AIMS is a prospective, longitudinal, multi-institutional clinical database of patients with RA managed by providers in clinical practice. A total of 72 US private and academic rheumatology practices participate in AIMS under Institutional Review Board (Advarra IRB Services, Columbia, MD, USA; IRB #PRO00044807) approval, with informed consent obtained from all enrolled patients. All patient data is deidentified, and management of clinical data is conducted in accordance with the Health Insurance Portability and Accountability Act (HIPAA) [18].

The MSRC identifies a molecular signature associated with inadequate achievement of ACR50 to TNFi therapy in patients with a clinical diagnosis of RA [15, 16]. The MSRC test underwent an update during the study period.

For patients tested prior to August 2021, the 23-biomarker MSRC included assessment of single-nucleotide polymorphisms (SNPs), gene expression features, anti-CCP, sex, BMI, PtGA of disease activity, and C-reactive protein (CRP) [19]. After August 2021, the MSRC test was updated and validated to a 19-biomarker version removing the SNP and CRP components and relying on gene expression, anti-CCP status, sex, BMI, and PtGA of disease activity [16].

The MSRC test produces a continuous result from 1 to 25. Pre-defined thresholds along this scale categorize patients as having very high, high, or no signal of inadequate response to TNF inhibitors at 6 months as measured by ACR50 criteria. Patients exceeding a threshold of 10.6 are identified as having a molecular signature of inadequate response. These response categories were optimized during analytical validation of the MSRC [16].

Gene expression feature analysis was performed in PAXgene Blood RNA tubes. Anti-CCP testing was conducted in serum separator tubes. Sample processing and anti-CCP testing was performed in the Scipher Medicine Laboratory (Durham, North Carolina; CAP# 8821838, CLIA# 34D2180776) on the Roche Cobas® system in accordance with standard operating procedures. RNA sequencing was performed at the Ambry Genetics Corporation (Aliso Viejo, CA) under Clinical Laboratory Improvement Amendments laboratory standard operating procedures, as previously described [16]. Algorithmic analysis of the MSRC was performed at the Scipher Medicine Laboratory. The MSRC test was analytically and clinically validated in both biologic-naïve and biologic-exposed patients with RA, demonstrating comparable performance characteristics between these patient populations [15, 16].

This study was designed to interrogate the AIMS database for physician use of the MSRC, b/tsDMARD treatment at baseline for each patient, and the subsequent treatment decisions made by physicians. Baseline characteristics, including previous csDMARD treatment, treatment at baseline, patient demographics, and baseline clinical variables were recorded (Table 1). The presence or absence of an inadequate response signature detected by the MSRC and the treatment prescribed after receiving MSRC results were documented for each. To evaluate alignment between treatment selection and MSRC test results, patients were characterized into two groups: (1) patients with a signature of inadequate response to TNFi who were prescribed a TNFi were noted to have received therapy that did not align with MSRC results; (2) patients with a signature of inadequate response to TNFi who were prescribed a non-TNFi as well as patients without a signature of inadequate response to TNFi were considered to have received a b/tsDMARD consistent with MSRC results.

Patients from the AIMS database who met the following inclusion criteria were included in this study: 18 years of age or older with a clinical diagnosis of RA by a rheumatologist; moderate or high disease activity at baseline [CDAI > 10]; naïve to or with a history of treatment with non-TNFi b/tsDMARD prior to enrollment; currently on a TNFi at the time of MSRC testing; must have baseline and treatment decision visit data available for analyses. Patients treated with non-TNFi b/tsDMARD at the time of MSRC testing were excluded. Patients were also excluded if (1) they were participating in other clinical studies and (2) patients’ physicians reported making a treatment decision prior to receiving MSRC test results.

Interventions consisted of the MSRC, described above. Two different versions of a questionnaire probed physician decision-making within the AIMS study, physicians completed: (1) version 1 of the questionnaire for patients enrolled between August 2020 and August 2021; and (2) version 2 for patients enrolled between August 2021 and February 2023. Improvements in questionnaire version 2 reflected an expansion of the MSRC intended use and in addition, versions of the questionnaire differed in how physicians were asked about use of the MSRC to guide treatment: version 1 used a “yes/no” question; version 2 asked if the MSRC was used with “yes/no” response, and category selection of how it was used. If the answer was “yes”, the options were “Start new therapy (new therapies include new products both within class and outside of class)”, “Continue existing therapy as originally prescribed”, and “Continue existing therapy, but modify dose and/or add an additional medication”. Both versions of the questionnaire documented reasons why the MSRC was not used to inform treatment decision-making.

The primary analyses evaluated observed physician behavior, reporting the percentage of patients who were prescribed and received a b/tsDMARD consistent with MSRC results, and the percentage of physicians who prescribed various treatments after receiving test results. Secondary analyses evaluated how physician self-reported perceptions of MSRC impact on decision-making, assessing the percentage of patients for whom their provider reported incorporating the MSRC as a precision medicine tool into their decision-making, and reasons for not incorporating test results into decision-making.

Statistical analyses were performed using R version 3.6.1 (www.rproject.org). Continuous variables were described using mean, standard deviation (SD), or median, interquartile range (IQR), and the number of non-missing observations. Categorical variables were summarized by providing the frequency counts and percentages, including a separate category for missing data. Normality is established using both Shapiro–Wilk test as well as visualization techniques such as Q–Q plot. Student’s t test and chi-squared tests were utilized for continuous and categorical variables, respectively, in the baseline table to determine differences between patient cohort subsets in primary analyses. For non-normally distributed continuous variables, Kruskal–Wallis test was utilized. All tests were conducted with a two-sided setting and, unless specified otherwise, the significance level was set at 0.05. Efforts were made to minimize the amount of missing data; no imputation of missing data was conducted.