Elicit: TNF-Blockers: Infection and Cancer Risks

Skip to main content

TNF-Blockers: Infection and Cancer Risks

Research reportView only

Create alertChat

May 5, 2026

Explore safety/mechanism links for TNF-blocker associated infections and malignancies

TNF-blockers have a clear mechanistic link to opportunistic infections through disruption of granuloma formation and immune surveillance (particularly for tuberculosis reactivation with monoclonal antibodies), but for malignancies the evidence indicates that baseline inflammatory disease severity rather than immunosuppression from TNF-blockade itself drives most cancer risk, with only skin cancers showing consistent modest increases attributable to treatment.

Abstract

TNF-blockers demonstrate mechanistically coherent infection risks linked to TNF’s essential role in immune surveillance and host defense. While overall infection risk shows only modest increases (OR 1.18-1.20) that become non-significant when adjusted for exposure time (IRR 1.01), opportunistic infections show consistent 90% increased risk (OR 1.90) and tuberculosis risk increases 3- to 4-fold (OR 3.3-3.5). The mechanism centers on disruption of granuloma formation, explaining why tuberculosis occurs exclusively with monoclonal antibodies that completely neutralize TNF and why 72% of infliximab-associated granulomatous infections occur within 90 days, consistent with reactivation of latent infections rather than increased susceptibility to new pathogens. Agent-specific differences are substantial, with infliximab carrying 3.25-fold greater granulomatous infection risk than etanercept, reflecting mechanistic differences between complete TNF neutralization and partial receptor blockade.

For malignancies, competing biological mechanisms—TNF’s dual roles in suppressing tumors through apoptosis versus promoting cancer through chronic inflammation—generate heterogeneous findings that resolve upon careful examination. Long-term observational studies show no overall increased malignancy risk (OR 0.90-0.95) and no evidence that longer exposure increases risk, while short-term RCT meta-analyses finding elevated risk (OR 3.3) likely reflect detection bias given inadequate latency periods for cancer development. Skin cancers represent the most consistent signal (OR 1.45 for non-melanoma skin cancer), amplified by concomitant methotrexate (RR 1.97). For lymphomas, standardized incidence ratios of 1.8-6.0 in rheumatoid arthritis must be interpreted against 2-fold baseline elevation from systemic inflammation itself, and pooled estimates controlling for this confounding show no significant TNF-blocker effect (OR 1.11). The mechanistic framework suggests that baseline inflammatory disease severity, rather than immunosuppression per se, drives most malignancy risk in these populations.

Methods

We analyzed 10 sources from an initial pool of 200, using 8 screening criteria. Each paper was reviewed for 7 key aspects that mattered most to the research question. More on methods

Records from Elicit search

n = 200

Papers screened using: TNF-Blocker Intervention, Safety or Mechanistic Outcomes, Study Design, Adult Population, Clinical Indication, TNF-Specific Focus, Adequate Sample Size, Original Human Clinical Data

n = 200

Papers screened out

n = 190

Papers included for extraction

n = 10

Press enter or space to select a node.You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.

Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

Paper search

We performed a semantic search across over 138 million academic papers from the Elicit search engine, which includes all of Semantic Scholar and OpenAlex.

We ran this query: “Explore safety/mechanism links for TNF-blocker associated infections and malignancies”

The search returned 200 total results from Elicit.

We retrieved 200 papers most relevant to the query for screening.

Screening

We screened in sources based on their abstracts that met these criteria:

We considered all screening questions together and made a holistic judgement about whether to screen in each paper.

Data extraction

We asked a large language model to extract each data column below from each paper. We gave the model the extraction instructions shown below for each column.

Extract comprehensive details about the TNF-blocker therapy studied, including:

Extract details about the study population relevant to TNF-blocker safety, including:

Extract all infection-related safety outcomes for TNF-blockers, including:

Extract all malignancy-related safety outcomes for TNF-blockers, including:

Extract any discussion of biological mechanisms linking TNF-blockers to infections or malignancies, including:

Extract patient or treatment factors that modify infection or malignancy risk with TNF-blockers, including:

Extract study design features relevant to interpreting TNF-blocker safety data, including:

Results

Characteristics of Included Studies

Study

Full text retrieved?

Study design

Patient population

Sample size (treatment/control)

TNF-blockers studied

Follow-up duration

E. Dommasch et al., 2011

Yes

Systematic review and meta-analysis of RCTs

Plaque psoriasis and psoriatic arthritis

4,598/2,313

Etanercept, infliximab, adalimumab, golimumab, certolizumab

Mean 17.8 weeks (range 12-30 weeks)

S. Minozzi et al., 2016

Yes

Systematic review and meta-analysis of RCTs and open-label extension studies

Rheumatoid arthritis, psoriatic arthritis, ankylosing spondylitis

14,766/7,994

Adalimumab, golimumab, infliximab, certolizumab, etanercept

1-36 months (RCTs), 6-48 months (open-label extensions)

M. Muller et al., 2020

No

Systematic review of observational cohort studies

Inflammatory bowel disease

298,717 (no control group)

Infliximab, adalimumab

Mean 7-80 months

Sean M. McConachie et al., 2018

No

Systematic review of meta-analyses and cohort studies

Inflammatory bowel disease

Not mentioned

Infliximab, adalimumab, certolizumab, golimumab

Not mentioned

R. Pereira et al., 2017

No

Observational cohort study

Immune-mediated inflammatory diseases

Not mentioned

Not specified

January 2000-December 2014

T. Bongartz et al., 2006

No

Meta-analysis of RCTs

Rheumatoid arthritis

3,493/1,512

Infliximab, adalimumab

At least 12 weeks

X. Mariette et al., 2011

No

Systematic review and meta-analysis of observational studies

Rheumatoid arthritis

Not mentioned

Not specified

Not mentioned

D. Solomon et al., 2012

Yes

Systematic review of observational cohort studies

Rheumatoid arthritis

Not specified

Infliximab, adalimumab, etanercept

Relatively short duration

S. Bonovas et al., 2016

Yes

Systematic review and meta-analysis of RCTs

Inflammatory bowel disease

9,003/5,587

Adalimumab, certolizumab, golimumab, infliximab, natalizumab, vedolizumab

1-24 months, average 6.5 months

R. Wallis et al., 2004

No

Registry-based study using FDA Adverse Event Reporting System

Not specified

Not mentioned

Infliximab, etanercept

January 1998-September 2002

toof

Pageof

The included studies encompassed multiple disease indications, with rheumatoid arthritis being the most frequently studied condition, followed by inflammatory bowel disease and psoriatic disease. Study designs varied substantially, including randomized controlled trials analyzed through meta-analysis, observational cohort studies, and registry-based surveillance. Follow-up durations ranged from 12 weeks to 80 months, with most RCT-based analyses having relatively short observation periods under 6 months.

