It took some digging, but here is the actual study. It is a real scientific study, but it also has some flaws.
Long term toxicity of a Roundup herbicide and a Roundup-tolerant genetically modified maize
As an aside, it really bugs me when blogs or media outlets do not provide a direct link to studies they paraphrase. Especially when they incorrectly paraphrase them (not the case here).
At first glance, the study looks good. Long (for rat) study duration, multiple treatment options, and lots of sampling timepoints with a variety of data collected at each timepoint.
Unfortunately, the researchers slipped up a bit in their design. I have read their rebuttals to the critiques, and I do understand that they were under cost constraints and also seemed very keen to compare their study to a previous 'landmark' study.
There may be additional factors that others can tease out, but the three things I don't like about the study are:
1) The control group was a bit small. With nine treatment groups of 10 rats per sex each, I'd want more than 10 of each as controls. I think I'd want more than 10/sex for the treatments too, especially considering the next point. Ten individuals sounds like a lot, but animals are not always as uniform as we'd like. Then again, there is huge pressure to reduce the number of individuals used in a study on the Animal Welfare side of it. Maybe they couldn't get approval for more rats and as it is, they probably spent $4000-$8,000 on rats alone. It might be overkill, but I would also like to see a control group that does not undergo isoflurane anesthesia, after the initial baseline blood/urine draw, and is only weighed and palpated for tumors until the study endpoint. Isoflurane shoudn't have any effect, but it is always nice to head off critiques at the pass.
2) The choice of rat breed was a bit iffy. The breed of rat used is known to suffer from various tumors and other cancerous conditions, with incidence increasing with age. This breed, coupled with the small cohort size, can result in misleading data. This critique came up in a few communications and the rebuttal was that carcinogenesis studies at the National Toxicology Program are standardizing to this breed of rat and that a 'sensitive' breed is preferred. While the Sprague Dawley is being evaluated by the NTP for a 'Default' breed, it is not being touted as THE breed to use for all studies. So, while it may be appropriate for their comparison study (90 days), that breed is not great for a 2 year study.
3) Food was choice and not portioned. Without controlling how much the rats ate, there is a problem differentiating between the effects associated with over-eating and your treatment. Some rats will stuff themselves silly if they favor the taste of the food. If they continually do this, then they can become overweight and/or suffer from other metabolic problems. Along similar lines, it is difficult to tell if a few rats are consuming most of the provided food/water, which may skew your data. Maybe this is nit picking. I guess they are only looking for any effect, and can let other studies figure out how to quantify any dose dependent responses.
I would be very interested if they could repeat the study with bigger cohorts, controlled food access, and less problematic rat breed.
It is definitely a good start to an area that should be explored at greater depth.