Renal Data System (USRDS), the national registry of persons with kidney failure treated by KRT, to conduct a case-control study nested in the recorded person-time between January 1, 2006, and September 30, 2015. Patients with end-stage kidney disease commonly qualify for benefits through Medicare, and the claims from these patients are available for study in the USRDS. We studied patients with comprehensive prescription drug coverage through enrollment in the prescription drug program, Medicare Part D, and receipt of a low-income subsidy. Participants were restricted to those with a documented KRT start date, demographic data, and a Medical Evidence Report from 1995 or later. For the examination of longer-term exposure, all participants were required to have at least 1 year (≥365 days) of Parts A and B coverage to allow for adequate time to determine participants’ comorbid conditions, at least 3 years (≥1,095 days) of both Part D and low-income subsidy coverage to retain access to a complete list of filled prescriptions, and at least 90 days of HD therapy before the index date to exclude patients with acute kidney injury requiring HD. As part of an analysis of short-term new-use exposure, a separate patient pool was identified with more than 18 months (≥547 days) of required Part D and low-income subsidy coverage before the index date. In this shortterm analysis of new users during these 6 months, only those without filled prescriptions of SSRIs during the preceding year (>6-18 months) before index were eligible. The Institutional Review Board at Baylor College of Medicine approved this study (H-36408). Because this study was conducted using deidentified data from the USRDS registry, individual-level informed consent was not indicated. Definition of Case Status We identified hip fracture events using the International Classification of Diseases, Ninth Revision (ICD-9) codes 820.xx or 821.xx. As an additional measure of specificity, cases were required to have the corresponding ICD-9 codes tied to a period of inpatient hospitalization and have one of the corresponding ICD-9 surgical procedural codes (78.55, 2

79.15, 81.52, 79.05, 79.25, or 81.40) within 7 days of diagnosis. We identified the date of hip fracture events as the index date and restricted cases and controls to participants without transplantation or with an index date before transplantation. Using incidence density sampling, we randomly selected 10 controls per case on each index date through M:N matching. This pooled strategy efficiently identifies controls without individually linking each control to a specific case, but rather the index date, allowing for multiple cases per strata. Selected controls for an index date could subsequently become cases at a later date, and controls could be selected more than once. Exposure of Interest SSRI use was identified from Part D claims for filled prescriptions during the time before the index date. We catalogued “any use” as having had at least 1 single prescription filled in the 3 years before the index date. We assigned “moderate use” to patients with ≥20% and <80% of the proportion of days covered by filled prescriptions. Finally, “high use” was determined by ≥80% of the proportion of days covered. In addition to these categories, we studied SSRI use as a continuous variable by determining its association with hip fracture events per additional year of medication use. Finally, in a separate analysis, after restricting cases and controls to those without any SSRI exposure for 1 year (between months 7 and 18 before the index date), we examined any “new use” in the 6 months before the index date. Covariates Age, sex, race (white, black, and other), Hispanic ethnicity, body mass index (BMI), census division, and duration of dialysis before the index date (vintage) were drawn from the Medical Evidence Form. With the requirement of at least 1 year of Parts A and B coverage, we intended on capturing comorbid conditions that affected hip fracture risk, namely hypertension, diabetes mellitus, coronary artery disease, cerebrovascular disease, peripheral vascular disease, arrhythmia, rheumatologic disorder, osteoporosis, depression, and tobacco use. We abstracted these comorbid conditions from Medicare billing claims in the institutional claims files and physician/supplier claims files of the USRDS (see Table S1 for specifications). Prior bisphosphonate use and steroid use were defined as any Part D claims for pharmacy-dispensed pills in the 3 years before the index date. Claims for filled prescriptions also helped identify other antidepressant use, including serotonin-norepinephrine reuptake inhibitors, norepinephrine-dopamine reuptake inhibitors, serotonin antagonist and reuptake inhibitors, tricyclic antidepressants, and tetracyclines. Statistical Analysis By fitting conditional logistic regression models, both unadjusted and multivariable adjusted, we estimated the association between hip fracture case (vs control) status AJKD Vol XX | Iss XX | Month 2019

