click below
click below
Normal Size Small Size show me how
EBHC definitions
Evidence Based Health Care Definitions
Question | Answer |
---|---|
CONSORT statement | "Consolidated Standards of Reporting Trials"; a checklist of info that should be included in publications of clinical trials; adopted by many medical journals; allows for standardization; |
which section of an RCT requires the most critique, as it could contain a fatal flaw, invalidating the results? | Method's |
Null hypothesis | statement of no difference |
Describe the 3 members of an IRB (investigational review board) | 1- representatives from various health disciplines; 2- specialists in a nonscientific area (law/ethics); 3- Lay person not affiliated w/institution (most important); |
Stratification | a sophisticated randomization procedure designed to achieve similarities in known and unknown baseline patient characteristics between the groups; |
single blinded | patient or researcher is blinded (doesn't know who's got rx vs placebo) |
Double blinded | both patient and researcher are blinded |
Triple blinded | patient, researcher, and 3rd group (ex- radiologists interpreting results) are all blinded; |
Double-Dummy | a term used when more than 1 placebo is needed to perform blinding; |
when is blinding the most important | the more subjective the end point, the more important blinding is. |
Placebo-controlled trials | When placebo is used to determine efficacy of a drug compared to the natural course of disease; often used to get a drug FDA-approved; |
Active-controlled trials | determine the efficacy of a drug compared to another drug; used when it is unethical to do nothing, or when comparative efficacy is desired; |
Noninferiority Design | shows that a difference between 2 drugs is small enough to conclude that the test drug has an effect not too much smaller than the active control. Design used when superiority is not anticipated; design and interpretation are different from normal RCT's; |
Parallel design | when group 1 gets the actual drug until the trial end, while group two gets the placebo until the trial end; |
Cross-over design | when group 1 and two both get the drug and placebo during a trial, but at different times; |
Advantage of cross-over design | each patient serves as his/her own control; won't have a table of baseline characteristics; eliminates concern of complete randomization, but introduces a new obstacle of timing, and therefore can't be used with diseases that get better/worse with time; |
Placebo Run-in | a period of time inwhich all patients receive a placebo before the active vs placebo phase begins; helps to screen out placebo-reactors (benefit from placebo alone) and non-compliant patients; |
Washout period | period of time during which no drug for the disease to be evaluated is given; used at the start of any trial or in the middle of a cross-over trial; gets rid of other drugs that may be in the system; |
Surrogate Markers | laboratory measurements or physical signs that directly measure how a patient feels, functions, or survives; ex- cholesterol, HbA1c; Changes are expected to reflect changes in clinically significant endpoints; |
Composite endpoint | a clinically meaningful endpoint that includes more than 1 event; it's usually the primary endpoint of event driven trials; used because individual events occur in small numbers; |
Power | ability to accurately detect differences between groups; describes if the sample size was large enough to detect small differences; should be done prior, but can be done after the study; anticipate dropouts; = 1-beta; |
alpha | the probability of a false positive; typically set at 5%; |
Beta | probability of a false negative; typically set at 20%; = 1-power; |
Delta | the amount of difference in order to be statistically significant; variable, and dependent on the primary endpoint; the smaller the delta, the larger the sample size needed; |
false positive | finding a statistically significant difference that is not due to the drug, but due to chance; equal to alpha (5%); |
false negative | finding a real difference that is not statistically significant, due to chance or small sample size; probability is equal to beta (20%) |
P-value | indicates whether a difference is statistically significant; determined by alpha; P should be less than alpha; |
Meaning of P-vallue < alpha | statistic is significant, and not due to chance (true difference); |
Meaning of P > alpha | statistic is not significant; due to chance; |
Intention to treat analysis | Analysis of all subjects in a trial regardless of whether they completed it or not; estimates effectiveness (how drug works in the real world) |
Per Protocal analysis | analysis of only patients who complete the entire trial; estimates Efficacy (how drug works under controlled conditions) |
Modified intention to treat | an intention to treat analysis that is modified however described, but comonly "involving all patients with at least 1 post-baseline measurement" |
nominal data | catagorical data; ex- MI, death, etc |
Ordinal data | data that is on an arbitrary scale that reflects ranking; ex- pain scale, or new york heart association classes for HF; |
Continuous (interval or ratio) | data that is on a scale in which measurements have an equal distance between points; ex- Blood pressure, cholesterol, weight; |
When to use parametric tests | when the endpoint is normally distributed (bell shaped) |
When to use nonparametric tests | when the data is not normally distributed (not bell shaped) |
When does the distribution (bell curve) apply? | When the data is continuous |
one tailed test | used when the direction of the relationship is known. |
two tailed test | used when the direction of the relationship between variables is not known; requires a stronger relationship to acheive statistical significance (preferred method); |
survival analysis | analysis for nominal endpoints that may occur at any point in time; different because time prior to event is considered (ex- survival time); patients who drop out or who don't experience are "censored"; |
Kaplan-Meier curve | a survival analysis that presents as a graph of cumulative survival over the trial period; |
Log-rank test | most common statistical test for "survival time" and "2 independent groups" |
Central tendencies | a way of reporting data/results including: mean, median, mode; |
Mean | the average of a data set |
Median | the mid point of a data set |
mode | the most common data point |
Range | a type of variability that describes the smallest to largest data points |
interquartile range | difference in scores at the 75th and percentiles; describes variability; |
Standard deviation | degree in which data points deviate from the mean; 68% describes 1 standard deviation; |
Ways to evaluate nominal data (central tendency & variability) | mode |
Ways to evaluate ordinal data (central tendancy & variability) | median, mode, range, interquartile range |
Ways to evaluate continuous data (central tendancy & variability) | mean, median, mode, range, interquartile range, standard deviation |
Confidence interval | a numerical range that contains the true value of the population a certain percentage of time; estimate of what would happen in the entire population; typically 95%; |
Interpretation of confidence intervals | when CI is crosses 0 or 1 (below or above), p-values are not statistically significant; when CI does not cross 0 or 1 (below or above), p values are statistically significant; |
relative risk | ratio of risk of an event occuring in one group compared to another group (usually placebo); |
Interpretation of relative risk (RR) | if RR < 1, drug decreases the risk of an event; If RR > 1, drug increases the risk of an event; If RR = 1, drug has no impact on the event; |
Relative risk formula (RR) | % of risk from the drug divided by the % of risk in the placebo population; want the number to be lower; |
Hazard Ratio (HR) | used in place of relative risk; used with survival data; want the number to be lower; |
Odds Ratio (OR) | used in place of relative risk; used in observational studies; want the number to be lower; |
Absolute risk reduction (ARR) | The percentage of patients who were spared the event because they received the drug; |
Formula for absolute relative risk (ARR) | % placebo that experienced the event - % drug that experienced the same event; want this number to be higher; |
Relative risk reduction (RRR) | the percentage of baseline risk that is removed as a result of the drug; percentage less likely that a group was to experience the event because they received the drug; |
Formula for relative risk reduction (RRR) | 1-RR x 100; want this number to be higher; |
Number needed to treat (NNT) | the number of patients who required treatment with the drug to prevent one event; time course must be included; |
Formula for number to treat (NNT) | 1/absolute risk reduction (ARR) as a decimal; ex- if ARR is 4%, (1/.04), NNT is 25 patients; want this number to be lower; |