Health Questions and Answers » Critical Review of Medical Literature

Should I read reviews on clinical topics?

rtrafaelmd — Thu, 23 Aug 2007 02:19:08 +0000

This depends on many factors:

Are you looking for basic knowledge or understanding of a disease process? If so, a clinical review may be sufficient and can provide the foundation for you to continue your reading on the topic.
Are you looking for the latest information? Clinical reviews may be outdated by the time of publication because the literature on which they are based was written before the review.
Is it a narrative or systematic review? In narrative reviews, the author selects the articles to include in the review and summarizes the topic based in part on his or her experience. In a systematic review, the author identifies articles through a search and includes or excludes the articles based on predefined criteria and summarizes the topic based on strength of the evidence from the included articles.

Does it matter who sponsors a study?

rtrafaelmd — Thu, 23 Aug 2007 01:56:16 +0000

Yes. Any direct involvement in a study by a sponsor, particularly one with a financial interest in the outcomes of the research (e.g., pharmaceutical industry), has the potential to influence the study. Sponsors should not have any input in study design, data collection, or how to report the results. Many research studies do not adhere to these standards. Disclosure of financial support is important and should alert the reader that there is the potential for introduction of bias into the study. Industry-sponsored studies may provide valuable information but must be reviewed carefully.

How do I interpret confidence intervals?

rtrafaelmd — Sun, 19 Aug 2007 04:03:34 +0000

A confidence interval is the expected range of results in the study population. A 95% confidence interval means that you would expect 95% of your results to fall within the specified range. A smaller range of values or less variance usually is found with larger sample sizes. A wide confidence interval could mean that some of the study results may not be clinically significant. Look at the upper and lower boundaries of the confidence interval and determine if both values still would hold clinical significance for you. If only the upper boundary value would have significance, there may not be an overall clinical benefit.

What does a P value refer to?

rtrafaelmd — Sun, 19 Aug 2007 04:00:58 +0000

The probability that the results of a study or the differences between study subsets occurred by chance. The most commonly used value, P < 0.05, means that there is less than a 5% probability that the study results occurred by chance. This is statistically significant, but not necessarily clinically significant. In an emergency room decrease by 1 minute in overall length of stay may be statistically significant (P < 0.05), but a 1-minute reduction in overall length of stay likely has no clinical relevance for physicians.

What should I look for when evaluating a chart review study?

rtrafaelmd — Sun, 19 Aug 2007 03:57:43 +0000

1. Trained chart abstractors
2. Explicit criteria for case selection and exclusion
3. Defined study variables
4. Standardized abstraction forms for data collection
5. Periodic meetings among researchers to resolve abstraction disputes
6. Monitored performance of abstractors
7. Blinded chart reviewers
8. Measures of interrater agreement

Do sample size and power matter?

rtrafaelmd — Sat, 18 Aug 2007 14:38:26 +0000

Power is the probability that the study will detect a treatment effect between the two experimental groups. The smaller the size of the treatment effect being studied, the larger the sample size should be. Many studies do not have a large enough sample size to detect a statistically significant difference and may report negative results when a significant difference may have been detected in an appropriate sample size. Without adequate power, the study results may be inconclusive.

What is blinding, and why is it important?

rtrafaelmd — Sat, 18 Aug 2007 14:34:56 +0000

A technique in which patients, physicians, researchers, and anyone else involved in the research study are unaware of whether patients are in the experimental or control group. This helps eliminate potential bias, unequal distribution of groups, differential administration of interventions, and distorted results and outcome assessments.