Epi

Treatment Effect

search

Treatment Effect

  • See Also
  1. Statistical Terms (Statistical Significance, P-Value, Confidence Interval, Clinical Significance, Statistical Heterogeneity)
  • Approach
  1. See Contingency Grid for description of setting up this grid
  2. Given
    1. We are studying whether an event is less likely to occur in an experimental group, then in a control group
  3. Create a hypothetical grid
    1. Label the grid top
      1. Event (or E+)
      2. No Event (or E-)
      3. Total patients
    2. Label the grid left
      1. Control (or T-)
      2. Experimental (or T+)
      3. Total patients
    3. Fill in the next row (Control Row)
      1. Event = a
      2. No Event = b
      3. Total = a + b
    4. Fill in the top row (Experimental Row)
      1. Event = c
      2. No Event = d
      3. Total = c + d
    5. Fill in the bottom row (Total row)
      1. Event = a + c
      2. No Event = b + d
      3. Total = a + b + c + d
  4. Calculations
    1. Control event rate (CER)
      1. Risk of event from control intervention = a / (a+b)
    2. Experimental event rate (EER)
      1. Risk of event from experimental intervention = c / (c+d)
    3. Relative Risk (RR)
      1. RR = EER / CER
      2. Hazard Ratio (HR) is a similar concept related to survival
    4. Relative Risk Reduction (RRR)
      1. RRR = 1 - RR
    5. Absolute Risk Reduction (ARR)
      1. ARR = CER - EER
    6. Number Needed to Treat (NNT)
      1. NNT = 1 / ARR
  1. See Contingency Grid for description of setting up this grid
  2. Given
    1. Example: Does CVA occur less often in Atrial Fibrillation with Dabigatran (experimental) than with Warfarin (control)
    2. Connolly (2009) N Engl J Med 361:1139-51 [PubMed]
  3. Create a grid
    1. Label the grid top
      1. CVA (or E+)
      2. No CVA (or E-)
      3. Total patients
    2. Label the grid left
      1. Warfarin (or T-)
      2. Dabigatran 150 mg (or T+)
      3. Total patients
    3. Fill in the next row (Warfarin Row)
      1. CVA = a = 199
      2. No CVA = b = 5823
      3. Total = a + b = 6022
    4. Fill in the top row (Dabigatran Row)
      1. CVA = c = 134
      2. No CVA = d = 5942
      3. Total = c + d = 6076
    5. Fill in the bottom row (Total row)
      1. CVA = a + c = 333
      2. No CVA = b + d = 11765
      3. Total = a + b + c + d = 12098
  4. Calculations
    1. Warfarin control event rate (CER)
      1. Risk of CVA on Warfarin = a / (a+b) = 199/6022 = 0.033 (3.3%)
    2. Dabigatran experimental event rate (EER)
      1. Risk of event from experimental intervention = c / (c+d) = 134/6076 = 0.022 (2.2%)
    3. Relative Risk (RR)
      1. RR = EER / CER = 0.022 / 0.033 = 0.667
    4. Relative Risk Reduction (RRR)
      1. RRR = 1 - RR = 1 - 0.667 = 0.333
    5. Absolute Risk Reduction (ARR)
      1. ARR = CER - EER = 0.033 - 0.022 = 0.011
    6. Number Needed to Treat (NNT)
      1. NNT = 1 / ARR = 1 / 0.011 = 90.9
  • References
  1. Hersh (2014) Evidence Based Medicine, AMIA’s CIBRC Online Course