There is a very high correlation between KS2 scores and GCSE results, both at school level and at pupil level - in fact these are used as the basis for any calculation of "value added" and for most internal school target-setting. Somewhere I've seen a graph showing the scatter of actual results but until I can track it down I've found this old technical document from dcsf
http://www.dcsf.gov.uk/performancetable ... docD.shtml
which (if I can disentangle the details) implies that (for example) a child with 3 level 5's at KS2 would be expected to obtain the equivalent of 56 GCSE points (based only on the best 8, so this could be 8 grade A) whereas a child with 3 level 3's would be expected to achieve around 23 GCSE points (roughly 8 grade D)
I think there is more recent analysis somewhere which looks at other measures such as the probability of a child gaining the "magic" 5 A*-C grades given particular KS2 results.
Obviously a lot of other factors affect the actual result of any individual child, and trying to establish the school's contribution to the progress made above or below the "expected average" is what lies behind first simple value added and later versions such as contextual value-added