Introduction

The crucial role of tumor necrosis factor alpha (TNF-α) in the pathogenesis of several immune-mediated inflammatory diseases (IMIDs), primarily those affecting joints, skin, gut and eye, has been widely established. The first evidence of the role of TNF-α in inflammatory arthritis pathophysiology derived from the human TNF transgenic mouse model [1]. The favorable effect of blocking TNF-α was also demonstrated in these experiments for the first time. It is now clear that TNF-α constitutes a common effector module of the inflammatory cascade that occurs in several IMIDs [2]. Following impressive results of randomized clinical trials in the mid-1990s, and since their first approval in 1998, agents targeting TNF-α (anti-TNFs) have been proven effective for the treatment of various IMIDs. During the first decade of their use, concerns for anti-TNF-α safety, that may arise from the inhibition of prophylactic actions of TNF-α in infection and cancer, were alleviated. Millions of IMIDs patients have been safely treated, while clinical experience regarding optimization of their use accumulated [3].

In the past 10 years, anti-TNFs indications have expanded to include non-radiographic axial spondyloarthritis (AxSpA), paediatric enthesitis-related arthritis, paediatric ulcerative colitis (UC), hidradenitis suppurativa (including adolescents) and non-infectious uveitis (including children) (Table 1). Moreover, the therapeutic potential of anti-TNFs has been also demonstrated in several disorders, such as Behcet’s disease (BD), sarcoidosis, Takayasu arteritis, idiopathic inflammatory myopathies, synovitis-acne-pustulosis-hyperostosis-osteitis (SAPHO) syndrome, polyarteritis nodosa, adenosine deaminase 2 deficiency (DADA2) and pyoderma gangrenosum [4,5,6,7,8,9,10,11].

Herein, we aimed to summarize new lessons learned during the second decade of anti-TNFs utilization in clinical practice. Besides a better understanding of the mechanisms of action of TNF-α inhibition that brought synovial fibroblast and gut myofibroblasts on stage, the long-term effects of anti-TNF-α treatment taught us that the natural history of chronic inflammatory diseases may be indeed altered [12,13,14]. Long-term effects of anti-TNFs on bone and gut pathophysiology, as well as their impact on systemic manifestations of chronic inflammation, such as atherosclerosis and cardiovascular (CV) events, have become evident and led to improved mortality rates [15].

Accumulating “real-world” data on additional millions of patients with various background conditions further established anti-TNFs’ excellent safety profile, which was extended to include pregnant women. Anti-TNFs safety regarding CV disease and cancer was further confirmed in patients with RA older than 50 years in comparison to newer drugs, such as Janus kinase (JAK) inhibitors [16]. The current COVID-19 pandemic revealed also unexpected beneficial effects of anti-TNF-α agents [17]; anti-TNF-α use protects against severe COVID-19 infection and hospitalization due to COVID-19 and is being examined for the treatment of COVID-19 [18,19,20]. Moreover, we have learned to better use TNF-α inhibition in order to apply treat-to-target strategy in IMIDs [21], including anti-TNF-α intensification, combination with immunomodulators, tapering, discontinuation, cycling, and/or switching to other biologics. While we have learned a lot on therapeutic drug monitoring (TDM), robust biomarkers of response remain to be established. The introduction of biosimilars had major socioeconomic impact, giving the opportunity to increased use of anti-TNF-α therapies worldwide. Finally, we briefly describe the future of anti-TNF-α therapy, including the next generation of anti-TNF-α drugs, as well as the possible combination of anti-TNFs with other biologic and targeted therapies.

For this narrative review, an exhaustive literature search was conducted by all authors independently, using PubMed via MEDLINE, Scopus and EMBASE databases. The included publication dates focused from 2010 to March 2022 in order to capture the advances during the second decade of anti-TNF-α use and COVID-19 pandemic, but earlier publications were also taken into consideration. The main terms used for database search were “TNF-α”, “TNF alpha”, “TNF receptor”, “anti-TNF-α”, “TNF inhibitor”, “COVID-19”, “safety”, but literature search was not only confined to these. This review was designed taking into account previously published considerations on narrative review writing [22].

Mechanisms of action of anti-TNFs in target organs: the central role of mesenchymal cells

Early studies have established that anti-TNFs, namely the anti-TNF-α monoclonal antibodies (mAbs) infliximab, adalimumab, certolizumab and golimumab, and the soluble TNF-α receptor etanercept (Fig. 1), work by neutralizing the activity of soluble TNF-α and preventing its binding to TNFα receptor 1 (TNFR1) and TNFR2. TNFR1 is constitutively expressed on the membrane of almost all cell types and is associated with proinflammatory and cytotoxic responses [23]. On the other hand, TNFR2 is expressed on immune cells, mainly monocytes and T cells, as well as on activated mesenchymal cells, mediating cellular activation, proliferation, migration and other pathogenic functions on synovial fibroblasts [23, 24].

