[Skip to Content]
[Skip to Content Landing]
Figure.
Subgroup Analysis for Risk of Encephalopathy by eGFR Category (Weighted Risk Difference and Weighted Risk Ratio)
Subgroup Analysis for Risk of Encephalopathy by eGFR Category (Weighted Risk Difference and Weighted Risk Ratio)

The 30-day risk of a hospital admission with encephalopathy, defined as main diagnosis of delirium, disorientation, transient alteration of awareness, transient cerebral ischemic attack, or unspecified dementia. To improve the specificity of this outcome, we only considered codes that were entered in the main diagnosis field of the database; this field contains a single diagnosis that most influences the patient’s length of hospital stay and/or is responsible for the greatest proportion of resource use. Inverse probability of treatment weighting on the propensity score was used to balance comparison groups on indicators of baseline health.12-14 The propensity score was estimated using multivariable logistic regression with 164 covariates chosen a priori (defined in eTable 7 in the Supplement). Patients in the reference group were weighted (propensity score/[1−propensity score]).12-14 This method produces a weighted pseudo sample of patients in the reference group with the same distribution of measured covariates as the exposed group.12,13 Weighted risk ratios and 95% CIs were obtained using modified Poisson regression,16 and weighted risk differences and 95% CIs were obtained using a binomial regression model with an identity link function. eGFR indicates estimated glomerular filtration rate.

Table 1.  
Baseline Demographic Characteristics of Older Adults With Chronic Kidney Disease Newly Prescribed Baclofen in Ontario, Canada (2007-2018)a
Baseline Demographic Characteristics of Older Adults With Chronic Kidney Disease Newly Prescribed Baclofen in Ontario, Canada (2007-2018)a
Table 2.  
Baseline Clinical Characteristics of Older Adults With Chronic Kidney Disease Newly Prescribed Baclofen in Ontario, Canada (2007-2018)a
Baseline Clinical Characteristics of Older Adults With Chronic Kidney Disease Newly Prescribed Baclofen in Ontario, Canada (2007-2018)a
Table 3.  
Risk of Encephalopathy in Older Adults With Chronic Kidney Disease Who Started a New Prescription for Baclofen (≥20 mg/d vs <20 mg/d)
Risk of Encephalopathy in Older Adults With Chronic Kidney Disease Who Started a New Prescription for Baclofen (≥20 mg/d vs <20 mg/d)
Supplement.

Table 1. Checklist of Recommendations for Reporting of Observational Studies Using the Reporting of Studies Conducted Using Observational Routinely Collected Health Data (RECORD) Guidelines

eTable 2. Coding Definitions for Demographic and Comorbid Conditions

eTable 3. Chronic Kidney Disease Equation (CKD-EPI) to Estimate the Glomerular Filtration Rate for Drug Dosing Adjustments

eTable 4. Summary of Case Reports of Baclofen-Induced Toxicity in Patients With Chronic Kidney Disease

eTable 5. Literature Search

eTable 6. Operating Characteristics of Hospital Diagnosis Codes Used to Define Primary and Secondary Outcomes

eTable 7. Variables Included in the Propensity Score Model

eTable 8. Dose Difference Within 30 Days After the Baclofen Dispense Date

eTable 9. Duration of Continuous Baclofen Dispensing During the Years 2007 to 2018

eTable 10. Baseline Characteristics of Older Adults With Chronic Kidney Disease Newly Prescribed Baclofen in Ontario, Canada (2007–2018)

eTable 11. Characteristics of 134 Patients With a Hospital Admission With Encephalopathy After Starting Baclofen

eTable 12. Frequency and Rate of Outcome Types

eTable 13. Prespecified Comparison of Baclofen Users and Nonusers

eTable 14. Baseline Characteristics of Older Adults With Chronic Kidney Disease Comparing High Dose Baclofen Versus Nonusers in Ontario, Canada (2007–2018)

eTable 15. Baseline Characteristics of Older Adults With Chronic Kidney Disease Comparing Low-Dose Baclofen Versus Nonusers in Ontario, Canada (2007–2018)

eTable 16. Post-hoc Survival Analysis in Older Adults With Chronic Kidney Disease Who Started a New Prescription for Baclofen at ≥20 mg/day vs <20 mg/day: Risk of Hospitalization With Encephalopathy

eTable 17. Post-hoc Analysis Accounting for the Correlation Between Patients Who Received a Prescription From the Same Physician

eTable 18. Post-hoc Analysis Using a Negative Control Exposure Where the Index Date Was Defined to be 90 Days Before the Baclofen Start Date

eTable 19. Post-hoc Analysis in Older Adults With Chronic Kidney Disease Who Started a New Prescription for Baclofen at ≥20 mg/day vs <20 mg/day: Risk of Heart Failure (a Negative Outcome)

eFigure 1. Flow Diagram of Cohort Build

eFigure 2. Post-hoc E-Value Analysis to Assess the Extent of Unmeasured Confounding That Would Be Required to Negate the Observed Results

eReferences.

