Tumor necrosis factor, tumor necrosis factor inhibition, and cancer risk

Abstract

Objective:

Tumor necrosis factor (TNF) is a highly pleiotropic cytokine with multiple activities other than its originally discovered role of tumor necrosis in rodents. TNF is now understood to play a contextual role in driving either tumor elimination or promotion. Using both animal and human data, this review examines the role of TNF in cancer development and the effect of TNF and TNF inhibitors (TNFis) on malignancy risk.

Research design:

A literature review was performed using relevant search terms for TNF and malignancy.

Results:

Although administration of TNF can cause tumor regression in specific rodent tumor models, human expression polymorphisms suggest that TNF can be a tumor-promoting cytokine, whereas blocking the TNF pathway in a variety of tumor models inhibits tumor growth. In addition to direct effects of TNF on tumors, TNF can variously affect immunity and the tumor microenvironment. Whereas TNF can promote immune surveillance designed to eliminate tumors, it can also drive chronic inflammation, autoimmunity, angiogenesis, and other processes that promote tumor initiation, growth, and spread. Key players in TNF signaling that shape this response include NF-κB and JNK, and malignant-inflammatory cell interactions, each of which may have different responses to TNF signaling. Focusing on rheumatoid arthritis (RA) patients, where clinical experience is most extensive, a review of the clinical literature shows no increased risk of overall malignancy or solid tumors such as breast and lung cancers with exposure to TNFis. Lymphoma rates are not increased with use of TNFis. Conflicting data exist regarding the risks of melanoma and nonmelanoma skin cancer. Data regarding the risk of recurrent malignancy are limited.

Conclusions:

Overall, the available data indicate that elevated TNF is a risk factor for cancer, whereas its inhibition in RA patients is not generally associated with an increased cancer risk. In particular, TNF inhibition is not associated with cancers linked to immune suppression. A better understanding of the tumor microenvironment, molecular events underlying specific tumors, and epidemiologic studies of malignancies within specific disease indications should enable more focused pharmacovigilance studies and a better understanding of the potential risks of TNFis.

Key words::

• Cancer
• Inflammation
• Malignancy
• TNF
• TNF inhibition
• Tumor

Introduction

Various lines of evidence demonstrate a relationship between the immune system and cancer risk in humans and animal models1^,2. In humans, an increased risk of various cancers is observed in patients with chronic inflammatory conditions3–5, primary (congenital) immune deficiencies6^,7, acquired deficiencies (e.g., HIV/AIDS)8–10, or with therapeutics used to suppress inflammation in patients with various inflammatory disorders or to support organ engraftment11–14. Additionally, emerging data demonstrate the potential for immunotherapy in the treatment of specific cancers15–17 and the strong prognostic value of characterizing the immune composition in the tumor microenvironment in predicting patient survival18^,19.

Collectively, several juxtaposing models have been proposed to describe the relationship between specific arms of immunity and cancer1. In the tumor immunosurveillance/immunoediting model, elements of the innate and adaptive immune system protect the host by eliminating transformed cells, with a mutual evolution of the immune response and the cancer as disease progresses20^,21. Host immunity can also protect the host by eliminating, or at least suppressing, infection with oncogenic viruses22^,23. However, chronic inflammation can contribute to cancer initiation and tumor progression by creating a permissive microenvironment for tumor invasion and metastasis24–27. Finally, chronic antigenic stimulation, as seen with autoimmunity or unresolved pathogenic infection, can drive lymphomagenesis1. Contextual relationships between specific immune alterations and specific cancers challenge the development of generalizations regarding cancer risk with immune alterations. This challenge is becoming more acute as therapeutic strategies evolve from broad-spectrum immune suppression, used in the treatment of grievous conditions such as organ transplantation, to limited immunomodulation targeting specific cytokines or immune subsets.

Tumor necrosis factor (TNF, cachectin, TNF-α) is a central proinflammatory cytokine involved in various inflammatory conditions, including autoimmunity; inhibitors of TNF (TNFis) can markedly reduce the inflammation and manifestations of disease in patients with these conditions28. TNF is a trimeric protein, expressed from a wide variety of cells, and exists in both soluble and membrane-associated forms. TNF was originally identified as a glycoprotein found in the serum of Bacillus Calmette-Guérin–infected infected mice given endotoxin (also known as lipopolysaccharide). When this serum was administered intravenously, necrosis of sarcomas that had been transplanted subcutaneously into mice was observed29^,30. Cloning of human TNF was reported 10 years later by several groups31^,32. TNF is now known to have potent cytotoxic and proinflammatory effects. The observations that TNF could induce inflammatory responses and was elevated at sites of inflammation led researchers to explore the potential therapeutic benefit of inhibiting TNF activity in vivo33^,34. A secreted cytotoxic protein expressed by lymphocytes was also being characterized, purified, and subsequently cloned35–37. This factor, called lymphotoxin (LT), also demonstrated tumor necrosis activity and had noticeable sequence identity to TNF38^,39. Both LT and TNF have been shown to bind and activate TNFR1 and TNFR2, inducing similar in vitro responses such as cytotoxicity in L929 cells38.

Both lymphotoxin (LT or LT-α, TNF-β) and TNF bind to the same two receptors: TNF receptor (TNFR) superfamily 1a (TNFRSF1a; p55 or p60, type I, TNFR1, CD120a, ubiquitous expression) and TNFRSF1b (p75 or p80, type II, TNFR2, CD120b, expression is restricted to immune cells)28^,40, inducing similar in vitro responses such as cytotoxicity in L929 cells38. Soluble TNFRs are shed forms of the extracellular portion of the receptor40^,41 that bind TNF and inhibit its activity, thus regulating the bioavailability of soluble TNF40.

There are five TNFis that are currently licensed for use in the United States: three (adalimumab, golimumab, and infliximab) are monoclonal antibodies42–44, one (certolizumab) is a PEGylated Fab' fragment45, and one (etanercept) is a soluble receptor Fc fusion protein46. In addition to blocking TNF, only etanercept also inhibits LT; the properties of these agents are outlined in Table 1. Since the approval of the first marketed TNFi in 1998, globally there have been over 4 million patient-years of exposure to etanercept alone47. The broad clinical administration of these therapies, which are used chronically in a variety of inflammatory conditions (Table 1), combined with the identified association of systemic immune suppression (i.e., with congenital immunodeficiencies or acquired/induced profound immunosuppression) and increased cancer risk, has promoted interest in the potential long-term risks to patients treated with TNFis.

Table 1. Summary of currently approved TNFi therapies in the United States.

Agent	Structure	Target (s)	Approved Indications*
Adalimumab44	Fully human IgG1	TNF	Active AS, moderate to severe CD, moderate to severely active polyarticular JIA, active PsA, moderate to severe PsO, moderate to severely active RA, moderate to severe UC
Certolizumab45	PEGylated Fab' fragment	TNF	Patients with AS, moderate to severe CD, PsA, severe RA
Etanercept46	Human soluble p75 TNF Receptor linked to human Fc	TNF/LT	Active AS, moderate to severe JIA in patients ≥2 y, active PsA, moderate to severe PsO, moderate to severe RA
Golimumab42	Fully human IgG1	TNF	Active AS, active PsA, moderate to severe RA, moderate to severe UC
Infliximab43	Chimeric mouse-human IgG1	TNF	Active AS, moderate to severe adult and pediatric CD (≥6 y), active PsA, chronic severe PsO, moderate to severe RA, moderate to severe adult and pediatric UC

AS = ankylosing spondylitis; CD = Crohn’s disease; Fab' = fragment antigen binding; IgG1 = immunoglobulin G1; JIA = juvenile idiopathic arthritis; LT = lymphotoxin; PsA = psoriatic arthritis; PsO = psoriasis; RA = rheumatoid arthritis; TNF = tumor necrosis factor; TNFi = TNF inhibitor; UC = ulcerative colitis.

