Predictors of Infection in Rheumatoid Arthritis Patients Using Anti-Tumor Necrosis Factor Agents

Background: Rheumatoid arthritis (RA) is an incurable autoimmune disease that can cause permanent joint damage and loss of function. Anti-tumor necrosis factor (antiTNF) agents inhibit the function of tumor necrosis factor (TNF), which leads to a reduction in the progression of joint damage due to inflammation. However, an increased risk of serious infections in RA patients using anti-TNF agents has been observed in previous studies. This increased risk may be due to the immunologic disturbance of the RA disease process itself, the immunosuppressive properties of anti-rheumatic drug therapies, or co-existing risk factors for infection present in RA patients. Herein, our aim is to assess potential predictors of hospitalized infection in RA patients using anti-TNF agents. Objective: Our objective is to determine if patients with RA who are prescribed the anti-TNF agents; adalimumab, etanercept, or infliximab, are at an increased risk of having a serious infection. In addition, we sought to identify potential predictors of an increased risk of infection in RA patients using anti-TNF agents. Methods: A nested case-control study was conducted using de-identified data from the Clinformatics DataMart (OptumInsight, Eden Prairie, MN), an administrative health claims database from a large national private insurer. An initial cohort of 78,657 patients with ≥1 RA diagnosis was identified. Patients were included from this initial cohort based on age, enrollment eligibility, number of RA diagnoses, exposure to an anti-TNF agent, and excluded based on certain comorbidities. A final RA cohort sample of 15,181 patients was formed. A follow-up period of 1 year was selected to analyze serious infections requiring hospitalization; these events were identified with a comprehensive set of ICD-9 codes for serious infections requiring inpatient admission. Patients were classified as cases if they experienced a serious infection during the 1year follow up. The final selected cases and controls were matched on a 1:1 ratio based on gender, region, and RA cohort entry date (quarter, year). A total of 155 cases and 155 controls were identified. Both univariable and multivariable conditional logistic regression models were built to produce a final multivariable predictive model. Results: Among, RA patients using anti-TNF agents, those with recent prednisone use were 1.873 times more likely to have a hospitalized infection (95% confidence interval [CI] 1.015-3.458). Patients with comorbid diabetes were 2.963 times more likely to experience a hospitalized infection (95% CI 1.445-6.078) and patients with comorbid chronic obstructive pulmonary disease (COPD) were 9.233 times more likely to experience a hospitalized infection (95% CI 2.755-30.947). Lastly, patients with a previous history of infection were 8.984 times more likely to have a hospitalized infection (95% CI 1.895-42.595). No associations between anti-TNF agent (adalimumab, infliximab, or etanercept) or incident/prevalent anti-TNF use and hospitalized infection were observed. Conclusion: The use of specific anti-TNF agents was not independently associated with an increased risk of hospitalized infection in RA patients. Predictors associated with hospitalized infection in RA patients using anti-TNF agents included recent prednisone use, comorbid diabetes, comorbid COPD, and previous history of infection.

psoriatic arthritis, plaque psoriasis, and ankylosing spondylitis. 6 The introduction of anti-TNF drugs beginning in the late 1990s revolutionized the treatment of rheumatoid arthritis (RA) as they have been shown to be very effective at reducing inflammation in RA patients. [7][8][9] Tumor necrosis factor (TNF) plays an important role in host cell defense as a proinflammatory cytokine involved in systemic inflammation. 10 This inflammatory response against harmful bacteria and viruses is controlled and regulated by the body under normal conditions. In patients with RA, macrophages overproduce TNF leading to inflammation of the joints. Anti-TNF agents inhibit the function of TNF by preventing TNF from binding to its receptor, which reduces the progression of joint damage due to inflammation. 10,11 An uncommon, yet significant side effect associated with anti-TNF drug use is the increased risk of serious infections. 12,[14][15][16] The prescribing information for the anti-TNF agents; adalimumab, etanercept, and infliximab contains black box warnings ordered by the FDA indicating the increased risk of serious infection leading to hospitalization or death and recommending discontinued use if a patient develops a serious infection. [17][18][19] This increased risk of infection may be due to the immunologic disturbance of the RA disease process, the immunosuppressive properties of anti-  The objective of this study was to assess risk factors for serious infections requiring hospitalization in RA patients using anti-TNF agents. This study will provide clinicians and RA patients with more knowledge on which factors are predictive of an increased risk of infection which may help reduce and prevent such infections. Furthermore, the use of a large, national administrative claims database representative of the privately insured population in the United States will produce more generalizable results on potential risk factors for infection among patients with RA in the United States.