1.
SymphonyHealth.  Top 200 Drugs-2016.https://symphonyhealth.prahs.com/wp-content/uploads/2017/04/Top-200-Drug-List-2016.pdf. Accessed August 15, 2019.
2.
Ghanavatian  S, Derian  A.  National Library of Medicine website. Baclofen (updated October 1, 2019). In: StatPearls. Treasure Island, FL: StatPearls Publishing.https://www.ncbi.nlm.nih.gov/books/NBK526037/. Accessed October 25, 2019.
3.
Vlavonou  R, Perreault  MM, Barrière  O,  et al.  Pharmacokinetic characterization of baclofen in patients with chronic kidney disease: dose adjustment recommendations.  J Clin Pharmacol. 2014;54(5):584-592. doi:10.1002/jcph.247PubMedGoogle ScholarCrossref
4.
Murphy  D, McCulloch  CE, Lin  F,  et al; Centers for Disease Control and Prevention Chronic Kidney Disease Surveillance Team.  Trends in prevalence of chronic kidney disease in the United States.  Ann Intern Med. 2016;165(7):473-481. doi:10.7326/M16-0273PubMedGoogle ScholarCrossref
5.
Statistics Canada.  Population estimates on July 1st by age and sex, by Canadian province and territory (number, both sexes). Summary. Tables. 2016.http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/demo31a-eng.htm. Accessed August 15, 2019.
6.
Vandenbroucke  JP, von Elm  E, Altman  DG,  et al; STROBE Initiative.  Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration.  PLoS Med. 2007;4(10):e297. doi:10.1371/journal.pmed.0040297PubMedGoogle Scholar
7.
Langan  SM, Schmidt  SA, Wing  K,  et al.  The reporting of studies conducted using observational routinely collected health data statement for pharmacoepidemiology (RECORD-PE).  BMJ. 2018;363(363):k3532. doi:10.1136/bmj.k3532PubMedGoogle ScholarCrossref
8.
Institute for Clinical Evaluative Sciences.  ICES Data and Privacy.https://www.ices.on.ca/Data-and-Privacy. Accessed August 15, 2019.
9.
Levey  AS, Stevens  LA, Schmid  CH,  et al; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration).  A new equation to estimate glomerular filtration rate.  Ann Intern Med. 2009;150(9):604-612. doi:10.7326/0003-4819-150-9-200905050-00006PubMedGoogle ScholarCrossref
10.
Garg  AX, Mamdani  M, Juurlink  DN, van Walraven  C; Network of Eastern Ontario Medical Laboratories (NEO-MeL).  Identifying individuals with a reduced GFR using ambulatory laboratory database surveillance.  J Am Soc Nephrol. 2005;16(5):1433-1439. doi:10.1681/ASN.2004080697PubMedGoogle ScholarCrossref
11.
Zhou  Z, Rahme  E, Abrahamowicz  M, Pilote  L.  Survival bias associated with time-to-treatment initiation in drug effectiveness evaluation: a comparison of methods.  Am J Epidemiol. 2005;162(10):1016-1023. doi:10.1093/aje/kwi307PubMedGoogle ScholarCrossref
12.
Sato  T, Matsuyama  Y.  Marginal structural models as a tool for standardization.  Epidemiology. 2003;14(6):680-686. doi:10.1097/01.EDE.0000081989.82616.7dPubMedGoogle ScholarCrossref
13.
Brookhart  MA, Wyss  R, Layton  JB, Stürmer  T.  Propensity score methods for confounding control in nonexperimental research.  Circ Cardiovasc Qual Outcomes. 2013;6(5):604-611. doi:10.1161/CIRCOUTCOMES.113.000359PubMedGoogle ScholarCrossref
14.
Austin  PC.  An introduction to propensity score methods for reducing the effects of confounding in observational studies.  Multivariate Behav Res. 2011;46(3):399-424. doi:10.1080/00273171.2011.568786PubMedGoogle ScholarCrossref
15.
Austin  PC, Stuart  EA.  Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.  Stat Med. 2015;34(28):3661-3679. doi:10.1002/sim.6607PubMedGoogle ScholarCrossref
16.
Zou  G.  A modified poisson regression approach to prospective studies with binary data.  Am J Epidemiol. 2004;159(7):702-706. doi:10.1093/aje/kwh090PubMedGoogle ScholarCrossref
17.
Statistics Canada.  Migration: interprovincial, 2011/2012. http://www.statcan.gc.ca/pub/91-209-x/2014001/article/14012-eng.htm. Accessed August 15, 2019.
18.
Zou  GY, Donner  A.  Extension of the modified Poisson regression model to prospective studies with correlated binary data.  Stat Methods Med Res. 2013;22(6):661-670. doi:10.1177/0962280211427759PubMedGoogle ScholarCrossref
19.
Haneuse  S, VanderWeele  TJ, Arterburn  D.  Using the E-value to assess the potential effect of unmeasured confounding in observational studies.  JAMA. 2019;321(6):602-603. doi:10.1001/jama.2018.21554PubMedGoogle ScholarCrossref
20.
Lipsitch  M, Tchetgen Tchetgen  E, Cohen  T.  Negative controls: a tool for detecting confounding and bias in observational studies.  Epidemiology. 2010;21(3):383-388. doi:10.1097/EDE.0b013e3181d61eebPubMedGoogle ScholarCrossref
Views 6,360
Citations 0
Original Investigation
November 9, 2019