*Approved in adults unless otherwise indicated.

In this review, we analyze the current understanding of cancer risk with TNFis based on (a) information relating to TNF and tumorigenesis in experimental models, (b) understanding of the relationship between TNF, immunity, and cancer (TNF and tumor immune surveillance, TNF and antiviral immunity, TNF promoter polymorphisms and cancer risk), and (c) the accumulated clinical experience of TNFis over the last 16 years, notably in the treatment of rheumatoid arthritis (RA). A topic search was completed in MEDLINE and MEDLINE In-Process & Other Non-Indexed Citations on the Ovid database system. The search strategy consisted of using MeSH descriptors for tumor necrosis factor-alpha, lymphotoxin-alpha, cancer, cancer proliferation, malignancy, and immunosurveillance. Relevant clinical literature was searched through Ovid using the following search string: ((etanercept or Enbrel or Remicade or infliximab or adalimumab or Humira or golimumab or Simponi or certolizumab or Cimzia) AND (malignancy or malignancies or neoplasm malignant or neoplasm cancer or malignant tumor or malignant neoplasm or cancer or tumor malignant or neoplasm or unclassified tumor) and recurrence)).

Tumor necrosis factor and tumorigenesis in experimental studies

Antitumor activities of tumor necrosis factor

As mentioned above, TNF was originally identified as a glycoprotein found in the serum of Bacillus Calmette-Guérin–infected mice given endotoxin (also known as lipopolysaccharide). This serum could induce necrosis of sarcomas that had been transplanted subcutaneously into mice when administered intravenously29^,30. The structurally related protein LT has also demonstrated tumor necrosis activity39.

Experiments using purified TNF or cell lines expressing TNF demonstrate that in the models of established tumors, TNF can induce tumor necrosis. High concentrations of TNF are capable of killing tumor cells in in vitro systems and in animal models48–50. Based on this potential for TNF to induce tumor cell death, the utilization of TNF as a cancer therapy has been explored in the clinical setting. Local administration by isolated limb perfusion of TNF in combination with the alkylating agent melphalan or with doxorubicin was effective treatment for patients with metastatic melanoma and unresectable soft-tissue sarcomas51–54. Similar results have been observed for liver metastases of colorectal tumors in which isolated hepatic perfusion of TNF and melphalan was introduced via laparotomy55. The predominant effects of TNF delivered directly into tumors by isolated perfusion involve the tumor vasculature51^,56. Specifically, the procoagulant effects of TNFl50 may cause tumor necrosis through thrombus formation within the vasculature57. Overall, the available experimental data indicate that TNF is at most weakly cytotoxic or cytostatic to malignant cells, and the presence of metabolic inhibitors, which suppress downstream TNF signaling factors, is required for TNF-induced apoptosis to proceed58.

TNFR1 and TNFR2 are connected to multiple intracellular signaling pathways. TNFR1 can induce expression of genes involved in inflammation, survival, and proliferation through activation of kinase signaling cascades that ultimately activate nuclear factor kappa B (NF-κB)59. Alternatively, through the formation of the death-inducing signaling complex, TNFR1 can trigger cell death through apoptosis or necroptosis (programmed necrotic cell death; Figure 1)59. On balance, TNF is a poor inducer of apoptosis unless accompanied by inhibition of NF-κB, RNA synthesis, or protein synthesis60–62. TNFR2, expressed on a more restricted subset of cells, activates pathways implicated in cell survival proliferation and migration and modulates regulatory T-cell function63. Multimerization of transmembrane TNF (caused by association of surface-expressed receptors or through binding to bivalent antibodies) may also induce intracellular signaling of TNF-expressing cells (i.e., reverse signaling), leading to apoptosis, cell activation, or cytokine suppression28.

Figure 1. TNFRSF1a signal transduction. Activation of TNFRSF1a results in NF-κB activation and ultimately inflammation, angiogenesis, and cell survival. The apoptotic pathway is suppressed by FLIP. A decrease in NF-κB activation results in reduction of FLIP, which then favors signaling through the dissociated intracellular DD complex, resulting in caspase-dependent apoptosis. DD = death domain; FADD = Fas-associated death domain; FLIP = FADD-like IL-1β–converting enzyme; JNK = Jun kinase; NF-κB = nuclear factor kappa B; TNF = tumor necrosis factor; TNFRSF1a = TNF receptor superfamily 1a; TRADD =  TNF receptor–associated death domain; TRAF = TNF receptor–associated factor. Reprinted (with modification) from: Micheau O, Tschopp J. Induction of TNF receptor I-mediated apoptosis via two sequential signaling complexes. Cell 2003;114:188 © 2003 with permission from Elsevier.

Although originally identified as a serum factor associated with hemorrhagic necrosis of tumors induced by endotoxin29, TNF is now understood to play a contextual role in driving either tumor elimination or promotion26^,58^,64–70.

Impact of blocking tumor necrosis factor on tumorigenesis

The impact of blocking TNF on tumorigenesis has been examined using a number of models corresponding to different stages of tumor growth and different types of tumors, as summarized in Table 2.

Table 2. Role of TNF in tumorigenesis in nonclinical tumor models.

Model	Manipulation of TNF	Outcome	References
Protumorigenic Effects of TNF
Chemically induced skin tumors	TNF^−/− mice	Decreased tumor formation	71^,72
UVB-induced skin tumors	TNFR1^−/− or TNFR2^−/− mice	Decreased tumor formation	74
Urethane-induced pulmonary tumors	TNF^−/− or TNFi	Decreased tumor formation	73^,75
Choline-deficient, methionine-supplemented  diet induced liver tumors	TNFR1^−/− mice	Decreased tumor formation	76
PDAC in SCID mice	TNFi	Decreased tumor formation, decreased metastasis	77
	TNF administration	Increased PDAC growth and metastasis
Chemically induced ulcerative colitis and  colonic tumors	TNFi	Decreased tumor formation and associated neutrophil/macrophage infiltrates	78
Lung carcinoma cell lines	TNF^−/− mice	Decreased malignancy	79
Antitumorigenic Effects of TNF
Methylcholanthrene-induced sarcoma	Systemic TNF administration	Tumor necrosis	29^,30
Syngeneic mouse melanoma	Systemic TNF administration	Tumor growth inhibition	48
Human xenografts in nude mice	Intratumoral TNF injection	Tumor regression	49
Co-culture of myoblasts and prostate  carcinoma cell lines	TNFi	Recovery of tumor growth blocked by myoblast-derived TNF	80

PDAC = pancreatic ductal adenocarcinoma; SCID = severe combined immunodeficiency; TNF = tumor necrosis factor; TNFi = TNF inhibitor; TNFR = TNF receptor; UVB = ultraviolet B.

