Background: Clinical analgesic trials typically report response as group mean results. However, research has shown that few patients are average and most have responses at the extremes. Moreover, group mean results do not convey response levels and thus have limited value in representing the benefit-risk at an individual level. Responder analyses and numbers-needed-to-treat (NNT) are considered more relevant for evaluating treatment response. We evaluated levels of analgesic response and Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) score improvement and the associated NNTs.
Methods: This was a post-hoc analysis of a 6-week, randomized, double-blind study (N = 387) comparing etoricoxib 90 mg, etoricoxib 120 mg, naproxen 1000 mg, and placebo in AS. Spine pain and BASDAI were measured on a 100-mm visual analog scale. The number and percentage of patients achieving ≥30% and ≥50% improvement in both BASDAI and spine pain were calculated and used to determine the corresponding NNTs. Patients who discontinued from the study for any reason were assigned zero improvement beyond 7 days of the time of discontinuation.
Results: For etoricoxib 90 mg, etoricoxib 120 mg and naproxen 1000 mg, the NNTs at 6 weeks compared with placebo were 2.0, 2.0, and 2.7 respectively for BASDAI ≥30% improvement, and 3.2, 2.8, and 4.1 for ≥50% improvement. For spine pain, the NNTs were 1.9, 2.0, and 3.2, respectively, for ≥30% improvement, and 2.7, 2.5, and 3.7 for ≥50% improvement. The differences between etoricoxib and naproxen exceeded the limit of ±0.5 units described as a clinically meaningful difference for pain. Response rates and NNTs were generally similar and stable over 2, 4, and 6 weeks.
Conclusions: For every 2 patients treated with etoricoxib, 1 achieved a clinically meaningful (≥30%) improvement in spine pain and BASDAI beyond that expected from placebo, whereas the corresponding values were approximately 1 in every 3 patients treated with naproxen. Use of NNTs and responder analyses provide additional, complementary information beyond population mean responses when assessing efficacy compared to placebo and amongst active therapies.