Rapid identification of inflammatory arthritis and associated adverse events following immune checkpoint therapy: a machine learning approach

Abstract

Introduction: Immune checkpoint inhibitor-induced inflammatory arthritis (ICI-IA) poses a major clinical challenge to ICI therapy for cancer, with 13% of cases halting ICI therapy and ICI-IA being difficult to identify for timely referral to a rheumatologist. The objective of this study was to rapidly identify ICI-IA patients in clinical data and assess associated immune-related adverse events (irAEs) and risk factors.

Methods: We conducted a retrospective study of the electronic health records (EHRs) of 89 patients who developed ICI-IA out of 2451 cancer patients who received ICI therapy at Northwestern University between March 2011 to January 2021. Logistic regression and random forest machine learning models were trained on all EHR diagnoses, labs, medications, and procedures to identify ICI-IA patients and EHR codes indicating ICI-IA. Multivariate logistic regression was then used to test associations between ICI-IA and cancer type, ICI regimen, and comorbid irAEs.

Results: Logistic regression and random forest models identified ICI-IA patients with accuracies of 0.79 and 0.80, respectively. Key EHR features from the random forest model included ICI-IA relevant features (joint pain, steroid prescription, rheumatoid factor tests) and features suggesting comorbid irAEs (thyroid function tests, pruritus, triamcinolone prescription). Compared to 871 adjudicated ICI patients who did not develop arthritis, ICI-IA patients had higher odds of developing cutaneous (odds ratio [OR]=2.66; 95% Confidence Interval [CI] 1.63-4.35), endocrine (OR=2.09; 95% CI 1.15-3.80), or gastrointestinal (OR=2.88; 95% CI 1.76-4.72) irAEs adjusting for demographics, cancer type, and ICI regimen. Melanoma (OR=1.99; 95% CI 1.08-3.65) and renal cell carcinoma (OR=2.03; 95% CI 1.06-3.84) patients were more likely to develop ICI-IA compared to lung cancer patients. Patients on nivolumab+ipilimumab were more likely to develop ICI-IA compared to patients on pembrolizumab (OR=1.86; 95% CI 1.01-3.43).

Discussion: Our machine learning models rapidly identified patients with ICI-IA in EHR data and elucidated clinical features indicative of comorbid irAEs. Patients with ICI-IA were significantly more likely to also develop cutaneous, endocrine, and gastrointestinal irAEs during their clinical course compared to ICI therapy patients without ICI-IA.

Keywords: big data; electronic health records; immune checkpoint inhibitor-induced inflammatory arthritis; immune checkpoint inhibitors; immune-related adverse events; machine learning.

Figure 1

Methods diagram. (A) Manual adjudication of ICI cohort (N=2451 patients) for immune checkpoint inhibitor-induced inflammatory arthritis (ICI-IA) cases (N=89 patients). From the ICI cohort, 871 random patients without ICI-IA (ICI-NoArthritis) were adjudicated for all irAE. (B) Electronic health record data extraction of all diagnosis (ICD-9-CM, ICD-10-CM), medication (RxNorm), laboratory test (LOINC), and procedure (CPT) codes. Individual code occurrences were modified to specify whether they occurred before or after ICI initiation, and dichotomized to presence/absence of the code. Data was extracted for the full ICI cohort of 2451 patients. (C) Logistic regression (LR) and random forest (RF) machine learning models were trained on the EHR codes to identify ICI-IA. (D) Feature importance was analyzed to characterize ICI-IA patients in the EHR. Multivariate logistic regression was used to calculate odds ratios for development of ICI-IA given cancer and ICI regimen, as well as development of non-arthritis irAEs given ICI-IA versus ICI-noIA.

Figure 2

Model performance (area under the receiver operating characteristic curve, AUROC) versus percentage of the top features used to develop random forest and logistic regression models. Models maintain high performance with decreasing percentage of top features included before dropping performance with fewer than 0.1% of features or 31 features (vertical dotted line).

Figure 3

Key EHR codes in the machine learning models and association with ICI-induced inflammatory arthritis (ICI-IA). Left: the random forest model’s feature importance for identifying ICI-IA patients. Right: odds of the patient having an EHR code if they developed ICI-IA versus if they did not, by Fisher Exact test. Error bars are 95% confidence intervals. ‘ICI-IA’ codes are those directly relevant to ICI-IA. ‘irAE’ codes are those potentially describing other irAEs. ‘Other’ codes are those describing other parts of the patient medical history. The top codes are predominantly ICI-IA relevant codes, with a high concentration of relevant codes occupying the topmost importance. The top irAE related codes are endocrine (cortisol, thyroid function tests, estradiol), myositis, and cutaneous (medication order for triamcinolone and pruritus). The majority of the top codes are positively associated with ICI-IA. Codes are labeled by name, code vocabulary, code, and temporal modifier (before or after ICI therapy initiation).