While systemic or intratumoral injection of supraphysiological levels of TNF can cause inhibition of tumor growth, studies examining animal models of tumorigenesis demonstrate that eliminating TNF or its receptors through genetic manipulation (TNF^−/− or TNFR^−/− mice) is markedly inhibitory to tumor induction and growth.

Suganuma et al.71 and Moore et al.72 have reported that TNF knockout (TNF^−/−) mice were more resistant to tumor formation than TNF^+/+ mice when evaluated in skin carcinogenesis models with chemical induction/promotion. TNF^−/− mice are protected from urethane−induced tumorigenesis73, and blocking TNF bioactivity with the p75 TNF receptor Fc fusion protein etanercept blocked tumor initiation/promotion and tumor progression in urethane-dosed mice through reduction of tumor microvascular density and cell proliferation75. The tumor-promoting activity of TNF was also demonstrated in a benzo[a]pyrene-7,8-diol-9,10-epoxide skin carcinogenesis model81. Endogenous TNF enhanced the malignancy of the lung carcinoma cell line 3LL cells introduced into TNF^+/+ versus TNF^−/− mice as the result of the induction of the promalignant gene secretory leukocyte protease inhibitor SLP177–79^,82^,83. Studies with TNFR-deficient mice show similar results. TNFR1/2^−/− mice were found to have significantly fewer tumors in response to ultraviolet B irradiation73. TNFR1^−/− mice were also resistant to liver tumor formation in a model in which oval cell proliferation is induced by a choline-deficient, methionine-supplemented diet76.

TNF was identified as a critical macrophage-derived factor for promoting an epithelial-mesenchymal transition in a model of invasive colon cancer82. Inhibition of TNF by etanercept or the antibody infliximab exerted strong antitumoral effects in a pancreatic ductal adenocarcinoma (PDAC) model in severe combined immunodeficiency (SCID) mice, illustrating the important role of tumor-derived TNF in tumor growth77. Administration of etanercept to wild-type (WT) mice after treatment with azoxymethane and dextran sulfate sodium (a model of ulcerative colitis leading to the development of colonic tumors) markedly reduced the number and size of tumors and reduced colonic infiltration by neutrophils and macrophages78. TNF also facilitates the clonal expansion and dominance of mutated Janus kinase 2 cells in myeloproliferative neoplasms83.

Taken together, these data indicate that in various models TNF can stimulate or promote tumor formation, whereas the lack or inhibition of TNF can reduce tumor formation.

Tumor necrosis factor and metastatic processes in experimental models

Several lines of evidence demonstrate that TNF may promote metastasis of certain tumor cell types. As mentioned earlier, TNF treatment dramatically enhanced PDAC growth and metastasis in a mouse model, whereas TNFi treatment with infliximab or etanercept showed strong antitumoral effects, decreasing liver metastases and volumes of recurrent tumors77. Similarly, in a separate in vitro study, TNF treatment of cultured colon-26 cells enhanced metastatic properties involving production of metallomatrix protein-9, adhesion, migration, and invasion84–87. Chinese hamster ovary cells overexpressing TNF had a greater ability to invade and metastasize in mice compared with cells not overexpressing TNF86, and metastatic lymphoma cells have been shown to have an even greater metastatic potential when coadministered with recombinant mouse TNF87. Inhibition of TNF using an anti-TNF fusion protein (human TNFR1 extracellular portion and human immunoglobulin G1 [IgG1] Fc) decreased metastasis to the lung in a B16-BL6 mouse melanoma model, demonstrating a direct relationship between TNF and the metastatic process80^,88. Myoblasts restrict prostate cancer growth and limit metastasis formation through paracrine TNF secretion80.

Taken together, the results of these studies indicate that TNF can promote metastasis and that TNFi treatment in rodent models decreases metastatic potential.

Different mechanisms connecting tumor necrosis factor, immunity, and cancer

Tumor necrosis factor and tumor immune surveillance

In the immunosurveillance model, the host immune system protects against tumor growth and development21. Innate and adaptive immune cells, primarily natural killer (NK), cytotoxic T-lymphocyte (CTL), and natural killer T (NKT) cells, are the key mediators of host protection from cancer, working in concert with key cytokines and mediators (e.g., interferon [IFN]-γ, perforin, granzymes) to eliminate tumors. However, tumors can downregulate antigen presentation to evade the immune system or express cytokines (e.g., tumor growth factor [TGF]-β, interleukin [IL]-10) to induce immunosuppressive cells (e.g., myeloid-derived suppressor cells, tumor-associated macrophages, and regulatory T cells). This model is supported by mechanistic data evaluating tumor growth in various animal models with immunodeficiency (e.g., recombination activation gene [RAG]1^−/−, RAG2^−/−, signal transducer and activator of transcription [STAT]1^−/−, SCID, nude21). In addition, inflammatory cell infiltration within primary tumor samples from patients with colorectal, breast, and other cancer types shows that the presence of CD8+ and Th1 T cells is associated with good prognosis, and while the presence of Th17 and regulatory T cells in predicting patient outcome can vary between tumor types, generally poor outcomes have been reported18^,19. Evidence supporting the importance of a strong adaptive antitumor response is emerging from clinical data on approved immunotherapies, which augment antitumor immune responses, particularly dendritic cell (DC) and CD8+ T-cell interactions15^,16. In addition, a variety of tumor-derived factors contribute to the emergence of immunosuppressive networks, including vascular endothelial growth factor, IL-10, TGF-β, prostaglandin E(2), soluble phosphatidylserine, soluble Fas, soluble Fas ligand, and soluble major histocompatibility complex (MHC) class I–related chain A proteins. These factors can suppress protective host immune responses and promote tumor growth, invasion, and metastasis89^,90.

The specific role of TNF in regulating factors of immune surveillance of tumors has been studied in a variety of models (Table 3). Studies of TNF^−/−, LT^−/−, TNFR^−/− mice provide information on the key roles of these genes in the immune system and immune surveillance, which help identify their potential roles in tumorigenesis. In several rodent tumor models, TNF was associated with a protective immunosurveillance. Tumor rejection in TNF^−/− mice was reported to be inhibited with the highly antigenic fibrosarcoma MC57X91. In this model, tumor rejection is dependent on animals mounting an immune response to tumor antigens (i.e., mice are immunized with irradiated tumor cells before implantation), and thus failure to reject may be the result of the blunted immune response in TNF^−/− mice. The administration of a TNF antibody to WT mice also resulted in impaired tumor rejection in this model, and the administration of recombinant TNF (rTNF) to TNF^−/− mice resulted in tumor rejection91. In addition, in a modified model of metastatic lung cancer using the 3LLA-9 Lewis lung carcinoma cell line, in which cells transfected with a lymphocytic choriomeningitis virus antigen can be rejected because of an immune response to the transfected antigen, tumor rejection failed to occur in TNF^−/− mice100. Using the methylcholanthrene (MCA)–induced sarcoma as a chemically induced cancer model sensitive to effects on innate and adaptive immune surveillance, TNF^−/− mice showed increased susceptibility compared with TNF^+/+ mice101. However, TNFR1^−/−, TNFR2^−/−, and TNFR1/2^−/− animals did not show increased development of MCA-induced sarcoma compared with WT littermates102. Tumor-draining lymph node effector T cells cultured with TNF mediate increased regression of established mouse tumors103. In a model of LT deficiency, impaired NK function was associated with increased tumor growth and metastasis104.

