Background Trial reports are much more detailed on benefit than on harm. The Outcome in Rheumatology (OMERACT) Initiative has suggested a simultaneous analysis of the occurrence of benefit and harm at the individual patient level rather than at the group level [1].

Objectives To apply the OMERACT analysis strategy to the TEAR trial comparing combination to step-up strategies in early rheumatoid arthritis (RA) [2].

Methods The OMERACT method creates 3 levels for benefit (good, moderate or no response) and for harm (no adverse events [AEs], non-serious adverse events and serious adverse events [SAEs]). The outcome of each patient is expressed as a pair of values expressing the level of both benefit and harm, summarized in a contingency table. Patients achieving a good response without serious adverse events are labeled as “unqualified success” [3], and patients experiencing no response but at least one adverse event as “unmitigated failure”.

TEAR randomized 755 early RA patients to 1 of 4 treatments in addition to methotrexate: immediate etanercept; immediate hydroxychloroquine plus sulfasalazine; and the same treatments delayed, and only applied in case of inadequate response at 24 weeks. In the primary study report, analysis at 6 months showed a better response in the immediate treatment groups compared to the delayed groups, but no differences in safety between the 4 treatment groups. For this post-hoc analysis, initially 3 benefit/harm categories were made by applying the EULAR response criteria and the adverse event reports. Patients dropping out prematurely for “side effects or any other medical issue” were coded as adverse events. Subsequently the data were further collapsed into 2×2 tables by combining the AE and SAE categories for harm, and the “good” and “moderate” categories for benefit; and similarly, by combining the “moderate” and “non” categories.

Results Response data at 6 months was available for 693/755 patients. Only 4% of patients experienced a serious adverse event. Overall, 15% patients experienced an unqualified success and 15% an unmitigated failure (Table). The occurrence of benefit and harm were not correlated. Proportions were similar across the four treatment arms. Combining the counts of good and moderate responders increased the numbers of patients with unqualified success, as expected (32%), and combining those of moderate and non-response increased the numbers with unmitigated failure (38%). Results were similar at one and two years' follow up (data not shown).

Conclusions Reporting the occurrence and benefit and harm together enhances the understanding of the patient experience. However, in this trial the occurrence of benefit and harm showed no correlation, and no differences between treatment groups were identified. The possible correlation between benefit and harm needs to be studied in more trials.

References

1. Boers M, et al. J Clin Epidemiol 2010;63:627-32.

2. Moreland LW, et al. Arthritis and Rheumatism 2012;64:2824-35.

3. Mancini GB, et al. Circulation 1999;99:377-83.

Disclosure of Interest None declared

DOI 10.1136/annrheumdis-2014-eular.1954