Original Investigation and prior SSRI use. Estimates of the association with varying categories of use were expressed as conditional odds ratios (cORs) with corresponding 95% confidence intervals (CIs). Patient characteristics and comorbid conditions described in Table 1 were incorporated into a multivariable analysis. We further examined SSRI exposure among subgroups and performed interaction analyses with several characteristics, including age, sex, race, ethnicity, and BMI. A series of sensitivity analyses were undertaken. We performed an analysis with the inclusion of the use of available vitamin D receptor activators (calcitriol, doxercalciferol, and paricalcitol) and calcimimetics (cinacalcet). We also reassessed our model after excluding participants with bisphosphonate use or a history of osteoporosis. Last, we excluded patients with combined exposure to both SSRI and non-SSRI antidepressant use. In this data set, we re-examined the association with SSRI use while also estimating the association with the use of nonSSRI antidepressants. Only 2 variables included in the multivariable-adjusted analysis were incomplete: BMI (1.9%) and cause of kidney failure (for <10 patients); therefore, only 1,048 of the 54,032 (1.9%) records were incomplete, mainly arising from missing BMI. Most missing values appeared in the controls (2.0% of controls had missing BMI in comparison to 1.5% of cases). However, we believe that missing BMI was not related to itself and that we have sufficient information in the other variables to be able to correctly recover the distribution of BMI. We therefore assumed data to be missing at random. We used multiple imputation using chained equations to create and analyze 10 imputed data sets.17 Imputation occurred under fully conditional specification in which we used truncated regression to impute BMI and multinomial logistic regression for cause of kidney failure. We ran multiple imputation when the model included all variables in the adjusted analysis model (including indicator for case/control), as well as the exposure variable with 4 categories (no use and <20%, 20%-<80%, and ≥80% of days covered by filled prescriptions, respectively). All analyses were applied to each imputed data set separately to estimate parameters and their standard errors, which were then combined using Rubin’s rules.18 We also performed a complete case analysis as sensitivity analysis. Analyses were conducted using SAS software (version 9.3; SAS Institute Inc) and StataMP (version 14; Stata Corp). Results Patient Selection and Baseline Characteristics Between January 1, 2009, and September 30, 2015, we identified 4,912 patients with hip fracture (cases) who satisfied all criteria (Fig 1), and accordingly, 49,120 control patients M:N matched on index date (Table S2). Compared with controls, cases were older (mean age AJKD Vol XX | Iss XX | Month 2019

difference, 9.3 years) and had a larger proportion of nonHispanic whites (40% vs 25%). On average, controls had higher BMI (mean difference, 1.9 kg/m2) and more controls were obese (29% vs 25%) and severely obese (13% vs 7%). Despite having less exposure to dialysis (mean difference of 0.8 year), cases had greater proportions of almost all comorbid conditions. Patients with hip fracture had greater proportions of overall antidepressant use and SSRI use before the index date (52.6% vs 41.5% and 37.2% vs 27.6%). Slightly more than 10% of cases and 6.3% of controls had at least 80% of days covered by filled prescriptions for SSRIs. Non-SSRI use was comparable to SSRI use, with 26.2% of all patients having filled any prescriptions for antidepressants other than SSRIs. The prevalence of diagnosed depression was 18.5% of the overall population, of which 14.9% did not have exposure to any antidepressant in the 3 years before the index date. Of those without a diagnosis of depression, 19.4% had any exposure to SSRIs and 20.2% had any exposure to other antidepressants. Last, 12.1% of the overall population had exposure to both SSRIs and other antidepressants during the 3-year window. Between mid-2007 and September 2015, we identified 4,523 cases from the separate patient pool examining the short-term risk associated with new users of SSRIs. With 10 controls M:N matched on index date, we randomly selected 45,230 controls (Fig S1). Cases and controls in this short-term analysis had a distribution of covariates similar to the long-term analysis (Table S3). Associations Between SSRI Use and Hip Fracture In unadjusted analysis, the cOR for hip fracture associated with any SSRI use was 1.52 (95% CI, 1.43-1.62); the multivariable adjusted cOR was 1.25 (95% CI, 1.171.35). The unadjusted cORs for low (<20% of days covered), moderate (≥20%-<80% of days covered), and high (≥80% of days covered) SSRI use were 1.26 (95% CI, 1.15-1.38), 1.60 (95% CI, 1.47-1.75), and 1.83 (95% CI, 1.65-2.02). With multivariable adjustment, the corresponding cORs were 1.20 (95% CI, 1.08-1.32), 1.31 (95% CI, 1.18-1.43), and 1.26 (95% CI, 1.12-1.41), respectively (Table 2). When examined as a continuous variable, the unadjusted cOR for each additional year of SSRI use was 1.22 (95% CI, 1.19-1.26), which was attenuated to 1.07 (95% CI, 1.04-1.11) after multivariable adjustment. Formal testing for interactions with age, sex, BMI, race, or ethnicity revealed that both age and sex modified the effect of SSRIs on hip fracture risk quantitatively, but not qualitatively (Fig 2). After excluding patients with any exposure to SSRIs and non-SSRI antidepressants, the multivariable-adjusted cORs with any SSRI use and any non-SSRI use were 1.22 (95% CI, 1.13-1.33) and 1.12 (95% CI, 1.03-1.23), respectively. Analysis excluding participants with bisphosphonate use or 3

Original Investigation Table 1. Characteristics of Hip Fracture Cases and Controls Variable Age, y Female sex Ethnicity/race Non-Hispanic black Non-Hispanic white Non-Hispanic other Hispanic white Hispanic nonwhite BMI (kg/m2; missing in 1,043) Underweight Normal Overweight Obese Severely obese Dialysis vintage, y Census division New England Middle Atlantic East North Central West North Central South Atlantic East South Central West South Central Mountain Pacific Comorbid conditions Hypertension Diabetes mellitus Coronary artery disease Peripheral vascular disease Cerebrovascular disease Heart failure Arrhythmia Osteoporosis Rheumatologic disorder Tobacco use Depression History of steroid use History of bisphosphonate use History of PPI use Other antidepressant use Tricyclic antidepressant Serotonin-norepinephrine reuptake inhibitor Serotonin antagonist and reuptake inhibitor Norepinephrine-dopamine reuptake inhibitor Tetracycline Primary cause of kidney failure (missing in <10) Diabetes mellitus Glomerulonephritis/vasculitis Interstitial/pyelonephritis Hypertensive/large vessel disease Cystic/hereditary/congenital Neoplasm/tumor Miscellaneous

