Our aims were to examine the stability of self-rated anxiety and depression symptoms and the predictors for change in case-level status after 4 years in a general population sample. Prospective cohort study. Based on the total score on the Hospital Anxiety and Depression rating scale (HADS-T) in HUNT 2 (1995–1997), three groups were identified: Level 3 (n = 654, score ≥ 25 points), Level 2 (n = 654, score 19–24 points), and Level 1 (n = 1,308, score < 19 points). The groups were followed up with a mailed questionnaire after 4 years. Among the 1,326 (53% response rate) who participated in the follow-up, 816 (62%) had not changed symptom level. The number of participants that had crossed the HADS-T caseness level (19 points) was the same in both directions. In non-cases at baseline (Level 1), lack of friends (OR 2.34, 95% CI 1.28–4.27, P = 0.006) and previous episodes of depression (OR 2.90, 95% CI 1.76–4.78, P < 0.001) predicted HADS-T caseness at follow-up, while higher educational level (OR 0.66, 95% CI 0.46–0.96, P = 0.028) protected from developing caseness level of anxiety and depression. In HADS-T cases (Levels 2 and 3) at baseline, previous episode(s) of depression (OR 0.36, 95% CI 0.19–0.68, P = 0.002) and being unemployed (OR 0.58, 95% CI 0.34–1.00, P = 0.050) predicted HADS-T caseness at follow-up, whereas a higher educational level (OR 1.83, 95% CI 1.24–2.70, P = 0.002) was associated with remission from HADS-T caseness after 4 years. Though symptom fluctuation was considerable, conventional HADS-T caseness (≥19 points) was a reliable and valid predictor for high long-term symptom stability of anxiety and depression in our general population sample.