Real world data and the art of maintenance therapy

Abstract

The goal of treatment optimisation is to deliver an effective therapy to those patients who are likely to benefit from it. The issue of how long such treatment should last is, of course, a difficult one. If a patient is cured, then further treatment is unnecessary and will provide toxicity with no additional disease control. Likewise, if a patient is not responsive to a given therapy then persistence with a regimen that is ineffective may likewise give no benefit. The issue becomes one of balancing control of the disease with toxicity. There are twin challenges in setting up and running clinical trials of maintenance therapy. The relatively better outcomes mean that event rates are lower, and reliable data requires large numbers of patients – which makes recruitment a challenge. Randomisation also requires uncertainty at the patient level in order to make randomisation ethical, and patients or clinicians may not be sufficiently comfortable with either continuing or stopping treatment. These challenges make the use of real-world data an attractive option. In this issue, Hill et al (2019) describe the results of a multicentre experience of maintenance rituximab following bendamustine-rituximab (BR) therapy. The improvement in outcomes seen with BR over previous therapeutic regimens means the question of maintenance becomes open once more. The findings of this real-world experience are that patients treated with maintenance who have a partial remission (PR) have better outcomes than those not given maintenance, but there was a lack of significant difference in those achieving a complete remission (CR). This lack of significant benefit was also seen in a validation cohort of CR patients. The use of real-world data in such a setting is attractive, but the results need to be interpreted with some caution. The decision to give or not to give maintenance was one made by the clinician and patient, presumably based upon a combination of patient factors, disease response and patient wishes. Those patients given maintenance therapy might be expected to be different from those not given maintenance, but it is not entirely clear how they might differ. There is some evidence that patients in this study who were given maintenance were younger and fitter; and indeed, the rates of maintenance therapy differ between the initial and validation cohorts, which might imply that the clinician or centre is an important factor. This sort of unquantifiable confounding can obfuscate the benefits of therapy. Green and Byar (1984) gave the example of radiotherapy for thyroid cancer: because radiotherapy was given to patients with more aggressive disease, a simple analysis of treated versus not treated would give the erroneous impression that radiotherapy was actively harmful. A failure to take confounders into account can thus lead to incorrect conclusions. So, what does this mean for the use of rituximab maintenance in follicular lymphoma? One issue with the condition is that, even with a large study, the numbers of progression events here are relatively small, meaning that it is not appropriate to perform multivariable analyses. But even if one could, it would not be possible to control for all the potential confounders. These data provide an indication, and probably pose a question rather than give an answer. In light of these data, it is unlikely that clinicians would feel comfortable about randomising to stop therapy in patients with only a PR. However, the closeness of the curves in the CR cases, taken together with the lack of difference in the fatal adverse event rate between the maintenance and no maintenance groups means that the question remains a pertinent one for patients in CR after BR therapy. However, a noninferiority trial to rule out any clinically relevant negative effect of stopping maintenance would need to be an order of magnitude larger than this study and is therefore unlikely to be feasible. We must instead rely upon accumulating data from clinical practice, and a synthesis of all such data in this setting, to provide a more nuanced indication as to whether improving frontline therapy removes the need for maintenance in some patients with follicular lymphoma. The paper of Hill et al (2019) gives us a much-needed steer in identifying that the matter is in doubt.