Cases (n = 4,912) 71 ± 12; 72 [63, 80] 2,861 (58%)

Controls (n = 49,120) 61 ± 14; 62 [52, 72] 25,422 (52%)

P <0.001 <0.001

1,439 (29%) 1,939 (39%) 412 (8%) 1,084 (22%) 38 (0.8%) 28 ± 8; 26.7 [23.1, 31.9] 209 (4%) 1,660 (34%) 1,373 (28%) 1,234 (26%) 362 (7%) 5.3 ± 3.5; 4.5 [2.7, 7.1]

23,976 (49%) 12,316 (25%) 3,220 (7%) 9,119 (19%) 489 (1%) 30 ± 9; 28.4 [24.0, 34.8] 1,616 (3%) 13,428 (28%) 12,548 (26%) 14,324 (30%) 6,235 (13%) 6.1 ± 3.8; 5.3 [3.5, 8.1]

150 (3%) 503 (10%) 633 (13%) 257 (5%) 1,026 (21%) 380 (8%) 828 (17%) 274 (6%) 861 (18%)

1,518 (3%) 5,807 (12%) 6,436 (13%) 2,192 (4%) 11,760 (24%) 4,170 (8%) 8,342 (17%) 1,829 (4%) 7,066 (14%)

<0.001 <0.001 <0.001 <0.001 0.1 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.9 <0.001 0.7 0.01 <0.001 0.07 0.8 <0.001 <0.001

4,836 (98%) 3,764 (77%) 2,854 (58%) 2,073 (42%) 1,326 (27%) 3,188 (65%) 1,548 (32%) 32 (0.7%) 181 (4%) 756 (15%) 1,237 (25%) 1,481 (30%) 296 (6%) 3,426 (70%)

47,840 (97%) 32,275 (66%) 22,200 (45%) 15,629 (32%) 9,640 (20%) 26,300 (54%) 10,643 (22%) 141 (0.3%) 2,030 (4%) 8,190 (17%) 8,694 (18%) 14,677 (30%) 1,211 (2%) 31,168 (63%)

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.1 0.02 <0.001 0.7 <0.001 <0.001

489 (10%) 414 (8%) 536 (11%) 149 (3%) 413 (8%)

4,645 (9%) 3,057 (6%) 4,294 (9%) 1,415 (3%) 2,692 (5%)

0.3 <0.001 <0.001 0.5 <0.001

2,777 (57%) 260 (5%) 134 (3%) 1,335 (27%) 85 (2%) 44 (0.9%) 277 (6%)

23,304 (47%) 5,056 (10%) 1,236 (3%) 14,692 (30%) 1,308 (3%) 328 (0.7%) 3,191 (7%)

<0.001 <0.001 0.4 <0.001 <0.001 0.07 0.02

Note: Values for continuous variables expressed as mean ± standard deviation or median [interquartile range]; for categorical variables, as count (percentage). Abbreviations: BMI, body mass index; PPI, proton pump inhibitor.

4

AJKD Vol XX | Iss XX | Month 2019

Original Investigation Maintenance Hemodialysis Patients Pool (N=557,387) Starting hemodialysis treatment on or before 9/30/2015 In Continental United States Sex, race, Hispanic not missing or unknown No preemptive kidney transplantation With Medical Evidence Report, version 1995 or later Eligible case Eligible controls ↓ | Subjects with any hip fracture claim | (N=44,001) | | ↓ 1st hip fracture cases (N=24,393) | Inpatient hip fracture claim followed by | surgery claims within 7 days | -

↓ Risk set: for each index date, i.e., the date with hip fracture cases - in year 1/1/2009-9/30/2015 - Index date in an in-patient hemodialysis period - Index date before the first kidney transplantation - Age 18-99 on the index date - With ≥90 days in-center hemodialysis prior to the index date - With ≥365 days Medicare Parts A and B coverage prior to the index date - With ≥1095 days Part D coverage and LowIncome Subsidy coverage prior to the index date - Without any hip fracture claim before or on the index date

↓

↓ Eligible hip fracture cases during hemodialysis, the 1st claim date as the index date (N=4,912) - in year 1/1/2009-9/30/2015 - Index date in an in-patient hemodialysis period - Index date before the first kidney transplantation - Age 18-99 on the index date - With ≥90 days in-center hemodialysis prior to the index date - With ≥365 days Medicare Parts A and B coverage prior to the index date - With ≥1095 days Part D coverage and Low-Income Subsidy coverage prior to the index date

↓

Random select 10 controls per case from risk set with M:N matching on each index date (4,912 cases and 49,120 controls)

