How do you improve the reliability of RCTs
(6 answers)
1) Randomised Treatment Groups (can be after splitting sample group into different stratas if you want to compare within certian prognostic variables)
2) Placebo
3) Blind Clinical Team (so they dont know which group the patient belongs to)
4) Pay attention to the reasons for any withdrawal from trial
5) Ensure trial size is sufficient to reach calculated statistical power. (Also, should stop trial once sig results is attained on basis of ethics)
6) Crossover studies (each patient serves as their own control)
What is Lead TIme
The time between early detection of a disease (usually via screening) and the presentation of a diseae (symptomatic)
Why is Lead Time Bias important?
When assessing if screening for a disease positively impacts morbiidty and mortality, we must aim to show that it truly improves survival (and this improvement in survival time is MORE THAN the lead time)
i.e It is far more important to see if early detection of a disease (reduce lead time) affects morbidity and mortality, vs just getting an earlier diagnosis.
E.g Huntington Dz - people generally die at age 65, typically symptoms appear at 50yrs. U can get a genetic test at birth to diagnose it but it does not prolong their survival (lead time is 50yrs)
How do we ensure that we accurately assess the impact/benefits of screening?
1) Take into account lead time bias
2) Measure outcome in all of the population selected for screening and not only in those members who actually undergo investigation
3) A satisfactory diagnostic test must be available (must detect cases in sufficient numbers and at acceptable cost, and it must not carry side effects that outweigh the benefits of screening)
What is positive predictive value?
True Predictive value of a positive result - is the probability that a person who reacts positively to the test actually has the disease. (True positive, PPV = A/(A+B))
What is the yield of a screening service
it is the number of cases identified whose prognosis is improved as a result of their early detection.
For breast cancer screening it has been found that identifying one case requires examining 170 women by palpation and mammography and taking nine biopsy specimens
What is the sensitivity of a test
Sensitivity (true positive rate) proportion of true positives, correctly identified.
Sensitivity = A /(A+C)
*Picture has slight error
What is the specificity of a test
Specificity (true negative rate) proportion of true negatives, correctly identified.
Specificity = D/B+D
What is the Negative Predictive Value
True Predictive value of a negative result - is the probability that a person who reacts negative to the test actually does not have the disease. (True negative, NPV = D/(C+D))
What is a type 1 error
False positive
What is a type 2 error
False negative
What is a Case Control study
2 comparison groups, one with disease and one without, usually used to investigate disease outbreak/rare diseases
What is a cohort study
2 comparison groups chosen from a population group who have NOT YET had the disease. Good for extimating relative risk or study dose response ratios. Hard to do if disease incidence is low
What is Relative RIsk?
Ratio of the incidence of the disease in those exposed to that of those unexposed to the risk factor.
Attributable RIsk
Absolute Risk that can be attributed to risk factor.