Clinical Trials- measurement Flashcards
Outcome measures
-Categorical
-Continuous
Categorical (qualitative) measures
- Nominal: named categories (ex. dead/alive)
- Ordinal: ordered categories (ex. cancer stages)
Continuous (quantitative) measures
1.Discrete: only integers (counts)
- Continuous: any numerical value (rectal temp, blood glucose)
Clinical effect size
Difference between means and medians. Need to determine if it is clinically significant
-evaluate with continuous outcomes. Put into means and medians
-also should have 95% CI to measure uncertainty
ex. Blood glucose of two groups. Find avg in each group. Difference between them= clinical effect size
P value chance vs effect
Tells us chance of getting result BUT does not tell us anything about its actual effect
Pinkeye study in Beef Cattle
-randomized at start; each groups were equal
-P value showing how much you would expect to see by chance.
-no difference between vaccines and placebo= fail to reject Ho
Measurement w/ categorical variables
-Many clinical trials that deal with disease use an outcome measure that is categorical
>Mortality: sick or not; morbidity: dead or alive
What needs to be explored when using categorical variables?
- How to measure the frequency of the health outcome event in trial
- How to measure the magnitude of the effect (impact) of the treatment or interventions
Relative risk
-index of strength of the association between the exposure and the disease (observational study)
vs.
-index of the magnitude of the clinical effect of treatment (clinical trials)
Relative risk eqn
RR= Risk in controls/risk in treated
RR >1
Means controls are at a greater risk of disease and the treatment groups is protected by treatment
RR<1
Means controls are at a reduced risk of disease and treatment may actually make disease worse
RR=1
Means there is no treatment effect
Attributable Fraction
-measure of clinical significance, often used with vaccines
“what proportion of disease in the untreated animals could have been prevented by vaccine/treatment?”
ex. 2 died in vaccine group; 4 died in control
AF=50%… 2 of control likely would have be saved
Absolute risk reduction
Amount by which your therapy reduces the risk of a bad outcome
ARR=RR(control)- RR (treatment)
Number needed to treat
Different way of expressing Absolute risk reduction or risk difference. Looking at the number of patients needed to treat to prevent one additional bad outcome
NNT=1/ARR
Odds Ratio
An estimate of relative risk
-tends to overestimate relative risk
-used more in observational studies, less likely in clinical trials
-Odds of exposure in the diseased group divided by the odds of exposure in the control group
Chi-squared test
-need Ho
-can look for expected values if there is no effect of treatment
-chi squared statistic which is compared with chi square table, to get P value
-if RR is not statistically significant then difference in risk is likely due to chance
Confidence intervals
-Number obtained when conducting a clinical trial a large number of times, that is 2 standard errors of the true mean
>will include true mean or true relative risk 95% of the time
**indicates the level of uncertainty around the measure of clinical effect
CI powers
Small study= low power, wide Cl
Large study= high power= narrow CI
Confidence interval crossing 1
implies no statistical difference for the estimate of relative risk
95% confidence interval
-useful in negative results
-P-value greater than 0.05, it doesn’t tell us anything about the power of the trial. Could be due to trial been too small. Would want CI to be small to indicated high power trial.
**width of CI gives idea of precision of RR estimate
Definition: 95% of the time, true value will lie within those numbers. Upper end will place the treatment as the most favourable