Figure 1. Selection of hip fracture cases and controls with low-income subsidy coverage for 3 years

osteoporosis resulted in unchanged cOR point estimates for any, low, moderate, and high use: 1.25 (95% CI, 1.16-1.35), 1.21 (95% CI, 1.09-1.35), 1.31 (95% CI, 1.18-1.44), and 1.24 (95% CI, 1.10-1.40), respectively. In the model including use of vitamin D receptor activators and cinacalcet, the multivariable-adjusted

associations with low, moderate, and high use were also largely unchanged, with cORs of 1.20 (95% CI, 1.091.33), 1.31 (95% CI, 1.19-1.44), and 1.26 (95% CI, 1.12-1.41). Results of the complete case analysis were also largely unchanged (Table S4). Finally, in the shortterm analysis of new users, unadjusted and adjusted

Table 2. SSRI Use in Hip Fracture Cases and Controls and Measures of Association SSRI No use Any use <20% PDC 20%-<80% PDC ≥80% PDC

Cases (n = 4912) 3,098 (63%) 1,814 (37%) 580 (12%) 741 (15%) 493 (10%)

Controls (n = 49,120) 35,488 (72%) 13,632 (28%) 5,267 (11%) 5,283 (11%) 3,082 (6%)

Unadjusted OR (95% CI) 1.00 (referent) 1.52 (1.43-1.62) 1.26 (1.15-1.38) 1.60 (1.47-1.75) 1.83 (1.65-2.02)

P — <0.001 <0.001 <0.001 <0.001

Adjusteda OR (95% CI) 1.00 (referent) 1.25 (1.17-1.35) 1.20 (1.08-1.32) 1.31 (1.19-1.44) 1.26 (1.12-1.41)

P — <0.001 <0.001 <0.001 <0.001

Abbreviations: CI, confidence interval; OR, odds ratio; PDC, proportion of days covered (sum of the number of days of drug supplied in prescriptions divided by the number of days of the interval; here: 1,095 days); SSRI, selective serotonin reuptake inhibitor. a Adjusting for demographics, body mass index, dialysis vintage, comorbid conditions, cause of kidney failure, and census region.

AJKD Vol XX | Iss XX | Month 2019

5

Original Investigation Test for Interaction 1.37 (1.22-1.54)

p=0.007 1.19 (1.08-1.31)

1.26 (1.15-1.37)

1.24 (1.09-1.42)

p<0.001

Odds Ratio Figure 2. Effect of age and sex on hip fracture risk attributed to selective serotonin reuptake inhibitor exposure.

cORs of hip fracture status with any SSRI use versus no use were 1.82 (95% CI, 1.59-2.08) and 1.43 (95% CI, 1.23-1.67), respectively. No significant interaction with age, sex, BMI, race, or ethnicity was discovered in the short-term analysis. Discussion Patients treated by maintenance HD are at high risk for hip fracture. At the same time, depression is common in this population; the prevalence of depression was estimated to be 39.3% when screening with a questionnaire and 22.8% when determined by clinical interview.19 Small trials of SSRI use have reported improvements in symptoms20 and even markers of nutrition21 in patients with kidney failure treated by KRT, but the sample sizes of these studies were insufficient to be confident in SSRI utility. Cognitive behavioral therapy serves as another potential mode of intervention that remains insufficiently explored. In a study of 59 patients crossover randomly assigned to cognitive behavioral therapy versus wait-list control, those receiving behavioral therapy experienced improvements in depression, quality of life, and interdialytic weight gain.22 Another small pilot study examined behavioral therapy that included mindful meditation and reported improvements in self-reported scales of depression.23 With the benefits of antidepressant therapy undetermined and the risk for undesired effects present, any association between antidepressant drug therapy and risk for hip fracture is of potential clinical importance. In a large nationally representative study, we found robust evidence linking SSRI use to hip fracture risk that was present throughout a number of sensitivity analyses. Previous observational studies in the general population detected increased risk that was dose dependent.24-27 In unadjusted analysis, the association was larger for increased exposure in terms of the proportion of days covered with

6

prescription medication. This graded association was tempered in adjusted analysis. The association between SSRI use and hip fracture may be caused by mechanisms that can be categorized into 2 groups: increased fall risk in patients with depression and specifically with SSRI treatment28 and longer-term alteration of bone metabolism by SSRIs.29 In the general population, patients with depression are known to have increased risk for fractures, and SSRIs increase hip fracture risk. We found that the association was more pronounced in patients with recent SSRI treatment initiation. This finding is not unique and the systematic review performed by Eom et al12 found the highest risk to be within 6 weeks of SSRI treatment initiation. It is possible that the treatment of the underlying depression gradually lowers the risk. However, with uncertainty surrounding the therapeutic utility in patients with advanced kidney disease,6,30 a more plausible explanation may be that the fall risk is greatest at SSRI initiation and this risk becomes attenuated over time, notwithstanding the persistent influence on bone metabolism. Macri et al31 studied the number of falls resulting in emergency department visits or hospitalizations within 90 days of the index date in a cohort of 16,037 new antidepressant users and 16,037 nonusers in a longterm care facility. Even while accounting for demographics, depression, cognition, and functional status, new users were almost twice as likely to have had a clinically significant fall.31 However, this fall risk was attributable to new users of all antidepressants and may explain the association uncovered between hip fracture status and non-SSRI use. More specifically, the Swedish cohort study Osteoporotic Fractures in Men also reported reduced muscle strength and an increased risk for falls with SSRI use.32 In efforts to uncover the mechanism of this fall risk, Hegeman et al33 demonstrated that