Infection Outcomes

Study

Any infection

Serious infections

Opportunistic infections

Tuberculosis

Other notable findings

E. Dommasch et al., 2011

OR 1.18 (95% CI 1.05-1.33)
IRR 1.01 (95% CI 0.92-1.11)

OR 0.70 (95% CI 0.40-1.21)
IRR 0.59 (95% CI 0.35-0.99)

Not measured

Not measured

Most common site: cellulitis

S. Minozzi et al., 2016

OR 1.20 (95% CI 1.08-1.34)

OR 1.41 (95% CI 1.16-1.73) fixed effects
OR 1.25 (95% CI 1.01-1.55) random effects

OR 0.94 (95% CI 0.33-2.64) fixed effects
OR 0.81 (95% CI 0.23-2.87) random effects

OR 3.53 (95% CI 1.58-7.85) fixed effects
OR 3.29 (95% CI 1.48-7.33) random effects

Risk increased with longer treatment duration

M. Muller et al., 2020

Not reported

Not reported

Not reported

Not reported

Focus on malignancy outcomes

Sean M. McConachie et al., 2018

Meta-analyses showed inconclusive association

Not specified

Registry data suggest independent risk

Not specified

Risk factors: older age, malnutrition, diabetes, combination therapy

R. Pereira et al., 2017

Not reported

IR 4.02/100 patient-years (95% CI 3.20-5.04)

Not reported

IR 0.28/100 patient-years (95% CI 0.12-0.66)
60% extrapulmonary

Most frequent site: gastrointestinal system
TB exclusively with monoclonal antibodies

T. Bongartz et al., 2006

Not reported

OR 2.0 (95% CI 1.3-3.1)
NNH 59 (95% CI 39-125) for 3-12 months

Not reported

Not reported

Not specified

X. Mariette et al., 2011

Not reported

Not reported

Not reported

Not reported

Study focused on malignancy outcomes

D. Solomon et al., 2012

Not reported

Not reported

Not reported

Not reported

Study focused on malignancy outcomes

S. Bonovas et al., 2016

OR 1.19 (95% CI 1.10-1.29)
NNH 26

OR 0.89 (95% CI 0.71-1.12)
OR 0.56 (95% CI 0.35-0.90) in low-risk bias studies

OR 1.90 (95% CI 1.21-3.01)
NNH 194

OR 2.04 (95% CI 0.71-5.89)

Specific pathogens: M. tuberculosis, JC virus, Nocardia, CMV/EBV, candidiasis, VZV, P. jirovecii, H. capsulatum

R. Wallis et al., 2004

Not reported

Not reported

239/100,000 for infliximab vs 74/100,000 for etanercept

144/100,000 for infliximab vs 35/100,000 for etanercept

3.25-fold greater risk with infliximab vs etanercept
72% of infections within 90 days for infliximab

toof

Pageof

The infection data revealed several consistent patterns. TNF-blockers demonstrated a modest but statistically significant increase in overall infections, with odds ratios ranging from 1.18 to 1.20. However, when adjusted for patient-years of exposure, this association became non-significant (IRR 1.01), suggesting that differences in follow-up time between treatment and control groups may account for some of the observed increase.

For serious infections requiring hospitalization or antimicrobial therapy, findings were more heterogeneous. Two meta-analyses of RCTs found increased risk (OR 1.25-2.0), while another found no significant increase overall (OR 0.89) but demonstrated reduced risk in low-risk-of-bias studies (OR 0.56). An observational study reported an incidence rate of 4.02 per 100 patient-years.

Opportunistic infections showed the most consistent signal, with a 90% increased risk (OR 1.90), though one analysis found neutral associations. Tuberculosis emerged as a particularly important concern, with 3- to 4-fold increased risk in rheumatologic conditions and notably occurring exclusively with monoclonal antibodies (infliximab, adalimumab) rather than the soluble receptor (etanercept). Registry data corroborated this finding, showing infliximab associated with 3.25-fold greater granulomatous infection risk compared to etanercept, with 72% of infliximab-associated infections occurring within 90 days of treatment initiation.

Specific pathogens identified included intracellular and granulomatous bacteria, fungi, and viruses, with particular documentation of Mycobacterium tuberculosis, JC virus, Nocardia, cytomegalovirus, Epstein-Barr virus, candidiasis, varicella-zoster virus, Pneumocystis jirovecii, and Histoplasma capsulatum. Gastrointestinal sites were most frequently affected for serious infections, while 60% of tuberculosis cases were extrapulmonary.

Malignancy Outcomes

Study

All-site malignancy

Non-melanoma skin cancer

Melanoma

Lymphoma

Other solid tumors

E. Dommasch et al., 2011

OR 1.48 (95% CI 0.71-3.09)
IRR 0.99 (95% CI 0.51-1.90)

OR 1.33 (95% CI 0.58-3.04)
70.6% of all malignancies

Not separately analyzed

OR 1.26 (95% CI 0.39-4.15) when NMSC excluded

Prostate and breast cancer reported

M. Muller et al., 2020

1.0% overall occurrence
No significant association in 10/11 studies

123/692 cases (17.8%)

Not specified

106/692 cases (15.3%)
One study found increased risk

Digestive malignancies: 120/692 (17.3%)

R. Pereira et al., 2017

IR 1.75/100 patient-years (95% CI 1.24-2.47)

Not specified

Not specified

Not specified

Not specified

T. Bongartz et al., 2006

OR 3.3 (95% CI 1.2-9.1)
NNH 154 (95% CI 91-500) for 6-12 months

Not separately analyzed

Not separately analyzed

Not separately analyzed

Dose-dependent relationship observed

X. Mariette et al., 2011

OR 0.95 (95% CI 0.85-1.05)

OR 1.45 (95% CI 1.15-1.76)

OR 1.79 (95% CI 0.92-2.67)

OR 1.11 (95% CI 0.70-1.51)

No evidence longer exposure increases risk

D. Solomon et al., 2012

Various estimates from different studies

OR 1.24 (95% CI 0.97-1.58) alone
RR 1.97 (95% CI 1.51-2.58) with MTX

Not specified

Risk estimates ranged 1.1-4.9
SIR 1.8-6.0 among TNFi users

Hematologic malignancies SIR 2.0-4.1

S. Bonovas et al., 2016

OR 0.90 (95% CI 0.54-1.50)
0.45% treatment vs 0.54% placebo

Not separately analyzed

Not separately analyzed

Not separately analyzed

Insufficient data on exposure/follow-up

toof

Pageof

Malignancy findings were substantially more heterogeneous than infection outcomes. For all-site malignancy, estimates ranged from protective (OR 0.90-0.95) to substantially increased risk (OR 3.3). An observational study in inflammatory bowel disease reported 1.0% overall malignancy occurrence with no significant association in 10 of 11 included studies.

Skin cancers emerged as the most consistent malignancy signal. Non-melanoma skin cancers showed significant increases in rheumatoid arthritis populations, with a pooled estimate of OR 1.45. In psoriatic disease, NMSC comprised 70.6% of all malignancies, though the overall odds ratio did not reach statistical significance. Melanoma risk was elevated with OR 1.79, though confidence intervals crossed one. The risk of NMSC was further amplified when TNF-blockers were combined with methotrexate (RR 1.97).

For lymphoma and hematologic malignancies, evidence was mixed. Observational studies in rheumatoid arthritis found standardized incidence ratios ranging from 1.8 to 6.0, while a meta-analysis of registries found no significant increase (OR 1.11). One systematic review of inflammatory bowel disease studies identified increased lymphoma risk in one of 11 studies, representing 15.3% of all malignancies in the TNFi-exposed population.

Importantly, one meta-analysis found dose-dependent malignancy risk, with higher TNF-blocker doses associated with more malignancies, yielding a number needed to harm of 154 for one additional malignancy within 6-12 months. However, another analysis found no evidence that longer exposure increased risk, and patients with previous malignancies showed higher risk of recurrence that was not further increased by TNF-blocker exposure.

Mechanistic Insights

TNF plays a critical role in immune surveillance and host defense, creating a theoretical framework for both infection and malignancy risks with TNF inhibition. Multiple biological pathways have been proposed to explain the observed safety signals.

