Low-Dose Methotrexate and Serious Adverse Events Among Older Adults With Chronic Kidney Disease

Key Points Question Is there a higher 90-day risk of serious adverse events among adults with chronic kidney disease (CKD) who start low-dose methotrexate vs hydroxychloroquine? Findings In this cohort study that included 4618 propensity-matched adults with CKD, the 90-day risk of serious adverse events among those who started low-dose methotrexate vs hydroxychloroquine was 3.55% vs 1.73%, a statistically significant difference. Meaning In this study, adults with CKD who started low-dose methotrexate had a higher 90-day risk of serious adverse events compared with those who started hydroxychloroquine.


Introduction
Low-dose methotrexate is a disease-modifying antirheumatic drug (DMARD) primarily used for the treatment of rheumatoid arthritis and psoriasis. 1,2It is also prescribed as an immunomodulator for patients with Crohn disease. 3The usual dose of methotrexate for these conditions is 5 to 25 mg/wk, 1,2,4 and occasionally, the maximum weekly dose is increased to 35 mg/wk. 1,2,4Low-dose methotrexate is also prescribed off-label for dermatomyositis, eczema, systemic lupus erythematosus, and systemic sclerosis. 5In the United States, 5.9 million methotrexate prescriptions were dispensed in 2019. 6thotrexate is primarily eliminated by the kidneys, with 80% to 90% excreted unchanged in the urine. 5,7To avoid toxic effects, product monographs and prescribing guidelines recommend that methotrexate be started at a low dose in patients with chronic kidney disease (CKD) (eTable 1 in Supplement 1).Recommended dose adjustments in CKD are largely based on clinical experience (eTable 1 in Supplement 1).A pharmacokinetic study showed that methotrexate clearance was slower in patients with lower levels of creatinine clearance, and its half-life was 2-fold higher in patients with a creatinine clearance less than 45 mL/min compared with those with a creatinine clearance greater than 80 mL/min (eTable 2 in Supplement 1). 8rious adverse events associated with low-dose methotrexate use include myelosuppression, serious infections, and hepatotoxic effects. 9While these risks have been reported in several randomized clinical trials, [10][11][12][13] there is limited information on whether these risks are amplified in patients with CKD.At least 50 case reports and 2 observational studies (one retrospective cohort study and one observational study using a spontaneous reporting system) of patients with CKD suggest that the risk of toxic effects with low-dose methotrexate (5-35 mg/wk) is substantial (literature search and summary of studies appear in eTables 3 and 4 in Supplement 1).In one cohort study of 120 patients with CKD, almost one-third developed toxic effects after starting methotrexate for rheumatoid arthritis (35 of 120 [29.2%]).This proportion increased in a stepwise manner across categories of patients with lower kidney function (P = .02).In another study of 88 patients with rheumatoid arthritis who started treatment with low-dose methotrexate, 3 times as many patients with CKD developed hematological toxic effects than those without CKD (33 of 88 [37.5%] vs 594 of 5560 [10.7%];P < .001)(eTable 4 in Supplement 1).
We conducted a population-based study to examine the risk of serious adverse events in older adults with CKD who started low-dose methotrexate.The primary objective was to assess the 90-day risk of a hospital visit with myelosuppression, sepsis, pneumotoxic effects, or hepatotoxic effects in older adults with CKD who started methotrexate at 5 to 35 mg/wk compared with those who started hydroxychloroquine (200 to 400 mg/d).The secondary objective was to separately examine the risk of toxic effects in patients with CKD starting 2 different doses of methotrexate (5 to <15 mg/wk and 15 to 35 mg/wk) compared with those starting hydroxychloroquine.

Study Design and Setting
This study was conducted using linked administrative health care databases in the province of Ontario, Canada (2008-2021).All Ontario residents (approximately 15 million) have universal access to hospital care and physician services through a government-funded single-payer system. 14