Table 3. Impact of TNF on immune surveillance (nonclinical models).

Model	Manipulation of TNF	Outcome	References
Increased Immune Surveillance by TNF
In vitro cytotoxicity against tumor cell lines	TNF^−/− mice	Impaired NK- or LAK-mediated cytotoxicity	91
Murine NK counts and activity against target cells	LT^−/− mice	Decreased NK cell development	92
In vitro cytotoxicity and NK activation in murine  DC and NK cultures	TNF^−/−, TNFR1^−/−, TNFR2^−/− mice, TNF gene transfer	DC membrane TNF mediates NK activity through TNFR2	93
CTL cytotoxicity assay following in vivo infection  (LCMV)	TNF^−/− LT^−/− mice	Attenuated antiviral CTL response	94
In vitro cultures of allospecific CTL	TNF^−/− mice	Decreased CTL-mediated cytotoxicity	91
In vivo tumor models	TNF^−/− mice	Decreased tumor rejection	91
Lack of Effect or Negative Impact of TNF  on Immune Surveillance
CTL cytotoxicity against allogeneic target	TNF^−/− mice	No impact on cytotoxicity	95
CTL cytotoxicity against allogeneic target	LT^−/− mice	No impact on cytotoxicity	96
CTL development in OVA-expressing tumor cells	TNFR2^−/− mice	Decreased activation-induced cell death of CTLs and increased control of tumor cell growth	97
Nonhuman primate toxicology study	Systemic TNFi	No changes in lymphocyte populations, DTH response, T-dependent antibody responses	47
MCMV infection model	Systemic TNFi	No lasting effect on viral infection	47
Cultures of myelodysplastic syndrome cells	Culture with TNF	Increased expression of immunoinhibitory molecule B7-H1	98
Murine tumor models	TNFR1/2^−/− mice	TNF drives myeloid-derived suppressor cell accumulation	99

CTL = cytotoxic T-lymphocyte; DC = dendritic cell; DTH = delayed-type hypersensitivity; LAK = lymphokine-activated killer; LCMV = lymphocytic choriomeningitis virus; LT = lymphotoxin; MCMV = mouse cytomegalovirus; NK = natural killer; OVA = ovalbumin; TNF = tumor necrosis factor; TNFi = TNF inhibitor; TNFR = TNF receptor.

TNF^−/− and/or LT^−/− mice have specific immune function changes, including impaired NK and lymphokine-activated killer (LAK) cytotoxicity against various tumor cell lines (L929, WEHI 164, P815 for LAK; YAC-1 for NK)91^,93–95^,105, supporting a role for TNF in tumor killing. Similarly, CTL responses against DCs pulsed with cellular extracts from allogeneic splenocytes were also decreased compared with TNF^+/+ littermates, while it was shown that TNF^−/− effector lymphocytes still exhibit both perforin and Fas ligand-based cytotoxicity91. Similar to LAK and NK activities, CTL cytotoxicity was restored upon incubation of TNF^−/− lymphocytes in vitro with rTNF during a cytotoxicity assay or upon treatment of TNF^−/− mice with rTNF91.

NK cell development from bone marrow precursors is suppressed in LT^−/− mice, and splenocytes from LT^−/− mice exhibit lower NK cytotoxicity in vitro against YAC-1 cells, possibly as the result of a lower content of splenic NK cells, while perforin is detectable in NK cells from LT^−/− mice104. In vitro cytotoxicity function against target cells P815 (MHC class I+, class II−) and TA3 (MHC class I+, class II+) is intact in LT^−/− mice96.

Characterization of the role of TNFR in immunosurveillance, using TNFR1^−/−, TNFR2^−/−, or TNFR1/2^−/− mice, has been limited and more complex to interpret; TNFR2^−/− CD8+ T cells are resistant to apoptosis, leading to increased protection against tumor growth97^,106.

Overall, the complete abrogation of TNF signaling may impair some components of immunosurveillance in rodents, although in some instances the disruption of TNF signaling may increase antitumor CTL activity (due to decreased apoptosis).

In different experimental settings, TNF was associated with a negative effect on immunosurveillance. TNF expression was significantly associated with positive lymph node stage and recurrence of colorectal cancer, possibly by induction of apoptosis of tumor-infiltrating CD8+ T cells107. TNF was demonstrated to favor proliferation of myelodysplastic syndrome blasts through induction of the immunoinhibitory molecule B7-H1, responsible for T-cell suppression via programmed cell death (PD-1) molecules98. In a murine model of melanoma, treatment with DCs matured by TNF led to an increase in the number of lung metastases, an effect associated with increased lung IL-10+ T cells and not observed in IL-10^−/− mice108. LT treatment was shown to decrease MHC I expression in HPV16-positive CaSki cells and to reduce CTL-specific lysis of CaSki cells109.

These seemingly contradictory results, which demonstrate both positive and negative roles of TNF in immunosurveillance, suggest that the role of TNF in immune surveillance is complex, contextual and reflects the pleiotropic nature of TNF signaling.

Specific effects of TNFis on some components of immune surveillance have been evaluated through animal studies. Cynomolgus monkeys administered 3 mg/kg etanercept twice weekly showed no changes in total lymphocyte/lymphocyte-subtype populations, nor were there any changes in markers for ‘activation’ or ‘memory’ phenotypes when using the following markers: CD3 (T lymphocytes), CD3/CD4 (helper T cells), CD3/CD8 (cytotoxic T cells), CD4/CD44 (memory CD4 cells), CD8/CD44 (memory CD8 cells), CD4/CD69 (predominantly activated helper T cells), CD8/CD69 (activated cytotoxic T cells), CD14 (monocytes), CD16 (NK cells), CD16/CD69 (activated NK cells), and CD20 (B lymphocytes)47. To assess the possible influence of etanercept on immune function, animals were immunized with an antigenic preparation for delayed-type hypersensitivity (DTH) response assessment, and with keyhole-limpet hemocyanin and tetanus toxoid to assess capacity for production and maintenance of a primary or secondary T cell–dependent antibody response47. The administration of etanercept had no apparent effect on the ability of cynomolgus monkeys to mount a DTH response to an antigenic mixture of candida/trichophyton/diphtheria, an antibody response to a novel antigen, or to a recall antigen.

Studies in mice have shown that etanercept can modulate the immune response to specific challenges. For example, in murine models of collagen-induced arthritis, suppression of the immune-based destruction of joint tissue is the basis for therapeutic activity. However, in these models, normal levels of T-cell responses to mitogens and type II collagen are found110. Together, these data show that pharmacologic inhibition of TNF is not associated with broad immunosuppression.

Tumor necrosis factor and antiviral immunity

The host response to viral infection is mediated by both innate (NK, NKT, and granulocyte) and adaptive (CTL) antiviral responses, which contribute to viral clearance. The production of proinflammatory cytokines such as TNF, IL-12, IL-18, and IL-6 and the cellular production of IFN-α both directly interfere with viral replication and promote antiviral immunity22^,23.