For infections, TNF-α antagonists suppress inflammatory pathways that are essential for immune defense. The disruption of granuloma formation represents a key mechanism for tuberculosis reactivation, as granulomas are crucial for containing mycobacterial infections. This mechanistic understanding explains why tuberculosis risk is elevated 3- to 4-fold and why TB occurs exclusively with monoclonal antibodies that more completely neutralize TNF, compared to the soluble receptor etanercept which may have different immunologic effects.

The clustering of granulomatous infections within 90 days of infliximab initiation is consistent with reactivation of latent infections, suggesting that TNF-blockade unmasks pre-existing but controlled pathogens rather than solely increasing susceptibility to new infections. This mechanistic insight has important implications for pre-treatment screening protocols.

For malignancies, competing mechanisms have been proposed. TNF may suppress tumor development through induction of apoptosis and suppressive effects on gene expression, suggesting that TNF-blockade could enhance cancer risk. Additionally, TNF serves as a key element of inflammatory responses whose inhibition may increase risk of infection-driven cancers, particularly viral malignancies.

Conversely, uncontrolled inflammation itself may potentiate cancer development, as evidenced by higher lymphoma rates in patients with greater systemic inflammation. TNF’s profound effects on angiogenesis, which is critical for tumor growth and metastasis, suggest that anti-TNF therapy could theoretically reduce cancer risk by suppressing both inflammation and angiogenesis. The observation that corticosteroids appear to reduce lymphoma risk supports this anti-inflammatory mechanism.

When TNF-blockers are combined with other immunosuppressants, synergistic immunosuppression may increase both infection and malignancy risks. This is particularly evident for non-melanoma skin cancers, where the combination of TNF-blockers with methotrexate nearly doubled the risk compared to TNF-blockers alone.

Risk Factors and Population Heterogeneity

Several patient and treatment characteristics modified infection and malignancy risk. Older age emerged as a risk factor for infections, though one meta-regression found no significant age association. Comorbid conditions including malnutrition and diabetes increased infection susceptibility.

Concomitant immunosuppressive therapy represented an important modifier, with evidence of synergistic effects when combining TNF-blockers with other systemic immunosuppressants. In psoriatic arthritis trials, 44.6% of patients received concomitant methotrexate, 5.5% other disease-modifying drugs, and 10.5% corticosteroids, while psoriasis trials generally excluded such combinations.

Disease-specific factors showed variable effects. For infections, Crohn’s disease patients demonstrated higher opportunistic infection risk compared to ulcerative colitis. The gastrointestinal system was the most frequent site of serious infections, potentially reflecting the intestinal inflammation in IBD populations. For malignancies, higher systemic inflammation levels increased lymphoma risk independent of treatment, introducing confounding by indication where sicker patients both receive TNF-blockers and have elevated baseline cancer risk.

Treatment-related factors included dose-dependent relationships for malignancies and duration-dependent increases in infection risk. However, longer TNF-blocker exposure did not increase malignancy risk in observational cohorts. Patients with previous malignancies had higher recurrence risk that was not further increased by TNF-blocker exposure, suggesting that baseline cancer risk may be more important than treatment effects in this subgroup.

Agent-specific differences were notable. Infliximab carried 3.25-fold greater granulomatous infection risk than etanercept, with tuberculosis occurring exclusively with monoclonal antibodies. This suggests that complete TNF neutralization by monoclonal antibodies may have different immunologic consequences than the partial blockade achieved by the soluble receptor etanercept.

Synthesis

The systematic review data reveal a complex safety profile for TNF-blockers that cannot be reduced to simple risk estimates. The apparent contradictions in findings—particularly for serious infections and malignancies—can be reconciled by considering methodological factors, population characteristics, and temporal dynamics.

For serious infections, the divergence between meta-analyses finding increased risk (OR 1.25-2.0) and those finding no increase or even protective effects (OR 0.56 in low-bias studies) reflects differences in study quality and outcome ascertainment. The protective finding in low-bias studies is paradoxical and likely represents residual confounding, where healthier patients both tolerate biologics better and have fewer baseline infection risks. Importantly, when infection rates are adjusted for differential follow-up time using incidence rate ratios rather than odds ratios, the association with any infection becomes non-significant (IRR 1.01), suggesting that the modest OR of 1.18-1.20 may partially reflect surveillance bias where treated patients have more clinical encounters and thus more infection detection opportunities.

The consistent signal for opportunistic infections (OR 1.90) and tuberculosis (OR 3.3-3.5) represents a mechanistically plausible effect. These findings are biologically coherent with TNF’s role in granuloma formation and align with the observation that 72% of infliximab-associated infections occur within 90 days, consistent with reactivation of latent infections rather than de novo acquisition. The exclusive occurrence of tuberculosis with monoclonal antibodies but not etanercept further supports a mechanistic relationship, as complete TNF neutralization disrupts granuloma maintenance more profoundly than partial receptor blockade.

For malignancies, the heterogeneity is even more pronounced, with estimates ranging from protective (OR 0.90-0.95) to substantially increased (OR 3.3). This variance can be explained by multiple factors:

Study Duration and Latency: The meta-analysis by Bongartz et al. finding OR 3.3 included only trials of 12 weeks or longer, representing short-term exposure inadequate to capture cancer development’s slow natural history. Observational studies with longer follow-up found no overall increased risk, and importantly, no evidence that longer exposure increased risk. This temporal pattern suggests that any short-term signal may reflect detection bias or unmasking of pre-existing cancers rather than carcinogenesis.

Malignancy Type Specificity: The heterogeneity resolves when examining specific cancer types. Skin cancers show consistent increases (OR 1.45 for NMSC), particularly when TNF-blockers are combined with methotrexate (RR 1.97). This likely reflects both immunosuppression-mediated reduction in skin cancer surveillance and the high baseline skin cancer rates in these populations. For lymphomas, observational studies in RA found SIRs of 1.8-6.0, but these must be interpreted against the 2-fold elevated baseline lymphoma risk in RA patients. When meta-analysis pooled data controlling for this baseline elevation, no significant TNF-blocker effect was found (OR 1.11).

Population-Specific Baseline Risk: The 1.0% malignancy rate in IBD patients with no significant association in 10 of 11 studies contrasts with findings in RA, where systemic inflammation itself increases lymphoma risk. This suggests that population-specific confounding by indication—where more severe disease necessitates TNF-blockers and independently increases cancer risk—may account for observed associations in some conditions but not others.

Dose-Response Relationships: The dose-dependent malignancy risk observed in the Bongartz meta-analysis provides mechanistic support for a causal relationship. However, this finding requires reconciliation with the absence of duration-dependent risk in observational cohorts. One explanation is that higher doses may unmask pre-existing cancers more rapidly without increasing cumulative carcinogenesis. Alternatively, dose may serve as a marker for disease severity, creating confounding by indication.

Agent-Specific Effects: The 3.25-fold greater granulomatous infection risk with infliximab versus etanercept demonstrates that not all TNF-blockers carry equivalent risks. This mechanistic heterogeneity—related to complete versus partial TNF neutralization—means that class-wide safety estimates may obscure important agent-specific differences.

In summary, TNF-blockers demonstrate a mechanistically coherent increased risk of opportunistic infections, particularly tuberculosis reactivation, that appears to be agent-specific and related to granuloma disruption. For malignancies, the preponderance of evidence from longer-term observational studies suggests no overall increased risk, though skin cancers may be modestly increased, particularly with combination immunosuppression. Short-term RCT signals likely reflect detection bias and insufficient latency periods for cancer development. Clinicians should focus on pre-treatment tuberculosis screening and skin cancer surveillance while recognizing that baseline disease severity and inflammation may be stronger cancer predictors than TNF-blocker exposure itself.