AJKD Vol XX | Iss XX | Month 2019

Original Investigation among paroxetine users, changes in postural balance control may account for SSRI users’ increased risk for falling. Several studies strengthen the argument that SSRIs themselves deliver an impact on fracture risk independent of the underlying disease, depression. Given that serotonin transporters are also found in osteocytes, osteoblasts, and osteoclasts,34-36 modulation of the serotonin system seems highly likely to affect bone metabolism. The prospective study Osteoporotic Fractures in Men observed 3,014 participants, 90 of whom endorsed SSRI use at the beginning of the study. In the overall population, Kristjansdottir et al32 revealed that lower serotonin values demonstrated somewhat higher bone mineral density and BMI, but more falls and lower muscle strength. The 90 SSRI users displayed lower serotonin levels, as has been found in multiple other animal studies and studies of humans37-41 and correspondingly had lower muscle strength, increased falls, and higher risk for nonvertebral osteoporotic fractures. SSRI use was also associated with lower hip bone mineral density, a finding that has been reproduced in several clinical studies but is in opposition to the inverse relationship between serotonin levels and bone density previously described from the overall population.32 As suggested by these findings, pathology does not involve a simple deficiency or surplus of measured serotonin, and the effect of serotonin depends on the location of origin and type of synthesizing cells involved.29 Gut-derived serotonin dampens bone growth. Yadav et al42 describe how LRP5 regulates gut-derived serotonin production, which in turn reduces osteoblast proliferation. By contrast, central nervous system–derived serotonergic activity increases bone formation by suppressing sympathetic input to the bone, which normally inhibits osteoblast proliferation.29 These studies examined serotonin metabolism in patients without kidney disease, which is known to influence serotonergic activity.43 The complex role of serotonin on bone is understudied in patients with kidney disease. However, in sensitivity analyses examining treatment for chronic kidney disease–mineral and bone disorder (calcimimetics and vitamin D receptor activators) and osteoporosis (bisphosphonates), our findings of increased fracture risk with SSRI use in HD patients suggests a direct effect on fracture risk, either through altered bone metabolism or increased falls. A systematic review and meta-analysis from Eom et al12 examined 7 case-control and 5 cohort studies and determined from pooled multivariable-adjusted analysis that SSRI use was associated with all fractures (OR, 1.38; 95% CI, 1.27-1.49) and specifically hip fractures (OR, 2.06; 95% CI, 1.84-2.30). One possibility for the tempered strength of association in our study is that patients with kidney disease already experience some amount of disordered serotonergic activity through changes in leptin,43 lessening the potential for SSRIs to influence brainderived serotonin. AJKD Vol XX | Iss XX | Month 2019

Relatively increased hip fracture risk among males (vs females) and those 65 years or older (vs <65 years) may be attributable to medication side effects or changes in hormonal regulation. Precedence exists for this distinction along the lines of sex and age. Wang et al44 reported that in postmenopausal women, serotonin level correlated with bone mineral density of the whole body, femur, and distal radius. Many SSRIs have been associated with androgen reduction,45,46 which can differentially affect male bone metabolism. SSRIs have been associated with orthostasis, corrected QT interval (QTc) prolongation, dizziness, neuropathy, and sedation. It is conceivable that an older patient population has less capacity to adjust to SSRIrelated side effects that could influence fall risk. However, given the number of subgroups examined, there is certainly the possibility that these findings exist due to chance. Patients with kidney disease are vastly underrepresented in studies evaluating the safety and side effects of SSRI use.47 Being highly protein bound (and therefore minimally dialyzable) and metabolized hepatically, these medications are commonly thought to be tolerated in a similar capacity to patients without kidney disease. However, many metabolites are cleared by the kidneys and uremia can affect protein binding, downregulation of hepatic enzymes, and drug transporters.48 Thus, patients should be monitored for unwanted cardiovascular side effects such as QTc prolongation, arrhythmias, and orthostasis, as well as noncardiovascular side effects, noncardiovascular side effects, such as central nervous system depression. Last, no known interactions exist between SSRIs and phosphate binders or vitamin D analogues. However, cinacalcet is a CYP2D6 inhibitor, which can theoretically decrease the metabolism of certain SSRIs, namely fluoxetine, paroxetine, and fluvoxamine. This interaction does not exist with the majority of SSRIs and is unlikely to explain the impact on hip fracture risk. With a sizable portion of SSRI exposure among Medicare recipients5 treated by maintenance HD, further studies are essential to elucidate the risks and benefits that accompany SSRI use in patients with kidney disease. Along with the residual confounding potentially present in any observational study, there are limitations with the reliance on diagnosis codes. We lacked understanding of participants’ frailty or alcohol use. We were also unable to capture the specific indication for medication use. Around 20% of the sample exhibited either any exposure to SSRIs or non-SSRIs despite lacking a diagnosis of depression. This finding either suggests that depression went uncaptured in a portion of patients or that SSRIs may have been prescribed for diagnoses other than depression, such as neuropathy. A series of sensitivity analyses were performed to further limit the effect of uncaptured residual confounding. In particular, the strength of association with non-SSRI use was less robust, suggesting that SSRIs may 7