Fig. 1
figure 1

Molecular structure of approved anti-TNFs. The five approved anti-TNFs, presented in chronological order of first approval. Etanercept is a fusion protein of extracellular domain (p75) of human TNFR2 and Fc fragment of IgG1; infliximab is a mouse/human chimeric monoclonal IgG1 anti-TNF-α antibody; adalimumab is a humanized IgG1 monoclonal anti-TNF-α antibody; certolizumab is a Fc-free Fab region of a recombinant humanized IgG1 monoclonal anti-TNF-α antibody, conjugated to PEG; golimumab is a human IgG1 monoclonal anti-TNF-α antibody. Anti-TNF-a: tumor necrosis factor alpha inhibitor, PEG: polyethylene glycol, TNFR2: tumor necrosis factor receptor 2

Full size image

Different TNF-α-mediated (auto)immune processes are involved in the initiation and perpetuation of IMIDs. The differential efficacy of anti-TNFs seen in clinical practice suggests that the mechanisms of therapeutic action are not distinct and may overlap. Briefly, anti-TNF-α treatment acts in the immune system as follows: (a) downregulating pro-inflammatory cytokines, chemokines, acute phase proteins and adhesion molecules expression, (b) increasing circulating regulatory T cells and (c) reducing the migration of inflammatory cells from blood to the inflamed tissue. Etanercept also neutralizes lymphotoxin, whereas anti-TNFs have cell-killing properties by directly binding to transmembrane TNF-α expressed in various cells. As Wu et al. recently showed, binding of anti-TNF-α to transmembrane TNF-α, affects intracellular signaling and, in addition to programmed cell death, may result either to suppression of cytokine production or cell growth arrest [25].

In addition to the effects on immune cells, anti-TNF-α mechanisms of action also include attenuation of vascular permeability and angiogenesis, as well as deactivation of epithelial, endothelial and mesenchymal cells. Recent advances in basic research have indeed shown that anti-TNF-α treatment works by interfering with TNF-α signaling in mesenchymal cells [26]. For example, mesenchymal-specific TNFR1 triggering is indispensable for arthritis development in acute and chronic TNF-dependent mouse models. While inhibitor kappa B kinase 2 (IKK2) in joint mesenchymal cells is necessary for cartilage destruction and bone erosion, in its absence synovitis still develops, as a result of local immunogenic synovial fibroblasts necroptosis. IKK2 deletion affects arthritic and anti-apoptotic gene expression leading to hypersensitization of synovial fibroblasts to TNF/receptor interacting serine/threonine kinase 1 (Ripk1)-mediated necroptosis via district mechanisms, depending on acute or chronic TNF-α signals. Moreover, Ripk3 is dispensable for TNF-mediated arthritis, yet it is required for synovitis in mice with mesenchymal-specific IKK2 deletion, clearly showing that TNFR1-IKK2-Ripk-mixed-lineage kinase domain-like (MLKL) signalling pathway orchestrates arthritogenic and death responses in synovial fibroblasts and that combinatorial inhibition of nuclear factor kappa beta (NF-κB) and MLKL/RIPKs may offer a therapeutic potential [26].

As has been demonstrated in TNF transgenic mice models, synovial fibroblasts and intestine myofibroblasts are activated early by TNF-TNFR1 signaling, produce several matrix degrading enzymes and are sufficient targets to induce TNF-driven inflammatory polyarthritis, Crohn's-like inflammatory bowel disease (IBD) and sacroiliitis [27]. Thus, mesenchymal cells are necessary targets of TNF-α in the development of spondyloarthritis-related disorders [27]. Importantly, mesenchymal cell activation by TNF-α has been shown to participate also in additional sites, such as in cardiac valvular cell activation and valve thickening [28]. This possibly explains several comorbidities (e.g. cardiac and lung pathologies) that may be associated with mesenchymal cell activation, interstitial inflammation and damage in TNF-α-driven rheumatic diseases.

Along these lines, the mesenchymal hypothesis to explain the pathogenesis of RA is gaining momentum and several elegant studies analyzing human RA synovial tissue at the single-cell level have produced exciting results [29]. Such studies indicated that synovial lining fibroblasts (Thy1-) were found to be predominantly responsible for driving articular damage, whereas sub-lining layer fibroblasts (Thy1 +) are mainly pro-inflammatory [30]. More recent evidence revealed an endothelial-cell-instigated Notch-mediated pathway in perivascular sub-lining synovial fibroblasts, that establishes a positional gradient for sub-lining synovial fibroblasts differentiating to lining synovial fibroblasts [31]. So far, studies in humans have not provided robust evidence on how synovial fibroblasts transition from homeostasis to pathology. Preliminary single-cell RNA and ATAC-seq chromatin profiling of synovial fibroblasts from the TNF transgenic mouse model indicated that development of TNF-α-driven arthritis primes the emergence of distinct pathology-associated synovial fibroblasts subtypes and revealed key transcription factors such as Bach1 and Runx1 to drive arthritogenesis [32]. Taken together, these recent studies provided deeper mechanistic understanding of the dynamics of synovial fibroblast subpopulations during pathogenesis of arthritis, paving the way to the discovery of mesenchymal cell-targeted therapeutics beyond anti-TNFs.

Effects of long-term TNF-α inhibition in IMIDs

Long-term effects on bone pathophysiology in inflammatory arthritides

Along with interleukin-6 (IL-6), TNF-α stimulates synovial fibroblasts to produce receptor activator of NF-κB ligand (RANKL) promoting osteoclasts’ differentiation and activation [33]. Moreover, TNF-α leads to increased synovial levels of Dickkopf-related protein 1 (DKK-1), an inhibitor of osteoblastogenesis [34]. As a result, TNF-α orchestrates a process of local (expressed as bone erosions and juxta-articular osteoporosis) and systemic bone loss in patients with inflammatory arthritis [12, 35, 36]. Indeed, patients with rheumatoid arthritis (RA), psoriatic arthritis (PsA) and AxSpA exhibit increased rates of systemic osteoporosis and osteoporotic fractures. On the other hand, bone loss in PsA and AxSpA is accompanied by pathological bone formation, in which TNF-α contributes in a lesser degree than other cytokines, such as IL-17 and IL-23 [37].