References

Sean M. McConachie, S. Wilhelm, A. Bhargava, P. Kale-Pradhan\ (2018).Biologic-Induced Infections in Inflammatory Bowel Disease: The TNF-α Antagonists. The Annals of Pharmacotherapy

R. Pereira, Raquel Faria, P. Lago, T. Torres\ (2017).Infection and Malignancy Risk in Patients Treated with TNF Inhibitors for Immune-Mediated Inflammatory Diseases. Current Drug Safety

T. Bongartz, A. Sutton, M. Sweeting, I. Buchan, E. Matteson, and 1 more\ (2006).Anti-TNF antibody therapy in rheumatoid arthritis and the risk of serious infections and malignancies: systematic review and meta-analysis of rare harmful effects in randomized controlled trials. Journal of the American Medical Association (JAMA)

X. Mariette, M. Matucci-Cerinic, K. Pavelka, P. Taylor, R. V. van Vollenhoven, and 5 more\ (2011).Malignancies associated with tumour necrosis factor inhibitors in registries and prospective observational studies: a systematic review and meta-analysis. Annals of the Rheumatic Diseases

E. Dommasch, K. Abuabara, D. Shin, Josephine C. Nguyen, A. Troxel, and 1 more\ (2011).The risk of infection and malignancy with tumor necrosis factor antagonists in adults with psoriatic disease: a systematic review and meta-analysis of randomized controlled trials. Journal of American Academy of Dermatology

D. Solomon, E. Mercer, A. Kavanaugh\ (2012).Observational studies on the risk of cancer associated with tumor necrosis factor inhibitors in rheumatoid arthritis: a review of their methodologies and results. Arthritis & Rheumatism

S. Minozzi, S. Bonovas, T. Lytras, V. Pecoraro, Marien González-Lorenzo, and 12 more\ (2016).Risk of infections using anti-TNF agents in rheumatoid arthritis, psoriatic arthritis, and ankylosing spondylitis: a systematic review and meta-analysis. Expert Opinion on Drug Safety

S. Bonovas, G. Fiorino, M. Allocca, T. Lytras, G. Nikolopoulos, and 2 more\ (2016).Biologic Therapies and Risk of Infection and Malignancy in Patients With Inflammatory Bowel Disease: A Systematic Review and Network Meta-analysis. Clinical Gastroenterology and Hepatology

R. Wallis, M. Broder, J. Y. Wong, M. Hanson, D. Beenhouwer\ (2004).Granulomatous infectious diseases associated with tumor necrosis factor antagonists. Clinical Infectious Diseases

M. Muller, F. D'amico, S. Bonovas, S. Danese, L. Peyrin-Biroulet\ (2020).TNF inhibitors and risk of malignancy in patients with inflammatory bowel diseases: a systematic review. Journal of Crohn's & Colitis

Download BIBDownload RISDownload TXT

Report

Status

Gather sources

200 sources found

Details

Screen sources

10 sources included

Details

Extract data

70 data points extracted

Details

Generate report

Save PDF

BIBLaTeX, ZoteroRISZotero, MendeleyTXTAPA BibliographyPDFPDFDOCXMicrosoft Word

Chat

Got some follow-up questions?

Sign up or sign in to chat with this report.

Back

Risk of infections using anti-TNF agents in rheumatoid arthritis, psoriatic arthritis, and ankylosing spondylitis: a systematic review and meta-analysis

S. Minozzi, S. Bonovas, T. Lytras, V. Pecoraro, Marien González-Lorenzo, A. J. Bastiampillai, E. Gabrielli, A. Lonati, L. Moja, M. Cinquini, V. Marino, A. Matucci, G. Milano, G. Tocci, R. Scarpa, D. Goletti, F. Cantini

Expert Opinion on Drug Safety·

2016·

271 citations

SourceDOI

Plain textPDF

Searching for PDF

Unable to find PDF from source

Back

TNF-Blocker Details

- Specific agents: adalimumab, golimumab, infliximab, certolizumab, etanercept - Dosage and dosing frequency: Not specified - Duration of treatment/exposure: 1 to 36 months - Route of administration: Not specified - Monotherapy or combination therapy: Both, but specific combinations not detailed - Dose-response relationships: Sensitivity analyses conducted using highest doses in multi-arm trials

Patient Population

- Primary condition being treated: Rheumatoid arthritis, psoriatic arthritis, ankylosing spondylitis - Sample size for TNF-blocker vs control groups: 14,766 in treatment groups, 7,994 in control groups - Age demographics: Mean age ranged from 30 to 62 years - Disease duration and severity: Not explicitly detailed - Prior treatment history: Conventional disease-modifying anti-rheumatic drugs - Baseline risk factors for infections or malignancies: Patients with histories of infections such as latent tuberculosis were excluded - Immunosuppressive co-medications: Conventional disease-modifying anti-rheumatic drugs

Infection Outcomes

- Types of infections studied: serious infections, opportunistic infections, tuberculosis, any infection - Incidence rates or event counts: Not explicitly provided - Risk estimates: - Serious infections: OR: 1.41 (95% CI: 1.16, 1.73) and OR: 1.25 (95% CI: 1.01, 1.55) - Tuberculosis: OR: 3.53 (95% CI: 1.58, 7.85) and OR: 3.29 (95% CI: 1.48, 7.33) - Opportunistic infections: OR: 0.94 (95% CI: 0.33, 2.64) and OR: 0.81 (95% CI: 0.23, 2.87) - Specific pathogens or infection sites: Tuberculosis - Time to infection onset: Not mentioned - Severity and clinical outcomes of infections: Not mentioned

Malignancy Outcomes

Not mentioned (the paper does not discuss malignancy outcomes or safety related to TNF-blockers)

Mechanistic Insights

- Role of TNF in immune surveillance and host defense: TNF-blockers increase the risk of infections, particularly serious infections and tuberculosis, indicating a role in immune surveillance. - Impact on specific immune cell functions: Disruption of granuloma formation affects macrophages and control of tuberculosis. - Effects on cytokine networks and immune signaling: Implied by disruption of granuloma formation, but not explicitly discussed. - Hypotheses about why certain infections or cancers are more common: Increased risk of tuberculosis due to granuloma disruption. - Differences in mechanism between TNF-blocker agents: Etanercept may have a different impact on immunity compared to other TNF-blockers. - Discussion of immune suppression vs immune modulation: TNF-blockers disrupt normal immune functions, leading to increased infection risk.

Risk Factors

- Age-related risk differences: No significant association found. - Disease-specific risk factors: Not explicitly mentioned. - Concomitant immunosuppressive therapy effects: Not mentioned. - Prior infection or malignancy history: Not mentioned. - Geographic or environmental factors: Not mentioned. - Genetic or biomarker predictors: Not mentioned. - Dose-dependent risk relationships: Higher doses may increase risk, but results are not significantly different. - Duration-dependent risk relationships: Longer treatment durations increase the risk of infections.

