|
Year |
2005 |
|---|---|
|
Question Number |
86. |
|
Category |
Orthopaedic Science |
Question
A prospective experimental study is being conducted to compare the relative efficacy of drug A and drug B in treating patients with osteoarthritis of the knee. Patients are randomly allocated to one of two treatment groups for 3 months and then after 1 month of no treatment, the patients are switched to the other drug treatment for 3 months. This specific design is susceptible to bias if
1- the clinical outcomes are determined in a double-blind fashion.
2- information is lacking on timing of exposure and outcome.
3- there are carryover effects from the first drug treatment.
4- there is differential loss to follow-up in the two treatment groups.
5- the symmetry of unknown confounders cannot be maintained.
Figures
Answer (click for answer)
PREFERRED RESPONSE: 3
This question seems to be getting at being able to understand different statistical terms such as bias, confounding, and chance effects. The Kocher article states that all research studies are susceptible to invalid conclusions due to bias, confounding, and chance. Bias is the nonrandom systemic error in the design or conduct of a study. Examples of bias include selection bias, when a nonrandom sample of the population is used, and observational bias, when a measurement is affected by the characteristics of the study group itself. Confounding or a confounder is a variable that has independent associations with both the independent and dependent variables, thus potentially distorting their relationship. Frequent confounders in clinical research include age, gender, comorbidities, and socioeconomic status. For example, an association between knee laxity and ACL injury may be confounded by female sex since woman may have greater knee laxity and higher risk of ACL injury. Bias, confounding, and chance can be minimized in a study by study design and statistical analysis. Randomization minimizes selection bias and equally distributes confounders. Blinding can decrease bias, and matching can decrease confounding. Multivariate and stratified analysis can sometimes control for confounders. Finally, the effects of chance can be minimized by adequate sample size and use of appropriate levels of significance in hypothesis testing.
Therefore, looking at the choices in this question, #1 would decrease bias since it is talking about blinding. #2 would cause problems with the statistical analysis in this study because information is missing, but it has nothing to do with bias. #4 deals more with issues of chance rather than bias, and any effect from this differential loss to follow-up can be minimized by an adequate sample size. #5 is discussing confounding specifically, not bias. Therefore, the answer is #3 because it is the only choice that addresses issues of bias. Carryover effects from the first drug treatment would produce a nonrandom effect in the second half of the study when the switchover in drugs is made, and would result in bias.
Reference: RECOMMENDED READINGS: Buckwalter JA, Einhorn TA, Simon SR (eds): Orthopaedic Basic Science: Biology and Biomechanics of the Musculoskeletal System, ed 2. Rosemont, IL, American Academy of Orthopaedic Surgeons, 2000, pp 1-17.
Kocher MS, Zurakowski D: Clinical epidemiology and biostatistics: A primer for orthopaedic surgeons. J Bone Joint Surg Am 2004;86:607-620.