{"id":19209,"date":"2016-05-10T02:25:53","date_gmt":"2016-05-10T06:25:53","guid":{"rendered":"https:\/\/www.saratoga.com\/living-well\/2016\/05\/are-most-scientific-studies-false-and-biased-part-ii.html"},"modified":"2017-11-29T08:05:52","modified_gmt":"2017-11-29T13:05:52","slug":"are-most-scientific-studies-false-and-biased-part-ii","status":"publish","type":"post","link":"https:\/\/www.saratoga.com\/living-well\/2016\/05\/are-most-scientific-studies-false-and-biased-part-ii\/","title":{"rendered":"Are Most Scientific Studies False and Biased? Part II"},"content":{"rendered":"
In my homepage \nA 2005 review tilted, “Why Most “There is <\/p>\n So, Do We Ignore the Data?<\/b><\/p>\n No, no, and no…there’s These are some of 1. Read the In total, 78% of the news did not 2. Consider Subjects: Who Type of study: Intervention: What 3. Search the How are the Do the charts \n \u00b7Epidemiological \u00b7Biases \u00b7Design, analysis and interpretation of \u00b7Quasi-experimental study designs (http:\/\/www.ncbi.nlm.nih.gov\/pmc\/articles\/PMC1380192\/)\n<\/p>\n<\/blockquote>\n Where to Go 1. Physicians and practitioners need to be honest about interventions and 2. Consumers and patients need to be aware that some of the studies and
\nblog<\/a>, I discussed the flaws and biases in supplement and drug trials. This is
\npart II of the discussion.<\/p>\n
\nPublished Research Findings Are False,” provides a good summary of factors that
\ninfluence conclusions of studies (bold emphasis mine):<\/p>\n\n
\nincreasing concern that most current published research findings are false. The
\nprobability that a research claim is true may depend on study power and
\nbias, the number of other studies on the same question, and, importantly, the
\nratio of true to no relationships among the relationships probed in each
\nscientific field.<\/b> In this framework, a research finding is less likely
\nto be true when the studies conducted in a field are smaller; when effect sizes
\nare smaller; when there is a greater number and lesser preselection of tested
\nrelationships; where there is greater flexibility in designs, definitions,
\noutcomes, and analytical modes; when there is greater financial and other
\ninterest and prejudice; and when more teams are involved in a scientific field
\nin chase of statistical significance.<\/b> Simulations show that for most study
\ndesigns and settings, it is more likely for a research claim to be false than
\ntrue. Moreover, for many current scientific fields, claimed research findings
\nmay often be simply accurate measures of the prevailing bias. In this essay, I
\ndiscuss the implications of these problems for the conduct and interpretation
\nof research.” (http:\/\/journals.plos.org\/plosmedicine\/article?id=10.1371\/journal.pmed.0020124)<\/i><\/p>\n<\/blockquote>\n
\nways to use this information to make informed decisions about what the “evidence”
\nis actually saying. But, it takes a lot of detective work. Most of my research and looking into studies
\nhas come from my own review of studies reporting on biases and lessons learned
\nfrom my mentors and teachers. In no way am I methodology whiz, but I do have a
\nbasic grasp of why our model and interpretations need to be interpreted with caution.<\/p>\n
\nthe considerations that I always review when applying studies to my clients:<\/p>\n
\nactual study and be wary of media spin. In one cross-sectional analysis of 130
\nstudies of health news reported on google and found the following:<\/p>\n\n
\nprovide a full reference or electronic link to the scientific article. We found
\nat least one spin in 114 (88%) news items and 18 different types of spin in
\nnews. These spin were mainly related to misleading reporting (59%) such as not
\nreporting adverse events that were reported in the scientific article (25%),
\nmisleading interpretation (69%) such as claiming a causal effect despite
\nnon-randomized study design (49%) and overgeneralization\/misleading
\nextrapolation (41%) of the results such as extrapolating a beneficial effect
\nfrom an animal study to humans (21%). We also identified some new types of spin
\nsuch as highlighting a single patient experience for the success of a new
\ntreatment instead of focusing on the group results. (http:\/\/www.ncbi.nlm.nih.gov\/pmc\/articles\/PMC4608738\/)<\/i><\/p>\n<\/blockquote>\n
\nmethods used.<\/p>\n
\nwere the participants\/what population was studied (differences in gender\/ethnicity\/health
\nstatus)? What were their characteristics? What was the dropout rate? Who was
\nexcluded and why? What were the factors controlled for in the subjects? <\/p>\n
\nWas it observational and correlational study which look for relationships
\nverses cause-and-effect or was it a case-control randomized trial? Was there a
\ncontrol or was it a comparison trial? (Too much or too little control both have
\nweakness. For example, too much control prevents extrapolation of the
\nintervention to the real world and too little prevents interpretation that the
\nintervention caused the change.)<\/p>\n
\nis the form of intervention? Was it the appropriate dosage? How long was the
\nstudy? How was it taken? What was the placebo effect?<\/p>\n
\nresults for inconsistencies:<\/p>\n
\nresults reported? For example, is it the use of an odds ratio<\/a>, is it relative
\nor absolute risk? What is the NNT? Is the
\np-value of significance<\/a> truly reflective of compatible data with the
\nstatistical model? <\/p>\n
\nand statistics match the author’s conclusions?<\/p>\n
\n4. More can be found here for the geeks…<\/p>\n\n
\nstudy interpretation (http:\/\/www.ncbi.nlm.nih.gov\/pmc\/articles\/PMC3077477\/)<\/p>\n
\nto search for (http:\/\/www.ncbi.nlm.nih.gov\/pubmed\/18582622) <\/p>\n
\nmethod-comparison studies (http:\/\/www.ncbi.nlm.nih.gov\/pmc\/articles\/PMC2944826\/)\n<\/p>\n
\nfrom Here?<\/b><\/p>\n
\nbe transparent about what they have experience with. Both parties should look
\nup the NNT and find a few studies to examine if the intervention is new.<\/p>\n
\nstandard of care physicians are using could be flawed. Don’t just accept
\ntreatment that isn’t helping without studying the data or asking your doctor
\nfor more information. Most importantly, look for if you’re getting results with
\nan intervention (nutrient, herb, oil, supplement, medication) and use that in
\nyour basis for the final decision.<\/p>\n