Original Investigation confer a specific risk among antidepressants. In addition to the fall risk associated with several psychotropic medications, SSRIs may induce further risk with influence on bone metabolism. Further differentiation to specific medications and dosages was not feasible without substantially limiting the power of this study. Perhaps the most obvious limitation is the inability to incorporate markers of mineral and bone disorder. Hypocalcemia, hyperphosphatemia, and secondary and tertiary hyperparathyroidism are all influential on bone turnover and quality. Unfortunately, these serum values were not available in our database. Last, the requirement of Medicare coverage presents a potential selection bias linked with survivorship. In a large population-based study, we found higher use of SSRIs in patients treated by maintenance HD who had a hip fracture compared with patients drawn from the same risk set who served as controls. In light of similar associations repeatedly demonstrated in the general population, this finding draws attention to potentially unwanted consequences of what has largely been a unimodal approach to depression in patients with kidney failure treated by KRT. Particular attention should be paid early on for side effects potentially attributable to SSRIs (hyponatremia, orthostasis, QTc prolongation, and arrhythmias) that may contribute to fall risk. The European Renal Best Practice guideline suggests that among patients with chronic kidney disease stages 3 to 5, treatment effect should be reevaluated after 8 to 12 weeks to avoid “prolonging ineffective medication,”30 and we echo this sentiment for patients with kidney failure treated by KRT, for whom the efficacy of SSRIs remains mostly unproven. Thus, the moderate increase in hip fracture risk associated with SSRI use is worth considering, particularly if patients and practitioners do not perceive benefit from the medication after appropriate duration of use. Supplementary Material

Authors’ Affiliations: Section of Nephrology and Selzman Institute for Kidney Health, Baylor College of Medicine (CV, JN, JY, SDN, WCW); Renal Section, Michael E. DeBakey Veterans Affairs Medical Center, Houston, TX (CV, SDN); and Division of Nephrology, Stanford University School of Medicine, Palo Alto, CA (MEM-R). Address for Correspondence: Chandan Vangala, MD, MS, Baylor College of Medicine, Section of Nephrology, One Baylor Plaza, Houston, TX 77030. E-mail: [email protected] Authors’ Contributions: Study design: CV, JN, WCW; bibliographic search: CV; data acquisition: WCW; statistical analysis: CV, JN, MEM-R; interpretation of results: CV, JY, SDN, WCW. Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved. Support: Dr Vangala was supported by a gift from Dr and Mrs Harold Selzman. Dr Winkelmayer receives support through the endowed Gordon A. Cain Chair in Nephrology at Baylor College of Medicine. Financial Disclosure: Dr Winkelmayer reports having served as an advisor or consultant unrelated to the topic of this article to Akebia/ Otsuka, Amgen, Astra-Zeneca, Bayer, Daichii-Sankyo, Relypsa, and Vifor-Fresenius Medical Care Renal Pharma. Dr Navaneethan serves on independent event adjudication committees for clinical trials sponsored by Bayer and Boehringer Ingelheim (unrelated to the topic of this article). The remaining authors declare that they have no relevant financial interests. Other Disclosures: Data were provided through a data use agreement between the National Institute for Diabetes and Digestive and Kidney Diseases (NIDDK) and Dr Winkelmayer. An NIDDK officer reviewed this manuscript for privacy and approved of its publication. Owing to federal research regulations, all cell sizes less than 10 are suppressed. Disclaimer: Data reported here were supplied by the USRDS. Interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the US government. Peer Review: Received December 29, 2018. Evaluated by 3 external peer reviewers and a statistician, with editorial input from an Acting Editor-in-Chief (Editorial Board Member Deirdre Sawinski, MD). Accepted in revised form July 16, 2019. The involvement of an Acting Editor-in-Chief to handle the peer-review and decision-making processes was to comply with AJKD’s procedures for potential conflicts of interest for editors, described in the Information for Authors & Journal Policies.

Supplementary File 1 (PDF) Figure S1: Selection of hip fracture cases and controls with lowincome subsidy coverage for 18 months. Table S1: Code algorithms used to identify outcomes and covariables. Table S2: Number of matched controls in each stratum. Table S3: Characteristics of hip fracture cases and controls in shortterm analysis. Table S4: Adjusted ORs from conditional logistic regression models applied in the complete and multiple imputation setting.

Article Information Authors’ Full Names and Academic Degrees: Chandan Vangala, MD, MS, Jingbo Niu, MD, DSc, Maria E. Montez-Rath, PhD, Jingyin Yan, MD, PhD, Sankar D. Navaneethan, MD, MS, MPH, and Wolfgang C. Winkelmayer, MD, MPH, ScD.

