A researcher is investigating the association between coffee consumption and myocardial infarction in a cohort of 5000 Indian adults. The analysis shows a strong positive association (RR = 2.5, 95% CI 1.8–3.2). However, when smoking status is controlled for in a stratified analysis, the association weakens considerably (RR = 1.1, 95% CI 0.9–1.3). Which feature best distinguishes confounding from bias in this scenario?