Thursday - Survival analysis

17 important questions on Thursday - Survival analysis

What is survival analysis about?

When you die
Not only if, but also when event takes place is relevant

What are time-to-event variables?

From starting point till end point (event/failure)

What are examples of survival data?

Medical data
Epidemiology - age at certain infection, time from HIV to AIDS
Demography - time to death or time to birth of first child
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What is typical for survival studies?

Patients are not included at the same time

What happens when data is collected in a survival study?

Data is organized, people who died are indicated with D and people who either completed the study without death or were lost to follow up are indicated with a C (censored)

What is censoring (in survival studies)?

Censoring: precise time of event unknown

--> most relevant is right censoring, event took place at the right of the end of observation

Which method is used in survival analysis?

Kaplan-Meier method or cox regression

How do you make calculations with the kaplan-meier method?

S = p1*p2*...

Where we use the p for time points where the number of people in the study changes.
p = 1 - d/r (1-hazard)
--> d is number of events and r is number of people at risk

How does a typical Kaplan-Meier table looks like?

Time of death
Nr of death
Nr under observation (also censored subtracted!)
Death hazard
Survival hazard
Survival probability

What is important when estimating the median for survival analysis?

Censored people are not taken into account! Median can be estimated with the Kaplan-Meier curve

What is the difference between a local and global test in survival analysis?

Local = at fixed time point test the null hypothesis S(A) = S(B) (two groups compared)

Global = compares two or more time points over whole time range (compare curves) and test whether S(A) = S(B) for all time points

Which test is used in survival analysis for a global test?

Log-rank test

When is it not smart to use a log-rank test?

When hazards cross (crossing curves)

How it the hazard ratio calculated?

HR (t) = h2 (t) / h1 (t) = exp(ß1 + ß2 ...)

-> h1 and h2 are hazard rates
-> we assume that HR is a constant

What is the relation between survival function and hazard ratio?

S = exp(-H) = S0^HR

H = H0*exp(ß1*x + ß2*x etc)

What are differences between KM plot and curve based on cox model?

Cox: baseline hazard is the same for both groups (both curves)
--> curves look highly similar

What to do when you have non-proportional data? (instead of cox model)

-> determine with -log(S) if data is proportional

- different model (complex)
- stratification
- artificial censoring

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo