Background: Tumor Necrosis Factor Inhibitors (TNFis) have been associated with Legionnaires’ disease and Listeriosis in adverse event reporting and short-term cohort analyses. These studies are limited and are unable to distinguish between risk due to the severity of the underlying illness and risk due to TNFi.
Objective: We will use randomized control trial data for two TNFi (infliximab and golimumab) to compare the risk of Legionnaires’ disease and Listeriosis among those randomized to TNFi compared to those randomized to placebo.
Study Design: An individual patient-level meta-analysis will provide estimates of pooled odds ratio (OR) and sub-group analyses to identify high risk groups and biologic mediation by sex.
Participants: Only adult participants enrolled in infliximab and golimumab RCTs available through YODA will be included in this analysis. Underlying diseases of interest include rheumatoid arthritis, ankylosing spondylitis, ulcerative colitis, Crohn’s disease, and psoriasis. Only RCTs where patients were randomized to TNFi or placebo will be included in this analysis.
Main Outcome: The main outcome is a summary odds ratio estimate for the odds of Legionnaires’ disease (or Listeriosis) among participants randomized to TNFi compared to those randomized to placebo.
Statistical Analysis: Study-specific odds of Legionnaires’ disease and Listeriosis among those in treatment groups compared to placebo controls will be calculated. Then pooled odds ratio estimates and 95% confidence intervals will be estimated through random effects models.
Certain medications can also weaken an immune system’s response to infection. Tumor necrosis factor inhibitors (TNFi) are prescribed for certain autoimmune conditions (e.g. rheumatoid arthritis, ulcerative colitis, Crohn’s disease) and are an alternative to conventional disease modifying anti-rheumatic drugs (cDMARDs) like methotrexate. Licensure and use of TNFis has increased since 2000 to include a broader approval range and more generics. TNFi and TNF deficiencies have been associated with inability to control both Legionella and Listeria infections in multiple murine models(1–4).
In 2011, the FDA issued a Boxed Warning for TNFi, listing Legionnaires’ disease and Listeriosis as a potential adverse events (1,5,6). Legionnaires’ disease (LD) is a severe bacterial pneumonia caused by Legionella spp. with a case-fatality rate around 10%. Incidence of LD in the U.S. increased steadily since 2006, reaching almost 2.5 cases/100,000 population. Listeriosis is a rare gastrointestinal infection caused by Listeria monocytogenes, amounting to 1,600 confirmed infections per year in the US. Infections are usually linked to outbreaks and case fatality rates can be high (20%), particularly among high risk groups like the elderly, immunocompromised and pregnant women(7).
Two cohort studies conducted in France and Spain have identified an association between Legionnaires’ disease and TNFi, where patients with RA taking TNFi had a higher incidence of LD compared to the general French population comparison group (8,9). A similarly- designed cohort study in Spain found that incidence of Listeriosis was higher among RA patients on TNFi compared to the general European population (10). Unfortunately, the comparison group used in these previous studies (general population) does not provide insight into whether the increased risk is due to the TNFi or the underlying condition that led to the TNFi prescription.
RCT data offers the unique ability to compare TNFi treatment to placebo and the randomization process can efficiently address confounding related to underlying disease status. Through meta-analysis, we will examine if there is evidence for an association between TNFi and Legionnaires’ disease and Listeriosis. How this association may differ by sex, length of RCT follow-up and by alternative, broader disease outcomes (respiratory and gastrointestinal) will also be explored in secondary analyses. A search of available adverse outcome reports for infliximab and golimumab placebo control trials on clinicaltrials.gov revealed that Legionella pneumonia is a reported outcome among treatment groups and that pneumonia and gastrointestinal infections are frequently reported even among small RCTs.
The results of this meta-analysis will inform understanding of Legionnaires’ disease and Listeriosis incidence and provide insight on understudied immune-mediated risk factors for LD and Listeriosis. This study will also provide the necessary preliminary data to launch a Danish national healthcare registry study to explore the exact nature of the association between TNFi and Legionnaires’ disease and Listeriosis during the study period of 2000 to 2020.
