Rani Gereige, M.D., MPH

1
Appraisal of Studies on Prognosis Doctor How
Long Do I Have to Live???

• Rani Gereige, M.D., MPH
• Associate Professor
• University of South Florida

2
Learning Objectives

• Be able to assess the validity of a study about
  prognosis
• Be familiar with the type of study designs used
  for prognosis
• Be able to critically appraise an article about
  prognosis
• Be able to use the evidence of prognosis to make
  treatment and counseling decisions

3
Example

• A 28 weeks preemie with Grade II IVH
• What are possible prognosis questions the parents
  of this baby might have?
• What are prognosis questions YOU as the provider
  might have?
• And where do you begin to answer these questions?

4
What is Prognosis??

• The possible outcomes of a disease and the
  frequency with which they can be expected to
  occur, over what period of time
• Qualitative aspect What might happen?
• Quantitative aspect How likely it might occur?
• Temporal aspect Over what period of time?

5
Prognostic Factors

• Characteristics of a particular patient that can
  be used to more accurately predict eventual
  outcome.
• Demographic (age)
• Disease-Specific (Tumor stage)
• Comorbid conditions

6
Prognostic Versus Risk Factors

• Prognostic factors dont necessarily cause the
  outcome, just have a strong enough association to
  predict the development of the outcome, which
  patients do better or worse
• Different from Risk Factors - patient
  characteristics associated with the development
  of the disease in the first place.

7
Example

• IVH prognostic factors
• Grade of the IVH
• Age at occurrence
• IVH risk factors
• Gestational age
• Mechanical ventilation pressures

8
Example

• STROKE
• Age (younger patients may fare better)
• Disease-Specific Variables (Hemorrhagic versus
  Thrombotic)
• Co-Morbid Factors (Those with hypertension may
  fare worse, even if treated)
• Prognostic or Risk Factors???

9
Example

• STROKE
• Age (younger patients may fare better)
• Disease-Specific Variables (Hemorrhagic versus
  Thrombotic)
• Co-Morbid Factors (Those with hypertension may
  fare worse, even if treated)
• Prognostic Factors

10
Knowledge of Prognosis is Important to Who?

• Clinicians
• Helps clinician make the right diagnostic and
  treatment decisions
• Organizations
• Broader issues beyond the individual patient
• Patients
• Counseling patients

11
Studies About Prognosis

• Best is a systematic review of several prognosis
  studies

12
Individual Study Design

• Best Study Design
• Cohort Study
• It is usually impossible or unethical to
  randomize patients to different prognostic
  factors.

13
Study Design

• Ideal Cohort Study
• Well-defined sample of individuals representative
  of the population of interest
• Objective outcome criteria

14
Study Design

• Randomized Trials
• Rigorous randomized trials can generate
  information about prognosis (Two cohorts The
  treated cases and untreated controls)
• But
• patients entered into the trial are often not
  representative of the population with the disease

15
Study Design

• Case-Control Studies
• Potential for bias in selecting both cases and
  controls- selection bias
• Retrospective data collection about prognostic
  factors (memory/chart accuracy issues)
• Biases Selection, measurement, recall.
• Cannot provide Absolute Risk information, only
  Relative Risk
• Can be useful when the outcome is rare, or the
  required duration of follow-up is long

16
Back to Our Patient

• Study design
• You are able to find a study that prospectively
  followed a cohort of 1000 babies with grade II
  IVH from the time of diagnosis forward looking at
  their prognostic indicators
• How could the population in this study be refined?

