Transcript Slides 5
Biostatistics - Concepts Tudor Calinici – JPEMS 2014 Study types Based on the researchers role Observational Experimental Main domains of medical research Description of a health phenomenon Evaluation of a diagnostic test Evaluation of a therapeutic approach Evaluation of prognostic (risk or protective) factors Types of studies Based on study objectives Descriptive Analytical Statistics Descriptive Inferential Study types Based on the studied population Exhaustive Sampling Statistical objects Population Sample / representative sample Population vs. Sample Use sample because size accessibility Economical reasons Financial reasons Studies Source: www.socialresearchmethods.net Representative sample A good sample must provide an image about the population Representative sample quantitative condition qualitative condition Types of samples - groups Representative sample Exposed – non-exposed Case - control Study types Based on the duration of data collection • Cross-sectional longitudinal - retrospective or prospective Descriptive statistics Collecting data Characteristics – Variables Quantitative; Qualitative Descriptive statistics To present data Tables Charts Numeric characteristics Tables Frequency tables Count of Gender Treatment Total Metistatin 584 Placebo 614 Grand Total 1198 Contingency tables Gender Treatment f m Grand Total Metistatin 293 291 584 Placebo 321 293 614 Grand Total 614 584 1198 Charts The representation of the data Distribution For frequency tables For contingency tables Problem A randomized clinical trial (RCT) has been performed in order to test the efficacy of Metistatin (a statin - drug intended to lower blood cholesterol levels) versus Placebo for preventing major vascular events (nonfatal myocardial infarction, cardiac death, stroke) in diabetic patients. A number of 1198 patients with diabetes from several hospitals in England were included in the study sample. The study was conducted after the approval for marketing of this preparation. . Of interest for the study was whether Metistatin is superior to placebo (placebo - identical in appearance, size, taste to Metistatin, but lacking the active substance) in preventing major vascular events. A secondary objective has been to follow if Metistatin lowers cholesterol more than Placebo. Another question was “Is gender related to major cardiovascular events?” Every patient who entered the study was randomly allocated to one of two preparations: Metistatin 80 mg daily administered as a single dose, or placebo. The allocation of the two interventions was masked – the investigator was allocating for each eligible patient a code corresponding to one of the two interventions, found in opaque numbered sealed envelopes. The pharmacy of the hospital was making sure each patient would receive the correct intervention based on the codes. Neither the patients nor the physicians were aware about which intervention the patients were receiving. Subjects were followed under similar care circumstances for all diabetic patients from the respective hospitals. The study excluded subjects receiving other medications that alter cholesterol levels (prednisone, estrogen, progesterone, anabolic drugs, and cyclosporine), subjects allergic to Metistatin, obese subjects and subjects for whom transportation for followup would be difficult or impossible. Also, all subjects who entered the study signed an informed consent (the act by which the patient consents to enter the study, acknowledging their rights, all the risks and benefits posed by his/her participation in the study). Exercise 1 Describe the data collected in BD_Ten.xls file according to: Gender Treatment Major vascular event Treatment by gender Major vascular event by gender Quantitative characteristics Central Measures Mean, median, mode Spread Measures Range, Standard Deviation, Coefficient of variation Normal distribution Sqew Kurt Qualitative characteristics Prevalence Incidence Ratios Odd Ratio Risk Ratio Exercise 2 For qualitative variables compute the central measures, spread measures and decide if the data are normal distributed Inference Qualitative variables Chi Square test Quantitative variables T test for means Exercise Define research questions in terms of statistical hypothesis? Respond to the research questions!