Blockade of TNF-α has been proven beneficial on bone metabolism in inflammatory arthritis (Fig. 2) [38]. Anti-TNF-α therapy controls synovial inflammation and leads to significant reduction of joint erosions in RA and PsA [12, 13]. Importantly, peripheral arthritis remains radiologically stable in AxSpA patients under infliximab, especially in the hip [39], resulting also in reduced hip replacement surgery rate [40, 41]. Through ameliorating inflammatory joint destruction, anti-TNF-α therapy led to 40% reduction of joint replacement surgery in older RA patients [42], while the benefit was stronger when treatment was introduced early in the disease course [43].

Fig. 2
figure 2

Long-term beneficial effects of anti-TNFs. Beneficial effects of anti-TNFs in bones, joints, intestine and cardiovascular system. IBD: Inflammatory bowel disease

Full size image

Although initially thought that syndesmophytes formation was not hindered, but even accelerated, by anti-TNFs [44], newer long-term data have questioned this hypothesis [45]. The radiographic progression in axial disease is decelerated in patients that anti-TNF-α treatment sufficiently suppresses inflammation markers [46]. Systemic bone loss is also ameliorated by anti-TNFs in RA and AxSpA patients [12, 38, 47, 48]. Although an increase in bone mass and improvement in bone turnover markers is reported in anti-TNF-α -treated patients, less data are available for the possible impact on fracture occurrence [38, 48, 49].

Long-term effects on intestinal inflammation

The natural history of IBD is characterized by cumulative damage to the intestinal wall, which may lead to permanent injury and structural complications, such as strictures and fistulae in the case of Crohn’s disease (CD) and “lead pipe-like” dysfunctional colon in UC. Such abnormalities often necessitate surgical interventions and serve as objective indications of disease severity. There is now ample evidence that anti-TNF-α therapy has substantially decreased the incidence of complication-related hospitalizations by 50% and surgeries by 33–77% in patients with IBD, compared to placebo and the period before anti-TNF-α introduction [14, 50, 51]. In paediatric IBD, it was shown that adherence to recent guidelines with increased use of anti-TNFs resulted in significantly higher cumulative probability of a relapse-free and surgery-free course [52]. Taken together, those studies support an important role for TNF-α as a mediator of intestinal inflammation in IBD and emphasize the significance of its inhibition for the prevention of irreversible bowel damage, as well as the feasibility of this approach.

It should also be noted that, despite the expansion of available therapeutic options, anti-TNF-α agents remain the only agents that are indicated for the full spectrum of local and systemic manifestations in patients with IBD. Such indications include perianal fistulizing disease, acute severe UC and post-operative prevention of CD recurrence [53, 54]. More importantly, joint, skin and/or eye inflammation co-exists in at least one-third of patients with IBD and anti-TNFs represent the only biological therapy that is suitable for the treatment of such extra-intestinal manifestations [55].

Impact on atherosclerosis and cardiovascular events

It is known that IMIDs patients carry an increased risk for atherosclerotic CV events. Chronic systemic inflammation leads in acceleration of subclinical atherosclerosis [56]. In RA and AxSpA patients, low disease activity or remission induced by biologic disease modifying anti-rheumatic drugs (bDMARDs), mainly anti-TNFs, resulted in stabilization of subclinical atherosclerosis [57, 58], while a meta-analysis shows amelioration of atherosclerosis and arterial stiffness in anti-TNF-α-treated patients with inflammatory arthritis (RA, PsA, AxSpA) [59]. Anti-TNF-α therapy improves microvascular function in AxSpA patients [60] and subclinical myocardial inflammation in patients with inflammatory arthritis [61]. Additionally, myocardial infarction rates are reduced in RA patients under anti-TNF-α treatment, compared to those receiving conventional synthetic DMARDs (csMARDs) [62]. Last, anti-TNF’s beneficial effect seems to further extend to improvement of insulin resistance, functional ability and fatigue [63, 64].

Impact on mortality

The effect of anti-TNFs gets beyond the control of disease activity. In general, mortality of patients with IMIDs has decreased over the past two decades. Mortality rate of RA patients declined significantly only a few years after the introduction of anti-TNFs in clinical practice [65], while recent data suggest that patients with RA, PsA and AxSpA exhibit comparable mortality rate with the general population the past 5 years before the COVID-19 pandemic [15]. Regarding RA, lower number of deaths was observed in adalimumab-treated patients compared to the expected age- and sex-matched individuals from the general population [66]. The reduction in mortality of patients with IMIDs can be mainly attributed to diminished CV mortality, due to amelioration of subclinical atherosclerosis and atherosclerotic CV events [65].

