Evaluating an article on Prognosis

1
Evaluating an article onPrognosis

• Roberto Cardarelli, DO, MPH
• Center for Evidence-Based Medicine

2
Overview Critically Appraising the Evidence

• Prognosis
• Patients How long do I have?
• Treatment decision If I treat the patient, do I
  effect how long the patient lives?
• To screen What happens to the undetected 4 cm
  abd aneurysm?
• These questions have 3 elements
• Qualitative aspect Which outcomes could happen?
• Quantitative aspect How likely are they to
  happen?
• A temporal aspect Over what time period?

3
Prognosis

• Prognosis is an important factor in diagnostic
  and treatment decisions.
• It can affect whether a test is ordered at all,
  the choice of treatment, and when a patient will
  return for follow-up.
• Example, knowing that the prognosis for men over
  70 with a Gleason stage 1 or 2 prostate cancer
  without metastasis is the same as that for the
  general population would discourage you from
  recommending surgery or radiation for that
  patient.
• Studies of prognosis are often (and ideally)
  cohort studies.  In a cohort study, a large group
  of individuals with (and sometimes without) the
  condition in question are followed over time.
• Case-control studies are also used to identify
  prognostic factors, but they are subject to
  additional biases.

4
Determining relevance

• There are three major relevance criteria, true of
  any type of article
• Will this information, if true, have a direct
  bearing on the health of my patients and is it
  something they will care about?
• Is this problem common to my practice?
• Will this information, if true, require me to
  change my current practice?
• To answer the first question, look at the
  outcomes being prognosticated.
• In a study of congestive heart failure, are we
  predicting mortality, or wall motion? Quality of
  life, or hemoglobin A1C?
• Articles which tell us about the prognosis for
  important patient-oriented outcomes are probably
  worth reading, while those which only describe
  intermediate or surrogate outcomes such as lab
  tests or imaging abnormalities are generally not.
• The second and third questions can only be
  answered by the individual practitioner.
• A rural physician in a small southern town may
  have a very different spectrum of patients and
  problems than a suburban HMO physician.

5
Is This Evidence Valid?

• Was a defined, representative sample of patients
  assembled at a common (usually early) point in
  the course of their disease?
• Some patients recover quickly, other die quickly-
  These individuals would be missed.
• This is overcome when patients are included in a
  uniformly early time of disease, ideally when the
  disease manifests (inception cohort).
• If interested in prognosis of late stage disease,
  we then want a representative and well-defined
  sample of patients whom at a similarly advanced
  stage.

6
Is This Evidence Valid?

• Was the patient follow-up sufficiently long and
  complete?
• If too short, we have little to tell our
  patients.
• If long follow-up and few adverse effects, we are
  assured what we tell our patients.
• Complete- are all patients accounted for?
• Why did people drop out (these may be important
  outcomes)?
• What is sufficiently complete?
• 5 and 20 rule
• Sensitivity analysis

7
Is This Evidence Valid?

• 5 and 20 rule
• If
• If 20 lost to f/u Seriously threatens validity
• If 5 and
• Sensitivity analysis
• Worst case vs Best case scenarios
• Out of 100, 4 died and 16 lost to f/u Case
  fatality rate 4/84 (4.8)
• Worst case 20/100 (20)
• Best case 4/100 (4)
• Probably say the study not sufficiently complete.

8
Is This Evidence Valid?

• Were objective outcome criteria applied in a
  blind fashion?
• Cause of death is subjective (death
  certificates).
• Extremes of spectrum are easy to ID Recovered vs
  Death
• In between the extremes is difficult to assess
  residual pain, return to work, etc.
• Specific criteria needs to be setup and applied
  to all patients to minimize bias- to be
  OBJECTIVE!
• Investigators should be kept blind to pts
  characteristics and clinical outcome- to reduce
  bias.

9
Is This Evidence Valid?

• If subgroups with different prognoses are
  identified, was there adjustment for important
  prognostics factors and validation in an
  independent group of test set patients?
• This is for studies that state one group of
  patients have a different prognosis from others.
• Confounders
• Pts w/ A Fib Large vs normal sized atria and the
  risk of stroke
• Pt with larged sized atria already had a high
  of stroke
• Statistical method multiple regression analysis
• Stratified analysis Keep the groups separate
• Tests set Since initial prognostic factors may
  be by chance alone, check the power with another
  set of patients to ensure it is a valid finding.

10
Is this evidence important?

• How likely are the outcomes over time?
• Survival at a particular point in time
• 1-yr 5-yr survival
• Median survival
• Length of follow-up by which 50 of the study
  patients have died
• Survival curves
• At each point of time, the proportion of the
  original sample who have NOT yet had a specific
  outcome.

11

• A Not many patients had events at the end of
  the study.
• Either good prognosis or the study was too short
• B, C, D Serious disease only 20 still alive
  at 1 year
• Note 50 median survival is different for each
• Need sometimes all 3 things to get a full picture
  of prognosis.

12
Is this evidence important?

• How precise are the prognostic estimates?
• We want to determine how much the results could
  vary by chance alone.
• 95 Confidence intervals
• We are 95 sure that the population values lies
  within a range of values
• The narrower- the more precise and sure we are.

13
Can You Apply This Evidence to Your Patient?

• Are the study patients similar to yours?
• Will this evidence make a clinically important
  impact on our conclusions about what to offer or
  tell our patients?

14
Prognosis

• Is this evidence valid?
• Is this evidence important?
• Can you apply this valid, important evidence in
  caring for your patient?

15
Case

• Ms. Helen Goldbloom is a 50-year-old woman in
  your family medicine practice who is in remission
  from localized breast cancer diagnosed ten years
  earlier. Due to a strong family history of breast
  cancer, she was seen by the Medical Genetics
  service, and she was eventually found to have a
  BRCA1 mutation on genetic testing. You also
  follow her daughter Sharon, 29 years of age and
  also carries the BRCA1 mutation, who is in for
  her annual follow-up.

16
Case (continued)

• She is asymptomatic, but on physical exam you
  feel a new firm mobile nontender 2 cm lesion in
  the upper outer quadrant of her right breast with
  no associated skin changes. Exam is otherwise
  unremarkable.
• You refer your patient to a breast surgeon for
  further evaluation of the breast lesion. Fine
  needle aspiration is performed, and cytology
  shows adenocarcinoma of the breast. The patient
  calls you wishing to know whether the fact that
  her cancer is related to the BRCA1 gene puts her
  at higher risk of dying from her cancer than if
  she did not have the gene. You tell her that you
  will research this issue, and see her back in one
  week to inform her of what you have found.

17
Develop your question
Rx
Educational Prescription
Pt Name Ms. Gold bloom-50 Date 9/13/04
Target Disorder Breast Cancer
Intervention (/- comparison) /- BRCA 1 gene
mutation
Outcome Higher risk of dying from Breast CA
Discuss Search strategy Search
results Validity Importance of the valid
results Can you apply this to your pt
18
Review the Article