DISCUSSION
This comparative effectiveness analysis addressed a clinically important question; which is more effective for RA patients who had failed their first anti-TNF therapy, switching to another alternative anti-TNF therapy or commencing RTX? Switching to RTX was found to be significantly more effective as measured by both clinical effectiveness (EULAR response) and patient-reported physical function (achieving a MCID in HAQ score) 6 months after switching.
To our knowledge, this is the first study to compare physical function after switching to either RTX or a second alternative anti-TNF therapy in RA patients who had failed their first anti-TNF therapy. A previous observational study by Finckh et al compared the change in DAS28 scores after switching to either RTX or alternative anti-TNF (12). However, the sample size was small: 50 patients switched to RTX and 66 patients switched to an alternative anti-TNF therapy compared to 387 and 941 patients, respectively, in the current analysis. A recent update of this study was of a larger sample size (155 and 163 patients, respectively) (20). However, the focus of this update was on subgroups of the patients according to the reason for switching or the number of failed anti-TNF therapies. The results of the current analysis come into agreement with the results of those previous observational studies that reported superior response in patients who switched to RTX (12, 20). This superior response to RTX may be explained by the new mechanism of action offered by RTX.
There was a higher rate of achieving disease remission in the patients who switched to an alternative anti-TNF therapy; however, it was not statistically significant. This might be due to the lower DAS28 score at the start of therapy in this group of patients. In the UK, retreatment with RTX is allowed no more frequently than every 6 months. Therefore, there is also the possibility that the DAS28 or HAQ scores recorded at 6 months were at the beginning of a disease flare. However, this was not supported by the overall finding of a better overall improvement for both outcomes in the RTX group at 6 months.
Previous analysis from the BSRBR had shown that, 6 months after initiating the first anti-TNF therapy in RA patients, the mean change in DAS28 score was −2.1 with 18% of the patients achieving a good EULAR response and 50% achieving a moderate response (21). In comparison to the results of the current analysis, it was noticed that the response to the second alternative anti-TNF therapy (mean change in DAS28 score −1.17, 13.5% good responders and 33.8% moderate responders) was relatively lower than that of the first anti-TNF therapy.
After failing the first anti-TNF therapy, the patients in both study arms were of similar physical function. The improvements in HAQ score 6 months after switching was small and did not reach the MCID in both treatments, which may suggest an irreversible physical disability in those patients who already have failed 1 anti-TNF therapy.
The BSRBR was established primarily to study the short- and long-term safety of biologic therapy, and as such was not developed to specifically address questions of treatment effectiveness. Nevertheless, the regular collection of DAS28 and HAQ scores allowed this secondary outcome to be analyzed. One limitation is that DAS28 and HAQ scores were not routinely collected at the time a subsequent biologic therapy was started, with the exception of those patients reregistered in the RTX cohort. Therefore, not all patients had a DAS28 and/or HAQ score recorded within the 3 months prior to the switch or at 6 months following the treatment switch, which explains the relatively lower proportion of patients included in these analyses compared to the total available. However, the very large sample size of the BSRBR had allowed running this analysis by matching the dates of switching, and the baseline or the followup data of the BSRBR resulted in a final sample size that is currently the largest to date addressing this specific clinical question.
On the other hand, based on the same limitation of not collecting a full “baseline” data set at a time of switching, the current analysis could not model the available data to study predictors of response to the second anti-TNF therapy. In addition, it was not recorded whether patients had experienced primary or secondary failure of their first anti-TNF therapy and, therefore, this question also cannot be addressed at this time. However, recently within the BSRBR, we have studied those factors associated with response to RTX (22). One important clinical question is whether patients who are rheumatoid factor (RF) negative should preferentially receive a second anti-TNF therapy over RTX. Interestingly, RF status was only a very weak predictor of change in DAS28 score and was not associated with EULAR response or DAS28 remission in our data set (22).
As the data were collected in the routine clinical setting, where clinic appointments may not fall precisely at timed intervals as per a strict study protocol, some allowances also needed to be made in terms of the timing of data collection. Therefore, we allowed a range of time in which to include study outcomes. As patients in the anti-TNF cohort would not have baseline data rerecorded at the time of the switch, baseline data on comorbidities were carried forward from the start of the first anti-TNF cohort for patients who switched to a second anti-TNF therapy (RTX switchers would have had this data recorded at the time of reregistering). It is possible that some patients may have developed new comorbidities during their first anti-TNF treatment time. However, a sample of 10% of the patients was checked (from the followup forms) to see whether new comorbidities had developed during this time, and 93% of these patients had not developed a new comorbidity.
Although there were more patients with ≥1 comorbidities in the arm of patients who switched to RTX rather than in those who switched to a second anti-TNF therapy, the response was higher in RTX patients. This may suggest that the comorbidity overall was not necessarily acting as a confounder. Unfortunately, low numbers within each type of comorbidity limited the ability to explore the role of any individual comorbidity as a potential confounder.
This analysis was limited to study the patients who had failed only 1 first anti-TNF therapy. In clinical practice, patients may switch their therapy after failing 1, 2, or even 3 anti-TNF therapies. Future analysis that considers multiple switchers is likely to be of interest, although sample sizes will be much smaller. Within those patients studied, multiple switching in the patients who switched to anti-TNF was noticed.
A limitation of this analysis might be that the anti-TNF patients were combined together (to maximize the power), which did not allow for the analysis of subgroups of patients separately. For instance, the proportions of patients switching to alternative treatments for a particular reason may differ in response from those switching for other reasons. However, our analysis allowed for adjustment for the reason for switching in the propensity models.
After the recent approvals of abatacept and tocilizumab, the options of available therapies to treat patients who have failed their first anti-TNF therapy increase. A future comparison of effectiveness after switching to abatacept, tocilizumab, RTX, or a second alternative anti-TNF therapy is likely to be of interest.
Finally, one of the limitations of this comparative analysis is inherited from the fact that this is an observational study and the patients were not randomized to switch to either RTX or a second alternative anti-TNF therapy. Therefore, there may have been certain characteristics that led physicians to choose one therapy over another. Some patients in this analysis would not have had the choice, as most of the anti-TNF switchers would have switched at a time that RTX was not available. However, to overcome this potential source of bias, the analyses were adjusted for propensity scores that considered differences in the baseline characteristics of the patients. A randomized controlled trial would help clarify some of these issues, but the importance of observational data in identifying these knowledge gaps is also recognized.
In conclusion, the results of this comparative analysis suggest that switching to RTX may be of more benefit than switching to a second alternative anti-TNF therapy in RA patients who have failed their first anti-TNF therapy. The benefits of RTX were superior in both achieving EULAR response and achieving a MCID in HAQ scores. These results suggest that, in clinical practice, for patients who had failed a first anti-TNF therapy, it may be better to start RTX at this point rather than switching to a second anti-TNF therapy. Future analysis that considers the response in multiple anti-TNF switchers, or specifically looking at the reason for the switching, is likely to be of interest.