Safety signals with chronic TNF-α inhibition

During the past decade, no new issues have arisen from long-term extension studies and real-life cohorts regarding anti-TNFs safety [66]. On the contrary, several points have been clarified. The hypothesis that anti-TNFs might make patients susceptible to lymphoma development has been rejected [67,68,69,70,71]. As shown from several meta-analyses, anti-TNF-α use is possibly not associated with an increased long-term malignancy risk in patients with inflammatory arthritis and IBD [71,72,73,74]. In addition, anti-TNFs exhibit a lower hazard ratio for cancer development compared to tofacitinib in RA patients aged over 50 years [16]. As for non-melanoma skin cancer (NMSC), some studies have shown that its incidence is increased in anti-TNF-α -treated patients [68, 75]. However, data from two large registries did not confirm these findings; in contrast, they showed that patients with RA, regardless of anti-TNF-α use, are in increased risk for NMSC development, compared to the general population [73, 76]. In RA patients, data for melanoma risk in anti-TNF-α -treated patients are conflicting, but the risk is possibly comparable to patients not exposed to anti-TNFs [68, 77]. Importantly, anti-TNF-α use in patients with malignancy history has not been associated with an increased risk for malignancy recurrence or new primary cancer development [78]. Moreover, the risk is not influenced by the time interval between initial cancer diagnosis and anti-TNF-α initiation [78]. There is currently no specific recommendation about when to start an anti-TNF-α in a patient with malignancy history. However, treatment with anti-TNFs should be avoided in patients with active malignancy.

As for active tuberculosis incidence, the introduction of screening methods prior to anti-TNF-α initiation led to a significant decline in (de novo or reactivation) tuberculosis cases in these patients [79]. Regarding other chronic infections, anti-TNF-α use is not related with an increased exacerbation rate of human immunodeficiency virus (HIV) infection [80, 81], while chronic hepatitis B and C reactivation rates remain low during anti-TNF-α treatment [82, 83]. In addition, skin psoriasis, an anti-TNF-α-related paradoxical adverse reaction [84], was proven to be induced also by other biologic agents used in IMIDs [85]. Moreover, although elevation of serum uric acid levels after 3 months of anti-TNF-α therapy have been observed, its clinical significance remains to be elucidated [86].

Based on the role of RANKL in the immune system, safety concerns had been raised for the combination of anti-TNFs with denosumab. Accumulating data demonstrated that the co-administration of these agents does not increase the risk for infection, compared to monotherapy with anti-TNF-α [87] and to those receiving zoledronic acid instead [88].

A growing body of evidence is now available about different anti-TNFs use during pregnancy and lactation. Anti-TNF-α use during pregnancy in women with IMIDs has not been associated with significantly increased rates of congenital malformations [89]. According to the IBD Parenthood Project, ideal scheduling of the final dose of an anti-TNF-α biologic prior to delivery is dictated by the half-life of the drug [90]. In “EULAR points to consider for use of anti-rheumatic drugs before pregnancy and during pregnancy and lactation” a slightly different perspective for women with inflammatory arthritis is suggested [91]. To be noted, as certolizumab pegol does not cross the placenta, it may be continued throughout pregnancy [91, 92]. Anti-TNF-α agents may be resumed 24 h after vaginal delivery and 48 h after cesarean delivery if there are no pregnancy-related or infectious contra-indications. Importantly, all anti-TNFs can be continued during lactation [91] and paternal exposure to anti-TNFs during conception period has not been associated with adverse pregnancy outcomes in their partners [93].

Last, reassuring “real-life” data about demyelinating disorders development after anti-TNF-α initiation have recently been seen the light [94]. It is still debatable if some sporadic cases are a coincidence or anti-TNF-α adverse events [95]; anti-TNFs’ use is either way contraindicated in patients with known concomitant demyelinating disorders.

Anti-TNFs in the COVID-19 era

The safety profile of anti-TNF-α emerged as an important issue for patients with IMIDs during the COVID-19 pandemic. Despite initial concerns due to fear of immunosuppression, only a small proportion of patients discontinued anti-TNF-α during the COVID-19 pandemic [96]. Experimental data suggest that anti-TNF-α-treated patients might be susceptible to be infected by SARS-CoV-2 [97]. Indeed, a threefold increased risk for COVID-19 infection was noted in patients with rheumatic diseases under anti-TNFs in a recent nationwide survey [18]. However, it seems that these patients do not carry an increased risk for serious COVID-19 disease [17, 18]. On the contrary, anti-TNF-α use was associated with reduced risk for hospitalization in SARS-CoV-2-infected patients with rheumatic diseases [98, 99], psoriasis [100] and IBD [101].

COVID-19 is an infection with a clear prothrombotic component [102,103,104]. TNF-α is elevated in patients with severe COVID-19, contributes in COVID-19 cytokine storm and acts as a major mediator in inflammation‐driven capillary leak syndrome that leads to lung injury [20, 104]. On this basis, the therapeutic potential of anti-TNF-α in COVID-19 infection emerged [20]. Ongoing trials examine the use of infliximab in serious COVID-19 infection, showing promising results, and etanercept has shown some effectiveness in case reports [105]. Interestingly, a trial with adalimumab was terminated early [106].

Importantly, anti-SARS-CoV-2 antibody development after vaccination are not hindered by anti-TNF-α use [107], as has been reported with rituximab, in patients with rheumatic diseases [108, 109]. Anti-SARS-CoV-2 antibody levels are comparable between PsA and AxSpA patients under anti-TNFs, anti-IL-17 and methotrexate [110, 111]. However, the sustainability of humoral response to COVID-19 vaccination in patients under anti-TNF-a agents is debatable, especially in IBD [112,113,114]. Thus, an earlier booster regimen might be crucial for these patients [112,113,114].

