Los sentimientos y las emociones en la Memoria Histórica

In this chapter, a model was constructed of the action of T N F -a inhibitors which sequester T N F -a and prevent it from binding to its receptor. The model was used to predict the effect of three anti-T N F -a agents (sTNFR2, etanercept and infliximab) th a t have been used as therapy for RA.

Because the model we used is one where T N F -a is being constantly produced at a fixed rate rather than exogenously added at the beginning of an experiment, our results are not directly comparable with those from Aderka et aZ.(1992). However, there is good correspondence with the results of a clinical trial conducted by Suffredini et aL(1995), where endotoxin-challenged volunteers were adm inistered etanercept. There are differences between our model and the situation Suffredini and co-workers studied, most notably the fact th a t Suffredini et a/. (1995) is a better model for runaway cytokine production characteristic of SIRS rather th an chronic equilibrium production characteristic of RA. However, the most essential feature, which is th a t T N F -a is being constantly generated, is invariant between the two models. Suffredini et a/.(1995) found th a t adm inistering 60 mg/m^ etanercept infusion caused plasma free T N F -a levels to fall. However, when both free and receptor-associated T N F -a was measured, peak cytokine levels rose dram atically in the etanercept group to more th an 25 times the peak for the placebo group. This is exactly the same effect th a t we found in our model (see Sections 4.5.1 and 4.5.2).

Our results differ from Aderka et a/.(1992) and R am anathan (1997), b ut are probably more valid in modelling the effect of an inhibitory drug used in vivo. This is because R am anathan (1997) used a simpler model which assumes th a t cytokine and inhibitor are in equilibrium, and does not consider receptor-bound T N F -a. As a result, increasing sTNFR2 concentration always up-regulates bioactive T N F -a in the model of R am anathan (1997), while in our model the situation is more complex.

Our results show th a t the use of T N F -a inhibitors causes the accumulation of a large pool of non-bioactive T N F -a bound to inhibitors. This pool is only slowly cleared from the circulation. However, due to the low dissociation rate constant of drug molecules, the T N F -a bound to the inhibitor molecules remains sequestered until clearance and is not available for receptor binding. Hence these drugs do not appear to have the danger of being a slow release “reservoir” th a t can increase the bioactive concentration of T N F -a over time. Our results concur with studies by Brennan et al.(1996) which found th a t

approxim ately half the T N F -a found in the RA synovial joint is not bioactive due to sequestering by naturally produced sTNFR2. This is despite the fact th a t sTNFR2 is generated in extremely low levels compared to the level of drug during a normal dosing regime of infliximab or etanercept. Although the sTNFR2 concentration is low, over time a large pool of non-bioactive T N F -a accumulates th a t is very slowly cleared.

The model (4.1)-(4.4) contained a number of simplifying assumptions which we felt justified in making:

a. Repeated exposure to antibodies or other drugs can cause the body to produce its own antibodies against the drug, and decrease its therapeutic potential over time. In the case of infusions of chimeric antibodies, this reaction is called the hum an anti-chimeric antibody (HACA) response. HACA response has been observed following repeated infliximab treatm ent. However, when infliximab is combined with m ethotrexate, the incidence of HACA response is reduced to levels th a t are extremely low, and in many cases undetectable (Maini et ai, 1998; Kavanaugh et al, 2000). Since infliximab must be prescribed together with m ethotrexate in the UK (British National Formulary, September 2001, p. 485) and USA (M arkham and Lamb, 2000), the HACA response will not be significant and is not modelled. For etanercept, HACA response was only detected sporadically in a small num ber of patients, and has not been correlated to a clinical response (Moreland et ai, 1999).

b. The rate of inhibitor delivery cua is dependent on the difference in inhibitor con­ centration in serum and the receptor com partm ent. This in tu rn depends on the pharmacokinetics of the drug from the point it is delivered to the body. To be more accurate, uja should be modelled as a function of tim e or inhibitor concentration. However, as we mentioned earlier, we have chosen not to fully model the pharm a­ cokinetics of the inhibitor in the various com partm ents of the body. Instead, the model focuses on the dynamic interactions between receptor, ligand and inhibitor. Hence, we have assumed th a t uJa is a zero order rate constant.

c. A further complication is th a t monocytes themselves can shed TN FR2 in response to ligand binding (Lantz et al, 1990), which can contribute to th e level of inhibitor. However, as noted earlier, the concentration of sTNFR2 in synovial fluid is two orders of m agnitude lower than th a t of drug during therapy, so this should have a negligible effect. The inadequacy of naturally produced sTN FR2 to bind T N F -a in the RA synovial joint supports the thesis of Feldmann and his co-workers, th a t

RA represents a state of imbalance between pro- and anti-inflam m atory cytokines, w ith the balance heavily shifted towards the pro-inflammatory side (see for instance, Feldmann et aZ., 1996b).

