Tumour necrosis factor α (TNFα) blocking agents are now widely used in the treatment of aggressive rheumatoid arthritis (RA) and have proved to be effective and relatively safe.^1–,¹⁰

Earlier studies have shown the induction of autoantibodies during treatment with TNFα blocking agents.¹¹ Thus, antinuclear antibodies (ANA) have been described in 53–68% and anti-double stranded DNA (anti-dsDNA) antibodies in 5–10% of patients with RA treated with the TNFα blocking agent infliximab (Remicade).^1–3,5,11 In patients treated with etanercept (Enbrel) the same autoantibodies are seen in 11% and 5–13%, respectively.^7,8 Only limited data have been published about the induction of anticardiolipin antibodies (aCL) during TNFα blocking treatment. Thus, Elliott et al found aCL in one of 20 patients with RA treated with anti-TNFα (cA2) in an open trial lasting 8 weeks.⁴ Rankin et al measured the serological effects of repeated doses of the humanised anti-TNFα antibody CDP 571 in patients with RA and found that some patients develop positive aCL (IgG).¹² Ferraccioli et al showed variations in aCL titres over time in etanercept treated patients with concomitant bacterial infection, where lowering of titres was seen after treatment with antibiotics.¹³ In the normal population (healthy blood donors) aCL are found in 2–6%, and in an aging population in up to 12%.^14,15 In patients with RA aCL have been found at even higher frequencies.^16,17 aCL, in general, are associated with thromboembolic disease, but their clinical significance in RA is uncertain and their presence has been considered to be a non-specific marker of activation of the immune system.¹⁸

In this study we assessed the frequency of aCL in patients with RA treated with infliximab or etanercept, and their relation to clinical outcomes.

PATIENTS AND METHODS

RESULTS

Frequency of aCL increases during treatment with TNFα blockers

At baseline 9/65 (14%) infliximab and 10/56 (18%) etanercept treated patients had positive IgM and/or IgG aCL, at low to medium levels. After 3 months the frequencies of aCL positivity were 29% (p<0.05 compared with baseline) and 27%, respectively, and after 6 months 28% and 25%, respectively. Increases were seen for both IgG and IgM aCL (figs 1 and 2⇓⇓). Increasing age, a greater number of prior DMARDs, and a greater disease activity at baseline by DAS28 were positive predictors for development of aCL after 3 months of treatment. No correlation was seen between ANA positivity and the development of aCL (table 1⇓). The mean aCL IgM and IgG titres showed a small, non-significant increase over time (fig 3⇓).

View this table:

• View inline
• View popup

Table 1

 Predictors of aCL positivity at baseline and of seroconversion

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Percentage of aCL(+)patients at baseline and after 3 and 6 months of treatment with infliximab. Percentage positive for IgG aCL: p<0.05 for linear trend.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

 Percentage of aCL(+)patients at baseline and after 3, 6, and 12 months of treatment with etanercept.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

 (A) IgM aCL titres (IU/ml) in all patients at baseline and after 3, 6, and 12 months of treatment. (B) IgG aCL titres (IU/ml) in all patients at baseline and after 3, 6, and 12 months of treatment.

aCL are associated with worse clinical results

ACR responses and DAS values were studied in 44 infliximab treated patients. Twenty six of 30 (87%) aCL(−) patients were ACR20 responders while only 7/14 (50%) aCL(+) patients met the ACR20 criteria (p<0.05) (fig 4⇓). Assessment of the DAS28 response after 3 months showed that the aCL(−) patients had a mean (SD) DAS28 of 3.6 (0.15) compared with the aCL(+) patients who had a mean DAS28 of 4.3 (0.24) (p<0.05). However, baseline DAS28 values were also higher in the aCL(+) patients and there were no significant differences between the changes in DAS28 in the two groups (not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

 Percentage of ACR20 responders to treatment with infliximab in aCL(−) and aCL(+) patients. Comparison is by χ² test.

In etanercept treated patients, clinical responses were similar between aCL(+) and aCL(−) patients (data not shown).

aCL are associated with more frequent adverse outcomes

The frequency of treatment limiting infusion reactions in the aCL(+) patients was 17% compared with 5% in the entire TNFα registry (p<0.05) (fig 5⇓). The adverse effects were fevers, chills, urticaria, chest pain, dyspnoea, sickness, flushing with headache, tachycardia, hypotension, and anaphylactic reactions. Four patients had thromboembolic events (three receiving infliximab, one receiving etanercept). Two of these patients were aCL(+) and two aCL(−). Of the aCL(+) patients, one had a pulmonary embolus and the other a stroke. Thus, frequencies for thromboembolic events are 7% and 2.6% in the aCL(+) and aCL(−) patients, respectively, which is not statistically significant (relative risk 2.7, 95% confidence interval 0.8 to 4.6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

 Percentage of treatment limiting infusion reactions in aCL(+) patients compared with all patients in the registry (infliximab treatment only). Comparison is by χ² test.

