Model-based meta-analysis in ankylosing spondylitis: a quantitative comparison of biologics and small targeted molecules
作者单位：Clinical Research Center Beijing Children's Hospital Capital Medical UniversityNational Center for Children's Health School of Basic Medicine and Clinical Pharmacy China Pharmaceutical University
会议名称：《第八届全国治疗药物监测学术年会》
会议日期：2018年
学科分类：1006[医学-中西医结合] 1002[医学-临床医学] 10[医学] 100602[医学-中西医结合临床]
关 键 词：model based meta analysis ankylosing spondylitis biologics small targeted molecules
摘 要：Objective: Information on the comparative efficacy is important for drug development as well as drug therapy. Up to now, the relative efficacy of approved biologics and many agents under investigation in ankylosing spondylitis（AS） are still unclear. Several meta-analyses have been conducted to discover the potential difference among the available treatments for AS. Yet, no statistical difference among these treatments was observed and their relative clinical efficacy remains unknown. The common drawback of previous meta-analyses were that, primarily focusing on the single-end point data, they failed to take into account the varied time points of outcomes which many RCTs reported, thus leading to the inadequate utilization of the data. The objective of this study was to quantify the relative efficacy and time-course of various treatments measured by Ankylosing Spondylitis Assessment Study group response criteria 20（ASAS20）scores, change from baseline in Bath Ankylosing Spondylitis Disease Activity Index（BASDAI）and Bath Ankylosing Spondylitis Functional Index（BASFI）.Method: 34 double-blinded trials of 10 biologics and small molecules（adalimumab,certolizumab, etanercept, golimumab, infliximab, secukinumab, tocilizumab, sarilumab,tofacitinib, apremilast） encompassing 5,339 patients with AS were included in this analysis. Three mathematical models（ASAS20 model, ?BASDAI（change from baseline in BASDAI） and?BASFI（change from baseline in BASFI） model） with non-parametric placebo estimations were used to describe the longitudinal profile for the three efficacy measures.Baseline characteristics with relative rich information（percentage of male, disease duration, age, BASDAI, BASFI, CRP（C-reactive protein）） were included in our model as covariates to investigate their association with the drug effects. Model selection was based on the log likelihood ratio at an acceptance P value of 0.05.Results: The results detected significant differences among included treatments.Generally, the efficacy trend is similar across the three measures.Anti-TNF treatments were the most effective treatment as infliximab（5 mg/kg 0, 2, 6, q6 w） and iv golimumab provided the highest response,The lowest efficacy was observed in IL-6 inhibitors（tocilizumab, sarilumab）. Also,efficacy measured by ASAS20 showed an immediate achievement of maximum effect for most drugs. Among all the covariates investigated, the baseline BASFI value was included as covariate in ASAS20 model,and the percentage of male patients was included in ?BASDAI and ?BASFI model.Discussion:Our meta-analysis provided a quantitative method for efficacy comparison across approved treatments and with those under investigation. The characterization of all the three endpoint allows the evaluation of three different areas of drug impact, thus providing a comprehensive understanding of the drug efficacy. One advantage of longitudinal model-based meta-analysis over traditional meta-analysis is that it enables the estimation of