Administration of etanercept in an acute model of murine cytomegalovirus infection resulted in increased viral load in the liver in only the very early phase of infection (day 3), and the effects did not extend to later times; no infected cells were seen at day 7 in either treated or control groups47. However, the effects on a latent infection have not been examined, and at present, there are no established rodent viral models for potential human oncogenic viruses in which the human protein etanercept was examined.

In a clinical study, patients who received infliximab, adalimumab, or etanercept showed no difference in Epstein–Barr virus (EBV) viral load and no difference in numbers of EBV-specific IFN-γ–producing T cells (for both latent-cycle and lytic-cycle peptides) between baseline and treatment week 12111. Other clinical studies showed that TNFi therapy did not increase the EBV burden in patients with RA or ankylosing spondylitis112^,113.

Tumor necrosis factor promoter polymorphisms and cancer risk

Inflammation contributes to tumor development and growth through the expression of cytokines such as TNF and other inflammatory mediators, which establish a permissive environment for tumor growth and metastases, angiogenesis, increased cell survival, reduced cell death, altered differentiation, and altered response to hormones and chemotherapeutic agents24^,26^,27^,58^,114–117. Clinical data indicate that inflammation associated with increased TNF is a negative prognostic indicator of patient outcomes across a variety of tumor types. The TNF-308 A>G promoter region polymorphism, associated with two- to three-fold increased transcription and up to two-fold increased circulating TNF levels (for G/A vs G/G) in healthy humans118^,119, has been positively associated with various malignancies. Meta-analyses covering eight cancer classifications (cervical, breast, lung, hepatocellular, gastric, aerodigestive, non-Hodgkin lymphoma, and colorectal) and involving over 50,000 cases and 66,000 controls have established a statistically significant increased cancer risk (expressed as an odds ratio) for A allele polymorphisms ranging from 1.1 to >4, with some variability across allelic forms (i.e., A/A, G/A, or either) and subgroup factors (e.g., menopausal status, race)120–129. These data indicate a broad and complex role for TNF-308 G/A polymorphisms that regulate TNF transcription and production, although the relationship of these polymorphisms with genetic and other factors, such as ethnicity, that affect cancer risk remain poorly understood130.

Other TNF polymorphisms have been linked with an increased cancer risk (i.e., 238 G/A, 857 C/T, 863 A/C, 1031 C/T, 1210 C/T), although meta-analyses have not identified clear associations of these polymorphisms with cancer risk120^,123^,124^,127^,131^,132. In addition to TNF polymorphisms, the LT-α NcoI A/A genotype, and its associated high production of LT-α, are associated with an increased risk of developing esophageal adenocarcinoma133.

Whereas TNF polymorphisms have been associated with increased TNF production and cancer risk, increased systemic TNF from other causes also increases cancer risk. In B-cell chronic lymphocytic leukemia and hairy cell leukemia, overproduction of TNF by the malignant cells has been reported to result in more severe disease progression and decreased survival134^,135. Freedman et al. found that cells from patients with juvenile chronic myelogenous leukemia endogenously produced TNF, which played an important role in the pathogenesis of the disease by acting as an autocrine growth factor136. These effects were inhibited following treatment with an anti-TNF antibody. TNF also maintains TNFR1-dependent IL-17 production in CD4+ T cells, which leads to myeloid cell recruitment into the tumor microenvironment and enhances tumor growth in ovarian cancer in mice and humans137. Although Helicobacter pylori is considered to be a major factor contributing to gastric cancer, it was demonstrated that TNF mRNA and protein expressions were significantly increased in chronic gastritis, intestinal metaplasia, dysplasia, and gastric adenocarcinoma patients and that the degree of overexpression of TNF showed significant association with various stages of gastric carcinogenesis138. In the context of H pylori, the TNF-inducing protein (Tip-α) was identified as a carcinogenic factor that acts through strong induction of TNF139.

Overall, these data indicate that TNF in the inflammatory microenvironment may favor tumor growth in several models, and TNF expression is a poor prognostic factor for several tumor types in humans.

Taken as a whole, the studies summarized above indicate that TNF can activate pathways leading to three contextually different cellular responses: cell survival and activation/proliferation, transcription of proinflammatory genes, and cell death. These various effects are critical for controlling effectors of immune surveillance of tumors, tumor growth, and nonvirally-mediated lymphomagenesis as summarized below and illustrated in Figure 2.

• Although some models indicate that abrogation of the TNF pathway can reduce NK- and CTL-mediated cytotoxicity against tumor cells, other models indicate that CTL activity is intact when the TNF pathway is inhibited and, furthermore, that TNF can inhibit antitumor immunity through expansion of suppressor cells and induction of T-cell inhibitory molecules. Pharmacologic modulation of the TNF pathway (e.g., with etanercept) does not affect the number of circulating leukocytes (including NK cells and CTLs) in nonhuman primates and does not interfere with T-cell responses and antiviral immunity.

• TNF shows antitumor activity in vivo in vascularized tumors through disruption of endothelial cell adhesiveness and procoagulation effects. It is weakly cytotoxic against many tumor cell lines unless combined with antimetabolic agents.

• TNF (and inflammation in general) favors tumor growth and metastasis in several models, and available clinical data indicate that inflammation associated with TNF is a negative prognostic indicator of patient outcomes across a variety of tumor types. In particular, the TNF-308 G/A promoter region polymorphism, associated with a two- to three-fold increased transcription and up to two-fold increased circulating TNF levels in healthy humans, has been positively associated with various malignancies.

Figure 2. TNF and tumorigenesis. (A) TNF can impact multiple effectors of immune surveillance and exhibits either pro- (increased NK and CTL function), or anti- (increased suppressor cells, induction of PD-L1/B7-H1) surveillance activities. (B) TNF directly affects tumors through a complex contextual integration of signals driving a cellular response toward apoptosis/necrosis or survival, inflammation, and growth promotion. (C) TNF can initiate and progress autoimmunity, which can drive lymphoma development. AICD = activation-induced cell death; CTL = cytotoxic T-lymphocyte; DC = dendritic cell; IL = interleukin; MDSC = myeloid-derived suppressor cell; NK = natural killer; Th = T-helper lymphocyte; TNF = tumor necrosis factor. Adapted from: Aringer M, Smolen J. Complex cytokine effects in a complex autoimmune disease: tumor necrosis factor in systemic lupus erythematosus. Arthritis Res Ther 2003;5:174 © 2003 with permission from BioMed Central Ltd.

Tumor necrosis factor inhibitors: clinical trials and epidemiology data

While nonclinical models contribute to the understanding of the biological consequences of TNF inhibition, many of these models rely on rodent systems and/or are contrived, therefore their translatability to human disease is uncertain. The following section focuses on human data and reviews the overall risk of malignancy associated with TNFis, as well as the risk for specific tumor types.

Overall risk of malignancy in rheumatoid arthritis patients

Randomized clinical trials are underpowered and too short to determine the risk of rare events such as malignancy. Meta-analyses, large registries, and administrative database analyses have provided substantial evidence that malignancy rates are not increased overall with TNFis. Here, we focus on data from RA because this population is well defined regarding background rates of malignancy and has the largest amount of data examining the risk of malignancy associated with TNFis relative to other indications.

