NNT indicates the number of patients that would have to be treated to prevent one bad outcome or allow the gain rendered by treatment.

The concept is discussed also in the Therapy lecture

To understand NNT consider the following hypothetical example

1000 patients are randomized to receive either heparin or placebo. In the heparin group, adverse events (death, or MI ) occurred in  10% (50 of 500). For those treated with placebo, the adverse event rate was 20% (100 of 500). 

It is convenient to set up a contingency table to assist data analysis
 
 

In other terms, 
the Experimental Event Rate (EER) is a/(a + b)
and the Control Event Rate (CER)  is  c /(c + d)
 

• The Absolute Risk Reduction (ARR) is the difference

            ARR=CER-EER      AAR= 0.2 - 0.1 = 0.1
 

• The Relative Risk (RR) is the proportional reduction in risk

            RR= EER/CER        RR= 0.1/0.2 = 0.5

       "the relative risk of an adverse event in heparin treated patients is 50%"
 

• Commonly Relative Risk Reductions (RRR) are reported in the literature

                     "the risk of an adverse event was reduced by 50%"
 

• Pundits of Evidence Based Medicine (evidencists?) favor the Number Needed to Treat (NNT) as a "bottom line" measure of effect size.  NNT can be determined in two ways

                  "You need to treat 10 patients to avoid one adverse event" 

NNT is in some ways more intuitive and informative than RR or RRR.  However, a full appreciation of NNT's requires a little more reading