17
Is This Evidence About Prognosis Valid?

• Was a defined, representative sample of patients
  assembled at a common (usually but not
  necessarily early) point in the course of their
  disease?
• Ideally- Entire population who developed the
  disease
• Study sample fully reflect the spectrum of
  illness?
• INCEPTION COHORT Ideally when disease becomes
  manifest (except if we want to look at prognosis
  of late stages of disease)
• Look for inclusion and exclusion criteria, look
  for filters for patients to get to the study

18
Is This Evidence About Prognosis Valid?

• Was patient follow-up sufficiently long and
  complete?
• Follow-up length
• Short ? Few patients with outcome (not enough)
• Long ? Worry about loss to follow-up reasons
  (unavoidable? Unrelated to prognosis?
  Death/illness?)
• The greater the number of patients unavailable
  for follow-up, the less accurate the estimate
  regarding the risk of the adverse outcome

19
Is This Evidence About Prognosis Valid?

• Was patient follow-up sufficiently long and
  complete?
• Follow-up completeness
• Ideally ALL inception cohort patients followed
  till recovery or development of outcome
• Two ways to judge completeness
• The 5 and 20 rule
• lt 5 loss ? Probably little bias
• gt 20 loss ? Threat to validity
• Between 5-20 ? Intermediate
• Sensitivity Analysis Series of What if
  questions, worst and best case scenarios

20
Sensitivity Analysis
Assuming a study followed 100 women with breast
cancer
4 died
16 lost to follow-up
80 complete f/u
Crude case-fatality rate (CFR) 4 deaths/84 with
data 4.8
Worst case scenario All 16 lost died CFR 20
deaths/100 20 (Lost added to Num and denom)
Best case scenario All 16 lost did not die CFR
4 deaths/100 4 (Lost only added to denom)
21
Loss to Follow-up

• If the number of patients lost jeopardizes the
  validity of the study
• Look for the reasons for unavailability
• Compare important demographic characteristics of
  available vs unavailable patients. If they are
  similar, it may be less of a problem.
• If information about reasons for unavailability
  is not provided, the strength of inferences from
  the study are weakened.

22
Is This Evidence About Prognosis Valid?

• Were objective outcome criteria applied in a
  blind fashion?
• What constitutes an outcome?
• Extreme outcomes (death recovery) are easy to
  define
• In between outcome can be difficult to define
• Did investigators have specific objective
  criteria for each outcome?
• Are criteria used on ALL patients
• Are users of the outcome criteria kept blind?

23
Outcome Criteria

• Clear and sensible definition of adverse outcomes
  before the study starts
• Outcome events can be objective or can require
  limited or considerable judgment
• To minimize bias, individuals determining outcome
  should be blinded to whether patient has the
  prognostic factor in question. Not necessary
  with entirely objective outcome (e.g. death)

24
Is This Evidence About Prognosis Valid?

• If subgroups with different prognoses are
  identified, was there adjustment for important
  prognostic factors and validation in an
  independent group of test set patients?
• Reports that claim that one subgroup has a
  different prognosis from others.
• Adjustment can be done either (results section)
• The simple way (stratified analysis)
• Fancy way (Multiple regression analyses)

25
Is This Evidence About Prognosis Valid?

• A newly identified prognostic factor does not
  guarantee that it holds true in a similar
  subgroup
• Was there adjustment for important prognostic
  factors?
• Investigators should consider whether the
  clinical characteristics of the groups is similar
  and adjust the analysis for any differences found
• Investigators should adjust for differences in
  treatment
• Training set Derivation set Initial patient
  group where the prognostic factor was found
• Test set Validation set subsequent
  independent group of patients
• Method section
• Prestudy intention
• Second independent study

26
Assuming You Found the Study to be Valid
27
Is This VALID Evidence About Prognosis Important?

• How likely are the outcomes over time?
• Three ways of reporting it
• Survival at a particular point in time (1 year
  or 5 year survival)
• Median Survival (Length of F/U by which 50 of
  the study patients have died)
• Survival Curves/ Kaplan-Meier Curve ( of study
  population at each point in time that is free of
  the specified outcome)

28
Is This VALID Evidence About Prognosis Important?

• How precise are the prognostic estimates?
• Study is done on a sample
• Look at 95 C.I.
• Usually in text, tables/ graphs, or can calculate
  it
• The narrower the better, the more precise the
  estimate
• Usually C.I. Are narrower earlier in the study
  due to loss to follow-up