The specific aim of this project is to estimate the association between TNFi and Legionnaires' disease (and Listeriosis). Our specific hypothesis is that TNFi are associated with an increased risk of Legionnaires' disease and Listeriosis and that this risk is not attributed to the underlying disease. In order to best address this, placebo-controlled trials will be used so that all patients have (assumed) similar underlying illness severity but are randomized to treatment.
All RCTs for infliximab and golimumab where patients enrolled were restricted to ages 18 and older. Only trials where placebo controls were used will be included. This includes trials for the following underlying conditions: rheumatoid arthritis, ulcerative colitis, Crohn’s disease, ankylosing spondylitis, and psoriasis. Included are trials where enrollment is "despite methotrexate treatment".
Trials with standard of care controls or non-methotrexate combination therapies will be excluded. All trials among patients with juvenile arthritis or asthma will be excluded.
The main outcome is an adverse event code for Legionnaires’ disease or Listeriosis. These will be categorized as a binary outcome variable for our analysis.
The main independent variable would be treatment status. Patients randomized to TNFi (infliximab or golimumab) would be compared to the patients randomized to placebo. Patients recieving any dosage of TNFi will be grouped together.
Subgroup analysis where patients will be analyzed by sex and active ingredient (golimumab only and infliximab only analyses) will be explored. If statistical power allows, analysis by underlying condition of interest be explored. Specifically, we will examine the association among only patients with RA (n ~ 4300) and ulcerative colitis (n ~ 3700). Additionally, sensitivity analysis will be performed where combination treatment trials ("despite methotrexate trials") will be excluded.
Because Legionnaires’ disease and Listeriosis are relatively rare diseases, we believe it’s possible that very few cases may be identified as adverse events during case follow-up. Due to underdiagnosis of Legionnaires’ disease and Listeriosis we also suspect that some cases may be coded as bacterial pneumonia or gastrointestinal illness. Secondary outcomes: bacterial pneumonia (regardless of cause), shigellosis, and gastro-intestinal infections. We would also like to consider a control outcome that we don’t expect to differ between the two groups, such as fracture (which has been recorded as an adverse outcome in TNFi trials on clinicaltrials.gov).
The primary statistical analysis will be a meta-analysis to estimate the pooled odds ratios and 95% confidence interval through random and/or fixed effects models. We will first estimate trial specific odds ratios (or relative risks) with corresponding 95% CIs. Then pooled effect estimates and corresponding 95%CIs will be estimated through meta-analysis models using the R packages ‘meta’ and/or ‘metafor’. The choice between fixed or random effects models will depend on the level of between-study heterogeneity. In meta-analyses of rare events, fixed effects models are usually preferred over random effects since estimates of between study heterogeneity may not be accurate for rare events. Since we expect Legionnaires’ disease and Listeriosis to be a rare event, we anticipate using fixed effects models. To summarize the effect estimates and properly assign weights to each trial included in the analysis, the Peto method will be used. This method assumes an OR close to 1 and is better for rare outcomes in studies where arms are equivalent in size (expected in RCTs)12. We will have to reassess where the ORs lie but it is expected that they will not be much greater than 1, particularly for the secondary outcomes that are more composite, such as unspecified pneumonia and gastrointestinal infections. When no events are reported in one or both study arms, a continuity correction of 0.5 will assumed in each cell.
If outcomes are not rare (as is likely with the bacterial pneumonia and gastrointestinal infection secondary outcomes), we will use fixed (Mantel-Haenszel) or random effects (DerSimonian and Laird) models with inverse variance methods. We will calculate the I2 statistic in order to assess study heterogeneity in a manner comparable to other meta-analysis of RCT adverse events. An I2 less than 25% would indicate strong heterogeneity between RCTs and the need for a random effects model.
Forest plots of effect estimates and corresponding 95% confidence intervals along with pooled estimates will be presented as the results.
If the data allows, sub-group level pooled ORs by sex will also be estimated. Some diseases like RA are more prevalent among women but the outcome of Legionnaires’ disease is more prevalent among males so we would like to explore biologic modification by sex. Golimumab-only and infliximab-only sub-group analysis will also be conducted to understand if any TNFi and disease association is actually driven by a specific active ingredient.
