Non-Study Supplied Glucocorticoids
The FDA representatives and some committee members also raised concerns about the use of glucocorticoids in the study. In addition to the study-supplied glucocorticoids (in the prednisone group), many of the participants received additional glucocorticoids as part of tapering from their initial management, as a premedication (in the rituximab group) or for other indications.
‘We saw signs of efficacy in this study, but confounding factors put into question the degree of efficacy.’ —Dr. Chung
Most of the additional glucocorticoids were given in the first half of the study: 86% of the subjects given avacopan used glucocorticoids at some point during the first 26 weeks. However, the mean cumulative glucocorticoid dose per patient was considerably lower in the avacopan group compared to the dose in the prednisone-treated patients (1,349 mg vs. 3,655 mg).
The FDA criticized this, saying the investigators could not rightfully claim non-inferiority at six months because both treatment groups had had some steroid use. Dr Jayne notes, “It was not easy to fully defend against that, except to say that with avacopan you got by with much less steroid.”
Statistical Analyses
The FDA also contested the actual numbers and resulting statistical significance. The investigators had used a blinded adjudication committee to help get the most accurate data possible in terms of their endpoint analyses using the Birmingham Vasculitis Activity Score (BVAS). Potential issues that might have led to data scoring problems were predefined. This is standard practice in vasculitis and lupus trials because the available scoring tools can be difficult to use 100% consistently across investigators, in terms of their ability to accurately detect current disease activity.
The FDA representatives reran statistical analyses using the original scores of the investigators, claiming they “might be more reflective of real-world use.” For these data that had not been cleaned up by blinded adjudicators, the endpoint of remission at week 26 and sustained remission at week 52 didn’t change. But using this other dataset, the endpoint of superiority at week 52 no longer reached statistical significance, raising questions for some about the drug’s true efficacy.
Subgroup analyses also posed thorny questions. The patients receiving rituximab vs. cyclophosphamide were distributed equally between the avacopan and prednisone groups as part of the trial design. But the FDA and some panel members expressed reservations that when response to avacopan was analyzed by subgroup, the patients who had received cyclophosphamide (plus, later, azathioprine) had not shown overall treatment superiority (in contrast to analyses of the rituximab group or analyses including all patients).
However, it is important to note the trial was not designed to pick up such differences in terms of its statistical power, and differences in rituximab vs. cyclophosphamide may reflect other factors, such as the part of the world where they were prescribed.
