Researcher details biases in pediatric studies
On July 13, it was reported that researchers found 41% of studies in pediatrics had the potential for bias.This assessment was conducted using The Cochrane Collaboration tool. This tool is used to assess internal validity and the risk of bias directly in studies.
Dr. Crocetti is an Associate Professor and specializes in General Pediatrics and Adolescent Medicine and co-author of this study. He agreed to be interviewed about bias risk in pediatric RTCs.
Question: What types of biases can be found in pediatric research studies?
Dr. Crocetti: First of all, I would like to emphasize that what we measured was the risk for bias, not actual bias. In other words, we wanted to see if the researchers had taken proper precautions to minimize bias or if they reported these precautions clearly to eliminate any perception of bias.
This said, research studies can become biased in several ways. The most common ones are if patients are not randomized properly to different treatment groups, if the masking is sub-par, (i.e., if either a patient or a researcher is aware of which patient gets which treatment), and if either the patient or the researcher can easily guess which treatment the next patient might get based on the previous patient’s treatment. The last step basically means there should be NO predictability in what treatment is received by which patient.
The six types of errors that can lead to bias are described in the paper in detail.
1. Allocation Sequence Generation
This type of bias mainly consists of using numbers based on odd or even number birth dates, a rule based on date of admission, hospital, or clinic record number, clinician judgment, participant preference, or intervention availability or even seemingly random things such as drawing lots or tossing coins. Instead, to lower risk, researchers should use random number table that’s computer generated.
2. Allocation Concealment
The risk for bias is high when “any procedure in which participants or researchers could have foreseen allocation, [such as] assignment envelopes unsealed, see-through, or not numbered; any alternation procedure that would allow an individual to predict allocation.
3. Masking Procedures
The potential for bias exists when anything happens to remove the “mask” that is in place, such as a participant who is “unmasked” can introduce bias to other participants (i.e. a patient knows that he or she received the treatment vs. placebo).
4. Incomplete Outcome Data
“[The] reasons for missing data were likely related to outcomes.” This can happen, for example, if intervention received patients drop out of a study due to an inordinate amount of adverse affects. This is like the situation with GlaxoSmithKline when they failed to report data that affected the outcomes of the drug, Avandia.
5. Selective Outcome Reporting
This is self-explanatory. This happens when not all the predefined outcomes are discussed in the study’s report.
This covers a wide range of areas. Doctors should follow the protocols set forth for conducting RTCs and reporting the findings fully and accurately to avoid many types of bias potentials.
Question: How does bias potentially affect the outcome of a study?
Dr. Crocetti: Bias can affect research findings in several ways. It can erroneously show that a drug/therapy has greater effect than it actually does, or it can underestimate the true effect of a drug or a therapy. Erroneous results can, obviously, lead to erroneous conclusions and therefore erroneous recommendations about the benefits/risk of a drug or a procedure. All of these have the potential to influence clinical practice and patient care.
Question: What are some ways researchers can protect themselves against bias affecting the study's outcomes?
Dr. Crocetti: They should make sure they follow proper protocol and don’t fall into the six pitfalls described in the paper. Another important note: All studies should be registered in a public-access database for transparency. We found that studies that are registered and described in the public domain had less risk of bias.
Question: Should the studies you reviewed that had potential bias be redone? Why or why not?
Dr. Crocetti: No. We reviewed the risk for bias, not actual bias. And redoing studies would be not only cost-prohibitive, but often logistically impossible. The more important take-home lesson here is that pediatric researchers need to be more cautious about properly following and describing the techniques they use to prevent bias. The other important point is that consumers of clinical literature, such as doctors who read these studies, should learn to read them critically and review the methodology carefully. They should focus not only on the results/bottom line, but ask how the researchers reached the conclusions that they reached.