Optimization of anti-TNF-α use in IMIDs

Anti-TNFs as the basis for treat-to-target strategy

Implementation of bDMARDs in the clinical practice inevitably altered the therapeutic strategy in patients with IMIDs. Being the first bDMARD class applied, anti-TNFs made remission or low disease activity a feasible target. As a result, the “treat-to-target” strategy has been adopted in the management recommendations of several IMIDs [21]. This concept consists of the following: (a) baseline stratification of patients according to their individual risk for a complicated disease course; (b) pre-defined therapeutic targets that are predictive of a favorable long-term outcome of the disease; (c) modifications of the administered therapy until the target is met and, (d) periodical reassessment of the patient to ensure long-term maintenance of therapeutic target accomplishment. Physicians should maximize the possible control of disease activity to achieve remission or low disease activity as soon as possible. In patients with early CD, finetuning of anti-TNF-α therapy based on combinatorial clinical and biomarker endpoints resulted in better clinical and endoscopic outcomes than symptom-driven decisions alone [115]. A similar strategy has shown to improve disease outcomes compared to conventional follow-up in patients with RA [116,117,118] and PsA [119], while its value is still under debate in AxSpA [120]. For RA, physicians should also keep in mind that an anti-TNF-α without Fc portion of immunoglobulin (certolizumab pegol) might be more efficacious that other anti-TNF-α in patients with high rheumatoid factor levels, as was recently shown [121]. Moreover, in line with the “treat-to-target” strategy, patients with persistent monoarthritis, usually of the knee, might benefit from intra-articular anti-TNF-α infusion and etanercept is the agent with the better outcomes reported [122].

Anti-TNF-α cycling or switching to other biologics?

As said before, anti-TNFs are still considered the first-line bDMARDs used in several IMIDs, due to the established efficacy and safety profile. Nevertheless, a varying proportion of patients treated with a first anti-TNF-α agent will discontinue this therapy, mainly due to loss of response, rather than adverse effects [123, 124]. After failure of a first anti-TNF-α in a patient with RA, either a second anti-TNF-α or a drug with another mechanism of action can be chosen [21, 125]. However, in cases of primary failure, a switch to a non-anti-TNF-α bDMARD might be more efficacious than swapping to another anti-TNF-α [126], especially for seropositive RA patients (in favor of rituximab) and in those under bDMARD monotherapy (in favor of tocilizumab) [127]. Moreover, switching from monoclonal anti-TNFs (adalimumab or infliximab) to soluble anti-TNF-α receptor (etanercept) is associated with better clinical outcomes than the opposite [128]. Regarding AxSpA, recent studies imply that both a second anti-TNF-α and anti-IL-17 inhibitors are considerable options after failure of a first anti-TNF-α agent [129, 130]. For PsA, according to the last ACR guidelines, anti-TNF-α cycling is preferred over anti-IL-17 and anti-IL-12/23 inhibitor use after anti-TNF-α failure, except of the case of severe psoriasis or primary anti-TNF-α non-response [131]. More data are needed in psoriasis and IBD to guide the therapeutic strategy after anti-TNF-α failure [132, 133]. Eventually, the decision of second-line bDMARD treatment relies on the cause of anti-TNF-α non-response and should be individualized [134].

Therapeutic drug monitoring (TDM): Anti-drug antibodies & drug levels

In the era of “treat-to-target” approach and precision medicine, TDM, which consists of measuring drug levels and anti-drug antibodies (ADAs), is increasingly gaining attention [135]. TDM has been, almost exclusively, investigated for anti-TNF-α therapy, and applied in IBD to a much greater extent than in other IMIDs. A constantly expanding body of evidence indicates that TDM may be a valuable clinical tool for improving disease outcomes and can be applied in either reactive or pre-active fashion [136]. Reactive TDM calls for measuring trough levels and ADAs when a clinical endpoint is not met (“treat-to-target” approach). Pro-active TDM constitutes of dosage modifications until pre-determined serum drug levels are achieved. In case of treatment failure, either primary non-response or secondary failure to anti-TNFs, TDM may dictate the best therapeutic strategy, either by modifying the dosing scheme or switching out of anti-TNF-α therapy [135, 137].

Failure to respond to a biologic agent or loss of response after initial benefit are multifactorial phenomena. During induction, serum drug levels are affected by body mass index, gender, serum albumin, inflammatory burden, prior failure to bDMARDs, as well as tissue specific processes, such as loss in the stools, digestion by local proteases and recycling by neonatal Fc receptor. On the other hand, secondary loss of response may be mediated by either pharmacodynamic (mechanistic) or pharmacokinetic (immune or non-immune) factors.

Among the most recognized causes of secondary failure to anti-TNF-α therapy is the formation of ADAs [137]. This is influenced by several factors. Firstly, not all anti-TNFs have the same risk for ADA development; etanercept, compared to monoclonal anti-TNFs, exhibits low immunogenic potential [137]. Moreover, ADAs against mAbs (such as infliximab and adalimumab) have been associated with diminished drug effectiveness, whereas ADAs to etanercept are usually non-neutralizing [138]. Second concomitant therapy might alter this risk; methotrexate or thiopurine co-administration has been associated with decreased anti-TNF-α ADAs formation [139]. Additionally, the use of distinct antibiotics classes may also affect the risk of ADA formation in patients with IBD [140]. Importantly, ADAs against originator anti-TNF-α will cross-react with the relative biosimilar leading to treatment failure [137]. Although the precise clinical utility of TDM is still developing, patients who have higher serum drug levels always perform better in relation to a variety of disease outcomes, such as remission, drug survival and endoscopic scores [135].

Is anti-TNF tapering or discontinuation feasible?