The attribution of malignancy risk to TNFis is complicated by the increased risk of malignancy in RA patients relative to the general population, as observed in the Danish Biologics (DANBIO) register (standardized incidence ratio [SIR] = 1.27; 95% CI, 1.08–1.49)140 and the British Society for Rheumatology Biologics Register (BSRBR; SIR = 1.28; 95% CI, 1.10–1.48)141. This is further confounded by the fact that these patients are also exposed to other cytotoxic or immunosuppressive drugs.

Hematologic malignancy

Lymphoma risk is increased in RA patients at least two-fold overall, and increased disease severity is correlated with increased risk of lymphoma142. The RA population also has a higher rate of leukemia relative to the general population in the United States (SIR = 1.7; 95% CI, 1.2–2.4) and Sweden (SIR = 2.1 [95% CI, 1.7–2.5] and 2.2 [95% CI, 0.6–5.7] for prevalent and incident RA, respectively)143^,144. In particular, patients with RA have a higher risk of acute myeloid leukemia, with an SIR of 2.4 (95% CI, 1.79–3.15)145^,146. Taken together, these data show that patients with RA have an increased risk of leukemia and lymphoma relative to the general population.

Solid tumors

A decreased risk of colon and breast cancers in RA patients relative to the general population was reported in both the Spanish Registry for Adverse Events of Biological Therapies in Rheumatic Diseases (BIOBADASER) and the US National Data Bank (NDB) for Rheumatic Diseases143^,147. The observed decrease in colon cancer may be attributed to nonsteroidal anti-inflammatory drug use in this population148^,149. Similarly, decreased rates of prostate and female genital cancers were observed in RA patients in the BSRBR registry141. Lung cancer may be increased in patients with RA compared with the general population, although conflicting results have been observed147. Increased rates of smoking may explain increased lung cancer rates, because smoking is a risk factor for RA150. An increased risk of nonmelanoma skin cancer (NMSC) in RA (irrespective of treatment) has been reported151. Further, compared to patients with osteoarthritis, RA patients have an increased risk of developing NMSC (hazard ratio [HR] = 1.19; 95% CI, 1.01–1.41)152. Disease duration is also associated with NMSC (HR = 1.01; 95% CI, 1.00–1.02; p = 0.004)152.

Cancer risk associated with tumor necrosis factor inhibitors

Meta-analyses of pooled clinical trial data across combined approved indications do not indicate an overall increased risk of malignancy with TNFis140^,144^,147^,153–165 (Table 4). Moreover, increased duration of TNFi exposure does not appear to increase malignancy risk. Results from the DANBIO registry showed no overall increased risk of malignancy based on the cumulative duration of TNFi exposure140, a finding similar to that of an analysis of a large US observational study143. Additionally, individual TNFis do not show differences in risk; no significantly increased risk of malignancy was observed for the individual TNFis adalimumab, etanercept, or infliximab in the BIOBADASER 2.0 study166. In a study conducted using the Taiwan National Health Insurance Research Database, the risk of newly diagnosed cancer was compared between RA patients taking biologics (predominantly TNFis alone; 13.1% received rituximab) and matched subjects taking nonbiologic disease-modifying antirheumatic drug (nbDMARDs) only. The overall risk of cancer was significantly reduced in subjects receiving TNFis (HR = 0.63; 95% CI, 0.49–0.80)154.

Table 4. Relative risk of malignancies in patients with RA and other inflammatory conditions (key studies).

Study; Population	Tumor Type	TNFi	Reference Group	N	Risk or Incidence Ratio (95% CI)
Clinical Trial Meta-Analyses
Askling et al.155; all indications for TNFi	All	ADA, ETN, IFX	Control arms	22,904	HR = 1.30 (0.89–1.95)
Gottlieb et al.156; AS, RA, JIA, PsA	All	ETN	SEER database	13,977	HR = 1.00 (0.83–1.19)
Burmester et al.159; RA	All	ADA	SEER database	14,109	SIR = 0.93 (0.82–1.06) for RA
Database and Registry Analyses
BSRBR148; RA	Solid tumors	ADA, ETN, IFX	nbDMARD	15,015	HR = 0.83 (0.64–1.07)
DANBIO140; RA, AS, PsA, and  other arthritides	All	TNFi	nbDMARD	9696	HR = 1.02 (0.80–1.30)
BIOBADASER 2.0147; RA	All	TNFi	nbDMARD	3320	IRR = 0.48 (0.09–2.45)
SABER157; RA, PsO, PsA, IBD	All	TNFi	nbDMARD	29,555	HR = 0.94 (0.79–1.12) for any solid tumor in RA patients; HR = 1.25 (0.71–2.20) for any lymphoma in RA patients
Mariette et al.158; RA	All	TNFi	nbDMARD	>40,000	Pooled risk estimate = 0.95 (0.85–1.05)
Corrona160; RA	All	TNFi	Methotrexate	5327	HR = 0.29 (0.14–0.60)
ARTIS161; RA	Solid tumors	ADA, ETN, IFX	General Swedish population	4160	SIR = 0.9 (0.7–1.2)
ARTIS144; RA	Hematopoietic	ADA, ETN, IFX	General Swedish population	4160	SIR = 2.1 (1.1–3.8)
SSTAG162; RA	All	ETN and IFX	RA patients not exposed to TNFi	1557	SIR = 1.1 (0.6–1.8)
LORHEN163; RA	All	ADA, ETN, IFX	General population	1064	SIR = 1.09 (0.64–1.72)
NA database165; RA	All	ETN and IFX	Methotrexate	7830	HR = 0.99 (0.71–1.36)
RABBIT164; RA	All	TNFi	nbDMARD	5120	HR = 0.70 (0.44–1.12)
Taiwan National Health Insurance  Research Database154; RA	All	ADA, ETN	nbDMARD	22,130	HR = 0.63 (0.49–0.80)

ADA = adalimumab; ARTIS = Antirheumatic Therapies in Sweden; AS = ankylosing spondylitis; BIOBADASER = Spanish Registry for Adverse Events of Biological Therapies in Rheumatic Diseases; BSRBR = British Society for Rheumatology Biologics Register; CD = Crohn’s disease; ETN = etanercept; HR = hazard ratio; IBD = inflammatory bowel disease; IFX = infliximab; IRR = incidence rate ratio; JIA = juvenile idiopathic arthritis; LORHEN = Lombardy Rheumatology Network; NA = North America; nbDMARD = nonbiologic disease-modifying antirheumatic drug; PsA = psoriatic arthritis; PsO = psoriasis; RA = rheumatoid arthritis; RABBIT = Rheumatoid Arthritis Observation of Biologic Therapy; SABER = Safety Assessment of Biological Therapeutics; SEER = Surveillance, Epidemiology, and End Results; SIR = standardized incidence ratio; SSTAG = South Swedish Arthritis Group; TNFi = tumor necrosis factor inhibitor.

