Science doesn't change, the methodologies behind the studies do. One study might forget one variable and thus draw wrong conclusions, or perhaps there are confounders to the situation which skew the results.
An example just from something I was doing today was the effect of an intra workout CHO supplement. Previously, it was shown that only 30-60g of exogenous CHO per hour could be oxidised; however this is dependant on which glucose transporters are used. The older studies didn't realise this and thus lead to the recommendation of 30-60. It is now shown that by incorporating a GLUT-5 transporter you can increase oxidation up to 80-90g/h.
You also have to be VERY weary of the conclusions media and scientists draw from the conclusions; i.e. the most misinterpreted studies that I come into contact with are those that assume dietary fat causes obesity when this is not so. That is a product of eating too much. It just so happened that those people they studied ate more on a higher fat diet.
Little things such as a undetected vitamin defieciency can skew results on an unrelated study.
When these things are taken into account and a study is planned exactly the same using the same parameters, you will likely replicate the same results each and every time
If you read a study, you have to be able to pick it apart and understand it and crictically analyse it otherwise you fall victim to misinterpretation.