Data Sources
Data for this study were obtained from 8 health care databases housed at ICES. 16 The following data sets were linked using unique encoded identifiers and analyzed at ICES: the Canadian Institute for Health Information Discharge Abstract Database, the ICES-derived Physician Database, the National Ambulatory Care Reporting System database, the Ontario Drug Benefit Database, the Ontario Health Insurance Plan database, the Ontario Laboratories Information System database, the Ontario Mental Health Reporting System Database, and the Registered Persons Database.Data on hospital admissions and diagnoses were coded by trained personnel using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) system; personnel only consider physician-recorded diagnoses in a patient's medical record when assigning codes and do not review or interpret symptoms or test results.8][19][20][21] Except for prescriber data (8% missing, defined as a separate category) and neighborhood income quintile (0.2% missing, recorded as the middle quintile), the databases were complete for all variables in this study.The only reason for loss to follow-up was emigration from the province, which is less than 0.5% per year on average. 22The codes used to ascertain comorbidities and outcomes are detailed in eTable 5 in Supplement 1.We assessed comorbidities in the 5-year period before cohort entry and health care use in the 1-year period before cohort entry.We used a 120-day look-back period to ascertain use of other prescription drugs because the Ontario Drug Benefits program allows a maximum prescription duration of 100 days.

Patients
The primary cohort included adults aged 66 years and older who had an estimated glomerular filtration rate (eGFR) of less than 60 mL/min/1.73m 2 (excluding patients receiving dialysis and kidney transplant recipients) and who were newly dispensed oral methotrexate or hydroxychloroquine from an outpatient pharmacy between January 1, 2008, and July 31, 2021.The prescription fill date was the patient's cohort entry date (the index date); patients could only enter the cohort once.The age restriction was applied to ensure all patients in the study had at least 1 year of provincial universal prescription drug coverage before their index date.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 new race-free CKD Epidemiology (CKD-EPI) equation: 142 × min([serum creatinine concentration in μmol/L/88.4]/ĸ,1)α × max([serum creatinine concentration in μmol/L/88.4]/ĸ,1)-1.200× 0.9938Age ×1.012 [if female] ĸ = 0.7 if female and 0.9 if male; α = -0.241if female and −0.302 if male; min = the minimum of serum creatinine concentration/ĸ or 1; max = the maximum of serum creatinine concentration/ĸ or 1. 23 Ontarians with African ancestry represented less than 5% of the population in 2016. 24In Ontario, many older adults have at least 1 outpatient serum creatinine measurement in routine care each year, where a single value represents a stable, chronic value. 25Patients with no serum creatinine measurements in the year before the index date were excluded.
To ensure that patients were new users, we excluded those with any evidence of methotrexate or hydroxychloroquine use in the 180-day period before the index date and those with any evidence of other DMARD use (conventional or biologics) in the 30 days before the index date.We also excluded patients who were discharged from the hospital or emergency department within 2 days before the index date.(In Ontario, DMARD outpatient prescriptions are given on the same or next day after a hospital stay.)To ensure generalizability to usual prescribing, we excluded patients who started nonstandard doses (ie, methotrexate <5 mg/wk or >35 mg/wk; hydroxychloroquine <200 mg/d or >400 mg/d).A diagram summarizing the cohort creation steps is provided in eFigure 1 in Supplement 1.
The secondary cohorts included (1) patients who started methotrexate at 15 to 35 mg/wk and those who started hydroxychloroquine and (2) patients who started methotrexate at 5 to <15 mg/wk and those who started hydroxychloroquine.The 2 cohorts were not mutually exclusive (ie, hydroxychloroquine users were included in both cohorts).To better characterize the risk

Exposure
The primary exposure was low-dose oral methotrexate (5-35 mg/wk).An active comparator, oral hydroxychloroquine (200-400 mg/d), was chosen as the primary comparator to reduce the influence of indication bias.As a DMARD, hydroxychloroquine has some of the same indications as methotrexate, including rheumatoid arthritis, systemic lupus erythematosus, and dermatomyositis. 26We considered other DMARDs for the comparator (eg, leflunomide and sulfasalazine) 2,27-29 ; however, these drugs have been associated with the primary outcome (eg, hepatotoxic effects for leflunomide and myelosuppression for sulfasalazine). 28,29For the secondary objective, patients starting methotrexate at 15 to 35 mg/wk and 5 to less than 15 mg/week were compared separately with hydroxychloroquine users.

