Statistics from Altmetric.com
A recent paper in the Annals of the Rheumatic Diseases attempted to examine the “real world” experience of etanercept treatment by examining the incidence of flares of rheumatoid arthritis (RA) disease activity in a cohort of patients with RA who had started treatment with etanercept before September 1999.1 The number of flares and patients experiencing flares within the first year of etanercept treatment was compared with that seen in the same cohort of patients the year before they started etanercept. It is well documented that the withdrawal rate from disease modifying antirheumatic drug (DMARD) treatment in RA increases with the length of time the patient has been receiving the drug and that a number of these withdrawals relate to loss of efficacy.2–4 Therefore, it is reasonable to assume that the number of disease flares will increase the longer a patient with RA is receiving treatment, particularly if that treatment is failing to control the disease activity. The fact that treatment of this cohort of patients with RA was changed to etanercept suggests that their current DMARD treatment was failing to control their disease. Therefore it is likely that there would be an increased number of disease flares in this group before starting etanercept treatment. Although not stated in the paper, it is reasonable to assume that this cohort of patients with RA had been receiving their previous DMARD treatment for some time before changing to etanercept treatment. Therefore what the authors of this paper have compared is the number of RA disease flares in a cohort of patients with RA in their first year of etanercept treatment with the number of RA flares in their last year of (failing) DMARD treatment. The results are predictably in favour of the new treatment.
Would the authors have found the same results in favour of etanercept treatment if they had conducted a “fairer” comparison and compared the number of flares in this RA cohort during their first year of the previous DMARD treatment, especially if it was methotrexate, with the first year of etanercept treatment?
We thank Dr Smith for his interest in our recent publication. In his comments he states that it is well documented that the withdrawal rate from disease modifying antirheumatic drugs (DMARDs) in rheumatoid arthritis (RA) treatment is increased with increasing duration of use. He therefore suggests that our study design biases our results because by using patients who have changed their DMARD, we have selected those for whom the usefulness of that particular drug has been outlived. However, we contend that although the withdrawal rate from DMARDs increases with time, this may be a consequence of increasing disease severity and RA refractoriness, and not simply lost effectiveness of the drug.
The natural history of RA as a progressive and increasingly recalcitrant disease is also well documented. Furthermore, recent studies have shown better outcomes when DMARDs are introduced to naive patients, suggesting that those who have already been exposed to other DMARDs are even more difficult to control. Also, the very nature of the American College ofRheumatology “response” criteria is such that no absolute value is sought to demonstrate a treatment’s success, but rather it is the individual subject’s relative improvement compared with his or her baseline assessment. If Dr Smith’s assumption was made, this method would also be invalid.
Our design attempted to mimic these criteria in a “real world” situation. Although we agree that some bias was introduced by our design, we would also argue that in order to use subjects as their own controls, their overall status must be the same between the two periods of comparison. Comparing outcomes within the same subject during the first years of a new DMARD would eliminate the presumed problem introduced by Dr Smith’s comments, while potentially introducing others.
A more ideal study would be to compare outcomes (that is, flares) between two groups of patients with RA exposed to the same DMARD and matched for multiple variables, including disease duration, number of DMARDs previously used, similar comorbidity, etc. However, this would eliminate the “real world” setting.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.