Hematologic malignancies – lymphoma

Across clinical trials, observational studies, meta-analyses, and postmarketing analyses, the risk of lymphoma does not appear increased with TNFis relative to background therapy. Analyses that compared patients with RA treated with TNFis to the general population reported SIRs ranging from 3.5 to 5.8156^,167. However, patients with more severe disease are more likely to receive TNFis, thus complicating calculations of true risk of malignancy. Data do exist that compare the risk associated with TNFis to other RA medications. In a meta-analysis of 10 randomized controlled RA trials, the odds ratio (OR) for lymphoma for all TNFis combined was not statistically significant (2.14; 95% CI, 0.55–8.38) compared with patients with RA who received an nbDMARD or placebo168^,169. Similarly, Leombruno et al. found no significantly increased risk for lymphoma; the exposure-adjusted relative risk (RR) was 1.26 (95% CI, 0.53–3.01) for all TNFis combined168. There were also no major differences found for the individual TNFi agents, nor did high doses of TNFis confer a greater risk of lymphoma relative to patients who received nbDMARDs168. A pooled analysis of three registries reported no increased risk of lymphoma associated with TNFis in patients with RA compared with nbDMARDs (RR = 1.11; 95% CI, 0.70–1.51)158. In the Safety Assessment of Biological Therapeutics (SABER) study, the risk associated with TNFis versus nbDMARDs was also nonsignificant (HR = 1.25; 95% CI, 0.71–2.20)157. In an analysis of a Swedish registry of 757 patients with RA treated with TNFis compared with 800 treated with nbDMARDs, a trend for an increased risk was found; the adjusted HR for lymphoma was 5.0 (95% CI, 0.9–27.9; p = 0.06) with an SIR of 11.5 (95% CI, 3.7–26.9); however, the population examined in this study was smaller than in other studies162. In the Taiwan National Health Insurance Research Database Study, the SIR for lymphoma was higher in subjects receiving TNFis (6.13; 95% CI, 3.26–10.49) than in subjects receiving nbDMARDs (2.52; 95% CI, 1.56–3.85), which was significant154. Thus, there is no consistent evidence that TNFis increase the risk of lymphoma above the generally high background rate in the RA population.

Although hepatosplenic T-cell lymphoma was initially thought to be associated with TNFis in patients with inflammatory bowel disease, more recent evidence indicates an increased risk with TNFis when used in combination with a thiopurine (or with the thiopurine alone) but not with TNFi monotherapy170.

Hematologic malignancy – leukemia

The risk of leukemia was not found to be increased with biologic TNFis in an analysis of the NDB for Rheumatic Diseases (OR = 1.2; 95% CI, 0.5–3.1)143 and in a North American study (versus the general population; SIR = 1.3; 95% CI, 0.85–1.97)165.

Solid tumors

There does not appear to be an overall increased risk of solid tumors associated with TNFis compared with nbDMARDs. An analysis of the BSRBR for RA demonstrated no significant difference in proportion or RR for any of the common site-specific solid cancers in RA patients treated with TNFis compared with nbDMARDs148. Similarly, SABER reported no significant increase for any solid cancer associated with TNFis compared with nbDMARDs157, while the most recent analysis of the Corrona RA Registry found a decreased risk of solid cancer associated with TNFis compared with methotrexate (HR = 0.21; 95% CI, 0.07–0.64)160. Conversely, in the DANBIO registry, colon cancer was increased in patients treated with TNFis compared with nbDMARDs; however, the number of overall cases was low among the TNFi and nbDMARD groups140. Registry data have not indicated an increased incidence of lung cancer related to TNFi therapy above the already increased risk in the RA population143^,148; the risks of breast and lung cancers were reported to be decreased in patients treated with TNFis compared with methotrexate in the Corrona registry160.

Nonmelanoma skin cancer

Overall, there appears to be a potential increased risk of NMSC associated with TNFis compared with nbDMARDs in RA patients. Two meta-analyses of randomized clinical trials reported no increased risk of NMSC168^,169, and another reported a significant two-fold increase in the risk of NMSC associated with TNFis relative to comparator arms (HR = 2.02; 95% CI, 1.11–3.95)155. Registry data from the BSRBR and administrative data from the SABER study showed statistically nonsignificantly increased risks of NMSC with TNFis compared with nbDMARDs151^,157, and the Corrona registry showed a decreased risk of NMSC compared with methotrexate160. Conversely, a study from the NDB reported a statistically significantly increased risk of NMSC associated with TNFis, although this was relative to the general US population143. A pooled analysis of four registries reported a statistically significantly increased risk of NMSC (pooled estimate = 1.33; 95% CI, 1.06–1.60). This analysis included data from the three registries (Corrona registry, BSRBR, and Antirheumatic Therapies in Sweden study [ARTIS]) with results that were nonsignificant compared with RA patients treated with nbDMARDs and a previous result from the NDB, which was the only study with a significant result included in the analysis158^,171. A separate analysis from the NDB reported a trend toward an increased risk of NMSC associated with TNFis without concomitant methotrexate (HR = 1.24; 95% CI, 0.97–1.58) or when used in combination with methotrexate (HR = 1.97; 95% CI, 1.51–2.58)152. The potential association between TNFis and NMSC requires further observation.

Melanoma

There have been mixed findings with respect to the association of TNFis and melanoma, and the overall risk is difficult to determine. An analysis of 29,423 patients with RA in 63 randomized controlled trials found an OR for melanoma associated with TNFi agents of 1.08 (95% CI, 0.11–10.21)169. A separate meta-analysis reported a trend toward an increased risk of melanoma with TNFis based on the inclusion of two studies (pooled estimate = 1.79; 95% CI, 0.92–2.67)158. Infliximab and etanercept were both associated with a statistically nonsignificantly increased risk of melanoma in the NDB, with ORs of 2.6 (95% CI, 1.0–6.7) and 2.4 (95% CI, 1.0–5.8), respectively143. An analysis of the ARTIS registry found an increased risk of invasive melanoma compared with RA patients not treated with biologic drugs (HR = 1.5; 95% CI, 1.0–2.2) but no significantly increased risk of melanoma in situ (HR = 1.1; 95% CI, 0.5–2.1) associated with TNFis compared with other RA therapies172. Taken together, current evidence suggests a potential increased risk of melanoma overall, but the evidence is conflicting. However, although there does not appear to be an increased risk of in situ melanoma associated with TNFis, there may be increased risk of invasive melanoma.

Malignancy associated with viral infection

Patients treated with TNFis are at increased risk for certain viral infections such as herpes zoster173 and reactivation of hepatitis B174, raising the question of susceptibility to oncogenic viruses. Infection with certain viruses may lead to malignancy: EBV and lymphoma, HPV and cervical and anogenital cancers, polyomavirus and Merkel cell cancer, HHV-8 and Kaposi’s sarcoma. Certain types of HPV cause all cervical cancers; most anal cancers; and some vaginal, vulvar, penile, and oropharyngeal cancers175. However, there is no evidence for an increased risk of these viral infection–associated malignancies with TNFi. In one study, patients who received infliximab, adalimumab, or etanercept showed no difference in EBV viral load and no difference in numbers of EBV-specific IFN-γ–producing T cells (for both latent-cycle and lytic-cycle peptides) between baseline and treatment week 12111. Other clinical studies showed that TNFi therapy did not increase the EBV burden in patients with RA or ankylosing spondylitis112^,113. Lower rates of female genital cancer (including cervical) were reported in the BSRBR141, and there was no increase in female genital cancers in women with prior cervical carcinoma in situ treated with TNFis compared with those treated with nbDMARDs, although the number of patients was small176. Merkel cell carcinomas are rare, aggressive, cutaneous neuroendocrine tumors. Although rates of occurrence are not available, postmarketing case reports of Merkel cell carcinoma have occurred with TNFis, leading to mention of this development in the prescribing information for the majority of TNFis42^,43^,46.