Outcomes
All primary and secondary outcomes were prespecified.The primary outcome was the 90-day risk of a hospital visit (ie, an emergency department visit or a hospital admission) with myelosuppression (defined as a diagnosis of aplastic anemia, neutropenia, thrombocytopenia, or pancytopenia), sepsis, pneumotoxic effects, or hepatotoxic effects.These adverse events have been reported in several randomized clinical trials of low-dose methotrexate [10][11][12][13] and have been linked to myelosuppression, sepsis, pneumotoxic effects, and hepatotoxic effects in patients with CKD (eTable 4 in Supplement 1).
We combined these outcomes into a composite to increase statistical power because, individually, they occur infrequently in Ontario.The 90-day follow-up was defined based on a review of the literature (eTable 4 in Supplement 1); in these studies, the median (IQR) time to toxic effects after methotrexate initiation was 26 (IQR 8-60) days.Acute kidney injury was not included in the primary outcome because a recent randomized clinical trial did not support a causal association with low-dose methotrexate. 30e secondary outcomes were the components of the primary composite outcome, all-cause hospitalization, and all-cause mortality.Diagnostic codes for all outcome variables and information on their validation and interpretation are provided in eTable 6 in Supplement 1.

Statistical Analysis
Analyses were conducted using SAS version 9.4 (SAS Institute).Propensity score matching was used to balance comparison groups on indicators of baseline health, including all known indications for methotrexate use (including off-label indications). 31,32The propensity score was estimated using multivariable logistic regression with 140 covariates chosen (eTable 7 in Supplement 1) because they are known confounders or risk factors for study outcomes, including the year of cohort entry, which serve as a proxy for changes in health care practices and drug use during the study period.This selection was based on subject matter knowledge and previous literature. 31,32Using greedy matching, we matched each low-dose methotrexate user (1:1) to a hydroxychloroquine user based on the logit of the propensity score (within a caliper of ±0.2 SDs). 33In simulation studies, greedy matching with a caliper width of 0.2 SD produced less bias than optimal and nearest-neighbor matching. 34,35tween-group differences in baseline characteristics were compared using standardized differences in both the unmatched and matched samples (differences >10% were considered meaningful). 36Risk ratios and 95% CIs were obtained using modified Poisson regression, 37 and a sandwich variance estimator was used to account for the correlation within matched pairs.Risk differences and 95% CIs were obtained using a binomial regression model with an identity link function which also accounted for the correlation within matched pairs.We interpreted To comply with ICES privacy regulations to minimize the risk of identification, values of cells with 5 or fewer patients were reported as less than 6.
A prespecified subgroup analysis for the primary composite outcome by baseline eGFR (grouped into 3 categories: Ն60, 45-59, and <45 mL/min/1.73m 2 ) was conducted by including an interaction term in the model.To ensure that baseline health indicators were balanced between comparison groups, we recalculated the propensity score within each eGFR category. 38We then matched each methotrexate user 1:1 to a hydroxychloroquine user based on the logit of the propensity score (within a caliper of ±0.2 SDs) as in the primary analysis. 33 address the secondary objective, each group of low-dose methotrexate users (15 to 35 mg/wk and 5 to <15 mg/wk) was compared separately to hydroxychloroquine users on the risk of the primary outcome.Propensity score matching was performed separately for these comparisons.
We conducted the following post hoc analyses to assess the robustness of the main results and examine the potential for bias.We performed a survival analysis with a 90-day follow-up censoring on death (the proportional hazards assumption was met: methotrexate × follow-up time interaction term, P = .99).To assess the potential for indication bias, we reran the primary analysis using 2 different weighting methods 39 and we compared those who started methotrexate at 15 to 35 mg/wk with those who started methotrexate at 5 to less than 15mg/wk.To examine the potential for unmeasured confounding, we conducted ( 1) an E-value analysis to assess the extent of unmeasured confounding that would be required to negate the observed results 40 and (2) an analysis using a negative-control outcome, 41 defined as receipt of a hearing test during an outpatient or hospital visit.
To account for surveillance bias, we compared the proportion of patients in each group who received a test for complete blood count or liver function or who received a chest radiograph in follow-up.
Respectively, these tests may be used to assess myelosuppression, hepatoxic effects, and pneumotoxic effects (the components of the primary outcome).To control for potential surveillance bias, we repeated the primary analysis restricted to patients who received at least 1 of these tests in follow-up.We also repeated the primary analysis using (1) a more restrictive outcome definition (ie, only counting hospital admissions with the primary outcome, but not emergency department visits) and (2) a 30-day follow-up period.

Patients
The primary cohort included 6909 older adults with an eGFR of less than 60 mL/min/

Study Outcomes
The primary outcome, the 90-day risk of a hospital visit with myelosuppression, sepsis, pneumotoxic but not a hospital visit with sepsis or all-cause mortality (Table 2).

