Survival Analyses Flashcards
Survival Analyses
Model ‘TIME TO EVENT’ data
Directly incorporates time into the analysis to show level of decay over time
Survival Data are inherently censored meaning what?
Not fully observed, the time to an event is unknown
When is survival analysis useful?
When follow-up is either incomplete or variable
Examples of Censoring are what?
- Loss to follow-up
- Study withdrawal
- An event occurs outside of the study period
Left Censoring
Before the study begins
Right Censoring
After the study ends
Interval Censoring
Missing observations during the study period itself
What type of outcome is seen in Survival Analysis?
Binary/Dichotomous
Yes/No outcome
Do Linear and Logistic Regression include censoring?
NO
What is Hazard Ratio?
First derivative of survival function, captures instantaneous slope
What is Survival Function?
Kaplan Meier: what is reported in the RCT
1:1 Comparison
What is Hazard Function?
Instantaneous risk of event at a certain time
-Can be used when non-randomized aka controlling for a variable so the data will not be skewed
Survival Analysis overall can be used for what?
- Account for censoring
- Compare survival times between 2 or more groups
- Assess relationships between hazard ratios and several covariates
What can be involved in Survival Analyses?
- Descriptive Statistics
- Bivariate Statistics
- Multivariate Statistics
Descriptive Analysis includes what?
- Average Survival
- Average Hazard Rate
What are the two most common methods in estimating the Survivor Curve?
- Kaplan Meier KM
- Cox Proportional Hazards Regression (COX Regression)
Kaplan Meier KM Method
The KM survival curve is a graphical method of summarizing the probability of survival over time estimated from a sample
GENERATES A STAIR STEP survival curve
Inferring KM
Separate KM curves can be estimated so statistically significant differences between the groups can be calculated via LOG RANK TEST (type of chi-square)
Log Rank Test
Used to compare two survival curves
If there is a significant difference, one group will be viewed as having significantly greater survival times
What is the main limitation of the Log Rank Test?
UNIVARIABLE, meaning it does NOT account for confounding by other covariates or effect modification
Is confounding by other covariates a problem in randomized clinical trials?
NO, randomization is intended to uniformly distribute covariates across groups
What is the most common multivariable extension of the log-rank test?
COX Regression
Does COX Regression produce a stair step survival curve?
NO
COX Regression is the MAIN method for RCT trials since it is the most common what?
Multivariable extension of the log rank test
COX Regression builds on what?
Builds upon the concept of a hazard
Hazard is defined as what?
The risk that a specific event will occur at any given time
Hazard Rate
Number of events that occur per interval of time
Hazard Function
Collection of an individual’s hazard for an event over time
COX Regression mathematically what?
Separates the baseline hazard function from time-independent covariates
The KEY assumption for COX Regression is what?
Proportional Hazards, must test if the assumption holds via various statistical diagnostics
Proportional Hazards
Predictors have a constant proportional effect on the outcome
-If the assumption is violated, it means a given predictor usually has a time dependent proportional effect
Strengths of COX and KM
- A formalized multivariable approach to incorporate time-dependent censoring
- COX regressions can incorporate several covariates
- Hazard ratios are a type of relative risk measure
Hazard Ratio Meanings
- Hazard Ratio <1 = increased predictor leads to decreased hazard
- Hazard Ratio = 1 = no statistical difference
- Hazard Ratio >1 = increased predictor leads to increased hazard
Limitations of COX and KM
- Does not automatically accommodate predictor variables that change over time
