Table 2 How summary results can obscure situations where the typical patient receives no benefit or risks net harm
From: Assessing and reporting heterogeneity in treatment effects in clinical trials: a proposal
Assumption: Treatment reduces baseline risk by 25% but with a cost of 2 serious treatment-related adverse events per 1,000 patients per year | |||||
|---|---|---|---|---|---|
| Â | Control Event Rate (CER, %) | Experimental Event Rate (EER, %) | Relative Risk Reduction (RRR) | Absolute Risk Reduction (ARR) | Number Needed to Treat (NNT) |
   (% of study population) | Results over 5 years | ||||
Scenario #1 | Â | Â | Â | Â | Â |
Overall result (100%) | 8 | 7 | 0.125 | 0.01 | 100 |
   Average risk subjects (75%) | 4 | 4 | 0 | 0 | ∞ |
   High risk subjects (25%) | 20 | 16 | 0.2 | 0.04 | 20 |
Scenario #2 | Â | Â | Â | Â | Â |
Overall result (100%) | 9 | 7.75 | 0.14 | 0.0125 | 80 |
   Average risk subjects (75%) | 2 | 2.5 | -0.25†| -0.005†| -200 |
   High risk subjects (25%) | 30 | 23.5 | 0.22 | 0.065 | 15 |