What is the purpose of a longitudinal or follow-up study?
To characterize how an outcome is acquired and how it evolves over time.
What things about follow-up studies makes statistical analysis difficult?
1) Key outcome is the time between two events. 2) Patients begin study at different times. 3) Patients followed for different length of times 4) At time of analysis, outcomes will remain unknown in many patients.
What is the key outcome in follow-up studies?
The time between the starting and ending events.
What is a censored patient?
One who was either lost to follow-up or still did not have the outcome by the end of the study.
Why can't you just average all of the survival times of the participants in you follow-up study? Would the median work?
An example is the participant who joined later on did not only live 0.7 years, he was only in the study for 0.7 years. The mean survival value will actually give a lower number than the reality. Median only works if the lower half of your population are not censored patients, otherwise you don't know who lived longer and thus don't know where to order them to determine a median.
What value is fairly useful in determining incidence in survival data?
Person-years approach. Confirmed outcomes/ person-years to follow-up. This allows for censored observations to contribute to the denominator.
What assumption must be met when you use the person-years approach to survival data?
You assume that the risk of the outcome is constant over time.
What are the modern approaches to summarize survival data?
Life tables (Kaplan-Meier for exact data and Actuarial for huge data sets). They incorporate data on all subjects, they adjust for differing lengths of follow-up and do not require a constant outcome risk over time.
What does the Kaplan-Meier table calculate?
Risk of the outcome at each time point
Who is at risk at each time point in the Kaplan-Meier approach?
Subjects are at risk up to and including time of death. Subjects who drop out between two death times are incorporated into the calculation of the first death.
What are the three basic assumptions of the Kaplan-Meier life table?
People drop-out for random reasons (not because they are at greater or lower risk), risk for early recruits = risk for later recruits (treatment didn't change) and that events happen at the exact time indicated.
What is the three year survival of the subjects in this curve? At what point is 50% of your population still surviving?
At three years, 90% of your population is still surviving. 50% of your population will be surviving by about 4.5 years.
What is different about these two lymphomas?
DLBC lymphoma has rapidly decreasing survival until about 2 years, then you're good to go. Follicular lymphoma survival slowly, but steadily decreases over the years.
How should you compare different groups in survival curves?
Across the entire curve, not at particular points.
How do you determine if the data you got from a survival curve is statistically significant?
Log rank tests. These will tell you if each curve is significant via a p-value.
In what setting would you utilize Cox proportional hazards regression?
To get a summary of risk between two groups over time, to adjust for other variables and to analyze categorical factors in survival analysis.
What is a hazard ratio?
The instantaneous risk in exposed group/ hazard in unexposed group. This gives you the amount of risk for the outcome that increases or decreases for an increase or decrease in one unit of your exposure.
What does a hazard ratio of 2 mean when the outcome is death?
A patient in one group at a certain time point is 2x as likely to die at the next time point as a patient in the other treatment group.
Is the hazard ratio consistent over time in this data set?
No. The curves have a large divergence at different times that is not a constant hazard ratio.
Why do people who have had a broken hip have a greater chance of living longer?
Immortal time bias. They have lived long enough to break a hip.
Why did the invention of nuclear scanning increase median survival of cancer patients?
It allows us to find metastasis when they are smaller. This took the most sick people in the localized cancer group and gave them to the metastatic group where they are now the healthiest. This increases survival in both groups and is called stage migration.