In the bDMARDs era, adequate control of the disease activity has turned long-term remission in a realistic goal. As more patients manage to achieve sustained remission, a growing body of evidence suggests that anti-TNF-α treatment could be tapered or even discontinued in this group of patients (Table 2).

In the latest EULAR and ACR guidelines for RA management, bDMARD tapering is considered a feasible practice in patients with sustained remission [21, 141]. This principle derives mainly from studies with anti-TNFs. Several studies after 2010 examined this issue. Tapering of anti-TNF-α or csDMARDs leads in similar flare rates after 9 months, but data from 12 months shows numerical less disease relapses in patients who continued anti-TNF-α treatment [142]. Indeed, in RA patients receiving etanercept and methotrexate combination, withdrawal of methotrexate led to less RA flares than etanercept withdrawal [143]. Thus, tapering of csDMARD might be a preferable option compared to anti-TNF-α discontinuation. Half patients with RA under anti-TNF-α can successfully discontinue treatment after achieving remission [144, 145], while this is less feasible for patients with established RA [145]. Long-term drug-free remission can be currently achieved in only a minority of RA patients and in about 30% a reduced anti-TNF-α dose can be maintained [146]. However, anti-TNF-α re-initiation restores remission in most cases [147]. It seems that early initiation of anti-TNF-α along with methotrexate and tight disease control provides the greatest potential of treatment holiday [148]. Moreover, sustaining deep remission was an independent predictor of better outcomes after adalimumab discontinuation [147].

Data in AxSpA patients support that anti-TNF-α withdrawal carries a high risk for disease flare (76–100%) in a median time of 4 months after drug discontinuation [149]. In contrast, flare-free tapering can be successful in about 53–100% of AxSpA patients receiving etanercept, adalimumab or infliximab [149]. In addition, a recent study demonstrated that AxSpA patients treated only for 12 months with certolizumab pegol showed comparable rates of flare-free remission when continuing on full-dose (200 mg per 2 weeks) or on reduced-dose (200 mg per 4 weeks) maintenance treatment thereafter [150]. Despite fewer data are available in PsA patients, short-term studies show a trend toward dose tapering and against treatment discontinuation in patients with well-controlled disease [151], although 55% of PsA patients might remain on remission after anti-TNF-α discontinuation [152]. Lower disease activity is a predictor of remission after tapering in PsA patients, while concomitant csDMARDs treatment does not have an impact on flare risk [151]. Importantly, drug re-institution effectively controls disease activity [151]. Anti-TNF-α discontinuation in children with juvenile idiopathic arthritis (JIA) led to high disease relapse rates (80–88%) [153], while gradual dose-lowering of etanercept in these patients is safe in maintaining remission [154].

Regarding IBD, 1 year after anti-TNF-α discontinuation 42% of CD patients and 28% of UC patients experience a disease flare [155]. In line with the aforementioned data, only 54% from UC patients who discontinued infliximab were in remission one year after drug withdrawal [156]. In another study, 0%–54% of IBD patients who tapered anti-TNF-α dose experienced a disease flare [157]. As in other diseases, remission was induced again in 80% of IBD patients that restarted the anti-TNF-α agent [155]. It should be noted, however, that in most IBD studies, patients who discontinued anti-TNFs maintained treatment with thiopurines [158]. On the other hand, only scarce data are available for psoriasis [159], but 50% of patients will still be in remission 6 months after anti-TNF-α withdrawal, especially those that had a complete response after drug initiation [160]. Adalimumab tapering is associated with 60% complete skin clearance within a follow-up of at least 4 years [161].

Discontinuation of anti-TNFs has been also reported in patients with BD. In a study from our group, about 40% of patients with severe BD in remission for at least 3 years might achieve long-term relapse-free remission of the disease after anti-TNF-α discontinuation [162]. Importantly, anti-TNF-α re-treatment was effective and safe in 82% of the patients that relapsed after treatment discontinuation [162].

Taken together, in patients with sustained remission available data strongly encourage dose reduction or increase of dosing intervals over drug discontinuation, as the latter carries a substantial risk of loss of remission. However, in those whom anti-TNF-α was suspended, drug re-administration is usually effective in restoring low disease activity.

Biomarkers of response

Given anti-TNF-α cost and their possible immunosuppressive effect, biomarkers capable of predicting patient’s response would guide physician’s therapeutic decisions for patients with IMIDs. In this line, many studies have examined several candidate molecules in a variety of clinical scenarios. Baseline higher serum calprotectin levels have been found in patients with RA that respond to etanercept or adalimumab [163]. Tao et al. showed that transcription signatures from peripheral blood mononuclear cells (PBMCs), monocytes and CD4 + T cells, of RA patients were different between adalimumab and etanercept responders and non-responders [164]. Another group applied multi-omics analyses in PBMCs from RA patients at the start of anti-TNF-α treatment and showed high expression of epiplakin 1 (EPPK1) and T-cell inhibitor chitinase-3-like 1 (CHI3L1) genes in those who responded [165]. Interestingly, a meta-analysis regarding patients with CD showed a discrepancy between proposed genomic and transcription biomarkers [166]. Treatment with anti-TNFs affects the levels of circulating microRNAs in RA patients [167]. Some studies have shown that serum levels of miR-22, miR-23, miR-27a, miR-223 and miR-886 before anti-TNF-α initiation could serve as predictors of patient’s response [167].

