1. Lecture 5: The Natural History of Disease: Ways to Express Prognosis
Reading:
Gordis - Chapter 5
Lilienfeld and Stolley - Chapter 10, pp

2. Introduction
How can we characterize the natural history of disease in quantitative terms
That is: what is the prognosis?
Problems in defining disease
Determine when the disease begins
Histological confirmation
Determine stage of disease

3. Introduction
Screening tests and diagnostic tests characterize people as sick or well
Once diagnosed as sick – the question is:
How sick and what duration
cure or death

4. Introduction Quantification is important because:
Knowing severity is useful in setting priorities for clinical services and public health programs
Patients want to know the prognosis
Baseline prognosis is useful when evaluating new therapies

5. Prognosis
Prognosis can be expressed in terms of deaths from the disease or survivors with the disease.
Ways to express prognosis:
Case-fatality rate
Five-year survival
Observed survival rate
Life table analysis
Kaplan-Meier method
Median survival time

6. Case-fatality rate Case-fatality rate =
Number of people who die from the disease
Number of people with the disease
Given that a person has the disease
what is their risk of dying from that disease
Different than mortality rate (how?)
Case-fatality often used for acute diseases of short duration
In chronic disease, death may occur many years after diagnosis and the possibility of death from other causes becomes more likely.

7. Case-fatality rate: example
1000 new recruits get infected with disease X over a 15 day period
10 die within 5 days of diagnosis
Case-fatality rate: 10/1000 = 1%

8. Person-years of follow-up
Incidence rate
using person-time for denominator
Because with chronic diseases
the diagnosis are not clustered around a single event (like an industrial exposure)
Follow-up may differ and these differences can be “adjusted” by using person-time in the denominator

9. Person-years of follow-up
Assumptions with incidence rate:
Prognosis is the same over the entire follow-up period
That is:
Following 5 people for 2 years
will give the same information as
following 2 people for 5 years

10. Person-years: example

11. Five-year survival
Percent of patients who are alive 5 years after diagnosis.
Nothing magical about 5 years
Most deaths from cancer occur during this period (historically)
Convenient
However, changes in screening may affect the time of diagnosis

12. Five-year survival
Comparing 5-year survival among groups is only informative if the individuals began at a similar stage of disease
The interval between diagnosis and death may be increased not because of better treatment but because of earlier diagnosis
Lead time bias

13. Five-year survival
What if we want to examine the effects of a therapy that was introduced 2 years ago.
Do we wait for 5 years so we can use the 5-year survival rate?
We use life table analysis

14. From person-years example
In the previous example:
10.53 cases /100 PY / over 5 years
or 2.1 cases / 100 PY / per year?
13 cases / 100 PY / over 5 years
or 2.6 cases / 100 PY / per year?
8.55 cases / 1000 PY / over 5 years
or 1.7 cases / 100 PY / per year?

15. Life table analysis
Previous to now – when we used follow-up time we were describing the RATE at which disease occurred.
How do we assess the RISK of disease development using follow-up time?
Without making the assumption that risk is the same across all strata of time

16. Life table analysis
Calculate the probabilities (risks) of surviving different lengths of time
Using all of the data available
If follow-up is complete:
the easiest way is using the cumulative incidence
Follow-up is usually NOT complete
Therefore: LIFE TABLES and Kaplan Meier

17. Life table analysis without withdrew
Cumulative proportion surviving = Pr(survival time t) =
Pr(survival time t | survival time t-1) x Pr(survival time t-1)
So:
0.81 = 0.9 x 0.9
0.73 =0.901 x 0.81
0.66 = x 0.73

18. Life table analysis with withdrew
Withdrew or loss to follow-up
Effective number at risk = alive at beginning – ½ x withdrew So
375 = 375 – 0
175.5 = 197 – ½ x 43 …

19. Kaplan-Meier method
In the life table analysis, we predetermine the intervals (e.g., 1 year).
Kaplan-Meier method identifies the exact point in time when each death occurred
Each death determines the interval

20. Kaplan-Meier method: example
Patient 1
Patient 2
Patient 3
Patient 4
Patient 5
Patient 6
died
loss to follow-up
died
died
loss to follow-up
died
100 80 60 40 20
Percent
surviving

21. Life table analysis Assumptions in using Life Tables
No secular (temporal) change in the effectiveness of treatment or in survivorship over calendar time
Survival experience of those lost to follow-up is the same as the experience of those who are followed

22. Median survival time
The length of time that half of the study population survives
Two advantages over mean survival
Less affected by extremes (outliers)
Can be calculated before the end
To observe the mean survival – we need to observe all of the events

23. Generalizability of survival data
The cohort must be at a similar stage of disease
Patient data from clinics or hospitals may not be generalizable to all patients in the general population
Referral patients may not represent all sick individuals