Recurrent malignancy

The data are limited, and it is unknown whether there is an overall increased risk of malignancy recurrence with TNFis across all indications; current guidelines do not provide clear guidance regarding the use of TNFis in patients with recent malignancies. Current European League Against Rheumatism guidelines have no specific recommendations regarding the treatment of patients with prior malignancies, although rituximab (a non-TNFi biologic) may be preferred in this scenario177. The British Society of Rheumatology guidelines recommend caution in the use of TNFis in patients with previous malignancies178. The American College of Rheumatology recommends the use of any biologic agent in patients with a history of a treated solid malignancy or NMSC >5 years ago179; rituximab is recommended for patients with a more recent history (≤5 years) of a treated malignancy based on lack of evidence associating the use of this biologic DMARD with cancer179.

Safety registries provide insight regarding the risk of malignancy recurrence with TNFis. In the BSRBR, Dixon et al. reported an incidence rate ratio (IRR) of incident cancer in TNFi-treated patients with RA with prior malignancies of 0.45 (95% CI, 0.09–2.17)180. This study reported only incident malignancy occurring after the start of TNFi treatment (nbDMARD as control) and was not designed specifically to examine recurrence; in fact, many of the malignancies in those who reported prior malignancies could be classified as second primary tumors180. In the Rheumatoid Arthritis Observation of Biologic Therapy (RABBIT) study, malignancy recurrence rates were similar between patients with RA who received TNFis (45.5/1000 patient-years) and those who received nbDMARD therapy (31.4/1000 patient-years), with an IRR = 1.4 (95% CI, 0.5–5.5)164. Most (14 of 15) of the second tumors were true recurrences of the primary tumor; nine patients received TNFis and five received nbDMARDs164. The recurrent malignancies in patients treated with TNFis were breast (n = 4) and lung, bladder, liposarcoma, melanoma, and testicular cancers (n = 1 each)164. A pooled analysis of the RABBIT and BSRBR studies reported an estimated IRR of 0.62 (95% CI, 0.04–1.20)158. Data regarding cancer recurrence are also available from BIOBADASER 2.0, which compared patients with rheumatic diseases (including RA, juvenile idiopathic arthritis, ankylosing spondylitis, psoriatic arthritis) and psoriasis treated with TNFis to a cohort of patients not exposed to TNFis. In the TNFi-treated patients, 24 of 4694 were found to have a malignancy before TNFi exposure. The IRR for recurrent cancer in patients with prior malignancies was 5.22 (95% CI, 0.79–34.34), a nonsignificant trend that was complicated by small numbers147.

A recent analysis of the ARTIS registry evaluated cancer recurrence in patients with RA and at least one diagnosis of breast cancer before initiating TNFi treatment. Comparing TNFi-treated patients with biologics-naive patients and adjusting for nodal status, type of surgery, and chemotherapy at index cancer, the HR was 1.1 (95% CI, 0.4–2.8). The authors of this report indicated that these results support current clinical guidelines, which state that TNFis may be initiated in patients with a >5-year history of treated breast cancer181. Analyses of the BSRBR and ARTIS registries also specifically examined melanoma recurrence. In a small data set of patients with prior melanoma in the BSRBR, there were two definite (and one probable) melanoma recurrences in 17 patients exposed to TNFis and none in 10 patients not exposed to TNFis180. An analysis of the ARTIS registry reported a nonsignificant three-fold increase in the risk of a second primary melanoma associated with TNFis relative to nbDMARDs in patients with RA, although only a total of 13 cases were reported (3 TNFi and 10 nbDMARD)172.

Owing to the small number of cases reported, further analyses are needed to determine the recurrence risk, if any, of TNFi therapies in patients with prior malignancies.

Discussion and conclusions

Clinical and experimental animal data demonstrate that TNF can both promote and prevent tumor formation and that a complex, contextual integration of signals drives a cellular response toward apoptosis/necrosis or survival, inflammation, and growth promotion (Figure 1). Mechanistic data indicate that key players in TNF signaling that shape this response include NF-κB and Jun kinase, as well as other factors intrinsic to normal or transformed cells64^,182. In addition to direct effects of TNF on tumors, TNF can variously affect immunity and the tumor microenvironment. Whereas TNF can promote immune surveillance designed to eliminate tumors, it can also drive chronic inflammation, autoimmunity, angiogenesis, and other processes that promote tumor initiation, growth, and spread (Figure 2). The context-specific role of TNF in tumor regression or growth is evident in variable outcomes observed in rodent models (Table 2).

Human expression polymorphisms suggest that TNF is generally a tumor-promoting cytokine. Additionally, data from population studies, administrative databases, registries, meta-analyses, and observational studies demonstrate that overall rates of malignancy are not higher with TNFi therapy use among patients with RA, who have an elevated risk of certain malignancies compared with the general population. These data indicate that the risk of cancer associated with TNFi use is markedly different from the cancer risks experienced by individuals with primary (congenital) immune deficiencies, acquired deficiencies (e.g., HIV/AIDS), or those treated with therapeutics used to support organ engraftment, who have increased risks of both virally associated and nonviral cancers5–14.

Taken as a whole, the available data indicate that elevated TNF is a risk factor for cancer, whereas its inhibition among patients with RA is not generally associated with an increased cancer risk, particularly for the range of cancers linked with immune suppression. More information is required to fully understand the risk of TNFi use regarding risk of skin malignancies and risk of cancer recurrence. Although this review focused on RA, which has the most robust data, there are reports in other disease indications showing no increased rate in malignancy with TNFis in juvenile idiopathic arthritis183, psoriatic arthritis157, and Crohn’s disease or inflammatory bowel disease157^,184.

Because the immune system can both suppress and promote cancer, it is obligatory that an assessment of cancer risk among RA patients consider the risk of undertreating their inflammatory disease balanced against the use of immunosuppressive agents that impair immune surveillance. Such data, including outcomes with individual TNFis as presented here, are critical for informing patients and clinicians regarding the overall cancer risks they face given their disease and treatment options.

Developing an understanding of the cancer risk associated with TNFi use is confounded by a number of factors, including the nature of the underlying disease (and its severity), history of oncogenic viral infection, and the use of concomitant therapies1. Because more potent immunosuppressive agents are typically used in the treatment of more severe disease, dissociation of their effects as drivers of malignancy is especially challenging. As personalized medicine evolves and individual tumors are evaluated for markers, the role of TNF may be further understood.

Transparency

Declaration of funding

The preparation of the manuscript was funded by Amgen Inc.

Declaration of financial/other relationships

H.L., R.P., B.D.P., J.I., T.L.B., and M.H. have disclosed that they are employees of and own stock in Amgen Inc.

CMRO peer reviewers on this manuscript have received an honorarium from CMRO for their review work, but have no relevant financial or other relationships to disclose.

Acknowledgments

The authors acknowledge editorial assistance from Tim Peoples MA ELS CMPP of Amgen Inc., and Miranda Tradewell PhD of Complete Healthcare Communications, Inc., whose work was funded by Amgen Inc.