In IBD, search for predictive biomarkers is facilitated by the ability to obtain mucosal biopsies during endoscopy. This allows for the study of pre-treatment expression of local immunological and inflammatory mediators at the mRNA and protein levels and correlate them with subsequent response to treatment [168]. In one such study, non-response to anti-TNFs was associated with increased numbers of mucosal plasma cells and macrophages and elevated local expression of triggering receptor expressed on myeloid cells-1 (TREM-1), chemokine receptor type 2 (CCR2) and chemokine ligand 7 (CCL7) [169]. Another study with UC patients reported that the pretreatment mucosal expression of a combination of five genes predicted response to infliximab therapy [170]. In patients with colonic CD, another five-gene signature was found to discriminate between infliximab responders and non-responders [171]. Recently, the concept of molecular resistance via immunological pressure was proposed as an explanation for secondary loss of response to anti-TNFs. A significant upregulation of mucosal IL23p19, IL23R and IL17A expression was detected in patients who did not respond to anti-TNF-α therapy, in association with elevated number of mucosal T cells which expressed both TNFR2 and IL-23R and demonstrated high expression of inflammatory genes and resistance to apoptosis [172].

Recent studies have also shown that response to anti-TNFs may be genetically determined. The most important associations have been reported for patients with IBD who bear a specific polymorphism in the gene encoding the Fc fragment of the low affinity IgG3a receptor, who showed increased ADA formation, enhanced infliximab elimination and elevated risk of relapse after infliximab discontinuation [173]. In addition, in patients with CD, ADA formation may be determined by the HLA class II gene variant HLA-DQA501 [174].

The clinical applications of such findings are obvious as much they may identify patients who will benefit from combination of anti-TNFs with immunomodulators or who should better receive non-anti-TNF-α therapies. It should be noted, however, that the majority of these reports remains to be validated in additional patient cohorts and in prospective manner, which explains why none has reached clinical practice yet.

Socioeconomic perspectives on anti-TNF-α therapy: the case for biosimilars

Biosimilar drugs are agents with highly similar chemical characteristics and biological activity to original bDMARDs. Due to the financial success of originator anti-TNFs, especially adalimumab, infliximab and etanercept, several industries have developed biosimilars of anti-TNFs during the past decade, according to strict requirements [175,176,177]. Several biosimilars have shown clinically similar effectiveness, safety and immunogenicity compared to originator drugs and have been approved for the treatment of IMIDs [177]. As expected, healthcare systems exhibit substantial cost savings from the use of biosimilars not only due to their reduced price, but also by the consequent fall in the originator anti-TNFs cost [175]. Moreover, the use of anti-TNF-α biosimilars is expected to broaden anti-TNF-α availability, as their lower cost will mitigate the inequity in drug accessibility that has been reported [177]. Advances in bDMARDs manufacturing, especially in molecular engineering, allow changes in protein formulation that might improve originator drug’s properties, introducing the term “biobetter” [175, 176]. An example of such innovative strategy is CT-P13, an infliximab biosimilar, which was the first infliximab to be effectively administered subcutaneously [175].

On the other hand, biosimilars introduction in clinical practice was accompanied by several issues. First, biosimilars’ availability differs between countries, especially due to different regulation policies [176]. Importantly, switching from a reference bDMARD to a biosimilar can lead in subjective increase of disease activity or pain in some patients [177], in contrast to the results from double-blind switch randomized clinical trials [178]. This can be mainly attributed to the nocebo effect, which is based on a pre-existing negative perception of the patient for the administered drug [178]. Patients with IMIDs are more prone to nocebo behavior and data from biosimilar studies indicate that discontinuation rate due to subjective complains is not negligible [179]. As nocebo behavior can lead in several consequences, such as polypharmacy and worse disease outcomes, physicians need to recognize patients with risk factors for nocebo behavior and cautiously communicate biosimilars efficacy and safety, in order to prevent nocebo phenomena [179]. This has become more important in the COVID-19 era, during which nocebo behavior may be intensified [180].

The future of anti-TNF-α therapy

The next generation of anti-TNF-α drugs

Although the efficacy of anti-TNFs has been established for several IMIDs, a proportion of patients do not respond adequately and this might be partially attributed to pharmacokinetics. Typically, anti-TNF-α are administered subcutaneously or intravenously. However, the doses needed to achieve sufficient drug concentrations locally can lead to increased immunosuppression and other adverse events. To overcome this issue, new forms of anti-TNFs are under development.

Korkmaz et al. delivered anti-TNF-α-enriched microneedles on psoriasiform skin lesions in mice resulting in reduced epidermal thickness and inflammation mediators’ levels [181]. Moreover, a bispecific antibody combining a single-chain variable fragment (scFv) specific for synovium and a scFv of adalimumab successfully delivered adalimumab to the inflamed synovium in a mouse model, ameliorating local inflammation and indicating a new targeted therapy potential for RA [182]. On the other hand, oral anti-TNF-α agents could be suitable for diseases confined to gastrointestinal tract, such as IBD [183, 184]. However, oral administration of antibodies is hindered mainly by the proteolytic enzymes and the acidic environment of the stomach; thus, alternative delivery methods have been investigated for anti-TNFs [183]. Rectal administration (infliximab), tablet coating (V565-adalimumab), anti-TNF-α loaded nanoparticles [185,186,187], avian polyclonal anti-TNF-α, antisense oligonucleotides in several formulations, intracolonic microRNA and per os- or rectal-administered small interfering RNA that silence TNF-α expression, have been studied with varying efficacy [183]. Lactic acid bacteria have been utilized as mucosal delivering system and anti-TNF-α scFv-producing Lactococcus lactis administration have been shown to improve histopathologic findings in colitis models [188].

