Killing the mosquito to prevent the ulcer

Design

cluster randomised controlled field trial, six intervention and six control areas, with an ad hoc analysis of human notifications

Population

twelve urban geographic areas in a Buruli ulcer endemic part of Melbourne, Australia, January to March 2024

Primary outcome

Aedes notoscriptus mosquito population numbers, measured by egg counts

Effect

70% average reduction in egg counts vs control (P=0.0076); exploratory disease analysis 1 vs 6 cases, incidence rate ratio 0.167 (95% CI 0.0026 to 1.054)

Buruli ulcer, caused by Mycobacterium ulcerans, has spread through urban Melbourne, and Aedes notoscriptus is implicated in transmission. This randomised controlled field trial allocated six geographic areas to receive 100 autodissemination stations each - containing the larvicide pyriproxyfen and the entomopathogenic fungus Beauveria bassiana - for eight weeks from late January 2024, with six control areas receiving none.

The primary outcome was mosquito population. Egg counts fell by an average of 70% in intervention areas against controls (P=0.0076). The authors then looked, in an analysis they describe as ad hoc, at human Buruli ulcer notifications: allowing for a mean incubation period of 4.8 months, there was 1 case in intervention zones against 6 in control zones at the point of peak intervention effect, incidence rate ratio 0.167 (95% CI 0.0026 to 1.054, P=0.047). No such difference appeared in the equivalent period the year before or after. Mosquito suppression and disease reduction correlated closely (R squared 0.85).

The entomological result is the trial; the human result is a hypothesis. Seven cases in total cannot support a claim about disease prevention, and the confidence interval on that incidence rate ratio crosses 1.0 despite the P value - which should be read as the analysis straining at the limits of its data rather than as a positive finding. What the trial does establish is that a vector control method can work at neighbourhood scale, which matters wherever a skin disease has an arthropod vector.

• The demonstrated result is 70% fewer mosquito eggs; the effect on human disease is exploratory.
• Seven Buruli ulcer cases in total is not a basis for a prevention claim, whatever the P value says.
• Note the confidence interval on the disease outcome, 0.0026 to 1.054, includes no effect.
• Vector control at neighbourhood scale is the transferable idea, not this specific station or this specific pathogen.
• In endemic areas the clinical message is unchanged: a slowly enlarging painless nodule or ulcer in an exposed area needs a specific diagnosis, not empirical antibiotics.

The statistics, in plain English

A confidence interval of 0.0026 to 1.054 around an incidence rate ratio of 0.167 includes 1.0, meaning the data are compatible with no effect on disease, even though the reported P value is 0.047 - a mismatch that usually arises from different methods being used for the interval and the test on very sparse counts. With one case in one arm and six in the other, either number moving by a single case would change the conclusion. The 70% reduction in egg counts is the robust result: it is the prespecified primary outcome, measured across twelve areas, with a P value of 0.0076. An R squared of 0.85 between suppression and disease risk comes from a handful of areas and should not be read as strong evidence of a causal chain.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.