In a cohort study, the risk ratio for an exposure is 1.2 among women and 3.0 among men, and the difference persists after careful assessment of confounding. What should investigators consider?