Tumor Necrosis Factor Inhibitors (TNFi) are commonly prescribed for auto-immune diseases, but they can also place individuals at increased risk for severe infections. There is an FDA warning stating that TNFi could be associated with an increased risk of Legionnaires’ disease and Listeriosis. These studies could not clarify if the increased risk was due to TNFi use or the underlying autoimmune disease. This is optimally answered through an individual patient-level meta-analysis of TNFi clinical trials. These results can be used to make more targeted risk assessments when prescribing TNFI and also inform epidemiological understanding of Legionnaires’ disease and Listeriosis incidence.
This project will last approximately 1 year from data-acquisition to manuscript submission. The first month will be allocated to data cleaning and preparation, months 2-6 will be dedicated to statistical analysis, and, finally, months 6-12 will be dedicated to manuscript drafting, revisions among authors, and submission(s). Expected analysis start date is August 1, 2020, analysis completion date is January 1, 2021, manuscript first draft date is March 1, 2021, and first manuscript submission date/date results reported to YODA is May 1, 2021.
The intended product of this analysis is a high-quality manuscript to be submitted to peer-reviewed journals such as Clinical Infectious Diseases, Journal of the American Medical Association, Rheumatology, and American Journal of Epidemiology. The target audience are public health officials interested in understanding Legionnaires' disease and Listeriosis infectious disease trends, and clinicians such as rheumatologists who may have prescribing concerns.
1. Skerrett SJ, Bagby GJ, Schmidt RA, Nelson S. Antibody‐Mediated Depletion of Tumor Necrosis Factor‐α Impairs Pulmonary Host Defenses to Legionella pneumophila. J Infect Dis. 1997;176(4):1019-1028. doi:10.1086/516530
2. Virna S, Deckert M, Lütjen S, et al. TNF Is Important for Pathogen Control and Limits Brain Damage in Murine Cerebral Listeriosis. J Immunol. 2006;177(6):3972-3982. doi:10.4049/jimmunol.177.6.3972
3. Torres D, Janot L, Quesniaux VFJ, et al. Membrane tumor necrosis factor confers partial protection to Listeria infection. Am J Pathol. 2005;167(6):1677-1687. doi:10.1016/S0002-9440(10)61250-3
4. Nakane A, Minagawa T, Kato K. Endogenous tumor necrosis factor (cachetin) is essential to host resistance against Listeria monocytogenes infection. Infect Immun. 1988;56(10):2563-2569. doi:10.1128/iai.56.10.2563-2569.1988
5. Sorbello A, Diak P, Gilbert J, Boucher R. Pneumonia Due to Legionella pneumophila in Patients Treated With Tumor Necrosis Factor-α Antagonists. Infect Dis Clin Pract. 2011;19(6):399-405. doi:10.1097/IPC.0b013e31821bc8f9
6. Fujita M, Ikegame S, Harada E, et al. TNF receptor 1 and 2 contribute in different ways to resistance to Legionella pneumophila-induced mortality in mice. Cytokine. 2008;44(2):298-303. doi:10.1016/j.cyto.2008.08.015
7. Datta A, Burall L. Current Trends in Foodborne Human Listeriosis. Food Saf. 2018;6(1):1-6. doi:10.14252/foodsafetyfscj.2017020
8. Lanternier F, Tubach F, Ravaud P, et al. Incidence and Risk Factors of Legionella pneumophila Pneumonia During Anti-Tumor Necrosis Factor Therapy. Chest. 2013;144(3):990-998. doi:10.1378/chest.12-2820
9. Tubach F, Ravaud P, Salmon-Ceron D, et al. Emergence of Legionella pneumophila Pneumonia in Patients Receiving Tumor Necrosis Factor- Antagonists. Clin Infect Dis. 2006;43(10):e95-e100. doi:10.1086/508538
10. Peña-Sagredo JL, Hernández M V., Fernandez-Llanio N, et al. Listeria monocytogenes infection in patients with rheumatic diseases on TNF-alpha antagonist therapy: The Spanish Study Group experience. Clin Exp Rheumatol. 2008;26(5):854-859.
11. Mazumdar S, Greenwald D. Golimumab. MAbs. 2009;1(5):422-431. doi:10.4161/mabs.1.5.9286
12. Cai T, Parast L, Ryan L. Meta-analysis for rare events. Stat Med. 2010;29(20):2078-2089. doi:10.1002/sim.3964