Lecture 10 - Key principles in prognosis

Question 1

What are prognostic studies?

Answer 1

• Similar to cohort studies of risk
• Groups of patients with disease are assembled and followed forward in time
• Clinical outcomes are measured
• Conditions associated with a given outcome are prognostic factors

Question 2

What are prognostic factors?

Answer 2

Conditions associated with a given outcome

Question 3

Define prognosis

Answer 3

A prediction of the future course of disease following its onset

Question 4

What is the ideal study type of natural history prognosis and the phases involved?

Answer 4

• Inception cohort (study type)
• Follow-up phase
• Analysis phase

Question 5

What is Inception Cohort?

Answer 5

• Group of people assembled near the onset of disease / diagnosis
• Population-based series (no selection bias)
• Accepted definition of a case

Question 6

What are the characteristics of the follow-up phase?

Answer 6

• Complete follow-up
• Ascertainment of endpoints (uniform surveillance effort / unbiased application of accepted definition)
• Prospective measurement of predictors (avoids bias related to knowledge of outcome)

Question 7

What are the characteristics of the analysis phase?

Answer 7

• Primary hypothesis is stated in advance
• Limited number of secondary hypothesis (you want to limit ‘false positives’ through multiple comparisons and data dredging)
• Make sure appropriate statistical technique is used (stratification, standardization, multi variant, adjustment, sensitivity, analysis)
• Validation using independent data (‘gold standard’ test of predictive models)

Question 8

What are the two approaches for estimating effects of losses to follow-up?

Answer 8

• “5 and 20” rule (fewer than 5% loss probably leads to little bias / greater than 20% loss seriously threatens validity of study)
• Sensitivity analysis (more accurate, particularly when outcomes are infrequent)

Question 9

What is Sampling Bias?

Answer 9

Subjects are not representative of the population being studied, thus findings in those subjects would not be valid in the population being studied.

Example: People who volunteer for a trial of a new anti-smoking drug may be more motivated than the average smoker to quit smoking. Thus, findings amongst this group of unusually motivated people would not apply to the smoking public at large.