Background Published randomised controlled trials (RCT) and guidelines invariably present the benefits gained from implantable cardioverter defibrillators (ICD) in terms of relative risk reduction (RRR). It is well documented that patients have a better understanding of absolute risk reduction (ARR) and this figure, together with numbers needed to treat (NNT), should be used in the informed consent process. The fact that most trials present data over varying follow-up periods also adds to confusion.
Methods A Medline search was performed for RCT comparing ICD with medical treatment. Publications that included predominantly epicardial implants, which are not used in national guidelines or did not present survival data were excluded. Annual mortality for ICD and medically treated groups were recorded at 1, 2 and 3-year follow-up from raw data included in the text or by estimation from Kaplan–Meier survival curves and used to calculate ARR. When subgroup analysis data were available, similar calculations were made. Flow charts were constructed to present 1, 2 and 3- year ARR and NNT. Individualised risk reduction statistics could then be estimated based on patient characteristics.
Results Seven primary prevention RCT, three secondary prevention RCT and one secondary prevention meta-analysis were included. For primary prevention, ARR at 3-year follow-up ranged from zero (AMIOVERT and CAT) to 24.6% (MADIT, NNT 4); fig 1. Subgroup calculations showed a wide range of primary prevention ARR, from 0% in MADIT2 risk scores of 0 or >3, SCD-HeFT patients who were NYHA class 3 and MADIT 2 patients of black ethnicity, to 29% (NNT 3) in MADIT2 patients with a risk score of 2. For secondary prevention, ARR at 3-year follow-up ranged from 3.7% (CIDS) to 11.3% (AVID); fig 2. The greatest secondary prevention ARR was found in high-risk patients from CIDS (16.8%, NNT 6), whereas patients in the secondary prevention meta-analysis with left ventricular ejection fraction greater than 35% had a 3-year ARR of only 2.2% (NNT 45). In primary prevention trials, ARR increased with follow-up, whereas in secondary prevention trials there was an initial increase then a plateau after 24 months. In primary prevention trials there was a strong correlation between medical group mortality and ARR in the ICD group; however, subgroup analysis revealed patient groups with very high total mortality and no benefit from ICD. A striking finding was that only a third of MADIT2 and AVID patients were eligible for follow-up at 2 years and 10–14% at 3 years. MADIT, MUSTT, the SCD-HeFT NYHA class 2 subgroup and the MADIT2 risk score 2 subgroup are the only 2 and 3-year ARR figures with 95% CI smaller than the ARR.
Conclusion ARR from ICD varies significantly depending upon trial entry criteria and subgroup analysis. The relatively small number of patients followed for 2 years or more results in wide confidence intervals. Despite these limitations, simple flow charts that provide ARR and NNT at 1, 2 and 3-year follow-up and give individualised estimations of patient benefit may provide a valuable aid to physician education and the informed consent process.