Short version: In 2000, the CDC had a conference about possible effects of thiomersal in vaccines, commonly known as the Simpsonwood conference. At the conference, a team lead by Thomas Verstraeten reported results of their data analysis that suggested that vaccinated children had a higher level of numerous disorders. Other attendees pointed out that the team had not controlled for the level of contact that children had with medical professionals - parents who vaccinate their children are generally more likely to see doctors if they have concerns, and therefore more likely to have their children diagnosed with various disorders, even if the actual prevalence of those disorders is the same. The team agreed to reanalyze the data, taking account of this major confounding factor, and publish their results.
Although the initial results were not published (because they were flawed), they were publicly reported at a conference the same year. Three years later, the results of the reanalysis were published and, to the the surprise of no one sensible, showed no link between vaccination and negative health outcomes.
Despite the fact that the preliminary results were stated publicly back in 2000, in 2005 Robert F. Kennedy Jr. (an anti-vax wanker with no respect for pesky little matters like honesty) authored an article for Salon that contended that the CDC had covered up the initial Simpsonwood findings until they could manipulate the data in such a way that there would seem to be no connection. Periodically since then, some new nut will stumble on documents related to Simpsonwood that were released long ago and claim that they're new revelations of CDC perfidy. It's all a bunch of horseshit.