Prespecified Subgroup Analysis
The results of the subgroup analyses by baseline eGFR categories are shown in Table 3.The risk ratios and risk differences for the primary outcome increased significantly and progressively as eGFR declined (eg, eGFR <45 mL/min/1.73m 2 : RR, 2.79 [95% CI, 1.51-5.13];P = .003for additive interaction; P = .008for multiplicative interaction).The characteristics of low-dose methotrexate and hydroxychloroquine users within eGFR categories are presented in eTables 12 to 14 in Supplement 1.

Risk of the Primary Outcome in New Users of 2 Different Methotrexate Doses vs Hydroxychloroquine
Starting low-dose methotrexate at 15 to 35 mg/week vs hydroxychloroquine was associated with a higher 90-day risk of the primary composite outcome (52 of 1357 [3.83%]  Abbreviations: CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate.
a Unless otherwise specified, baseline characteristics were assessed on the cohort entry date.
b A value greater than 10 is interpreted as a meaningful difference.
c Income was categorized into fifths of average neighborhood income on the index date.
d The most recent eGFR measurement in the 7-to-365-day period before the index date.
e Baseline comorbidities were assessed in the 5-year period before the index date.
f Presence of kidney disease is a variable in the Charlson comorbidity index, so that all individuals have a minimum score of 2.
g Total number of health care visits and tests in the 12-month period before the index date.
h Medication use was examined in the 120-day period before the index date.

JAMA Network Open | Pharmacy and Clinical Pharmacology
Low-Dose Methotrexate and Adverse Events Among Older Adults With CKD  a Reference group was hydroxychloroquine.
b The propensity score was estimated using multivariable logistic regression with 140 covariates chosen a priori (eTable 7 in Supplement 1).Greedy matching was used to match each low-dose methotrexate user (1:1) to a hydroxychloroquine user based on the logit of the propensity score (within a caliper of ±0.2 SDs). 33Risk ratios and 95% CIs were obtained using modified Poisson regression 34 and risk differences and 95% CIs were obtained using a binomial regression model with an identity link function.
c The 90-day risk of a hospital visit with myelosuppression, sepsis, pneumotoxic effects, or hepatotoxic effects.The components of the primary outcome are not mutually exclusive, as such, some individuals experience more than 1 of the components of the primary outcome.To comply with ICES privacy regulations to minimize the risk of identification, specific values of cells with 5 or fewer patients were suppressed (reported as <6).As a result, the number of adverse events for hepatotoxic effects is not presented in the table.
d Defined as a diagnosis of aplastic anemia, neutropenia, thromobocytopenia, or pancytopenia.The components of the primary outcome are not mutually exclusive, as such, some individuals experience more than 1 of the components of the primary outcome.To comply with ICES privacy regulations to minimize the risk of identification, specific values of cells with 5 or fewer patients were suppressed (reported as NR).As a result, the number of adverse events for this outcome is not presented in the matched cohort in the table.a The propensity score was estimated using multivariable logistic regression with 140 covariates chosen a priori (eTable 7 in Supplement 1).Greedy matching was used to match each low-dose methotrexate user (1:1) to a hydroxychloroquine user based on the logit of the propensity score (within a caliper of ±0.2 SDs). 33Risk ratios and 95% CIs were obtained using modified Poisson regression 34 and risk differences and 95% CIs were obtained using a binomial regression model with an identity link function.
b The 90-day risk of a hospital visit with myelosuppression, sepsis, pneumotoxic effects, or hepatotoxic effects.
c To ensure that baseline health indicators were balanced between comparison groups, the propensity score was recalculated within each eGFR category. 35Each methotrexate drug user was then matched 1:1 to a hydroxychloroquine user based on the logit of the propensity score (within a caliper of ±0.2 SDs) as in the primary analysis. 33d P value for additive interaction.
e P value for multiplicative interaction.

JAMA Network Open | Pharmacy and Clinical Pharmacology
Low

JAMA Network Open | Pharmacy and Clinical Pharmacology Low
Those aged 65 years and older (approximately 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 or informed consent from participants.Study reporting follows recommended guidelines for pharmacoepidemiological studies that use routinely collected health data. 15-Dose Methotrexate and Adverse Events Among Older Adults With CKD

Table 1 .
Baseline Characteristics of Older Adults With CKD Newly Prescribed Low-Dose Methotrexate vs Hydroxychloroquine in Ontario, Canada, 2008-2021 a vs 16 of 1357 [1.18%]; risk

Table 2 .
Risk of a Hospital Visit With Myelosuppression, Sepsis, Pneumotoxic Effects, or Hepatotoxic Effects in Older Adults With CKD Within 90 Days of Starting a New Prescription for Low-Dose Methotrexate vs Hydroxychloroquine a Abbreviations: NA, not applicable; NNH, number needed to harm; NR, not reported.

