A dam project drove malaria near zero in one Amazon town. Then the program ended.

Between 2006 and 2020, the municipality of Altamira in the Xingu Basin of the Brazilian Amazon recorded 17,578 malaria cases [s1]. The distribution of those cases across the period is the story: more than 1,200 per year before construction of the Belo Monte Dam, below 100 during construction, and back above 700 by 2020 [s1].

A study published in GeoHealth on 9 July reconstructs that arc and asks what explains it [s1].

What happened, and when

A malaria control program was implemented in Altamira during construction of the Belo Monte Dam — an infrastructure project large enough that the associated public health effort eliminated locally acquired malaria cases in the municipality [s1]. When those efforts ceased, transmission returned, predominantly in rural clusters rather than in the urban centre [s1].

The researchers split the record into a before period (2006–2012) and an after period (2017–2020), analysing annual notified cases alongside geospatial variables: forest edge, population size, and travel time to urban centres [s1]. Because incidence was low in the later years, case data were clustered by the geographic proximity of health centres rather than analysed at individual facility level [s1].

What predicted where malaria came back

Using generalised additive models, the study found that malaria cases increased significantly with increases in forest edge habitat and with local population size — and that both associations held in the before and after periods [s1]. Neither distance to deforestation nor climatic variables explained variation in incidence [s1].

That negative result is worth pausing on. Climate is the variable most often invoked to explain shifting malaria in the Amazon, and here it did not carry explanatory weight once landscape structure and population were accounted for. Nor did proximity to deforestation as such. What mattered was forest edge — the boundary zone between cleared and standing forest — and how many people lived near it [s1].

Why the ending of a program matters more than its success

The finding the authors emphasise is not that the control program worked. It is how fast the gains reversed once it stopped [s1]. Their framing is that interruptions to control efforts, even localised ones, can rapidly reverse progress, and that understanding landscape epidemiology is critical to achieving and maintaining malaria elimination in Amazonian settings [s1].

This is the recurring structural problem in malaria elimination. Programs are commonly financed as projects — attached to a dam, a mining concession, a donor cycle, an outbreak response — and elimination is a permanent commitment. Altamira is a small, well-documented instance of what the gap between those two timescales costs: a municipality where local cases were eliminated during the dam-linked program and which was reporting more than 700 cases a year by 2020, predominantly in rural clusters [s1].

The limits of what this shows

This is an observational analysis of notified cases in a single municipality, and it inherits every constraint that comes with that. Notified cases depend on people presenting to health services and on those services testing and reporting — so an apparent rise can partly reflect changes in surveillance intensity, particularly when a well-resourced control program ends and its case-detection capacity ends with it. The study does not report a counterfactual municipality that never had a control program.

The clustering of case data by health centre proximity, adopted because of low incidence, means the spatial resolution of the analysis is coarser than the underlying notification data [s1]. And the gap between the two study windows — 2013 to 2016 is not part of either period [s1] — means the transition itself is characterised by its endpoints rather than observed continuously.

The models also cannot separate the mechanisms bundled inside "population size." More people in a forest-edge cluster means more hosts, but also more housing, more movement, and more occupational exposure, and the available reporting does not disentangle those.

What it means for the wider region

The practical implication of the Altamira record is about where residual effort should be aimed: not uniformly across a municipality, but at forest-edge settlements, and not for the duration of a project, but permanently. The study's own conclusion runs the same way — elimination requires both reducing transmission in endemic areas and maintaining those reductions where elimination has already been achieved [s1].

What to watch

Whether Altamira's post-2020 trajectory has continued upward, and whether the forest-edge relationship documented here reproduces in other Amazonian municipalities that have gone through infrastructure-linked control programs. The study covers data through 2020 only [s1].