This cohort study evaluated the fate of sound surfaces and inactive non-cavitated (INC) and active non-cavitated (ANC) caries lesions in a population-based sample of South Brazilian adolescents, in answer to the question: “Is lesion activity assessment a reliable criterion to diagnose a patient’s caries activity?” A total of 801 schoolchildren were examined at baseline (aged 12 years) and after a mean time interval of 2.5 years. Data collection included a questionnaire and clinical examination. Patients were classified as caries-free (patients without any lesion), caries-inactive (patients with only inactive lesions), and caries-active (patients with at least one active lesion). The primary outcome was caries progression (presence of cavity, underlying dentin shadow, filling, or extraction at the follow-up exam). Negative binomial regression models were used to estimate the risk for caries progression. The main predictor variable was status of the surface at baseline: sound, INC, or ANC. Progression rates of 1.0, 9.0, and 12.6% were found for sound surfaces, INC, and ANC, respectively. INC (incidence risk ratio [IRR] 5.37, 95% CI 4.22–6.83) and ANC (IRR 4.96, 95% CI 3.43–7.17) had greater risk for caries progression than sound surfaces. Similar risks for progression were found for ANC and INC (IRR 0.92, 95% CI 0.64–1.32). Progression rates were 0.6, 1.1, and 2.2% for caries-free, caries-inactive, and caries-active individuals, respectively (<i>p</i> &#x3c; 0.05). The risk for caries progression of sound surfaces was higher among caries-active adolescents (caries-free: IRR 2.78, 95% CI 1.63–4.72; caries-inactive: IRR 2.19, 95% CI 1.65–2.90). Caries-inactive patients behaved similarly to caries-free individuals (IRR 1.27, 95% CI 0.73–2.20). This study demonstrated the possibility of defining a patient’s caries activity profile based on lesion features.