Table 3 .
Subgroup Analysis for the Risk of the Primary Composite Outcome by eGFR Category a,b

Network Open | Pharmacy and Clinical Pharmacology Low
-Dose Methotrexate and Adverse Events Among Older Adults With CKD 26.Hydroxychloroquine: drug information.UpToDate.Accessed March 31, 2023.https://www.uptodate.com/contents/hydroxychloroquine-drug-information 27.UpToDate.Alternatives to methotrexate for the initial treatment of rheumatoid arthritis in adults.Accessed March 31, 2023.https://www.uptodate.com/contents/alternatives-to-methotrexate-for-the-initial-treatment-ofrheumatoid-arthritis-in-adults?search=rheumatoid%20arthritis%20treatment&topicRef=7491&source=see_link 28.UpToDate.Leflunomide: drug information.Accessed July 11, 2022.https://www.uptodate.com/contents/Austin PC.Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies.Pharm Stat.2011;10(2):150-161.doi:10.1002/pst.43336.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.-Dose Methotrexate and Adverse Events Among Older Adults With CKD Frequency of each component in the matched cohort eTable 12. Baseline characteristics of older adults with an eGFR>60 mL/min/1.73m2newly prescribed low-dose methotrexate (LD MTX) vs those newly prescribed hydroxychloroquine (HCQ) in Ontario, Canada (2008-2021) eTable 13.Baseline characteristics of older adults with an eGFR between 45 and <60 mL/min/1.73m2newly prescribed low-dose methotrexate (LD MTX) vs those newly prescribed hydroxychloroquine (HCQ) in Ontario, Risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease within 90 days of starting a new prescription for low-dose methotrexate at 5 to <15 mg/week vs a new prescription for hydroxychloroquine eTable 18. Baseline characteristics of older adults with chronic kidney disease newly prescribed low-dose methotrexate (LD MTX) at 5 to <15 mg/week vs those newly prescribed hydroxychloroquine (HCQ) in Ontario, Canada (2008-2021) eTable 19.Survival analysis in older adults with chronic kidney disease who started a new prescription for low-dose methotrexate-vs a new prescription for hydroxychloroquine: risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity eTable 20.Risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease within 90 days of starting a new prescription for low-dose methotrexate vs. a new prescription for hydroxychloroquine using Inverse probability of treatment weighting analysis eTable 21.Risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease within 90 days of starting a new prescription for low-dose methotrexate vs a new prescription for hydroxychloroquine using Fine stratification weighting eTable 22. Risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease within 90 days of starting a new prescription for low-dose methotrexate at 15 to 35 mg/week vs 5 to <15 mg/week eTable 23.Baseline characteristics of older adults with chronic kidney disease newly prescribed low-dose methotrexate (LD MTX) at 15 to 35mg/week vs 5 to<15 mg/week in Ontario, Canada (2008-2021) eTable 24.Risk of an outpatient or hospital visit with hearing disorders in older adults with chronic kidney disease within 90 days of starting a new prescription for low-dose methotrexate vs a new prescription for hydroxychloroquine eTable 25.Proportion of patients who received a test for a complete blood count, liver function, or a chest radiograph within 90 days of starting a new prescription for low-dose methotrexate vs. a new prescription for hydroxychloroquine eTable 26.Risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease who received at least one test within 90 days of starting a new prescription for low-dose methotrexate vs a new prescription for hydroxychloroquine eTable 27.Risk of a hospital admission with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease within 90 days of starting a new prescription for low-dose methotrexate vs a new prescription for hydroxychloroquine eTable 28.Risk of a hospital visit with myelosuppression, sepsis, pneumotoxicity, and hepatotoxicity in older adults with chronic kidney disease within 30 days of starting a new prescription for low-dose methotrexate vs. a new prescription for hydroxychloroquine eFigure 1. Study design diagram comparing use of methotrexate vs hydroxychloroquine and the risk of serious adverse events in patients with chronic kidney disease eFigure 2. E-value analysis to assess the extent of unmeasured confounding that would be required to negate the observed results eReferences.