Testing the usefulness of the number needed to treat to be harmed (NNTH) in benefit-risk evaluations: case study with medicines withdrawn from the European market due to safety reasons.

Objective: To explore the usefulness of number needed to treat to be harmed (NNTH), in benefit-risk assessments, by studying the agreement between NNTH values and withdrawals of medicines from European market due to safety reasons.

Methods: Medicines with data from longitudinal studies were included. Studies were identified from European Medicines Agency's Reports. Meta-analyses were performed to pool odds ratios (OR) with 95% confidence-intervals (CI). Published control event rates were applied to ORs to calculate NNTHs (95%CI) for selected adverse events.

Results: NNTH (95%CI) decreased from pre- to post-marketing for the eight medicines included: peripheral neuropathy (∞ vs. 12[non-significant; NS] with almitrine; heart valve disease with benfluorex (∞ vs. NNTH ranging from 7[4-13] to 7[5-9]); myopathy (-4096[NS] vs. 797[421-1690]), new-onset diabetes (113[NS] vs. 390[425-778]), bleeding (∞ vs. 517[317-1153]), and infection (∞ vs. 253[164-463]) with niacin-laropiprant; psychiatric disorders (12[7-34] vs. 9[5-24]) with rimonabant; myocardial infarction (MI) [-1305 vs. 270[89-4362]) with rofecoxib; MI (-510 vs. NNTH ranging from 152[55-4003] to 568[344-1350]) with rosiglitazone; cardiovascular events (∞ vs. 245[129-1318]) with sibutramine; and liver injury (∞ vs. 5957[NS]) with ximelagatran.

Conclusion: NNTH have potential of use as a supportive tool in benefit-risk re-evaluations of medicines and may help regulators to making decisions on drug safety.