\- Target trial emulation that will include Incident methotrexate users age ≥18 years in the REP between 1/1/2012-12/31/2022 (to allow at least three years follow-up through analysis date 12/31/2025) with ≥2 preceding codes for rheumatic disease
Treatment strategies:
* Guidelines group: monitor 3+ times in first 0-3 months, 3+ times in 3 to 12 months, and at least once every 4.5 months afterward.
* Reduced group: monitor 1-2 times in first 0-3 months, 1-2 times in 3 to 12 months, and at least once per year afterward (but not more than once every 4.5 months).
* Assignment procedures: We will use clone-censor-weighting (CCW) methods to explore the effect of both treatments in each patient. A negative control outcome such as trauma outcomes to assess for uncontrolled confounding.
* Time zero: start date of incident methotrexate use
* Follow up: Ends at methotrexate discontinuation (defined as not taking methotrexate for ≥6 months), death, loss to follow-up, or three years after baseline, whichever occurs first.
Data: Laboratory and prescription data were electronically available since 1/1/2010 and fully viewable since 1/1/2012. Therefore, this study will include all adults aged ≥18 years with incident low-dose (≤25mg/week) methotrexate use from 1/1/2012 on. To ensure the population would fall under the purview of the ACR methotrexate monitoring guidelines, we also required ≥2 codes (30 days apart) for a rheumatic condition within two years prior to index date of first methotrexate prescription, which we ascertained by prescription data and confirmed by manual chart review. A rheumatologist previously manually abstracted all methotrexate monitoring-related data including frequency as described previously (Reed et al., unpublished).
Outcomes:
* Primary outcome is end-organ injury including cirrhosis, severe cytopenia defined by blood product transfusion (RBCs, Plts, G-CSF), or death, all within the follow-up period above. The outcome will be reported as absolute risk per 100 person-years.
* For cirrhosis we will initially use a published code set of ICD-9 and ICD-10 codes with PPV \> 50% and then perform blinded, independent validation of outcome and date by chart review.
* Sample size/Power Calculation; A parallel, two-group design (where higher Poisson rates are considered worse) will be used to test whether the Group 2 (treatment) Poisson rate is non-inferior to the Group 1 (control) Poisson rate, with a non-inferiority ratio of 1.5 (H0: λ2 / λ1 ≥ 1.5 versus H1: λ2 / λ1 \< 1.5). The comparison will be made using a one-sided, two-sample, Poisson regression term Z-test using the variance calculation method with assumed true rates, with a Type I error rate (α) of 0.025. The dispersion is assumed to be 1. To detect a ratio of Poisson event rates (λ2 / λ1) of 1.25 (λ2 = 0.00413, λ1 = 0.0033), with a sample size of 1500 subjects in Group 1 and 1100 subjects in Group 2, with average exposure time 4, the power is 0.08092. The power was computed using PASS 2025, version 25.0.2. Anticipating a 20% dropout rate, 1875 subjects should be enrolled in Group 1, and 1375 in Group 2, to obtain final group sample sizes of 1500 and 1100, respectively.
* Missing data: Multiple imputation with predictive mean matching will be implemented for baseline data that are assumed missing at random (allows for inclusion of key predictors of both missingness and outcomes).26 Median/modal values of the multiple imputed values will be used for single imputation purposes.
Analysis plan:
* Estimate per-protocol effect via comparison of outcome under each treatment strategy using CCW estimation approach, adjusting for baseline confounders alone. Under the CCW estimation approach, we will only be able to estimate the per-protocol effect of reduced monitoring.
* noninferiority margin: We will survey all co-authors and use their mean answer for tolerable increased absolute risk per 100 person-years of end-organ injury or death to justify the benefits of reduced monitoring (e.g., reduced methotrexate avoidance, cost, time, iatrogenic harm) as the noninferiority margin. We will use one-sided alpha of 0.025
* Normally distributed continuous variables will be compared using t-tests, non-normally distributed continuous variables using the Wilcoxon rank sum tests, and categorical variables using the chi-squared or Fisher's exact tests (for variables with low cell sizes), as appropriate.
* Sensitivity analyses will be conducted for negative control outcome, patients on methotrexate monotherapy, no preceding cirrhosis, an eGFR of \< 45 mL/min, based on creatinine with CKD-EPI equation calculation, defining cytopenia using CIRT definition of Hgb \<8, plt \<50, or WBC \<2.9, serum albumin of \< 3.0 g/dL (30g/L)
