Designing Clinical Research Studies An overview S.F. O’Brien

To provide an overview of clinical research designs To describe key strengths and limitations of these. Objective

describe and summarize date e.g. mean, frequency make inference (use a probability model linking the data to a broader context) –to a population e.g. prevalence of disease based on a sample –causal inference e.g. treatment A is more effective than treatment B Statistics

Allocation of Subjects to Groups A random sample is selected from one population; units are then randomly assigned to different treatment groups. Random samples are selected from existing distinct populations. A group of study units is found; units are then randomly assigned to treatment groups. Collections of available units from distinct groups are examined. Causal inferences can be drawn Inferences to the populations can be drawn By RandomizationNot by Randomization Selection of SubjectsNot Random Random

Always consult a statistician before finalizing the research protocol.

Based on: significance level (usually 5%) power (usually 80%) scientifically or clinically meaningful difference standard deviation (obtained from preliminary data or the literature) Sample Size Estimates

Observational Studies: often the first step in understanding a phenomenon usually describe the phenomenon e.g. prevalence, risk factors, natural history of disease often smaller, simpler and less expensive than clinical trials not appropriate for inferring causal relationships Two Types of Studies

Clinical Trials: can infer causal relationships usually large studies, more difficult to implement some not appropriate for ethical or practical reasons

Ecological (Correlational) Studies uses data at the population level rather than the individual level build mathematical models to examine relationships between population data. requires a sound hypothesis – what you expect the relationships to be need to partner with a statistician

e.g. estimating the risk of malaria in blood donors based on immigration and travel data, and malaria prevalence in these destinations. Advantages: - can be done relatively easily and quickly - uses population data already collected - low cost Disadvantages: - Relies on numerous assumptions. - no data on individual so no way to know if the people with the risk factor actually get the disease.

Case Reports and Case Series makes observations about patients (or donors) with defined clinical characteristics is a simple description of one or more cases without a reference or comparison group e.g. case reports on Chagas transmission by transfusion

Advantages: -can inform the transfusion community of first cases -useful for forming hypothesis Disadvantages: - can be chance happenings or subject to selection bias so difficult to generalize

Cross-sectional Studies describes the frequency of disease or a risk factor in a population. can characterize the disease, how frequently various symptoms occur. usually involves drawing a random sample from a defined population.

Advantages: -is representative of the defined population -can be reasonably quick to do Disadvantages: -may be difficult to get everyone in the sample to participate -not feasible for rare diseases (e.g. 1/1,000) due to large sample size required.

Case Control Studies identifies possible associations between disease and risk factors. is retrospective – identifies persons with disease and looks backwards at risk factors. important to select controls representative of the general healthy population and to collect data from cases and controls the same way. often matching criteria are applied to ensure a more comparative control group. often several controls are interviewed per case e.g. donor case control HCV

Advantages: -very useful to study risk factors in rare diseases -can study multiple risk factors Disadvantages: -does not estimate prevalence -cannot infer causal relationships -depends on the memory of subjects and is prone to errors in order of events, duration

Longitudinal Cohort Studies makes observations about a risk factor and development of disease is prospective in that risk factor date is collected first often two groups – those with risk factors and those without risk factors e.g. smokers and non-smokers

Advantages: -studies the natural history of disease -studies incidence and temporal association with risk factors Disadvantages: -some of the sample can be lost to follow-up -diagnostic methods may change (change the case definition) -study duration is often long -cannot infer causal relationship

Clinical Trials Key concepts: random selection – study participants are randomly selected from a list random assignment – study participants are randomly assigned to treatment groups placebo controlled – control group receives same treatment as the experimental group but without active ingredient. double blind – neither investigators nor subjects know which treatment they are receiving.

Parallel Designs: All treatment groups are starting at the same time experimental groupTest treatment control group Test placebo Crossover Designs: subjects serve as their own control treatment 1washouttreatment 2

Factorial Designs: - allow evaluation of treatment interactions e.g. 2 x 2 Factorial Design. Treatment A Placebo B Treatment A Treatment B Placebo A Placebo B Placebo A Treatment B Placebo B Treatment B Placebo A Treatment A

Final Comments There are a range of study designs to choose from. Choice of design depends on the research question. Need to balance the scientific quality of the study (generalizability) with practical issues. Always consult a statistician before you start the study!