Association of Baclofen With Encephalopathy in Patients With Chronic Kidney Disease

Author Affiliations
  • 1Institute for Clinical Evaluative Sciences, London, Ontario, Canada
  • 2Department of Epidemiology and Biostatistics, Western University, London, Ontario, Canada
  • 3Division of Nephrology, Department of Medicine, Western University, London, Ontario, Canada
JAMA. Published online November 9, 2019. doi:https://doi.org/10.1001/jama.2019.17725
Key Points

Question  Is there an increased risk of encephalopathy among patients with chronic kidney disease (CKD) initiating treatment with a higher vs lower dose of baclofen?

Findings  In this retrospective cohort study that included 15 942 patients with CKD, the 30-day incidence of encephalopathy among those prescribed baclofen at greater than or equal to 20 mg per day compared with less than 20 mg per day was 1.11% vs 0.42%, a difference that was statistically significant.

Meaning  Baclofen use among patients with CKD, particularly at higher doses, may be associated with an increased risk of encephalopathy.

Abstract

Importance  At least 30 case reports have linked the muscle relaxant baclofen to encephalopathy in patients with chronic kidney disease (CKD).

Objective  To compare the 30-day risk of encephalopathy in patients with CKD and newly prescribed baclofen at greater than or equal to 20 mg per day vs less than 20 mg per day. The secondary objective was to compare the risk of encephalopathy in baclofen users vs nonusers.

Design, Setting, and Participants  Retrospective population-based cohort study in Ontario, Canada (2007-2018) using linked health care data. Participants comprised 15 942 older adults (aged 66 years or older) with CKD (defined as an estimated glomerular filtration rate [eGFR] <60 mL/min/1.73 m2 but not receiving dialysis). The primary cohort was restricted to patients who were newly prescribed baclofen; participants in the secondary cohort were new users and nonusers.

Exposures  Prescription for oral baclofen greater than or equal to 20 mg per day vs less than 20 mg per day.

Main Outcomes and Measures  Hospital admission with encephalopathy, defined as a main diagnosis of delirium, disorientation, transient alteration of awareness, transient cerebral ischemic attack, or unspecified dementia within 30 days of starting baclofen. Inverse probability of treatment weighting on the propensity score was used to balance comparison groups on indicators of baseline health. Weighted risk ratios (RRs) were obtained using modified Poisson regression and weighted risk differences (RDs) using binomial regression. Prespecified subgroup analyses were conducted by eGFR category.

Results  The primary cohort comprised 15 942 patients with CKD (9699 [61%] women; median age, 77 years [interquartile range, 71-82]; 9707 [61%] patients started baclofen at ≥20 mg/d and 6235 [39%] at <20 mg/d). The primary outcome, hospitalization with encephalopathy, occurred in 108/9707 (1.11%) patients who started baclofen at greater than or equal to 20 mg per day and in 26/6235 (0.42%) who started baclofen at less than 20 mg per day; weighted RR, 3.54 (95% CI, 2.24 to 5.59); weighted RD, 0.80% (95% CI, 0.55% to 1.04%). In subgroup analysis, the absolute risk increased progressively at lower eGFR (weighted RD eGFR 45-59, 0.42% [95% CI, 0.19%-0.64%]; eGFR 30-44, 1.23% [95% CI, 0.62%-1.84%]; eGFR <30, 2.90% [95% CI, 1.30%-4.49%]; P for interaction, <.001]). In the secondary comparison with 284 263 nonusers, both groups of baclofen users had a higher risk of encephalopathy (<20 mg/d weighted RR, 5.90 [95% CI, 3.59 to 9.70] and ≥20 mg/d weighted RR, 19.8 [95% CI, 14.0 to 28.0]).

Conclusions and Relevance  Among older patients with CKD who were newly prescribed baclofen, the 30-day incidence of encephalopathy was increased among those prescribed higher doses compared with lower doses. If verified, these risks should be balanced against the benefits of baclofen use.