DISCUSSION

In our patients we found a statistically significant increase in the frequency of IgM and/or IgG aCL positivity among patients treated with infliximab for 3 months or longer. Similar results were seen for etanercept, suggesting that this may be a class effect of the TNFα blocking agents. Previous studies on treatment with TNFα blocking agents and development of autoantibodies have mainly focused on ANA and anti-DNA. Thus, we believe that this is the first systematic description of increased aCL frequency in patients treated with TNFα antagonists.

A possible criticism of the study could be that patients treated with TNFα blocking agents might have reduced or discontinued glucocorticoid treatment and this might in itself allow the development of autoantibodies. Supporting this, a small but statistically significant decrease in mean glucocorticoid dosage was seen in these patients (not shown). However, the absolute change in dosage was small. Moreover, Lockshin et al showed that glucocorticoid treatment even at high doses did not change the presence of aCL nor did it lower the titres.²³

One possible explanation for the induction of aCL positivity in patients treated with TNFα blocking agents is that the down regulation of TNFα induces up regulation of interleukin 10, which in turn activates autoreactive B cells²⁴ and thus induces autoantibody production. Alternatively, lower levels of TNFα might lead to a generalised increase in Th2 activity, which even through other cytokine pathways might leads to B cell activation²⁴ and autoantibody formation. Ferraccioli et al have shown a relationship between clinically overt infections and aCL formation.¹³ Thus, a third explanation might be that TNFα blockade allows a subtle increase in bacterial stimuli from the environment, leading to increased aCL formation.

Increasing age, greater number of prior DMARDs, and more severe disease activity were predictors for the development of aCL in this study, but obviously, these factors are not independent.

We found that the development of aCL positivity during treatment with infliximab had prognostic significance because it was associated with worse clinical results and more frequent serious infusion reactions. However, it is unclear how the association between aCL positivity and clinical outcome is mediated. One could speculate that the precise immunopathologies of RA are different from patient to patient, and that in some patients autoreactive B lymphocytes might have a more causal role. Such patients might therefore be less likely to respond to TNFα blockade and also more likely to produce aCL. Indirect evidence for the existence of such a patient group with more B cell dependent disease has come from clinical studies with the B cell ablative monoclonal antibody rituximab.^25,26

In this study we found only a few patients with anti-dsDNA antibodies. However, it should be noted that we only measured IgG anti-dsDNA and not the IgM isotype, which in earlier studies was shown to be preferentially induced by TNFα blocking treatment.^{1–3,6–8,11} Among our patients only one developed IgG anti-dsDNA antibodies (both by an ELISA and by the Crithidia luciliae test) during treatment with infliximab. This patient developed a lupus-like syndrome with pleuropericarditis.

Our study may have some practical clinical implications. Although aCL in general are associated with an increased risk for thromboembolism, the small number of events in these patients did not allow us unequivocally to identify an increased risk. Continued long term follow up of patients treated with biological agents, such as is continuing through the Swedish ARTIS and STURE registries, is therefore of critical importance.

An important clinical question is whether all patients treated with the TNFα blocking agents should be screened for aCL. In this study, aCL positivity predicts less benefit from treatment with infliximab and a higher risk for a treatment limiting infusion reaction. Therefore, knowledge of aCL serology might, in principle, lead to a change in the risk/benefit assessment and might affect the decision about which antirheumatic treatment to choose. This may become a more germane consideration if a significantly greater number of therapeutic choices become available, which may well be the case in a number of years.

Another question is what should be done if a patient develops aCL during treatment with TNFα blocking agents. We believe that such patients should be treated as all subjects with fortuitously detected aCL, and that the aCL titres should be followed up periodically if found to be positive. However, this does not necessarily mean that aCL(+) patients need to stop the TNFα blocking treatment.

In future studies, it might be interesting to look more closely at the general immunoglobulin production and even the production of antibodies against infliximab and etanercept in our patient group. It would also be interesting to see if the aCL(+) patients would respond better to B cell ablative therapy. We are also following up our patient cohort for a longer period of time to see if more patients develop these autoantibodies, the longer the treatment continues.

In summary, aCL are induced in patients with RA treated with the TNFα blocking agents infliximab and etanercept. A statistically significant increase in frequency is seen for IgM and IgG aCL already after 3 months of treatment with infliximab. Etanercept treated patients show the same pattern, suggesting a class effect of the TNFα blocking agents. aCL positivity in patients treated with infliximab predicts worse clinical outcomes and more frequent treatment limiting infusion reactions. aCL are an important class of autoantibodies associated with TNFα blocking therapy.