Study Methodology

- Study design: Systematic review and meta-analysis of RCTs and open-label extension studies - Duration of follow-up: 1-36 months for RCTs, 6-48 months for open-label extension studies - Control group characteristics: Placebo or no treatment, multi-interventional therapies - Methods for outcome ascertainment and adjudication: Independent reviewers with consensus resolution - Sample size and power: 22,760 participants in RCTs, 2,236 in open-label extension studies - Handling of missing data or loss to follow-up: Excluded trials with zero-event data, intention-to-treat analysis - Risk of bias considerations: High or unclear risk of bias in many studies, particularly incomplete outcome data

ABSTRACT Introduction: Five anti-tumor necrosis factor (anti-TNF) agents have received regulatory approval for use in rheumatology: adalimumab, golimumab, infliximab, certolizumab, and etanercept. Apart from their well-documented therapeutic value, it is still uncertain to what extent they are associated with an increased risk of infectious adverse events. Areas covered: We conducted a systematic review and meta-analysis of published randomized studies to determine the effect of anti-TNF drugs on the occurrence of infectious adverse events (serious infections; tuberculosis; opportunistic infections; any infection). We searched Medline, Embase, and the Cochrane Library up to May 2014 to identify eligible studies in adult patients with rheumatoid arthritis, psoriatic arthritis, or ankylosing spondylitis that evaluated anti-TNF drugs compared with placebo or no treatment. Expert opinion: Our study encompassed data from 71 randomized controlled trials involving 22,760 participants (range of follow-up: 1–36 months) and seven open label extension studies with 2,236 participants (range of follow-up: 6–48 months). Quantitative synthesis of the available data found statistically significant increases in the occurrence of any infections (20%), serious infections (40%), and tuberculosis (250%) associated with anti-TNF drug use, while the data for opportunistic infections were scarce. The quality of synthesized evidence was judged as moderate. Further evidence from registries and long-term epidemiological studies are needed to better define the relationship between anti-TNF agents and infection complications.

1.Introduction

Five anti-tumor necrosis factor (anti-TNF) agents have received regulatory approval for clinical use in rheumatology: adalimumab, golimumab, infliximab, certolizumab, and etanercept. Adalimumab and golimumab are fully human monoclonal antibodies; infliximab is a chimeric monoclonal antibody with a murine variable region; certolizumab is a humanized Fab fragment conjugated to polyethylene glycol; and etanercept is a fusion protein of two TNFR2 receptor extracellular domains and the Fc fragment of human immunoglobulin 1 [1].

Apart from their well-documented therapeutic value for several diseases including rheumatoid arthritis (RA), psoriatic arthritis (PsA), and ankylosing spondylitis (AS), it is still uncertain to what extent therapy with anti-TNF drugs may be associated with an increased risk of infectious adverse events (AEs). Post-marketing surveillance and observational studies provided the first indication that TNF inhibitors might be associated with an increased risk of serious infections [2,3]. Subsequently, a systematic review and meta-analysis of randomized controlled trials (RCTs), published in 2006, identified a statistically significant rise in the risk of infectious AEs in RA patients treated with infliximab and adalimumab (the odds ratio [OR] for serious infections was 2.0, with 95% confidence interval [CI]: 1.3 to 3.1) [4]. However, observational studies have been inconsistent on this issue with reports of both increased risk [5][6][7][8][9] and of no increased risk [10][11][12].

Considering the conflicting results published in the literature, the high number of RCTs that have been performed since 2006, and the increasing use of TNF inhibitors as induction or maintenance treatment for adult patients with RA, PsA, or AS, we conducted a systematic review and meta-analysis of published trials to determine the occurrence of infectious AEs associated with use of anti-TNF agents.

2.1Protocol and registration

Our study protocol [13] is registered on PROSPERO, the international prospective register of systematic reviews. The current systematic review and meta-analysis was performed according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) statement [14].

2.2Data sources and search strategy

A comprehensive search of MEDLINE and EMBASE bibliographic databases was conducted through May 2014. The following search terms were used: adalimumab, golimumab, infliximab, certolizumab pegol, etanercept, anti-tumo(u)r necrosis factor(s), tumo(u)r necrosis factor(s), tumo(u)r necrosis factor alpha antibody(ies), tumo(u)r necrosis factor antibody(ies), anti-TNF, TNF, biologic(al) agent(s), or biologic(s), combined with rheumatoid arthritis, psoriatic arthritis, or ankylosing spondylitis. The search was limited to RCTs and humans. No language, date, or publication status restrictions were applied. We also searched the Cochrane Library for any RCT included in the Cochrane Central Register of Controlled Trials, and for any systematic review on the subject.

Results were exported and compiled into a common reference database using EndNote. References were then de-duplicated to derive a unique set of records. Two investigators independently examined the search results and screened the titles and abstracts to exclude clearly irrelevant reports. The full text of the selected articles was critically evaluated for eligibility, and their reference lists (and of relevant reviews and meta-analyses) were manually scanned to identify further eligible studies. Experts were consulted for additional evidence, but there was no search for unpublished studies or data.

2.3Study selection and data extraction

We considered RCTs or open label extension (OLE) studies that evaluated an anti-TNF agent (adalimumab, certolizumab, etanercept, golimumab, or infliximab) as induction or maintenance therapy for adults with RA, PsA or AS, and reported the occurrence of infectious AEs. Eligible outcomes were: any infection, serious infections (infections that require antimicrobial therapy and/or hospitalization), tuberculosis, or opportunistic infections. We included studies that evaluated an anti-TNF therapy compared with placebo or no treatment, or multi-interventional therapies where the effect of anti-TNF treatment could be separated out (i.e. add-on to conventional disease-modifying anti-rheumatic drugs).

In case of multiple publications from the same study, we selected the most updated one and extracted the data for the maximum follow-up time. OLE studies were eligible if they represented an extension of previous RCTs, and reported infectious AEs according to the group to which the patients were originally randomized.

Data extraction was undertaken by independent reviewers. Any discrepancy was resolved by consensus, referring back to the original article. The following data were extracted from each study: first author's last name, journal and year of publication, trial's acronym, study design and duration, number of participants, disease studied (RA, PsA, or AS), patient characteristics (age, concomitant treatments, duration of disease), intervention parameters (drug, dose, administration), and numbers of participants with events (serious infection; tuberculosis; opportunistic infection; or any infection) reported for the intervention and control groups.

2.4Assessment of risk of bias

We evaluated the risk of bias (RoB) in included studies using the Cochrane Collaboration's tool [15,16], which addresses the following domains: sequence generation; allocation concealment; blinding; incomplete outcome data; and other sources of bias, such as extreme baseline imbalances in prognostic variables, selective crossover bias (i.e. subsequent anti-TNF treatment in the control groups), etc. These items were considered for the RoB assessment and were classified as "adequate" (low RoB), "inadequate" (high RoB), or "unclear". We considered only the information that was available in the full-text publications. Studies with adequate procedures in all domains were considered to have a low RoB; ones with inadequate procedures in one or more domains were considered to have a high RoB; and those with unclear procedures in one or more domains were considered to have unclear RoB.

Discrepancies among reviewers were discussed, and agreement was reached by consensus.

On the other hand, OLE studies have higher RoB than the original trials. The study populations are no longer randomly allocated, they are not blinded, and usually represent only a proportion of the participants recruited in the original trial (e.g., those with an adequate drug response and tolerance during the original study period). Therefore, we decided a priori that the items related to sequence generation, allocation concealment, masking of participants, personnel and outcome assessors, and incomplete outcome data, should be rated as high RoB in the OLE study assessment.

2.5Data synthesis and analysis

OR was the metric of choice in all comparisons. Study-level ORs and their 95% CIs were calculated by reconstructing contingency tables based on the number of participants randomly assigned and the number of participants with the events of interest (analysis by intention to treat). When no events occurred in one group of the trial, we used a continuity correction that was inversely proportional to the relative size of the opposite group. In particular, the continuity correction for the treatment group was 1/(R+1), where R is the ratio of control group to treatment group sizes. Similarly, the continuity correction for the control group was R/(R+1). This methodological approach outperforms the use of a constant continuity correction of 0.5 in a setting of sparse data and imbalanced study groups [17]. Trials reporting zero-event data for both study groups were excluded from the analysis.

