Epi

Number Needed to Screen

search

Number Needed to Screen, Number Needed to Treat, Absolute Risk Reduction, Relative Risk Reduction

  • See Also
  1. Screening Test
  2. Contingency Grid or Cross Tab (includes Statistics Example)
  3. Bayes Theorem (Bayesian Statistics)
  4. Fagan Nomogram
  5. Experimental Error (Experimental Bias)
  6. Lead-Time Bias
  7. Length Bias
  8. Selection Bias (Screening Bias)
  9. Likelihood Ratio (Positive Likelihood Ratio, Negative Likelihood Ratio)
  10. Negative Predictive Value
  11. Positive Predictive Value
  12. Pre-Test Odds or Post-Test Odds
  13. Receiver Operating Characteristic
  14. Test Sensitivity (False Negative Rate)
  15. Test Specificity (False Positive Rate)
  16. U.S. Preventive Services Task Force Recommendations
  • Definitions
  • Event Rate (Event Probability)
  1. Event Rate: (Number Events) / (Number Total Patients)
  2. Calculate for both intervention and control groups
  1. RR = (Intervention Event Rate)/(Control Event Rate)
  • Definitions
  • Relative Risk Reduction (RRR)
  1. Relative Risk Reduction or RRR = 1 - (Relative Risk)
  2. RRR = (Absolute Risk)/(Control Group Event Probability)
    1. Proportional decreased risk for the treatment group compared with the control group
    2. Interpretation requires the actual risk numbers for context
  3. Relative Risk and Relative Risk Reduction can be misleading
    1. Relative Risk Reduction is out of context of the actual values and exaggerates benefit
      1. In one study of Statins in Hyperlipidemia, coronary disease deaths were measured
        1. Control group 7.9/100 (0.079)
        2. Statin group 5.5/100 (0.055)
        3. Shepherd (1995) N Engl J Med 333:1301-7 [PubMed]
      2. Benefit to the patient is more easily interpreted with ARR and NNT (but RRR makes headlines)
        1. Relative Risk Reduction (RRR) = (0.079-0.055)/0.079 = 0.024/0.079 = 30%
        2. Absolute Risk Reduction (ARR) = (0.079-0.055) = 0.024 = 2.4%
        3. Number Needed to Treat (NNT) = 1/0.024 = 42
        4. Henley (2000) Fam Pract Manag 7(5):59-60 +PMID: 10947349 [PubMed]
    2. When applied to an individual patient
      1. Relative Risk Reduction reflects the effect of an intervention on a population
      2. Effect is not guaranteed for the individual patient
      3. Number Needed to Treat (NNT) is a better measure of the chance of a positive effect on the individual
        1. http://www.thennt.com/thennt-explained/
  • Definitions
  • Absolute Risk Reduction (ARR)
  1. ARR = (Intervention Event Rate) - (Control Event Rate)
  2. Preferred over Relative Risk Reduction (RRR) since it reflects ACTUAL risk difference
  3. Refers to the actual difference in risk between the treatment group and the control group
    1. Result <0 (negative) is seen with effective intervention
    2. Result >0 (positive) is seen with a harmful intervention (worse than control)
  • Definitions
  • Number Needed to Screen (NNS) or Treat (NNT)
  1. Number of patients screened to prevent one death
  2. Assumes certain period of time (e.g. one year)
  3. NNS or NNT = 1/(Absolute Risk Reduction)
  4. NNT = 100/abs(rateTreatment - rateControl)
    1. Where Intervention is intended to prevent event X (e.g. Myocardial Infarction)
    2. Where rateTreatment = Rate of event X in treatment group (e.g. 5%)
    3. Where rateControl = Rate of event X in a control group (e.g. 10%)
    4. Example: NNT = 100/(5-10)=-20
      1. Twenty persons would need to undergo intervention, to prevent one event X
      2. In this case, the result is negative, consistent with Number Needed to Treat
      3. If the result were positive, the finding would be consistent with number needed to harm
  • Resources
  1. Visual Tool
    1. http://www.nntonline.com
  • References
  1. Cook (1995) BMJ 310:492-4 [PubMed]
  2. Daya (1999) Evid Based Obstet Gynecol 1:103-4
  3. Mcquay (1997) Ann Intern Med 126(9):712-20 [PubMed]