29
Survival Curves
100
100
50
50
20
0
0
A.
B.
6
9
12
6
9
3
3
12
100
100
50
50
20
20
0
0
C.
D.
3
12
12
6
6
9
9
3
Months of follow-up
Months of follow-up
30
Survival Curve
31
(No Transcript)
32
Exercise What is the Median overall survival
time? What is the 5 year relapse-free survival
rate? What is the 2-year relapse free mortality
rate? Which patients have a better prognosis
(Overall and relapse-free survival)? HER-negative
or HER-positive breast cancer patients?
33
6
34
Can We Apply This VALID, IMPORTANT Evidence About
Prognosis to Our Patient?

• Are the study patients similar to our own?
• Look at demographics and sample characteristics
• Will this evidence make a clinically important
  impact on our conclusions about what to offer or
  tell our patient?
• Good prognosis if untreated ? Discuss no Tx
  option
• Poor prognosis if untreated ? More likely to
  treat
• Valid evidence is always a good source of
  information

35
5
36
Examples
37
Example 1

• Children admitted with febrile seizure
• Parents want to know the risk of having more
  seizures in the future
• From reviewing the literature
• Risk in population-based studies 1.5-4.6
• Risk in clinic-based studies 2.6-76.9
• What are possible reasons for the wide difference?

38
Filter Bias

• Children in the clinic-based studies may have
  other neurologic problems predisposing them to
  recurrence
• Assuming your patient has no neurologic
  problems, which risk do you believe?

39
4
40
Example 2 - Framingham Study

• Study findings Rate of stroke in patients with
• Atrial Fibrillation fibrillation Rheumatic
  Heart Disease 41 per 1000 person-years
• Atrial Fibrillation but without Rheumatic Heart
  Disease is very similar
• However, it was noted that patients with
  Rheumatic Heart Disease were on average much
  younger than those without Rheumatic Heart Disease

41
Correction

• Consider the risk in young people separately from
  that of older people with versus without RHD
• Once adjustment for age was done (also for gender
  and HTN status), rate of stroke was six fold
  greater in patients with RHD and Atrial
  Fibrillation than in patients with Atrial
  Fibrillation who did not have RHD

42
3
43
Example 3 - Alzheimers disease

• The son of an elderly woman asks What are the
  chances that my mother will still be alive in 5
  years?
• A high-validity study found
• Prognosis In patients with dementia, 5 years
  after presentation to clinic, 50 died (5050)
• Son asks how come he knows his uncle who is 65
  year old who was diagnosed 10 years ago and still
  living? He is surprised that his mothers chance
  is so high.

44
Alzheimers Disease

• Based on the high-validity study patients with
  dementia died earlier if they were
• Older pats
• With severe dementia
• With behavioral problems
• With hearing loss

45
2
46
Heart Murmur

• A 45 year old woman new to your practice
• She is well, on no medications, unremarkable PMH
• Cardiac exam reveals a murmur suggestive of
  Mitral Valve Prolapse (MVP) with Mitral
  Regurgitation (MR)
• Remaining exam unremarkable
• Wondered if she should be concerned

47
Heart Murmur Study

• Inception cohort with asymptomatic MVP
• F/U 97 at 5.4 years
• Outcome Mortality and cause of death
• Findings
• After adjusting for age, sex, and comorbid
  conditions, moderate-to-severe MR and EFlt50 were
  found to be independent predictors of
  cardiovascular mortality.
• Median F/U of 5.4 years, mortality was 11.5.
  Mod-to severe MR (Hazard ratio 1.8, 95 CI
  1.03-3.0) and EF lt 50 (Hazard ratio 2.3, 95 CI
  1.05-4.4) were independent predictors of CV
  mortality

48
Back to Your Patient

• Transthoracic echo showed MR and EF gt 65
• Given her age (lt50 years old) and absence of
  other prognostic factors, we can reassure her
  that she is low risk for mortality and CV
  morbidity and her outcome (with respect to MVP)
  is similar to the general population

49
1
50
The End!!!
or is it??