We used two techniques to calculate the pooled effect estimates: the fixed-effects model (Mantel- Haenszel approach [18]) and the random-effects model (DerSimonian & Laird approach [19]). In the absence of heterogeneity, the fixed-and the random-effects model provide similar results. When heterogeneity is found, the random-effects model might be more prudent, though both techniques may be biased.

After the overall meta-analysis, we conducted subgroup analyses by the anti-TNF drug (adalimumab, golimumab, infliximab, certolizumab, or etanercept) to investigate potentially different effects on risk.

To assess any association between dose of TNF inhibitors and risk of infectious AEs, we also conducted a sensitivity analysis using (from the multi-arm trials) only the data referring to the intervention arms exposed to the highest doses. When no events were reported, the highest dose arm was merged with the second, the third highest, etc., in order to produce a group with at least one event, and include the study in the particular analysis. Full data were used from the two-arm trials.

Finally, we performed meta-regression analyses to investigate the impact of certain trial characteristics on the effect estimates. We converted all ORs by logarithmic transformation to achieve more symmetrical distributions. The natural logarithm of the OR was the dependent variable, and (i) age of participants enrolled and (ii) duration of follow-up were entered as covariates. This analysis was an indirect way to deal with aspects such as the possibility of effect modification by age, and to examine for increasing or decreasing risks with increasing duration of drug use, a feature associated with causal relationships. We applied a weighted regression model, so that the more precise studies have more influence in the analysis.

Regarding the open-label extension studies, their usefulness for generating reliable and valid data has been repeatedly challenged in the literature [20,21]. To avoid any biased or spuriously precise results, we did not include OLEs in our primary meta-analyses but synthesized their data separately.

Selective outcome reporting or publication bias was assessed using the funnel plot, Begg's test [22], and Egger's test [23]. The between-study heterogeneity was evaluated using Cochran's Q test [24] with a 0.10 level of significance. We also calculated the I-squared statistic [25], which describes the percentage variation across studies that is due to heterogeneity rather than chance. Negative values of I-squared were put equal to zero, so that I-squared lies between 0% and 100%. An I-squared value less than 40% was considered as indicative of "not important heterogeneity" and a value over 75% as indicative of "considerable heterogeneity" [26].

The quality of the meta-analytic evidence for each of the outcomes was assessed using GRADE (Grading of Recommendations Assessment, Development and Evaluation) [27].

For all statistical analyses, we used Stata 11 software (Stata Corp., College Station, Texas, USA), and the R software environment [28], version 3.1.1, and the "meta" package for R [29], version 3.8-0. All p-values are two-tailed. For all tests (except for heterogeneity), a p-value less than 0.05 was regarded as statistically significant.

2.6Role of the funding source

This study was supported by an unrestricted grant from Pfizer Italia. The funding source had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit it for publication.

3.1Search results

A summary of the evidence search and selection process is shown in Figure 1 , and in the Appendix.

Seventy-one RCTs met the eligibility criteria and reported the occurrence of infectious AEs (serious infection, tuberculosis, opportunistic infection, or any infection) during the study period; we were thus able to conduct a post hoc analysis of these trials, calculate ORs for the outcomes of interest according to the intention-to-treat principle, and incorporate them in the meta-analyses. These 71 RCTs evaluated infliximab (n=16), adalimumab (n=22), golimumab (n=11), certolizumab pegol (n=7), or etanercept (n=15) as induction or maintenance treatments for adult patients with RA (n=46), PsA (n=9), or AS (n=16). A total of 22,760 individuals participated in these trials; 14,766 in treatment groups, and 7,994 in control groups. Substantial imbalance was observed between the treatment and control group sizes within studies (median ratio, 2:1; maximum ratio, 7:1). The mean age of participants ranged from 30 to 62 years, and follow-up times from 1 to 36 months, between studies. A total experience of approximately 15,100 person-years was reached (8 months per patient, on average). The publication dates of these trials ranged from 1999 to 2014. A summary of the trials' characteristics is given in Table 1 .

Though most of the OLE studies were excluded because they did not report infectious AEs per study arm, we identified seven eligible OLEs [42, 56, [101][102][103][104][105] involving 2,236 participants. Their follow-up times ranged from 6 to 48 months. The outcomes of interest were reported according to the group to which patients were originally randomized, so we could calculate ORs and synthesize the evidence.

3.2Risk of bias in the studies included in meta-analysis

Random sequence generation: 11 of 71 RCTs (15.5%) reported adequate methods for sequence generation and were judged to be at low RoB. Information in 60 trials (84.5%) was insufficient to permit judgement (i.e. unclear RoB).

Allocation concealment: 33 of 71 RCTs (46.5%) reported adequate methods for allocation concealment (low RoB). In 38 trials (53.5%) information was insufficient (unclear RoB).

Blinding of participants and personnel: 20 of 71 RCTs (28.2%) were double-blind, three were not (4.2%), while for the other 48 trials (67.6%) information was insufficient.

Blinding of outcome assessment: In 26 of 71 RCTs (36.6%) the outcome assessment was blind, one was not blind (1.4%), while for the other 44 trials (62.0%) information was insufficient.

Incomplete outcome data: 49 of 71 RCTs (69.0%) were judged to be at high RoB. Fifteen trials (21.1%) were judged to be at low RoB, while for seven (9.9%) information was insufficient.

Other sources of bias: Nine studies (12.7%) suffered selective crossover across groups (high RoB).

Overall, the assessment indicated high RoB across 51 of the 71 included RCTs (71.8%) and unclear RoB for 19 RCTs (26.8%). Quality assessment items are presented in Figure 2 . Regarding the OLE studies' assessment, all studies were rated as high RoB.

3.3.1Serious infections

Fifty-eight RCTs [30][31][32][34][35][36][37][38][40][41][42][43][44][45][46][47][48][49][51][52][53][54][55] [56] [57] [58] [60] [61] [62] [63] [64][65][66][67][68][69][70][71][72][73][74][75][76]78,79,81,82,[84][85][86][87][88][89]91,94,95,99,100] involving 20,796 adult patients with RA, PsA, or AS, evaluated anti-TNF drugs and reported the occurrence of serious infections (2.6% in treatment groups; and 2.0% in control groups). Exposure to anti-TNF agents was associated with increased risk of serious infectious AEs, under both a fixed-effects model (OR:

1.41, 95% CI: 1.16, 1.73) and a random-effects model (OR: 1.25, 95% CI: 1.01, 1.55) (Table 2 ). The ORs with their 95% CIs for the individual trials, and the pooled results are presented in a forest plot (Figure 3 ).

Cochran's Q test had a p-value of 0.78 and the corresponding I-squared statistic was 0%, both indicating very little variability between studies. In contrast, the p-values for the Begg's test (p=0.06) and the Egger's test (p=0.09) suggested a possible bias. Indeed, there was a funnel plot asymmetry, with the left corner of the pyramidal part of the funnel missing (Figure 4A ). Small studies reporting relative risks lower than the unity are probably missing, and thus the estimated pooled OR for serious infections may have been overestimated.

The subgroup analysis investigating potentially different effects on risk by the anti-TNF drugs (adalimumab, golimumab, infliximab, certolizumab, or etanercept; Figure 3 ) was statistically significant under the fixed-effects model (p=0.05), while it was not under the random-effects model (p=0. 14), suggesting that etanercept and golimumab might have a better safety profile for serious infections, with point effects estimates closer to 1.0 (Figure 3 ).

To assess any association between higher doses of TNF inhibitors and risk of serious infections, we conducted a sensitivity analysis. The results did not materially change (fixed-effects, OR: 1.46, 95% CI:

1.19, 1.79; random-effects, OR: 1.36, 95% CI: 1.09, 1.69) (Table 2 ).