Study Design and RA Cohort Definition
A nested case-control study was conducted using data from the Clinformatics TM DataMart (OptumInsight, Eden Prairie, MN), an administrative health claims database from a large national private insurer. The database contains de-identified patient level data across multiple categories including medical claims, pharmacy claims, and administrative data. We selected a nested case-control design because it is efficient for the study of rare outcomes and good for assessing multiple exposures. The outcome of interest in this study, hospitalized infection, is relatively rare with rates of infection ranging from 6-7% during extended follow-up time periods. 22,24 Data from the study time period of January 1, 2010 to December 31, 2013 was used in the analyses.
For inclusion into the RA cohort, both cases and controls had to be 18 to 63 years old at the time of their entry into the RA cohort (for a timeline of the RA cohort, please see Figure 1).

Identification of Cases and Controls
Cases and controls were derived from the aforementioned RA cohort (for a timeline of the nested case-control design, please see Figure 2). Cases were defined as patients who had an inpatient hospital admission code in any position for infection in the 12-months following the cohort entry date.

Drug Exposure Determination and Potential Predictors
Cases and controls were considered exposed to an anti-TNF agent if they had ≥1 pharmacy or medical claim for adalimumab, etanercept, or infliximab in the 90-days prior to the index date. Patients could be exposed to ≥1 anti-TNF agent during this 90-  Collinearity was assessed in the final model to determine if any of the independent variables were highly correlated with one another. A variance inflation factor (VIF) ≥10 and/or a tolerance (TOL) value ≤0.10 would indicate variables with possible collinearity. 29 The odds ratios (OR) for hospitalized infection risk for each risk factor was calculated with 95% confidence intervals and a p-value <0.05 was considered statistically significant. Statistical analyses were performed using SAS® Version 9.4.

Study Population Characteristics
There were 176,745 patients initially identified as having ≥1 RA diagnosis during the study time frame of January 1, 2010 to December 31, 2013 (for study population flow chart, see Figure 3). All RA patients initially identified were age 18-63 years old.

Univariable Analysis
In univariable logistic regression, we identified covariates having a statistically significant association with the occurrence of hospitalized infection. Results of the univariable analysis of potential risk factors are presented in Table 2

Multivariable Analysis
Multivariable conditional logistic regression models were built to examine the association between anti-TNF drug use and the occurrence of serious infection requiring hospitalization. Results of the final model are presented in Table 3. Although our final eligible RA cohort was relatively large, a fairly small number of hospitalized infection events (1.7%) were identified resulting in a smaller sample size for our nested case-control design. Our smaller sample size contributed to wide confidence intervals of the effect estimates from our final multivariable model and may have limited our power to detect an association between anti-TNF use and additional potential predictors. If we were able to identify a substantially larger sample size, we could have potentially identified more predictors, but the strength of the association related to many additional predictors may be weaker. Our study was able to identify strong and significant predictors of infection, particularly in patients with comorbid COPD and previous history of infection despite our small sample size and resultant wide confidence intervals.

RA patients with
In addition, the nested case-control design means that causality cannot be explicitly inferred between anti-TNF use and hospitalized infection as not all relevant risk factors are captured and recorded within the dataset. Therefore, it is not possible to measure and control for all factors that may have influenced the occurrence of hospitalized infection.
Furthermore, previous studies have found measures of RA disease severity to be significant predictors of hospitalized infection in patients using anti-TNF drugs.
However, the Clinformatics TM DataMart does not directly capture indicators of disease severity. We included proxies of disease severity as potential predictors in our analyses, such as the number of rheumatologist clinic visits from RA cohort entry date to index date, prednisone use, and presence of comorbidities. In addition, there is no evidence that confounding by indication is a concern according to the type of anti-TNF agent. The 2015 American College of Rheumatology (ACR) recommendations do not indicate a particular order of preference when prescribing an anti-TNF agent to a patient based on disease severity. 31 Future research may be directed towards further analyses using disease risk scores as a stronger proxy for disease severity.
Strengths of our study include the use of validated ICD-9 codes with proven high positive predictive values for disease to identify cases of hospitalized infection in administrative data, which may reduce potential misclassification bias. In addition, the use of a large privately insured population allows for more generalizable results for RA patients using anti-TNF agents in the US as opposed to studies that use data sources outside of the US. However, this also means that the results cannot be In our patient cohort, we were unable to identify a significant association between the increased risk of hospitalized infection and specific anti-TNF use in RA patients.
Significant predictors of hospitalized infection in RA patients using anti-TNF agents included recent prednisone use, diabetes, COPD, and previous history of infection.
The impact of these findings suggest that careful monitoring of these specific patient populations may be important in reducing the occurrence of hospitalized infection among patients using anti-TNF agents.