Combinatorics of Malaria Eradication

Malaria infecting a mosquito

I’ve got at least 3 interesting blog posts worth of material on TCS applications for fighting malaria, but I haven’t had time to pen even one of them. Here is an abbreviated version:

Malaria is a major disease, something like the #3 infectious disease globally, and the #1 cause of both death and disability in many parts of Southern Africa.

The Gates Foundation is leading the charge to attempt to eradicate malaria from the world, and many national governments and NGOs are also involved in the fight.

There is a history of malaria eradication attempts, and the historic lesson is this: don’t start a fight with malaria unless you’re going to win.

The weapons people use to fight malaria are few enough to list: bednets, often treated with insecticide, to protect vulnerable people and kill mosquitoes; pesticides, now including DDT, to kill mosquitoes; and drugs, for preventing infection in vulnerable people, and for treating people when they become infected.

Bednet distribution has huge logistics issues, and logistics has been a source of great inspiration for combinatorial optimization, but seems to me that here the open questions are much more about business models and marketing than, say, solving TSPs or or the multiple vehicle routing problem.

DDT has had horrible environmental effects, so its use is quite contentious. Many policymakers believe that it was the most successful tool in removing malaria from more-developed regions.

It seems to me that it is the drugs which inspire lots of theory problems (try quoting that out of context!). The WHO recommends a class of drugs called ACTs, Artemisinin Combination Therapies, and it’s the word “combination” that first made my ears perk up. Artemisinin is derived from wormword, the same as absinthe, and is used to treat malaria in Traditional Chinese Medicine. According to the Institute of Medicine report on malaria Saving Lives, Buying Time, it turned into a commercial medicine because of an appeal for help from Ho Chi Minh to Zhou En Lai during the Vietnam War. They don’t mention if Chairman Mao was directly involved in isolating the active compounds. But I digress…

One theory problem comes from the emergence of drug-resistant strains of malaria. Artemisinin is currently a very effective treatment. But Chloroquine used to be a very effective treatment, too, until a Chloroquine-resistant strain of malaria arose. And the same thing is happening with another previously effective treatment, Sulfadoxine-Pyrimethamine. If Artemisinin stops working, there isn’t really anything to replace it with.

This is where the combination therapy comes in. By using two drugs (with different mechanisms) together, perhaps we can slow the emergence of drug resistance. And by using three drugs together… oh, wait, if we have three drugs, A, B, and C, should we use them all together, or should we use A and B and save C for the future?

This was supposed to be an abbreviated version, so I’ll wrap up here. If you are ready to know how epidemiologists grapple with this sort of thing, look at the PNAS paper Benefits of using multiple first-line therapies against malaria by Boni, Smith, and Laxminarayan. I can’t find a copy of this online for free, and that’s a shame, but I’ll put one here, and see if the national academy comes after me.  The mathematical details are all in the supplementary materials, which is available as a free download from PNAS.

1 Comment

Filed under global health, combinatorial optimization, TCS

One response to “Combinatorics of Malaria Eradication

  1. Maciej Boni sent me the PNAS paper that I was whining about not being available for free online, and also another recent commentary on malaria modeling, which is published by PLoS Medicine who do make all their material freely available.

    The PLoS paper contains some useful advice for any TCS people interested in really getting into disease modeling:

    The keys to a sound and understandable modeling conclusion are appropriate design, working within the model’s assumptions, a careful analysis of the model’s sensitivity to these assumptions, and a clear statement of the model’s limitations.