Meta-regression analysis, using the age of participants and the duration of follow-up as covariates, did not reveal any significant association (univariate analysis: age, p=0.72; duration of follow-up, p=0.21; and in multivariate analysis: age, p=0.69; duration of follow-up, p=0.20). [42, 56, 101,[103][104][105] did not provide any further evidence for the association between anti-TNF drugs and serious infections. The calculated summary effect estimate was not statistically significant assuming either a fixed-effects model (OR: 1.33, 95% CI: 0.77, 2.29) or a random-effects model (OR: 1.19, 95% CI: 0.68, 2.07). We found no evidence of selective outcome reporting or publication bias, or heterogeneity among the OLE studies (Table 2 ).

3.3.2Tuberculosis

Nineteen RCTs [33,43,46,48,51,52,54, 57, 59 ,62,67,73,75,81,84,93,94,99,100] involving 8,320 adults with RA, PsA, or AS, evaluated anti-TNF drugs and reported the occurrence of tuberculosis; it was 0.6% in the treatment groups (5,339 patients; 32 events), while no event was reported in the control groups (2,981 patients). Thus, continuity corrections (inversely proportional to the relative size of the opposite arm) were used in the analysis. Exposure to anti-TNF agents was associated with a statistically significant 3-fold increase in the risk of tuberculosis (fixed-effects model, OR: 3.53, 95% CI: 1.58, 7.85; random-effects model, OR: 3.29, 95% CI: 1.48, 7.33) (Table 2 ). The ORs with their 95% CIs from the individual trials, and the pooled results, are presented in Figure 5 . We found suggestive evidence of selective outcome reporting or publication bias, but no heterogeneity among studies (Table 2 ).

The subgroup analysis by the type of anti-TNF drug (Figure 5 ) did not reveal any difference among the drug-specific effect estimates (tests for subgroup differences: fixed effects, p=0.98; random effects, p=0.99). However, the power of this analysis is typically low, and, therefore, we cannot exclude clinically important differences between anti-TNF drugs treatment and progression of tuberculosis.

In the sensitivity analysis conducted to assess any association between higher doses of TNF inhibitors and risk of tuberculosis, the results did not materially change (fixed-effects, OR: 3.32, 95% CI: 1.54, 7.15; random-effects, OR: 3.23, 95% CI: 1.50, 6.98) (Table 2 ).

Meta-regression analysis, using the age of participants and the duration of follow-up as covariates, did not reveal any association (univariate analysis: age, p=0.89; duration of follow-up, p=0.98; and in multivariate analysis: age, p=0.87; duration of follow-up, p=0.93).

3.3.3Opportunistic infections

Only six RCTs [35,43, 63, 66,73,84] involving 3,886 adult patients reported the occurrence of opportunistic infections (0.3% in treatment groups; 0.3% in control groups). The association between anti-TNF drug use and the risk of opportunistic infections was neutral (fixed-effects, OR: 0.94, 95% CI: 0.33, 2.64; random-effects model, OR: 0.81, 95% CI: 0.23, 2.87). The ORs with their 95% CIs from the individual trials, and the pooled results, are presented in Figure 6 . We found no evidence of heterogeneity among the studies, selective outcome reporting or publication bias (Table 2 ).

The subgroup analysis, investigating potentially different effects on risk between types of anti-TNF drugs, did not reveal any difference among the drug-specific effect estimates. However, the power of this analysis was considerably low. Moreover, the sensitivity analysis did not suggest association between higher doses of the anti-TNF drugs and the risk of developing opportunistic infections (Table 2 ). 2 ). The ORs with their 95% CIs from the primary studies, and the pooled results, are shown in Figure 7 .

In

The Cochran's Q test had a p-value lower than 0.01 and the corresponding I-squared was 46%, both indicating important heterogeneity between studies. In contrast, the funnel plot (Figure 4B ), along with Begg's test (p=0.95) and Egger's test (p=0.75), showed no evidence of selective outcome reporting or publication bias.

The subgroup analysis by the type of anti-TNF drugs (Figure 7 ) did not reveal a significant difference among the drug-specific effect estimates (tests for subgroup differences: fixed effects, p=0.78; random effects, p=0.76).

The sensitivity analysis performed to assess any association between higher doses of TNF inhibitors and risk of any infection, confirmed the results reported above (fixed-effects, OR: 1.21, 95% CI: 1.11, 1.32; random-effects, OR: 1.22, 95% CI: 1.07, 1.39) (Table 2 ).

Importantly, in the meta-regression analysis, we obtained an estimate that was statistically significantly different from zero for the regression coefficient of the duration of follow-up (months) (coefficient=0.023; se=0.011; p=0.037; Figure 8 ). The between-trial heterogeneity was reduced by 17% when duration of follow-up was included as an explanatory variable in the model. The results did not substantially change when we included both the age of participants (p=0.88) and duration of follow-up (coefficient=0.023; se=0.011; p=0.039) in the model. This finding suggests that treatment with anti-TNF agents is associated with an increasing risk of infectious AEs, as duration of follow-up increases.

Synthesis of six OLE studies [42, 56, [101][102][103][104] with longer follow-up periods (range: 6 to 48 months) provided further evidence for the association between anti-TNF drug use and risk of any infectious AEs (fixed-effects, OR: 1.69, 95% CI: 1.31, 2.18; random-effects, OR: 1.56, 95% CI: 1.05, 2.33) (Table 2 ).

We found no evidence of heterogeneity among the OLEs, selective outcome reporting, or publication bias (Table 2 ).

3.4Quality of the evidence

In this meta-analysis, the quality of synthesized evidence is rated as "moderate" for the following reasons: (i) the evidence was derived from RCTs (randomized study design is considered the gold standard for assessing drugs); (ii) the meta-analytic effect estimates are precise (except for opportunistic infections); (iii) the results are consistent (heterogeneity was low or moderate across studies); and (iv)

the vast majority of the RCTs included in our study are characterized by high or unclear RoB, as assessed with the Cochrane Collaboration's tool (a fact that downgrades the quality of evidence). A moderate quality of evidence means that we are moderately confident in the effect estimate. The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

4.Conclusion

This systematic review encompassed data from 71 published RCTs involving 22,760 adult patients with rheumatologic disease (range of follow-up: 1 to 36 months), and from seven OLE studies with 2,236 patients (range of follow-up: 6 to 48 months). Quantitative synthesis of the available evidence supports the hypothesis that the use of anti-TNF drugs significantly affects the risk of infectious AEs. In particular, we found an increase in the occurrence of infections (20%), serious infections (40%), and tuberculosis (250%) associated with anti-TNF drug use, while the data for the opportunistic infections were scarce. Using the GRADE system [28], a summary of findings and strength of evidence is shown in Table 3 .

Over the last years, use of anti-TNF drugs by rheumatologic patients has been constantly increasing, and evidence on their safety continues to be collected [106]. Given the uncertainty on the effect of these agents on the risk of infectious AEs, we undertook a systematic review and meta-analysis on the topic.

Seventy-one RCTs with 22,760 adult patients met the eligibility criteria and reported the occurrence of infectious AEs, as a secondary (safety) endpoint. Study-level relative risk estimates were calculated, in accordance with the intention-to-treat principle, and were appropriately synthesized. Our results provide relevant evidence that the use of anti-TNF drugs significantly affects the risk of infectious AEs. Similar results were noted when we analyzed for higher drug doses or synthesized the seven eligible OLE studies. Thus, the findings of the present meta-analysis are in line with several observational studies reporting a significant rise in the risk of infectious AEs associated with anti-TNF drug use [5][6][7][8][9].