Introduction

Baclofen is a centrally acting γ-amino butyric acid agonist that was prescribed more than 8.3 million times in the United States in 2016.1 It is primarily used as a muscle relaxant for patients with spasticity, but it is also prescribed off-label for alcoholism, gastroesophageal reflux disease, nystagmus, and trigeminal neuralgia.2 Unlike other muscle relaxants that are primarily metabolized by the liver, baclofen is eliminated primarily unchanged in the urine, with a step-wise prolongation in the elimination half-life as kidney function declines.3 Chronic kidney disease (CKD) was estimated to affect 20% of older adults in 2011-2012,4 and at least 30 case reports have linked baclofen use to encephalopathy in patients with CKD within days of initiating the drug.

Despite these case reports, no clinical study has quantified the risk of baclofen-associated encephalopathy in patients with CKD. The objective of the present study was to examine the 30-day risk of hospitalization with encephalopathy in patients with CKD who were new users of oral baclofen. The primary objective was to compare the risk in patients prescribed greater than or equal to 20 mg per day vs less than 20 mg per day; a comparison of 2 dosing groups helps minimize concerns that the results are confounded by indication. The secondary objective was to examine the risk in these patient groups compared with nonusers (ie, patients with CKD with no evidence of baclofen use).

Methods
Study Design and Setting

This study was conducted using linked administrative health care databases in the province of Ontario, Canada (2007-2018). All Ontario residents (≈13 million) have universal access to hospital care and physician services through a government-funded single-payer system.5 Those aged 65 years and older (≈2.2 million) also receive universal prescription drug coverage. The use of data in this study was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a research ethics board. Study reporting follows recommended guidelines for observational studies that use routinely collected health data (eTable 1 in the Supplement).6,7

Data Sources

Patient characteristics, prescription drug use, covariate information, and outcome data were obtained from 8 health care databases at the Institute for Clinical Evaluative Sciences (ICES).8 The data sets were linked using unique encoded identifiers and analyzed at ICES (Canadian Institute for Health Information–Discharge Abstract Database, ICES Physician Database, National Ambulatory Care Reporting System, Ontario Drug Benefit Database, Ontario Health Insurance Plan Database, Ontario Laboratories Information System, Ontario Mental Health Reporting System, and the Registered Persons Database). Information on hospital admissions and diagnoses are coded by trained personnel using the International Classification of Diseases, 10th Revision (ICD-10) system; personnel only consider physician-recorded diagnoses in a patient’s medical chart when assigning codes and do not review or interpret symptoms or test results. Additional information on the databases, variable definitions, and administrative codes are provided in eTable 2 in the Supplement.

Patients

The primary cohort included adults aged 66 years and older with CKD (defined as having an estimated glomerular filtration rate [eGFR] <60 mL/min/1.73 m2 but not receiving dialysis) and who were newly dispensed oral baclofen from an outpatient pharmacy between January 1, 2007, and March 1, 2018. The prescription fill date was the patient’s cohort entry date (the index date); patients could enter the cohort only once. The age restriction was applied to ensure that all patients in the study had at least 1 year of prescription drug coverage. The GFR was estimated using the most recent outpatient serum creatinine measurement before the index date (using the isotope dilution mass spectroscopy–traceable enzymatic method), and eGFR was calculated using the chronic kidney disease epidemiology (CKD-EPI) equation (justification for use of this equation to guide baclofen dosing is provided in eTable 3 in the Supplement).9 In Ontario, many older adults have at least 1 outpatient serum creatinine measurement in routine care each year, in which a single value represents a stable (or long-term) value.10 Patients with no serum creatinine measurement in the year before the index date were excluded.

To ensure that patients were new baclofen users, those with any evidence of baclofen use (including combination drug prescriptions) in the 180-day period before the index date were excluded, as were those who were discharged from the hospital or emergency department within 2 days before the index date (in Ontario, patients who start a baclofen prescription during a hospital admission would have their outpatient prescription dispensed on the same day or the day after hospital discharge). Patients with an implausible baclofen dose (<5 mg/d or >80 mg/d) were excluded.

The secondary cohort included all patients in the primary cohort, as well as patients without any evidence of baclofen use (ie, nonusers). Nonusers were randomly assigned an index date (a simulated baclofen start date) that followed the same distribution of index dates as baclofen users.11

Exposure

The primary exposure of interest was oral baclofen of greater than or equal to 20 mg per day, which is the median dose reported in cases of baclofen toxicity in patients with CKD (eTable 4 in the Supplement; literature search in eTable 5 in the Supplement). An active comparator, oral baclofen at less than 20 mg per day, was chosen for the primary comparison to reduce the influence of indication bias. Other active comparators were considered (eg, other types of muscle relaxants including dantrolene and tizanidine); however, these drugs are prescribed infrequently in Ontario. For the secondary objective, each group of baclofen users (ie, those prescribed ≥20 mg/d and those prescribed <20 mg/d) were compared separately with nonusers (ie, patients with CKD with no evidence of baclofen use).

