Author(s)
Fekete I1, Hall S2, Bohn J2, Paradis AD2, Schneller S1, Gleißner E3, Coratti G4, Bovis F5, Nascimento A6, Povedano M7, Sormani MP5, Vazquez JF8, Mercuri E9, Kirschner J10, Patel SN3, Volmer T1
1SmartStep Consulting GmbH, Healthcare Strategy & Market Access Consulting, Hamburg, HH, Germany, 2Biogen, Cambridge, MA, USA, 3Biogen GmbH, Munich, Germany, 4Paediatric Neurology Unit, Catholic University, Rome, Italy, 5Department of Health Sciences (DISSAL), Section of Biostatistics, University of Genova, Genova, Italy, 6Department of Neurology, Neuromuscular Unit, Hospital Sant Joan de Déu, Esplugues de Llobregat, Barcelona, Spain, 7Functional Unit of Amyotrophic Lateral Sclerosis (UFELA), Service of Neurology, Bellvitge University Hospital, L’Hospitalet de Llobregat, Spain, 8Department of Neurology, Motor Neuron Unit, Hospital Universitario y Politecnico La Fe, Valencia, Spain, 9Department of Paediatric Neurology and Nemo Clinical Centre, Catholic University, Rome, Italy, 10Department of Neuropediatrics and Muscle Disorders, Faculty of Medicine, University of Freiburg; Department of Neuropediatrics, University Medical Hospital, Bonn, Freiburg; Bonn, Germany
With the GSAV* law, the German Federal Joint Committee (G-BA) has been authorized to mandate the collection of data on routine practice use of orphan drugs, drugs with conditional marketing authorization (MA), and drugs with MA under exceptional circumstances. Our aim is to highlight methodological challenges investigators will face when combining clinical practice data from different registries, using the example of the rare neuromuscular disorder 5q-associated spinal muscular atrophy (SMA). Challenges include missing data, variable data quality, imbalance in treatment groups, confounding, nonparametric distributions, and heterogeneity of patients. We will compare these challenges to the quality criteria highlighted in the IQWiG Rapid Report A19-43 [
Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWIG), 2020]. We present a pre-specified statistical workflow on how to organize a set of prospective observational data, and how statistical challenges in the analysis of registries could be addressed. We describe our utilization of mixed-effects models as the primary analysis [
Pinheiro, 2012] to address the statistical challenges of these data in line with European Medicines Agency (EMA) guidelines [
European Medicines Agency (EMA), 2010]. We suggest random splitting as an additional technique. We contrast our methods with those provided in the IQWiG Rapid Report, which favors a propensity score matching approach. The method of Conditional Inference Trees [
Hothorn et al., 2006], previously employed for the analysis of clinical registries [
Evaniew et al., 2019;
Martin et al., 2020], has not been specifically addressed in the current IQWiG methods paper, but will also be discussed in the German HTA context. There are a multiplicity of statistical models available to generate balanced comparisons of treatment arms in clinical registry data. These methods might improve analyses conducted with G-BA-mandated registries as part of the German HTA process. * Gesetz für mehr Sicherheit in der Arzneimittelversorgung
Conference/Value in Health Info
2020-11, ISPOR Europe 2020, Milan, Italy
Value in Health, Volume 23, Issue S2 (December 2020)
Code
PRO14
Topic
Methodological & Statistical Research, Organizational Practices
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics, Best Research Practices, Confounding, Selection Bias Correction, Causal Inference, Missing Data
Disease
Neurological Disorders