My understanding is that the study looked at change over time, so it's not as badly designed a study as a simple correlation, though it is still not ideal.