Observational designs in clinical multiple sclerosis research: Particulars, practices and potentialities.

Observational studies investigate a wide range of topics in multiple sclerosis research. This paper presents an overview of the various observational designs and their applications in clinical studies. Observational studies are well suited for making discoveries and assessing new explanations of phenomena, but less so for establishing causal relationships, due to confounding by indication (selection bias), co-morbidity, socio-economic or other factors. Whether observational findings are demonstrative, indicative or only suggestive, depends on the research question, whether and how the design fits this question, analytical techniques, and the quality of data. Observational studies may be cross-sectional vs. longitudinal, and prospective vs. retrospective. The term 'retrograde' is proposed to explicate that cross-sectional studies may obtain data that cover (long) preceding periods. Case reports and case series are usually based on accidental observations or routinely collected data. Cross-sectional studies, by simultaneously assessing clinical phenomena and external factors, enable the discovery and quantification of associations. In ecological studies the unit of analysis is population or group, and relationships on patient level cannot be established. A cohort study is a longitudinal study that investigates patients with a defining characteristic, e.g. diagnosis or specific treatment, by analyzing data acquired at various intervals. Prospective cohort studies use (some) data that are not yet available at the time the research is conceived, whereas in retrospective studies the data already exist. In a case-control study a representative group of patients with a specific clinical feature is compared with controls, and the frequencies at which an external factor, e.g. infection, has occurred in each group is compared; in a nested case-control study controls are drawn from a fully known cohort. Randomized controlled trial (RCT)-extension studies are informative because, due to RCT randomization, they are free from confounding by indication. Patient or disease registries are organised systems for the long-term collection of uniform data on a population that is defined by a particular disease, condition or exposure, with the purpose to study changes over time. In pharmacotherapeutic research, accidental observations of unexpected beneficial effects may lead to further research into a drug's efficacy in other conditions. Uncontrolled phase 1 studies investigate safety and dosing aspects. Observational studies are alternatives to RCTs when these are not feasible for ethical or practical reasons. Phase 4 observational studies play a crucial role in the evaluation of the effectiveness of treatments in daily practice, the validation of RCT-based side effect profiles, and the discovery of late occurring or rare, potentially life-threatening side effects. Combinations of multidisciplinary longitudinal data bases into large data sets enable the development of algorithms for personalized treatments. To improve the reporting of observational findings on treatment effectiveness, it is proposed that abstracts define the research question(s) the study was meant to answer, study design and analytical methods, and identify and quantify the patient population, treatment of interest, relevant outcomes and the study's strengths and limitations. The development of guidelines for Strengthening the Reporting of Observational Studies in Effectiveness Research (STROBER), as an extension of the guidelines used in epidemiology, is wanted.