The risk of determining risk with multivariable models
Article Abstract:
Multivariable analyses have become more frequent in medical research, but problems in their design may make the results inaccurate or misleading. Multivariable analysis is a statistical means of controlling for numerous factors. For example, a study might report that after controlling for age and level of formal education, population X has an abnormally high risk of developing disease Y. Of all articles published in the 'New England Journal of Medicine' and 'The Lancet' between 1985 and 1989, 451 used multivariable methods. Over the five-year period, the frequency of articles using these analyses increased from about 10% to 18%. In 44 randomly selected articles, there were several violations of methodological guidelines including over-fitting of data, nonconformity of variables to a linear gradient, non-proportional risk, no tests for interactions among variables, and unspecified coding and selection of variables. Improved reporting of results would make multivariate analyses more useful and easier to interpret.
Publication Name: Annals of Internal Medicine
Subject: Health
ISSN: 0003-4819
Year: 1993
User Contributions:
Comment about this article or add new information about this topic:
Clinical Epidemiological Quality in Molecular Genetic Research: The Need for Methodological Standards
Article Abstract:
Many research articles on molecular tests do not adhere to the standards for producing reliable research. Molecular tests are used in many genetic tests. A survey of 40 articles that discussed the molecular techniques that underly some tests found that only 5 complied with all seven standards identified for producing reliable research. Twenty-five did not comply with two or more standards and 22% did not comply with four or five standards. Examples of tests that were widely adopted before research showed them to be inaccurate include the carcinoembryonic antigen test, the dexamethasone suppression test for depression, and the indirect immunofluorescence assay for Lyme disease.
Publication Name: JAMA, The Journal of the American Medical Association
Subject: Health
ISSN: 0098-7484
Year: 1999
User Contributions:
Comment about this article or add new information about this topic:
Claims of Equivalence in Medical Research: Are They Supported by the Evidence?
Article Abstract:
Claims of equivalence in medical tests of drugs can not be relied on without careful examination of the test criteria as many claims incorrectly apply statistical results, have inadequate sample sizes. These errors can lead to false claims, inconsistencies, and harm to patients. A total of 88 papers which satisfied basic criteria were selected for evaluation. Of these, only 45 intended to study equivalence. In 67% of the reports, equivalence was claimed after tests for comparative superiority failed, and in 10% there was no statistical evaluation. The sample size in 25% of the reports was only 20 patients or fewer.
Publication Name: Annals of Internal Medicine
Subject: Health
ISSN: 0003-4819
Year: 2000
User Contributions:
Comment about this article or add new information about this topic:
- Abstracts: The risks of polypharmacy: More medicines than are clinically indicated are often prescribed for older patients
- Abstracts: Unlocking the doors: nursing a male patient who decides he is really a woman presentsparticularl challenges. When the patient is in a special hospital, there are additionalobstacles
- Abstracts: Long-term efficacy and safety of atazanavir with stavudine and lamivudine in patients previously treated with nelfinavir or atazanavir
- Abstracts: An evaluation of germinating beans as a source of vitamin C in refugee foods. Effects of blueberry and cranberry juice consumption o the plasma antioxidant capacity of health female volunteers
- Abstracts: Growth hormone replacement in healthy older men improves body composition but not functional ability. Postmenopausal Hormone Therapy Increases Risk for Venous Thromboembolic Disease