Statistics for the board

Document Sample
Statistics for the board Powered By Docstoc
					Statistics for the board

    September 14 th 2007
 Jean-Sebastien Rachoin MD
 Rule #1 : Statistics is cool
 Rule # 2 : Statistics is easy
 Rule # 3: P value is not every thing
 Statistics for the board
 Statistics questions : 2 -3 / modules
 Cardiology question 31 total questions
 GI questions 28 questions
 Heme-onc questions 27 questions
                    For the board

                       Disease    Disease
                       Positive   Negative
     Test +                  A       B
      Test -                 C       D

 Sensitivity : (A) /(A+C)
 Susceptibility :( D) / (D+ B)
       Classical Board question
 A new screening test is being developed by
  Amgar . Inc for early detection of CKD.
Which of the following assays would be
A- Sensibility 90 % Specificity 20 %
B-Sensibility 80 % Specificity 50%
C-Sensibility 70 % Specificity 60 %
D-Sensibility 60 % Specificity 90 %
 Amgar inc. is developing a test to confirm HIV
  status in patients with a positive ELISA test.
  Whish assay should be used?

A- Sensibility 90 % Specificity 20 %
B-Sensibility 80 % Specificity 50%
C-Sensibility 70 % Specificity 60 %
D-Sensibility 60 % Specificity 90 %
 A few years ago the AUA decided to change the
  threshold for Abnormal PSA value from >10 to >4 .
 What happened to the sensibility and the specificity
  of the test?
A-They were not affected
B-They both increased
C-The specificity increased and the sensibility increased
D- We need the prevalence of the disease
E-The sensibility increased and the specificity
 Screening test : high sensibility

 Confirmatory test : high specifity.
                   For the board

                        Disease        Disease
                        Positive       Negative
      Test +               A                 B
      Test -               C                 D
 Positive predictive Value : (A) /(A+B)
 Negative predictive value :( D) / (D+ C)
         Classical Board question :
 As the prevalence of a disease decreases in a
  population what happens to the Positive predictive
A- Stays the same and the Sensibility decreases
B-Increases and the Negative predictive values
C-Decreases and the Negative predictive value
D-Neither the Positive or the negative predictive values
  are affected
 A 40 year-old female wants to have a mammogram.
  The prevalence of Breast cancer in her population is
  estimated to 20 in 1000. The sensibility of the
  mammogram is 70 % and the specificity is 90 % . If
  her mammogram is positive what is the chance she
  has breast cancer?
                     For the board

                          Disease         Disease
                          Positive        Negative
      Test +                 A                B
       Test -                C                D
 Positive LR : (sensibility ) /(1-specificity)
 Negative LR :( 1-sensibility) / (specificity)
 LR are used to assess the pre and post test
  probability of the presence of a disease.

 Post-test Probability = LR X Odds

 Always remember Odds is not Probability:

 Probability = Odds /(Odds+1)
 Odds =Probability /(1-Probability)
 Step 1-A test has a sensibility of 60 % and a
  specificity of 90 % what is the Positive LR ?
E-can not calculate need more information
 Step 2: the pretest probability of a disease is
  20 % what are the odds of its occurrence?
What is the most compelling physical finding for the
DIAGNOSIS of DVT in a patient?

                           Sensitivity         Specificity

      Erythema                20%                 95 %

       Swelling               70%                 70%

         Pain                 2%                  99%

         SOB                  4%                  80%
                  Relative risk
                         Exposure   Nonexposed

    Disease +                A          B
   No disease                C          D

 Relative Risk : A/B
<1 the exposure is protective
>1 the exposure is a risk factor
                       No treatment        Treatment

    Disease +                A                  B
   No disease                C                  D

 Relative risk reduction: A-B / (A)
 Absolute risk reduction :A-B ( in percent )

 Number needed to treat (Favorite board question )
  1/ ARR
Classical Board question
 A new treatment for terminal heart failure is
  tried in a phase III trial . Mortality at 2 years is
  60 % with traditional medication and 35 %
  with this new drug .
What is the NNT?
Classical Board question
 A new experimental chemotheray drug is
  being tested for metastatic pancreatic
  adenocarcinoma. Patient who receive the
  drug have a 35 % survival rate at 1 year
  those who don’t have a 15 % survival rate.
What is the NNT
A- 4
Classical Board question
  A new treatment for breast cancer is being tested . In
  Population A 180 receives the treatment and 72
  survive at 5 years . Of the 240 patients who do not
  receive the treatment 48 survive .
What is the NNT for this treatment?
A- 20
E- need more information can not calculate.
300 subjects are randomized to trial for a drug .
  Placebo group had a mortality of 90%.
  Treatment group 75 % . How many patients
  need to be treated to save a life?
E-we need to know the exact number of
  patients in each group.
        Classical Board Question
  A metanalysis is just published an the benefits of N-
  acetylcystein in patients undergoing cardiac
  catheterisation . The risk of developping Contrast
  nephropathy with NAC is found to be 0.7 [ 0.5-1.1] p
  value : 0.05
  What can you say about the effect of NAC and
  cardiac catheterisation?
A- There is no proven effect
B- it is protective
C-It is a risk
D- We need more information on the type of contrast
              What about p?
 “ P” is not a magical number
 P tells you whether or not an observed effect
  can be explained by chance .
 It measures the probability that the results
  due to chance alone .
Concept of CI
 Confidence interval (CI) is the interval whithin
  which exists a high probability of finding the
  true value .

 Eg : Odds-ratio : 2.0 (CI 95 % : [1.5-3])

 On a question with a CI that encompasses 1
  for a risk or 0 for an effect there is no
  statistically significant relationship even is p
Important type of studies
 Observational /Retrospective

 For rare disease : case-series.

 Intent to treat : is a method of adjusting for
  bias caused by participants leaving the study
  because of treatment.
 Randomized – controlled trials

 Multicenter trial refers to the number of
  centers not the type of study.

 Systematic review (Cochrane)

 Meta-analysis: quantitative combination of
  multiple studies.
 The best level of evidence is Systematic
  reviews or RCT.

 Evidence of Class A or I is considered to be
  supported by good EBM evidence

 Evidence of class D or IV : expert opinion is
Thank You