One surprising result of the model is th a t infliximab, largely due to its lower affinity for T N F -a compared to etanercept, poorly inhibited T N F -a levels in comparison with etanercept. This conflicts with the excellent results infliximab has shown when used on RA patients in clinical trials. One reason could be due to the fact th a t T N F -a levels in the RA synovium show large inter-patient variations. For instance, an assay of the synovial fluid of RA patients by Saxne et a/. (1988) found th a t half (6 out of 12) had T N F -a levels below th e detection limit of the assay (50 pg/m l), while one showed T N F -a concentration above 7000 pg/m l. Indeed, the general cytokine p attern in synovial fluid shows large inter­ patient differences (Lettesjd, 1998). This may be related to the stage of the disease or may be caused by genetic variations within patients, as even healthy individuals show large variations in their T N F -a production (Yaqoob et aZ., 1999). Infliximab may be efficacious for patients with low to m oderate levels of T N F -a. Evidence for this also comes from the reported responses to anti-T N F -a drugs which also show considerable heretogeneity. In m ulti-centre trials w ith infliximab (Lipsky et aZ., 2000), about half of patients failed to respond according to the American College of Rheumatology criteria for 20% or better improvement. On the other hand, about a th ird showed an excellent response according to a more stringent 50% or better improvement.

A nother possibility is th a t infliximab may be able to inhibit T N F -a activity in other ways besides preventing receptor binding. For instance, infliximab is able to bind strongly to transm em brane T N F -a, a mechanism th a t results in the destruction of the T N F -a pro­ ducing cell and thus ultim ately a decrease in inflammation (Scallon et aZ., 1995; Feldmann

et aZ., 1997). Results of the simulation suggest th a t this effect may be responsible for a greater portion of the therapeutic efficacy of infliximab than previously suspected. Al­ though etanercept is also able to bind to transm em brane T N F -a, th e resulting complex is much less stable. As a result Scallon et aZ. (2002) found th a t at saturatin g concentrations, infliximab bound four times as much transm em brane T N F -a as etanercept.

The challenge now is to develop DMARDs which can perm anently restore non-pathological equilibrium to the cytokine network in the synovium. Our model suggests why infliximab and etanercept cannot bring about lasting effect when therapy is term inated. It does not seem to be due to lack of efficacy, because in the case of etanercept at least, a sufficient

dose achieves near total suppression of T N F -a in the receptor com partm ent. Following this suppression, there are three possible outcomes depending on th e mechanism driving T N F -a production in the synovial joint:

a. If T N F -a is being driven by a non-autonomous source of stim ulation, like factors secreted by T lymphocytes or fibroblasts, then anti-T N F -a inhibitors can only re­ duce the equilibrium level of T N F -a if they are constantly delivered to the receptor com partm ent. The rate of T N F -a, cj, production, is unaffected. If treatm ent is ever term inated, T N F -a concentration will return to its original level.

b. A second possibility is th a t T N F -a inhibitors can actually affect the rate at which T N F -a is being produced. This may happen if T N F -a induces its own production in these cells. Ulfgren et al.(2000) has found th a t infliximab therapy reduces T N F -a synthesis in the synovium of RA patients. Hence cj is a function of c, and decreasing the T N F -a level will decrease the rate at which new T N F -a is produced. Even so, the simulation predicts th a t anti-T N F -a drugs may not be able to perm anently suppress T N F -a. As Figure 4.4 shows, once cj drops sufficiently, the inhibitors will make up for the decreased production of T N F -a by releasing the ligand they have bound. This drives T N F -a autocrine stim ulation again, so the cytokine network returns to its old levels of T N F -a production.

c. Lastly, inhibition of T N F -a may actually up-regulate cell production of TN F-a. This is because T N F -a causes an increase in levels of T N F -a inhibitors produced by monocytes and macrophages, such as IL-10 (Katsikis et aZ., 1994) and P G E2 (Bachwich et al, 1986; Kunkel et al, 1988). If this is the case, then there is a negative feedback mechanism th a t will always seek to return T N F -a to its equilibrium level. Of course the hope is th a t there is also a non-pathological equilibrium where T N F -a is being produced at a much lower level. In order to reach it however, the network would have to leave the basin of attraction of the pathological equilibrium. This could potentially be quite large since RA is able to p erp etu ate itself over a wide range of cytokine concentrations.

A nti-T N F -a therapy has been attem pted for a group of disorders involving excessive in­ flam m atory response called systemic inflammation response syndrome (SIRS). SIRS is an um brella term covering occurrences of systemic inflam m atory host response to a variety of conditions, including infection, trau m a and burns (Members of the American College

of Chest Physicians/Society of Critical Care Medicine Consensus Conference Committee, 1992). SIRS is called sepsis when it is the response to infection from microbial pathogens such as bacteria. A local inflammatory response towards a focus of infection or injury results in the release of a cascade of cytokines and other m ediators (such as eicosanoids, adhesion factors and complement molecules) in order to bring the insult under control. However, if it does not occur rapidly enough, some of the m ediators may escape into the circulation and trigger a systemic response (Bone, 1991). Often th e systemic response persists long after the initial trigger itself has cleared. The systemic response to the insult th a t triggers SIRS can lead to multiple organ dysfunction, and is an increasingly common cause of death due to factors such as the spread of AIDS, increased use of invasive devices in medicine, and prolonged survival of critically ill patients.