Outcomes

All primary and secondary outcomes were prespecified. The primary outcome was the 30-day risk of a hospital admission with encephalopathy, defined as a main diagnosis of delirium, disorientation, transient alteration of awareness, transient cerebral ischemic attack, or unspecified dementia (unclear diagnosis of dementia). Transient ischemic attack was included as one of the encephalopathy outcomes since baclofen-related toxicity in patients with CKD has been characterized in some case reports by symptoms (visual disturbances, numbness, slurred speech) similar to those observed in transient ischemic attack (eTable 4 in the Supplement). The outcome and time frame were defined based on a review of the literature (studies summarized in eTable 4 in the Supplement); in these studies, the median time to encephalopathy after baclofen initiation was 2.5 days (interquartile range [IQR], 1-4). To improve the specificity of this outcome, we only considered ICD codes that were entered in the main diagnosis field of the database; this field contains a single diagnosis that most influenced the patient’s length of hospital stay, that was responsible for the greatest proportion of resource use, or both. Alternative definitions of encephalopathy were examined in sensitivity analyses. Secondary outcomes were hospitalization with delirium as the main diagnosis, hospitalization for any cause, and all-cause mortality. Diagnostic codes for all outcome variables and information on their validation and interpretation are provided in eTable 6 in the Supplement.

Statistical Analysis

Inverse probability of treatment weighting on the propensity score was used to balance comparison groups on recorded indicators of baseline health, including known indications for baclofen use (including off-label indications).12-14 The propensity score was estimated using multivariable logistic regression with 164 covariates chosen a priori (defined in eTable 7 in the Supplement). Patients in the reference group were weighted (propensity score/[1−propensity score]).12-14 This method produces a weighted pseudo sample of patients in the reference group with the same distribution of measured covariates as the exposed group.12,13 Between-group differences in baseline characteristics were compared using standardized differences in both the unweighted and weighted samples15 (differences >10% were considered meaningful). Weighted risk ratios and 95% CIs were obtained using modified Poisson regression,16 and weighted risk differences and 95% CIs were obtained using a binomial regression model with an identity link function. Two-tailed P values of less than .05 were interpreted as statistically significant. Because of the potential for type I error due to multiple comparisons, findings of the secondary, subgroup, and sensitivity analyses should be interpreted as exploratory. All variables in this study were complete except for baclofen prescriber specialty (7% missing; defined in a separate category) and patient income quintile (0.3% missing; recoded as the middle quintile). Emigration from the province, which occurs at a rate of 0.5% per year, was the only reason for lost follow-up designation.17

Prespecified sensitivity analyses were conducted to examine 2 alternative definitions of encephalopathy: (1) any hospital admission or emergency department visit with encephalopathy (ie, all relevant diagnostic codes considered, not just those entered in the main diagnostic field), and (2) hospital admission with receipt of an urgent computed tomography scan of the head. A prespecified subgroup analysis by baseline eGFR (grouped into 3 categories) was conducted by including an interaction term in the model. To address the secondary objective, each group of baclofen users (≥20 mg/d and <20 mg/d) was compared separately with nonusers on the risk of encephalopathy; inverse probability of treatment weighting was performed separately for these comparisons.

Five post hoc sensitivity analyses were conducted to assess the robustness of the main results: (1) a survival analysis (with 30-day follow-up censoring on death) that met the proportional hazards assumption (nonsignificant high dose × follow-up time interaction term; P = .06); (2) an analysis that accounted for the correlation between patients who received a prescription from the same physician18; (3) an E-value analysis to assess the extent of unmeasured confounding that would be required to negate the observed results19; (4) an analysis using a negative control exposure20 (in which the index date was defined to be 90 days before the baclofen start date); and (5) an analysis using a negative control outcome20 (in which the outcome was defined as hospitalization with heart failure). Analyses were conducted using SAS statistical software, version 9.4 (SAS Institute Inc).

Results
Patients

The primary cohort included 15 942 older adults with an eGFR of less than 60 mL/min/1.73 m2 (median age, 77 years; 61% women) who were newly dispensed baclofen at an outpatient pharmacy. The flow diagram for the cohort build is shown in eFigure 1 in the Supplement. Baclofen prescriptions were dispensed by 4128 unique pharmacies and prescribed by 4977 unique physicians (86% were primary care physicians). Overall, 66% of patients had an eGFR between 45 and 59 mL/min/1.73 m2, 27% between 30 and 45 mL/min/1.73 m2, and 7% had an eGFR less than 30 mL/min/1.73 m2. The median prescribed dose of baclofen in each eGFR category was 20 mg per day (IQR, 10-30).

Of 15 942 patients prescribed baclofen, 9707 (60.9%) started at greater than or equal to 20 mg per day (median, 30 mg/d [IQR, 20-30]) and 6235 (39.1%) started at less than 20 mg per day (median, 10 mg/d [IQR, 10-10]). While the baclofen product monograph indicates that the initial dose may be increased by 5 mg every 3 days until a desired clinical response is reached, no change in follow-up was observed in the initial median daily dose for either group (eTable 8 in the Supplement). In those prescribed greater than or equal to 20 mg per day, the median duration of continuous baclofen dispensing was 15 days (IQR, 10-30), and in those prescribed less than 20 mg per day, the median duration was 30 days (IQR, 14-30) (eTable 9 in the Supplement).

