From that perspective, surely you can draw no conclusion of causality from this data, or any conclusion you can draw has little predictive power about the direction of effect for a given sample.