Evaluating Prognosis Flashcards
What studies are used to estimate prognosis?
Cohort studies
RCTs
X and Y axis to evaluate prognosis
X: Time
Y: proportion surviving i.e. not attaining an ‘event’
What is displayed at each point on a graph?
Cumulative probability of an individual remaining free of event/outcome
What is the name of the curve for evaluating prognosis?
Kaplan-Meier Survival Curves
What does it mean if observations are censored?
Not all subjects will have reached the outcome by that period of time
What does left censored mean?
Observations began after some events have occurred (missed past events)
What does right censored mean?
Observation was terminated before events could be observed (missed future events)
Why can we not calculate the mean survival time from studies?
Due to censoring we do not know the actual survival times beyond the end of the study period
What is the median survival time?
The survival time of 50% of observed sample
How can one compare survival rates of two samples?
Log-rank test
What type of test is log-rank test?
Non-parametric
What can log-rank test not be used for?
To account or adjust for effects of other variables
How does one study multiple variables in survival studies?
Cox regression analysis
What type of test is cox regression analysis?
Semiparametric
Why is cox regression analysis called proportional hazard
It assumes that the effects of multiple variables are constant over a period of time
What does log-rank test inform us?
Whether significant differences are present between two survival curves
Does not measure magnitude
What tells us the size of difference between two survival curves?
Hazard ratio
What can Mantel-Haenszel test be used for in survival studies?
To study effect of two binary independent variables on outcome
Method used if one categorical IV and continuous DV
Kaplan-Meier curve with log-rank test
Method used if >1 categorical IV and continuous DV
Cox’s proportional hazard
Method used if 2 categorical IV and continuous DV
Mantel-Haenszel test
Questions to ask re validity of results in a prognostic study
Was sample representative?
Were patients similar to each other with respect to prognostic factors?
Was follow-up sufficiently complete?
Were outcome criteria objective and unbiased?
How to determine if sample of patients is representative
If they go through a filter, likely they will be different from population of interest
How to test if patients were similar to each other with respect to prognostic factors
If study includes patients at different stages of the disease, survival estimates of the group will not be applicable to each group member
Questions to ask about the results in prognostic studies
How likely are outcomes over time?
How precise are the estimates of likelihood?
How does one analyse the likelihood of outcomes over time?
Kaplan-Meier curve
How does one analyse the preciseness of estimates of likelihood>
Survival curves are more precise in earlier periods, indicated by narrower confidence bands around left-hand parts of curve
Why are survival curves more precise at earlier periods?
Due to losses to follow-up
Questions to ask when considering if results from prognostic study can be applied to patient care
Were study patients and clinical context similar to mine?
Was follow-up sufficiently long?
Endpoints of survival curves
Death during trial Loss to follow-up Still alive at end (right censor) Adverse event Discontinuation of treatment
How long does recruitment take for prognostic studies?
2-3 years
At what point in a prognostic study is there likely to be significant attrition?
If minimum follow-up is 5 years
What is the focus in survival curves?
Time to end-point
What does poor median survival mean in a survival curve?
Worse early prognosis but better late prognosis
What does representative median survival mean?
No comparable differences in early and late periods of disease
What does impressive median survival imply?
Good early prognosis but poor later prognosis