Characteristics of patients who started baclofen at greater than or equal to 20 vs less than 20 mg per day are shown in Table 1 and Table 2 (the full set of 164 characteristics is shown in eTable 10 in the Supplement). Before weighting, all standardized differences were less than 10% except for age, sex, living in a long-term care residence, and dementia. After weighting, the 2 groups were balanced on these and the other 160 variables, including type of prescriber, recorded indications for baclofen use, comorbidities, baseline eGFR, and concurrent medications (eTable 10 in the Supplement).

Hospitalization With Encephalopathy

The primary outcome, hospitalization with encephalopathy, occurred in 108 of 9707 patients (1.11%) who started baclofen at greater than or equal to 20 mg per day and in 26 of 6235 patients (0.42%) who started at less than 20 mg per day. The median time from starting baclofen to hospitalization was 3 days (IQR, 2-5) for patients who started baclofen at greater than or equal to 20 mg per day and 8 days (IQR, 3-12) for patients who started baclofen at less than 20 mg per day. Additional descriptive characteristics of these patients are shown in eTable 11 in the Supplement. Aggregate event rates for outcome type (ie, delirium, disorientation, transient cerebral ischemic attack, transient alteration of awareness, and unspecified dementia) are shown in eTable 12 in the Supplement.

Starting baclofen at greater than or equal to 20 mg per day vs less than 20 mg per day was associated with a higher 30-day risk of hospitalization with encephalopathy (weighted risk ratio, 3.54 [95% CI, 2.24 to 5.59]; weighted risk difference, 0.80% [95% CI, 0.55% to 1.04%]). Starting baclofen at greater than or equal to 20 mg per day vs less than 20 mg per day was also associated with a higher risk of hospitalization with delirium and all-cause hospitalization, but not all-cause mortality (Table 3).

Prespecified Sensitivity and Subgroup Analyses

Results were consistent when the outcome was defined as (1) any hospital admission or emergency department visit with encephalopathy; and (2) hospital admission with receipt of an urgent computed tomography scan of the head (Table 3). The results of the subgroup analyses by baseline eGFR categories are shown in the Figure. The weighted risk ratios and risk differences for encephalopathy increased progressively as eGFR declined; however, only the additive interaction was statistically significant (P < .001; P = .88 for multiplicative interaction).

Risk of Encephalopathy in Baclofen Users vs Nonusers

A hospital admission with encephalopathy occurred in 165/284 263 (0.06%) patients with no evidence of baclofen use. In comparison with nonusers, both groups of baclofen users had a higher risk of encephalopathy (<20 mg/d and ≥20 mg/d): (weighted risk ratio for patients with <20 mg/d, 5.90 [95% CI, 3.59 to 9.70]; weighted risk difference, 0.35% [95% CI, 0.18% to 0.51%]) (weighted risk ratio for patients with ≥20 mg/d, 19.8 [95% CI, 14.0 to 28.0]; weighted risk difference, 1.06% [95% CI, 0.85% to 1.27%]) (eTable 13 in the Supplement; characteristics of users and nonusers in eTable 14 and 15 in the Supplement).

Post Hoc Sensitivity Analyses

Results were consistent when the data were analyzed using a Cox proportional hazards regression (eTable 16 in the Supplement) and in an analysis that accounted for the correlation between patients who received prescriptions from the same physician (eTable 17 in the Supplement). The E-value for the risk ratio was 6.54, and the lower confidence bound for the primary outcome was 3.91, indicating that substantial unmeasured confounding would be needed to reduce the observed association or its 95% CI to the null (eFigure 2 in the Supplement). Study results were also supported by sensitivity analyses that used a negative exposure control (eTable 18 in the Supplement) and a negative outcome control (eTable 19 in the Supplement).

Discussion

In this study of older adults with CKD, those who started a prescription for baclofen at greater than or equal to 20 mg per day had a significantly greater risk of hospitalization with encephalopathy compared with those who started baclofen at less than 20 mg per day. Results were consistent in multiple sensitivity analyses and when alternative definitions of encephalopathy were analyzed. In a secondary comparison with patients who had no evidence of baclofen use, both groups of baclofen users (<20 mg/d and ≥20 mg/d) had a greater risk of encephalopathy.

Many patients benefit from using baclofen as a muscle relaxant or for several off-label indications including alcoholism, gastroesophageal reflux disease, nystagmus, and trigeminal neuralgia.2 This study was not designed to answer the question of whether the potential benefits of baclofen outweigh its risks, and clinicians will need to judge this on a patient-by-patient basis.

