I don't know about South America, but in most parts of Asia, the data is unreliable because a) poor rural people don't always get treatment, and even when they do, the diagnostics aren't always correct or even reported, and b) data collection is poor in rural areas - both the quality and quantity.
So I'd take any statistical claim with a grain of salt.
Farmers in the US have been using it for years as well. We have much better data on them. While farmers cancer numbers are not the same as the general US population, when you control for other known differences in farmer behavior roundup isn't an issue.
So I'd take any statistical claim with a grain of salt.