Literature Appraisal Effectiveness of Therapy
Measures of treatment effect Statistical significance Odds ratio Relative risk Absolute risk reduction Number needed to treat
Measures of treatment effect Outcome (death) YesNo Controlab Experimentcd
200 Total
Total in each group
Die 10 Total in each group After 1 year
(75) Die 10(90) (Survive) Total in each group + + After 1 year
Risk & Relative Risk
(75) Die 10(90) (Survive)Risk? (a proportion) What is the Total in each group + + After 1 year
(75) Die 10(90) (Survive)Risk Total in each group + + After 1 year
(75) Die 10(90) (Survive)Risk Total in each group + + After 1 year
(75) Die 10(90) (Survive)Risk Total in each group + + After 1 year
(75) Die 10(90) (Survive)Risk % + + After 1 year
(75) Die 10(90) (Survive)Risk % + + After 1 year
(75) Die 10(90) (Survive)Risk % % + +
(75) Die 10(90) (Survive)Risk % % + +
(75) Die 10(90) (Survive)Risk Risk ratio or Relative risk (RR) = % %
(75) Die 10(90) (Survive)Risk Risk ratio or Relative risk (RR) = % %
(75) Die 10(90) (Survive)Risk Risk ratio or Relative risk (RR) = % %
Odds & Odds Ratio
Odds The ratio between the amounts staked by parties in a bet, based on the expected probability either way. The balance of advantage or superiority.
(75) Die 10(90) (Survive)Risk Odds? What about + +
(75) Die (90) (Survive)Risk Odds 10+ +
(75) Die (90) (Survive)Risk Odds 10 1to3 + +
(75) Die 10(90) (Survive)Risk Odds 1to3 + +
(75) Die 10(90) (Survive)Risk Odds 1to3 1to9 + +
(75) Die 10(90) (Survive)Risk Odds 1to3 1to9+ +
(75) Die 10(90) (Survive)Risk Odds 1to3 1to9 Odds Ratio (O.R.) = + +
(75) Die 10(90) (Survive)Risk Odds 1to3 1to9 Odds Ratio (O.R.) = + +
(75) Die 10(90) (Survive)Risk Odds 1to3 1to9 Odds Ratio (O.R.) = + + 3
(75) Die 10(90) (Survive)Risk Odds 1to3 1to9 3 Odds Ratio (O.R.) = + + Risk ratio or Relative risk (RR) = 2.5
Measures of treatment effect influence clinicians decisions Clinicians: – more inclined to treat if the results are presented as relative risk – less inclined to treat if the results are presented as absolute risk reduction Forrow et al. Am J Med 1992;92:121
Control group event rate (CER) = Deaths / Controls Experiment group event rate (EER) = Deaths / Treated Absolute risk reduction (difference) ARR=CER- EER Absolute Risk Reduction
Die 10 Risk % % Absolute Risk Reduction (ARR) = CER EER
Die 10 Risk % % Absolute Risk Reduction (ARR) = %
Number Needed to Treat NNT
Clinical value of measures of treatment effect Number Needed To Treat The odds ratio etc. not easy to understand, especially for patients. The number needed to treat (NNT) to prevent an adverse event is a more clinically relevant measure of the consequences of treatment Sackett DL. EBM 1996; 1: Sinclair JC. J Clin Epidemiol 1994; 47: 881-9
Number Need to Treat (NNT) Out of 100 patients treated 10 died compared to 25 in the placebo group and 15 extra survived. Therefore: To get 1 more patient to survive, 6.7 (100/15) have to be treated.
100/15 1/ 0.15 NNT = 1/ ARR
Die 10 Risk % % Absolute Risk Reduction (ARR) = % NNT= 1/ARR= 1/0.15= 6.7
Q.E.D.
MAGPIE Of the patients treated (5015) 40 fitted compared to 96 in the placebo group (5055) In % Mg 0.8% vs Placebo 1.9% Therefore: ARR 1.8 – 0.8 = 1.1% (11 per 1000) To get 1 more patient to survive, 91 (100/1.1) have to be treated. = NNT