This population-based study of 15 942 older adults confirms and extends the findings of 30 international case reports linking baclofen use with encephalopathy in patients with CKD (eTable 4 in the Supplement). The findings of the present study are generalizable. The study was conducted in the setting of usual clinical care and included a representative sample of older adults with CKD in Ontario, Canada, where older adults have universal prescription drug coverage. Inverse probability of treatment weighting was used to help ensure that comparison groups were similar on baseline characteristics; however, even before weighting, groups were balanced on 98% of measured characteristics. Several sensitivity analyses were conducted and all supported the main findings. In particular, the magnitude of the E-values suggests that the observed association is unlikely to be explained by unmeasured confounding.

Limitations

This study has several limitations. First, the observational study design precludes reaching causal conclusions about the association between baclofen and encephalopathy, and the study requires replication before definitive conclusions can be reached. Second, despite the use of accurate information on baclofen dispensing, it was not possible to know the proportion of patients who took their medication as prescribed. Third, the patients studied were aged 66 years and older (likely to be at higher risk of encephalopathy), and so generalizability of the study findings to younger patients (who may be less prone to adverse drug events) is uncertain, although approximately half of the patients described in the case report studies were younger than 66 years. Fourth, the benefit-risk ratio of baclofen use was not assessed in this study. Fifth, the use of administrative data meant that the primary outcome definition was restricted to diagnostic codes recorded for the patients’ hospital stays, and granular data on patient symptoms were lacking. For example, baclofen use has been associated with several manifestations of encephalopathy, including confusion, drowsiness, and decreased consciousness, but occasionally vertigo, visual disturbances, numbness, nystagmus, and slurred speech. Many of these symptoms would not warrant hospital admission, and thus the overall incidence of baclofen-associated encephalopathy may be underestimated in this study. Sixth, information on serum baclofen levels was not available in the laboratory database, which precluded any corroboration in this study that excessive serum concentrations of baclofen were the cause for encephalopathy. Seventh, some misclassification in the outcome of encephalopathy is expected in this study because the codes are likely insensitive; however, there is no reason to believe that misclassification occurred differentially between exposure groups.

Conclusions

Among older patients with CKD who were newly prescribed baclofen, the 30-day incidence of encephalopathy was increased among those prescribed higher compared with lower doses. If verified, these risks should be balanced against the benefits of baclofen use.

Back to top
Article Information

Corresponding Author: Flory T. Muanda, MD, PhD, ICES Western, Victoria Hospital, 800 Commissioners Rd, Room ELL-215, London, ON N6A 5W9, Canada (flory.muanda-tsobo@lhsc.on.ca).

Accepted for Publication: October 9, 2019.

Published Online: November 9, 2019. doi:10.1001/jama.2019.17725

Author Contributions: Drs Muanda and Garg had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Muanda, Blake, Garg.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Muanda, Weir, Sontrop, Moist, Garg.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Muanda, Dixon, McArthur, Garg.

Obtained funding: Garg.

Administrative, technical, or material support: Muanda, Bathini, Chauvin, Garg.

Supervision: Garg.

Conflict of Interest Disclosures: Dr Moist reported receipt of personal fees (honoraria) from Otsuka and Janssen outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by the ICES (Institute for Clinical Evaluative Sciences) Western site. ICES is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). Core funding for ICES Western is provided by the Academic Medical Organization of Southwestern Ontario (AMOSO), the Schulich School of Medicine and Dentistry (SSMD), Western University, and the Lawson Health Research Institute (LHRI). The ICES Kidney, Dialysis and Transplantation team, at the ICES Western facility, is supported by a grant from the Canadian Institutes of Health Research (CIHR). Dr Muanda is a recipient of a CIHR and Mitacs postdoctoral award. Dr Garg is supported by the Dr Adam Linton Chair in Kidney Health Analytics and a Clinician Investigator Award from the CIHR.

Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The opinions, results, and conclusions are those of the authors and are independent from the funding sources. No endorsement by ICES, AMOSO, SSMD, LHRI, CIHR, or the MOHLTC is intended or should be inferred. Parts of this material are based on data and information compiled and provided by the Canadian Institutes of Health Information (CIHI). However, the analyses, conclusions, opinions, and statements expressed herein are those of the authors, and not necessarily those of CIHI.

Additional Information: This research was conducted by members of the ICES Kidney, Dialysis and Transplantation team, at the ICES Western facility.