On the surface, the cytokine networks involved in RA and SIRS appear almost identical. TNF-O! is well-established as the key mediator and inducer of other m ediators in SIRS, just as it has been in RA (see Zhang and Tracey, 1998a; R einhart and Karzai, 2001 for reviews of the evidence). Ju st as in RA, a sequential cascade has been suggested, with T N F-o and IL-1 at its apex (Blackwell and Christm an, 1996).

Yet anti-TN F-o therapy for human patients in SIRS has been pointedly unsuccessful compared to its results in RA. A recent survey of clinical studies utilising this therapy (Reinhart and Karzai, 2001) found th a t a number of studies reported positive improve­ ments to mortality, b ut the benefits were relatively minor (none in excess of 4% decrease in m ortality). On the other hand, there were a number of studies th a t actually reported dele­ terious effects, including two th a t reported increases in m ortality in excess of 10%. Trials of other anti-inflammatory agents in SIRS have been similarly unsuccessful; no therapy aimed at a single inflammatory mediator in SIRS has been safe and effective (reviewed in N atanson et al.^ 1994; Zeni et uA, 1997).

A number of reasons have been suggested for the failure of T N F -a inhibitor and anti­ inflammatory therapy in general to improve m ortality in SIRS. T N F -a plays a role in mobilising host defense to infection (Echtenacher et al, 1990), so suppressing it in cases of infection may block both its harmful and protective effects (Quezado et uL, 1995). There have also been suggestions th a t intervention is unable to sufficiently suppress pro- inflammatory response because it only deals with one out of several m ediators of inflam­ mation with overlapping and compensatory functions (Suffredini et uL, 1995; Blackwell and Christm an, 1996). Zanetti et a/. (1992) found th a t anti-T N F -a antibodies decreased

IL-1 and IL-6 levels as well as m ortality in LPS-challenged mice. However, in mice infected with Gram-negative bacterial peritonitis, a better model of sepsis in humans, anti-T N F-a antibodies neither decreased pro-inflammatory cytokine levels nor afforded any protection from lethality. This suggests th a t the cytokine cascade in sepsis at least does not have a sole cytokine (T N F-a) at its apex.

However, there may be a more mechanistic reason to do with the way anti-T N F -a works to sequester T N F -a. The dynamics of T N F -a concentration levels in th e presence of an inhibitor in our model could help to explain the mixed results of anti-T N F -a ther­ apy in sepsis. Both pro-inflammatory and anti-inflam m atory cytokines are up-regulated during SIRS (Gogos et ai, 2000). Bone (1996) has suggested th a t it may be too sim­ ple to characterise sepsis as runaway pro-inflammatory cytokine production. Instead, he sees two distinct phases in the systemic response to infection and other insults. The first is SIRS, involving a pro-inflammatory cytokine response. However, SIRS triggers an anti-inflammatory response in an attem p t by the body to restore a non-pathological equilibrium. This he term ed the com pensatory anti-inflam m atory response syndrome or GARS, which may be inadequate, may overcompensate, or may (in th e case of a recovery) successfully restore homeostasis. A further complication is th a t SIRS actually involves multiple sites of inflammation, which may be out of phase with each other in term s of the kind of response they are undergoing (Bone, 1991).

RA is a chronic disease th a t can be m athem atically represented as a dynam ical system in equilibrium, b ut SIRS appears to be characterised by rapidly varying cytokine levels. In RA, anti-T N F -a therapy succeeds in reducing T N F -a levels to a new, lower equilibrium, resulting in improvement to symptoms though not perm anent cure. In SIRS however, the carrier effect of T N F -a inhibitors is much more uncertain and dangerous. W hen T N F -a levels decrease during the com pensatory stage of GARS, the inhibitors may actually sab­ otage the b od y’s own attem p t to restore homeostasis (equilibrium). This happens when they release T N F -a at the very point th a t the immune system is attem p ting to suppress excessive T N F -a levels. In fact, as Figure 4.2 shows, even if the immune system succeeds in switching off T N F -a production altogether, the drug inhibitors would still ensure th a t T N F -a levels remain at a potentially dangerously high level. On th e other hand, a lo­ cal, well-measured and tim ed intervention could m oderate both the peaks and troughs of T N F -a levels, preventing both runaway T N F -a production and overcompensation by GARS. This could account for the very mixed results during clinical trials of T N F -a in­ hibitors, with some reports of improvement (albeit small ones) to m ortality rates, and

others reporting an adverse effect.

This underscores the fact th a t SIRS involves interactions between several cytokines, both pro- and anti-inflam m atory (Dinarello, 1997; Seely and Christou, 2000). It is not well- characterised by a model of a single cytokine (TN F-a) in isolation from other factors. This is implicitly w hat most attem pts at anti-T N F -a therapy are based on. W hen such a model is explicitly represented in m athem atical terms, as we have done here, th e limitations of this approach become much more obvious. A successful therapy will need to take into account host anti-inflam m atory reaction as well (Bone, 1996). To draw an accurate picture of what is happening during SIRS, and to predict the form, tim ing and duration of therapy