Progranulin, an endogenous protein that binds both types of TNFR, and preligand-binding assembly domain (PLAD), a part of the TNFR extra-cellular domain, play a crucial role in the TNF-α signaling pathway. Therapeutic administration of either progranulin or soluble versions of TNFR1 or PLAD in mouse models led to beneficial results [189,190,191], opening the door to another type of non-antibody anti-TNF-α drugs [23, 192]. As said above, TNFR1 and TNFR2 mediate different signaling pathways [23]. Blockade of TNFR1 leads in activation and proliferation of Tregs and ameliorates progression of collagen-induced arthritis [193, 194]. In this line, a TNFR2 agonist was associated with Tregs expansion and suppression of cytotoxic CD8 + T cells, exhibiting an anti-inflammatory potential [195]. Thus, based also on preclinical data, specific targeting of TNFR1 or TNFR2 by antibodies, small molecules or aptamers may represent a safer and more effective treatment in IMIDs in the future [196,197,198]. For example, zafirlukast, a leukotriene inhibitor, downregulates TNFR1 pathway by PLAD-mediated TNFR1 dimerization [199].

Another approach of TNF-α pathway blockade that has been recently examined is the development of orally active small molecules (such as UCB-9260) that destabilize TNF-α trimer, impeding TNF-α-TNFR1 binding and signaling [200]. In addition, several natural products that exhibit a TNFα inhibitory effect are being identified using in silico methods [201]. Whether selective TNFR targeting or non-biologic anti-TNFs block TNF-α pathway in a more refined way than the current marketed biologic anti-TNFs, remains to be seen [23].

Combination of anti-TNFs with other targeted therapies

The current therapeutic unmet needs require the constant revisiting of established trends in IMIDs’ pharmacotherapy. Although combination therapies have the theoretical potential of a synergistic anti-inflammatory effect, two important issues should be noted. First, as immunosuppression is expected to be higher in such cases, safety issues should be addressed and avoided. Second, a combination therapy should not be relied on the mere co-administration of two DMARDs, but, instead, must be carefully designed to combine independent and additive, and not similar and overlapping, modes of action.

Combination therapies carry the advantage of synergic action of both drugs, usually allowing lower doses to be administered. In a recent experimental study, the tyrosine kinase inhibitor dasatinib was co-administered with sub-therapeutic doses (only 10% of the standard dose) of infliximab, adalimumab, golimumab or etanercept in the human TNF-dependent Tg197 arthritis mouse model [202]. The combination therapy led to amelioration of clinical and histopathological findings of arthritis, which was more profound with the combination of low-dose infliximab and dasatinib, to a degree similar to the therapeutic infliximab dose.

In patients with refractory IBD, co-administration of anti-TNFs with vedolizumab, ustekinumab or natalizumab has shown promising results in case reports and case series and was not accompanied by a substantial increase in the incidence of serious adverse events, such as serious infections [203]. Interestingly, dual inhibition of TNF-α and IL-17 in experimental spondyloarthritis model has been shown to ameliorate local inflammation, joint erosions and new bone formation in spine [204], while the effect of double blockade in bone erosion was more potent than neutralizing TNF-α or IL-17 alone [205]. In this line, a bi-specific fusion protein against TNF-α /IL-17A [206,207,208] is being under investigation in clinical trials in psoriasis.

Moreover, also recently, a novel antibody (ABBV-337) was developed; adalimumab was conjugated with a glucocorticoid receptor modulator, targeting both transmembrane TNFR and cytoplasmic glucocorticoid receptor through endocytosis. This strategy allows the anti-inflammatory effects of glucocorticoids, minimizing the adverse events of systemic glucocorticoids use. In a phase 2a proof-of-concept study in RA patients, ABBV-337 exhibited statistically significant better response than adalimumab alone at week 12 [209].

Conclusion

The introduction of anti-TNFs altered the disease course of several IMIDs by achieving remission or low disease activity in a substantial proportion of patients. These agents exhibit beneficial effect on the basic pathophysiological process, while their use led in a significant reduction in all-cause mortality, at least in rheumatic diseases. Anti-TNFs' known safety profile has been confirmed, proving safe during the COVID-19 pandemic and expanded their use within pregnancy and lactation. Moreover, their use changed the management of these diseases, adopting the “treat-to-target” strategy, treatment de-escalation strategies have been successfully applied and TDM has optimized anti-TNF-α use. The implementation of anti-TNF-α biosimilars made anti-TNF-α treatment more accessible and improved anti-TNFs’ cost-effectiveness. Several biomarkers are currently under investigation and are expected to utilize as useful prognostic tools in anti-TNF-α-treated patients. Last, development of new anti-TNF-α forms and sophisticated TNFR targeting might optimize drug bioavailability, enhance their local efficacy and minimize adverse effects, while combination with other agents might refine disease outcomes using lower drug doses. Although a proportion of patients with IMIDs do not respond to anti-TNFs and newer therapeutic targets are seeing the light toward this direction, anti-TNFs remain in the front stage of our therapeutic armamentarium.

Table 1 anti-TNF-α indications by chronological order of first approval (either by Food and Drug Administration-FDA or European Medicines Agency-EMA)
Full size table
Table 2 Data on disease outcomes after anti-TNF-α tapering or discontinuation. LDA: low disease activity
Full size table