Research design

Whether you are conducting a small study or a large clinical trial, research design is important in minimising bias, maximising the reliability of your conclusions and providing robust and acceptable statistical evidence.

The basics

The null hypothesis
In clinical research your study should be designed to either prove or disprove a null hypothesis. There should be a clear and concise research question, and around this both a hypothesis and a null hypothesis should be formulated.

Statistics

Everyone should consult a statistician at the research planning stage.

Statistical calculations regarding sample size should be performed prior to commencing research. Sample size and statistical analysis are hugely important considerations if your research is to produce conclusions that are reliable, and also to recruit in a timely manner. You need to calculate the number of recruits you will need to provide acceptable chances of arriving at a wrong positive conclusion (often 5%), and a wrong negative conclusion (often 20%) (not the same) and decide whether or not you are likely to be able to recruit all of the patients within your project's timescale. Give yourself a little extra room for drop-outs and people who don't wish to participate.

Read more about statistics in research.

Statistical considerations are an important aspect of study design but many other influences on results are highly significant.

Clinical equipoise

According to Shaw and Peduzzi, Clinical equipoise, or genuine uncertainty within the expert medical community about what is the preferred treatment is a fundamental ethical requirement for the conduct of clinical trials. (Methodological issues in comparative effectiveness research: clinical trials, Peduzzi, 2010; Ethics in cooperative clinical trials, Shaw, 1970)

As an investigator it is important to remove as much bias from the process as possible. This includes the investigator assuming the null hypothesis to be correct until proven otherwise. In studies involving analysis by the investigator a lack of clinical equipoise might lead to bias in reporting, and in studies involving patients this might result in bias from patients due to both the placebo effect and a will to please the investigator.

Blinding

In interventional studies, participants should be blinded when possible. The best scenario is if both the investigator and patients are blinded, so that neither knows which treatment is being received. If measurements/readings are being taken by a different health care professional, these should be blinded too.

Peduzzi states that "similar to clinicians, patients often have preferences for a particular treatment, especially in studies of established therapies. These preferences should be considered in the design of a trial because they can affect recruitment and adherence (i.e., dropout)."

In some cases, such as in some surgical intervention studies it may not be possible to blind either the investigator or the patient. Other methods to minimise bias should be considered, for example by using a clinician/technician who is unaware of the treatment groups to take readings.

In retrospective research studies or service evaluations one member of the research team might enter data to be analysed onto a database, and give a code for treatments (example A and B), whilst the data is then analysed by a different member of the team, unaware of the meaning of the coding.

Randomisation

In interventional studies, randomisation of patients to groups receiving different treatments is an important technique preventing bias. There are various randomisation methods suitable for different types of study. In small sample sizes there might be a larger risk of higher prevalence of confounding factors in one group compared to another. Randomisation techniques can help reduce these risks.

Confounding factors are important and without careful consideration in how to minimise the effects of co-morbidities or other treatments, results from a study might have so much 'noise' that the treatment effect is unnoticeable. Peduzzi writes that "in designing clinical trials there is often a "single-disease" mentality that focuses mainly on the disease under study. However, in studies of patients with multiple comorbidities, as is often the case in the elderly, medical conditions other than the one under investigation also must be properly managed medically to avoid spurious treatment effects."

Kang describes how "although randomization appears to be a simple concept, issues of balancing sample sizes and controlling the influence of covariates are important. Various techniques have been developed to account for these issues, including block, stratified randomization, and covariate adaptive techniques." (Issues in outcomes research: an overview of randomization techniques for clinical trials, Kang, 2008)

Observational studies

According to Concato "whenever observational studies of therapy are considered, the problem of susceptibility bias or confounding is discussed. Confounding occurs when an extraneous factor is associated with both exposure and outcome, often leading to biased results". (Observational Methods in Comparative Effectiveness Research, Concato, 2010).

Observational studies should attempt to "identify confounding factors and account for them. This can be done in a variety of ways such as stratification, matching or regression analyses. " (Concato, 2010)

Questionnaire design

When using a questionnaire in a study, questionnaire design is very important, as is the method of analysing the results. This is an area that should be read into in depth. A starting point might be an article by Edwards - Questionnaires in clinical trials: Guidelines for optimal design and administration, available on Pubmed.

Further reading and references:

Kang M, Issues in outcomes research: an overview of randomization techniques for clinical trials, National Athletic Trainers' Association, Middle Tennessee State University, 2008.

Peduzzi P, Methodological issues in comparative effectiveness research: clinical trials, The American Journal of Medicine, 2010.

Knight KL, Study/experimental/research design: much more than statistics, National Athletic Trainers' Association, Brigham Young University, 2010.

Concato J, Observational methods in comparative effectiveness research, American Journal of Medicine, VA Clinical Epidemiology Research Center , 2010.

Edwards P, Questionnaires in clinical trials: guidelines for optimal design and administration, Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, 2010.

Noordzij M, Sample Size Calculations, Nephron Clinical Practice, Department of Medical Informatics, University of Amsterdam, 2011