Lessons for management of anaphylaxis from a study of fatal reactions

The unpredictability of anaphylactic reactions and the need for immediate, often improvised treatment will make controlled trials impracticable; other means must therefore be used to determine optimal management.

This study aimed to investigate the circumstances leading to fatal anaphylaxis.

A register was established including all fatal anaphylactic reactions in the UK since 1992 that could be traced from the certified cause of death. Data obtained from other sources suggested that deaths certified as due to anaphylaxis underestimate the true incidence. Details of the previous medical history, the reaction and necropsy were sought for all cases.

Approximately half the 20 fatal reactions recorded each year in the UK were iatrogenic, and a quarter each due to food or insect venom. All fatal reactions thought to have been due to food caused difficulty breathing that in 86% led to respiratory arrest; shock was more common in iatrogenic and venom reactions. The median time to respiratory or cardiac arrest was 30 min for foods, 15 min for venom and 5 min for iatrogenic reactions. Twenty-eight per cent of fatal cases were resuscitated but died 3 h-30 days later, mostly from hypoxic brain damage. Adrenaline (epinephrine) was used in treatment of 62% of fatal reactions but before arrest in only 14%.

Immediate recognition of anaphylaxis, early use of adrenaline, inhaled beta agonists and other measures are crucial for successful treatment. Nevertheless, a few reactions will be fatal whatever treatment is given; optimal management of anaphylaxis is therefore avoidance of the cause whenever this is possible. Predictable cross-reactivity between the cause of the fatal reaction and that of previous reactions had been overlooked. Adrenaline overdose caused at least three deaths and must be avoided. Kit for self-treatment had proved unhelpful for a variety of reasons; its success depends on selection of appropriate medication, ease of use and good training.