8

References 1. Beaubrun AC, Kilpatrick RD, Freburger JK, Bradbury BD, Wang L, Brookhart MA. Temporal trends in fracture rates and postdischarge outcomes among hemodialysis patients. J Am Soc Nephrol. 2013;24(9):1461-1469. 2. Mittalhenkle A, Gillen DL, Stehman-Breen CO. Increased risk of mortality associated with hip fracture in the dialysis population. Am J Kidney Dis. 2004;44(4):672-679. 3. Nair SS, Mitani AA, Goldstein BA, Chertow GM, Lowenberg DW, Winkelmayer WC. Temporal trends in the incidence, treatment, and outcomes of hip fracture in older patients initiating dialysis in the United States. Clin J Am Soc Nephrol. 2013;8(8):1336-1342. 4. Arneson TJ, Li S, Liu J, Kilpatrick RD, Newsome BB, St Peter WL. Trends in hip fracture rates in US hemodialysis

AJKD Vol XX | Iss XX | Month 2019

Original Investigation

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

patients, 1993-2010. Am J Kidney Dis. 2013;62(4):747754. Saran R, Robinson B, Abbott KC, et al. US Renal Data System 2018 Annual Data Report: epidemiology of kidney disease in the United States. Am J Kidney Dis. 2019;73(3)(suppl 1):SVII. Hedayati SS, Gregg LP, Carmody T, et al. Effect of sertraline on depressive symptoms in patients with chronic kidney disease without dialysis dependence: the CAST randomized clinical trial. JAMA. 2017;318(19):1876-1890. Palmer SC, Natale P, Ruospo M, et al. Antidepressants for treating depression in adults with end-stage kidney disease treated with dialysis. Cochrane Database Syst Rev. 2016;5: CD004541. Mehrotra R, Cukor D, Unruh M, et al. Comparative efficacy of therapies for treatment of depression for patients undergoing maintenance hemodialysis: a randomized clinical trial. Ann Intern Med. 2019;170(6):369-379. Khanassov V, Hu J, Reeves D, van Marwijk H. Selective serotonin reuptake inhibitor and selective serotonin and norepinephrine reuptake inhibitor use and risk of fractures in adults: a systematic review and meta-analysis. Int J Geriatr Psychiatry. 2018;33(12):1688-1708. Rabenda V, Nicolet D, Beaudart C, Bruyere O, Reginster JY. Relationship between use of antidepressants and risk of fractures: a meta-analysis. Osteoporos Int. 2013;24(1): 121-137. Liu B, Anderson G, Mittmann N, To T, Axcell T, Shear N. Use of selective serotonin-reuptake inhibitors or tricyclic antidepressants and risk of hip fractures in elderly people. Lancet. 1998;351(9112):1303-1307. Eom CS, Lee HK, Ye S, Park SM, Cho KH. Use of selective serotonin reuptake inhibitors and risk of fracture: a systematic review and meta-analysis. J Bone Miner Res. 2012;27(5): 1186-1195. Williams LJ, Berk M, Hodge JM, et al. Selective serotonin reuptake inhibitors (SSRIs) and markers of bone turnover in men. Calc Tissue Int. 2018;103(2):125-130. Tsapakis EM, Gamie Z, Tran GT, et al. The adverse skeletal effects of selective serotonin reuptake inhibitors. Eur Psychiatry. 2012;27(3):156-169. Briggs R, Carey D, McNicholas T, et al. The association between antidepressant use and orthostatic hypotension in older people: a matched cohort study. J Am Soc Hypertens. 2018;12(8):597-604.e591. Gebara MA, Lipsey KL, Karp JF, Nash MC, Iaboni A, Lenze EJ. Cause or effect? Selective serotonin reuptake inhibitors and falls in older adults: a systematic review. Am J Geriatr Psychiatry. 2015;23(10):1016-1028. Montez-Rath ME, Winkelmayer WC, Desai M. Addressing missing data in clinical studies of kidney diseases. Clin J Am Soc Nephrol. 2014;9(7):1328-1335. Van Buuren S, Brand JPL, Groothuis-Oudshoorn CGM, Rubin DB. Fully conditional specification in multivariate imputation. J Stat Comput Simul. 2006;76(12):1049-1064. Shirazian S, Grant CD, Aina O, Mattana J, Khorassani F, Ricardo AC. Depression in chronic kidney disease and endstage renal disease: similarities and differences in diagnosis, epidemiology, and management. Kidney Int Rep. 2017;2(1):94107. Blumenfield M, Levy NB, Spinowitz B, et al. Fluoxetine in depressed patients on dialysis. Int J Psychiatry Med. 1997;27(1):71-80. Koo JR, Yoon JY, Joo MH, et al. Treatment of depression and effect of antidepression treatment on nutritional status

