Abstract: . . .  8 Evidence Hierarchy Evidence Hierarchy SCIENCE Reviews Reviews Meta Meta - - analysis analysis Health Health Services Services Research Research Randomized Randomized Controlled Controlled Trials Trials Observational Observational Outcomes Outcomes Laboratory Laboratory Qualitative, Case Reports, Anecdotes Qualitative, Case Reports, Anecdotes PRACTICE PATIENTS POLICY Decision Points in Medicine Decision Points in Medicine • Patient • Practitioner • Scientist • Payer/Policy Jonas, WB. And Levin, JS. Essentials of Complementary andAlternative Medicine . Littincott Williams & Wilkins. 1999; p. 34. Page 9 9 Treatment of Allergies with Treatment of Allergies with Homeopathy Homeopathy A patient with seasonal allergies says that her daughter recommended a homeopathic remedy from a health food store for allergies. Her daughter said it worked great for her and had no side effects. Your patient says she would like to stop her conventional drugs, which make her sleepy. She wants to know if this is alright? Is she harming herself or wasting her money? Evidence Hierarchy Evidence Hierarchy Reviews . . .
. . .  Homeopathy Funnel Plot Standard error Odds ratio .01 .02 .05 .1 .2 .5 1 2 5 10 20 50 100 2 1 0 Meta-analysis of Clinical Trials of Homeopathy Funnel Plot after “Trim and Fill” Standard error Odds ratio .01 .02 .05 .1 .2 .5 1 2 5 10 20 50 100 2 1 0 Page 14 14 Summary Recommendations on Investigating and Dealing With Publication and Other Biases in a Meta-analysis Examining for bias • Check for funnel plot asymmetry using g raphical and statisti cal methods. • Use meta-re gression to look for asso ciations b etwee n k ey m easu res o f trial qu ality and trea tm ent effec t si ze . • Use meta-re gression to examin e oth er possibl e e xpl anations for heter og eneity. • If avail abl e, examin e a sso ciations b etwee n t rea tm ent effec t si ze a nd chang es in biologi cal markers o r pati ents ’ a dh ere nce to t rea tm ent. Dealing with bias • If th ere is evid ence of bi as, repo rt this with th e same promin ence as any combin ed estim ate of t rea tm ent effec t. • Consid er sensitivity analys es to est ablish wheth er th e e stim ated t rea tm ent effec t is robust to rea son abl e . . .
--2192,2,548,2357,10962