References
1.
SymphonyHealth.  Top 200 Drugs-2016.https://symphonyhealth.prahs.com/wp-content/uploads/2017/04/Top-200-Drug-List-2016.pdf. Accessed August 15, 2019.
2.
Ghanavatian  S, Derian  A.  National Library of Medicine website. Baclofen (updated October 1, 2019). In: StatPearls. Treasure Island, FL: StatPearls Publishing.https://www.ncbi.nlm.nih.gov/books/NBK526037/. Accessed October 25, 2019.
3.
Vlavonou  R, Perreault  MM, Barrière  O,  et al.  Pharmacokinetic characterization of baclofen in patients with chronic kidney disease: dose adjustment recommendations.  J Clin Pharmacol. 2014;54(5):584-592. doi:10.1002/jcph.247PubMedGoogle ScholarCrossref
4.
Murphy  D, McCulloch  CE, Lin  F,  et al; Centers for Disease Control and Prevention Chronic Kidney Disease Surveillance Team.  Trends in prevalence of chronic kidney disease in the United States.  Ann Intern Med. 2016;165(7):473-481. doi:10.7326/M16-0273PubMedGoogle ScholarCrossref
5.
Statistics Canada.  Population estimates on July 1st by age and sex, by Canadian province and territory (number, both sexes). Summary. Tables. 2016.http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/demo31a-eng.htm. Accessed August 15, 2019.
6.
Vandenbroucke  JP, von Elm  E, Altman  DG,  et al; STROBE Initiative.  Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration.  PLoS Med. 2007;4(10):e297. doi:10.1371/journal.pmed.0040297PubMedGoogle Scholar
7.
Langan  SM, Schmidt  SA, Wing  K,  et al.  The reporting of studies conducted using observational routinely collected health data statement for pharmacoepidemiology (RECORD-PE).  BMJ. 2018;363(363):k3532. doi:10.1136/bmj.k3532PubMedGoogle ScholarCrossref
8.
Institute for Clinical Evaluative Sciences.  ICES Data and Privacy.https://www.ices.on.ca/Data-and-Privacy. Accessed August 15, 2019.
9.
Levey  AS, Stevens  LA, Schmid  CH,  et al; CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration).  A new equation to estimate glomerular filtration rate.  Ann Intern Med. 2009;150(9):604-612. doi:10.7326/0003-4819-150-9-200905050-00006PubMedGoogle ScholarCrossref
10.
Garg  AX, Mamdani  M, Juurlink  DN, van Walraven  C; Network of Eastern Ontario Medical Laboratories (NEO-MeL).  Identifying individuals with a reduced GFR using ambulatory laboratory database surveillance.  J Am Soc Nephrol. 2005;16(5):1433-1439. doi:10.1681/ASN.2004080697PubMedGoogle ScholarCrossref
11.
Zhou  Z, Rahme  E, Abrahamowicz  M, Pilote  L.  Survival bias associated with time-to-treatment initiation in drug effectiveness evaluation: a comparison of methods.  Am J Epidemiol. 2005;162(10):1016-1023. doi:10.1093/aje/kwi307PubMedGoogle ScholarCrossref
12.
Sato  T, Matsuyama  Y.  Marginal structural models as a tool for standardization.  Epidemiology. 2003;14(6):680-686. doi:10.1097/01.EDE.0000081989.82616.7dPubMedGoogle ScholarCrossref
13.
Brookhart  MA, Wyss  R, Layton  JB, Stürmer  T.  Propensity score methods for confounding control in nonexperimental research.  Circ Cardiovasc Qual Outcomes. 2013;6(5):604-611. doi:10.1161/CIRCOUTCOMES.113.000359PubMedGoogle ScholarCrossref
14.
Austin  PC.  An introduction to propensity score methods for reducing the effects of confounding in observational studies.  Multivariate Behav Res. 2011;46(3):399-424. doi:10.1080/00273171.2011.568786PubMedGoogle ScholarCrossref
15.
Austin  PC, Stuart  EA.  Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.  Stat Med. 2015;34(28):3661-3679. doi:10.1002/sim.6607PubMedGoogle ScholarCrossref
16.
Zou  G.  A modified poisson regression approach to prospective studies with binary data.  Am J Epidemiol. 2004;159(7):702-706. doi:10.1093/aje/kwh090PubMedGoogle ScholarCrossref
17.
Statistics Canada.  Migration: interprovincial, 2011/2012. http://www.statcan.gc.ca/pub/91-209-x/2014001/article/14012-eng.htm. Accessed August 15, 2019.
18.
Zou  GY, Donner  A.  Extension of the modified Poisson regression model to prospective studies with correlated binary data.  Stat Methods Med Res. 2013;22(6):661-670. doi:10.1177/0962280211427759PubMedGoogle ScholarCrossref
19.
Haneuse  S, VanderWeele  TJ, Arterburn  D.  Using the E-value to assess the potential effect of unmeasured confounding in observational studies.  JAMA. 2019;321(6):602-603. doi:10.1001/jama.2018.21554PubMedGoogle ScholarCrossref
20.
Lipsitch  M, Tchetgen Tchetgen  E, Cohen  T.  Negative controls: a tool for detecting confounding and bias in observational studies.  Epidemiology. 2010;21(3):383-388. doi:10.1097/EDE.0b013e3181d61eebPubMedGoogle ScholarCrossref
×