AJKD Vol XX | Iss XX | Month 2019

22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

in chronic hemodialysis patients. Am J Med Sci. 2005;329(1):1-5. Cukor D, Ver Halen N, Asher DR, et al. Psychosocial intervention improves depression, quality of life, and fluid adherence in hemodialysis. J Am Soc Nephrol. 2014;25(1):196-206. Sohn BK, Oh YK, Choi JS, et al. Effectiveness of group cognitive behavioral therapy with mindfulness in end-stage renal disease hemodialysis patients. Kidney Res Clin Pract. 2018;37(1):77-84. Hung SC, Lin CH, Hung HC, Lin CL, Lai SW. Use of selective serotonin reuptake inhibitors and risk of hip fracture in the elderly: a case-control study in Taiwan. J Am Med Direct Assoc. 2017;18(4):350-354. Bolton JM, Metge C, Lix L, Prior H, Sareen J, Leslie WD. Fracture risk from psychotropic medications: a populationbased analysis. J Clin Psychopharmacol. 2008;28(4):384391. Richards JB, Papaioannou A, Adachi JD, et al. Effect of selective serotonin reuptake inhibitors on the risk of fracture. Arch Intern Med. 2007;167(2):188-194. Vestergaard P, Rejnmark L, Mosekilde L. Anxiolytics, sedatives, antidepressants, neuroleptics and the risk of fracture. Osteoporos Int. 2006;17(6):807-816. Carriere I, Farre A, Norton J, et al. Patterns of selective serotonin reuptake inhibitor use and risk of falls and fractures in community-dwelling elderly people: the Three-City cohort. Osteoporos Int. 2016;27(11):3187-3195. Lavoie B, Lian JB, Mawe GM. Regulation of bone metabolism by serotonin. Adv Exp Med Biol. 2017;1033:35-46. Nagler EV, Webster AC, Vanholder R, Zoccali C. Antidepressants for depression in stage 3-5 chronic kidney disease: a systematic review of pharmacokinetics, efficacy and safety with recommendations by European Renal Best Practice (ERBP). Nephrol Dial Transplant. 2012;27(10):3736-3745. Macri JC, Iaboni A, Kirkham JG, et al. Association between antidepressants and fall-related injuries among long-term care residents. Am J Geriatr Psychiatry. 2017;25(12):1326-1336. Kristjansdottir HL, Lewerin C, Lerner UH, et al. High serum serotonin predicts increased risk for hip fracture and nonvertebral osteoporotic fractures: the MrOS Sweden Study. J Bone Miner Res. 2018;33(9):1560-1567. Hegeman J, van den Bemt B, Weerdesteyn V, Nienhuis B, van Limbeek J, Duysens J. Unraveling the association between SSRI use and falls: an experimental study of risk factors for accidental falls in long-term paroxetine users. Clin Neuropharmacol. 2011;34(6):210-215. Bliziotes MM, Eshleman AJ, Zhang XW, Wiren KM. Neurotransmitter action in osteoblasts: expression of a functional system for serotonin receptor activation and reuptake. Bone. 2001;29(5):477-486. Westbroek I, van der Plas A, de Rooij KE, Klein-Nulend J, Nijweide PJ. Expression of serotonin receptors in bone. J Biol Chem. 2001;276(31):28961-28968. Battaglino R, Fu J, Spate U, et al. Serotonin regulates osteoclast differentiation through its transporter. J Bone Miner Res. 2004;19(9):1420-1431. Siesser WB, Sachs BD, Ramsey AJ, et al. Chronic SSRI treatment exacerbates serotonin deficiency in humanized Tph2 mutant mice. ACS Chem Neurosci. 2013;4(1):84-88. Trouvin JH, Gardier AM, Chanut E, Pages N, Jacquot C. Time course of brain serotonin metabolism after cessation of longterm fluoxetine treatment in the rat. Life Sci. 1993;52(18): Pl187-Pl192. Caccia S, Anelli M, Codegoni AM, Fracasso C, Garattini S. The effects of single and repeated anorectic doses of 5-

9

Original Investigation

40.

41.

42.

43.

10

hydroxytryptamine uptake inhibitors on indole levels in rat brain. Br J Pharmacol. 1993;110(1):355-359. Bismuth-Evenzal Y, Gonopolsky Y, Gurwitz D, Iancu I, Weizman A, Rehavi M. Decreased serotonin content and reduced agonist-induced aggregation in platelets of patients chronically medicated with SSRI drugs. J Affect Disord. 2012;136(1-2):99-103. Alvarez JC, Gluck N, Fallet A, et al. Plasma serotonin level after 1 day of fluoxetine treatment: a biological predictor for antidepressant response? Psychopharmacology. 1999;143(1): 97-101. Yadav VK, Ryu JH, Suda N, et al. Lrp5 controls bone formation by inhibiting serotonin synthesis in the duodenum. Cell. 2008;135(5):825-837. Pawlak D, Domaniewski T, Znorko B, et al. The impact of peripheral serotonin on leptin-brain serotonin axis, bone metabolism and strength in growing rats with experimental chronic kidney disease. Bone. 2017;105:1-10.

44. Wang Q, Chen D, Nicholson P, et al. The associations of serum serotonin with bone traits are age- and gender-specific. PloS One. 2014;9(10):e109028. 45. Hansen CH, Larsen LW, Sorensen AM, Halling-Sorensen B, Styrishave B. The six most widely used selective serotonin reuptake inhibitors decrease androgens and increase estrogens in the H295R cell line. Toxicol In Vitro. 2017;41:1-11. 46. Safarinejad MR. Evaluation of endocrine profile and hypothalamic-pituitary-testis axis in selective serotonin reuptake inhibitor-induced male sexual dysfunction. J Clin Psychopharmacol. 2008;28(4):418-423. 47. Hedayati SS, Yalamanchili V, Finkelstein FO. A practical approach to the treatment of depression in patients with chronic kidney disease and end-stage renal disease. Kidney Int. 2012;81(3):247-255. 48. Yeung CK, Shen DD, Thummel KE, Himmelfarb J. Effects of chronic kidney disease and uremia on hepatic drug metabolism and transport. Kidney Int. 2014;85(3):522-528.

AJKD Vol XX | Iss XX | Month 2019