In 2006, Bongartz et al. published the first meta-analysis on this topic [4]. They identified a two-fold increase in the risk of serious infections associated with anti-TNF drugs (OR: 2.0, 95% CI: 1.3, 3.1), and an even higher increase when the analysis was restricted to high-dose groups vs. placebo (OR: 2.3, 95% CI: 1.5, 3.6). Since then, several meta-analyses have been published with conflicting results.

Leombruno et al. [107] analyzed 18 RCTs involving over 8,800 RA patients treated over an average of 0.8 years. They did not identify an increased risk of serious infections (OR: 1.21, 95% CI: 0.89, 1.63).

However, high-dose therapy was associated with a two-fold increase in the risk of serious infections. In 2010, Bernatsky et al. [108] published a meta-analysis of seven observational studies involving RA patients. Anti-TNF therapy appeared to significantly increase the risk of serious infections (OR: 1.37, 95% CI: 1.18, 1.60). In 2011, Thompson et al. [109] conducted a meta-analysis of six RCTs and showed no increased risk of serious infection in patients with early RA receiving anti-TNF therapy (OR: 1.28, 95% CI: 0.82, 2.00). As compared to those studies, our meta-analysis uses a much broader evidence base, includes a large number of trials (n=71), and provides updated evidence that can be appropriately integrated into relevant clinical guidelines.

This study has some limitations. Firstly, our search was restricted to published studies and we did not search for unpublished/original data. Secondly, the trials included in our review are characterized by high or unclear RoB, as assessed with the Cochrane Collaboration's tool. This is of concern, because the quality of the meta-analysis depends on the quality of the primary studies; if they are biased, then the meta-analysis will be biased as well. However some studies were rated at low risk of bias, and the estimates from studies at low risk did not differ from studies at high risk of bias. This increases the overall confidence we have in the final estimates. Thirdly, because duration of follow-up was up to 36 months, risk estimates resulting from longer exposure to anti-TNF agents are not possible. Given the meta-regression finding that anti-TNF drug therapy is associated with higher risks of infectious AEs as duration of follow-up increases, evidence describing infection risk during longer durations of anti-TNF therapy is required. However, the present study also has merits. It was conducted using a rigorous and extensive bibliographic search that allowed the inclusion of all relevant published RCTs. Furthermore, the relatively precise meta-analytic effect estimates, the absence of significant between-study heterogeneity, and the stability of the results in the subgroup and sensitivity analyses, strengthen our confidence in the accuracy of our findings.

5.Expert opinion

The results of our meta-analyses raise concerns about the use of anti-TNF in patients with infectious diseases. Included trials carefully excluded patients with histories of infections such as latent tuberculosis. Even after careful selection of patients for inclusion, our meta-analysis provides definitive evidence that anti-TNF drugs can disturb physiological cytokine-mediated signaling. The tuberculosis meta-analysis is paradigmatic. All events were in the anti-TNF arms, while the control arms were tuberculosis free. There are several prospective and retrospective studies exploring the association between anti-TNFs and tuberculosis [110][111][112][113][114][115]. The annual incidence rate varied depending on the country observed and the anti-TNF drug administered. Therefore, it is important to establish accurate latent tuberculosis infection screening strategy before commencing anti-TNF therapy in patients with immune-mediated inflammatory diseases [116]. The implementation of recommendations for latent tuberculosis infection screening and (prophylactic) treatment before initiation of anti-TNF therapy might reduce the infection incidence. Risk stratification scores associated with host demographic and clinical features, and previous or current non-biologic therapies are warranted to support the decision to start a treatment and the safest biologic choice [117]. For all anti-TNF drugs, the tuberculosis incidence rate was consistently higher than that in the general population, but infliximab and adalimumab were associated with the highest incidence rates when compared to etanercept [118]. Despite all anti-TNF neutralize TNF resulting in disruption of the granuloma that normally compartmentalizes Mycobacterium tuberculosis, etanercept might have a different impact on immunity that may allow the granuloma to reconstitute itself, thus preventing bacillary dissemination [119]. Our review did not identify substantial differences among the anti-TNFs drugs. However, the power of these subgroup analyses is typically limited: up to two cases of tuberculosis were included in trials exploring etanercept and golimumab, an inadequate number to demonstrate a definite association between the use of the drug and reactivation tuberculosis. Additionally concomitant corticosteroid and methotrexate therapies might be important confounding factors, hiding differences on drug inflammatory mechanisms and safety profile.

Given the increased risk of reoccurrence of infections, rheumatologists should further consider that the number of patients experiencing these adverse events is higher in studies other then RCTs such that the clinical consequences of the treatment might be more severe. There is not clear hypothesis for assuming that harms are different in directions or magnitude of effects across diseases, so we did not group studies by disease. We hypothesized that there were not strong differences in the case mix of patients across populations included in the RCTs, so harms, overall, should have been fairly consistent across studies.

However there was some heterogeneity, so some differences between patient populations cannot be excluded.

In conclusion, synthesis of existing evidence from RCTs involving rheumatologic patients confirms that anti-TNF drug use significantly increases the risk of infectious AEs, especially the risk for serious infections and tuberculosis. Given the increasing use of anti-TNF agents in adult patients with RA, PsA, or AS, it is important to continue monitoring their safety profiles, through complementary sources of research data, such as registries and long-term epidemiological studies.

Twenty-four-week efficacy and safety results of a randomized, placebo-controlled study. Footnote. Ideally, the funnel plot should have a symmetrical shape with a wide base and a narrow peak.

The figure indicates that smaller trials reporting odds ratios lower than the unity are probably missing, and thus the pooled effect estimate for serious infections may have been overestimated. Tests of publication bias: Begg's p=0.07; Egger's p=0.09. Abbreviations. OR: odds ratio; CI: confidence interval. Abbreviations. OR: odds ratio; CI: confidence interval.

annex

Footnotes:

(i) the basis for the assumed risk is the overall event occurrence across RCT control groups, (ii) the corresponding risk is based on the assumed risk in the comparison group and the relative effect of the intervention, (iii) the relative effect and its 95% CI come from a fixed-effects meta-analytic model, (iv) a corresponding risk could not be estimated for Tuberculosis, because no event was reported in the control groups, (v) the overall quality of the synthesized evidence is "moderate" for the following reasons: Data was derived from RCTs (randomized study design is considered the gold standard for assessing drugs); the meta-analytic effect estimates are precise (except for opportunistic infections); the results are consistent (heterogeneity was low or moderate across studies); and all the RCTs included in meta-analysis are characterized by high or unclear RoB in several important quality domains, such as allocation concealment and incomplete outcome data (a fact that downgrades the quality of evidence). A moderate quality of evidence means that "we are moderately confident in the effect estimate. The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different".

(vi) explanations for Summary of findings Tables can be found at: www.thecochranelibrary.com/view/0/SummaryFindings.html\
Acknowledgements

AcknowledgementsWe are grateful to Dr Antonio Spadaro who participated in the early phase of this study.Unfortunately, he passed away before seeing the completion of the review.
Declaration of interestThis study was supported by an unrestricted grant from Pfizer Italia through a service agreement with
Health Publishing & Services Srl. Health Publishing & Services Srl supports research activities at the
IRCCS Galeazzi Orthopedic Institute and the IRCCS Mario Negri Institute for PharmacologicalResearch.Valentina Marino is an employee of Pfizer Italia.All the other authors declare no conflict of interest related to the article.

Failed to load PDF:

Paper sources

Abstract screening pilot

Abstract screening results

Extraction pilot

Extraction results

Research report
\