INTRODUCTION: Knowledge, once acquired, degrades over time. Exams that contain questions related to previously acquired knowledge ('retrieval practice questions') may promote retrieval practice and spaced learning, and subsequently prevent knowledge loss. To investigate this hypothesis, we compare the score of retrieval practice questions to regular questions in exams of a two-year (bio)medical study program. METHODS: The two-year "Mechanisms of Health and Disease"-program for biomedical sciences and medical students in Nijmegen (the Netherlands) contains 14 spaced exams of 80 questions each. The percentages of correct-, false-, and non-answers were compared between regular questions and retrieval practice questions. Using Pearson correlations between question scores and exam scores (RiT-values), the impact of retrieval practice questions on the internal consistency of exams was determined. Mixed model analyses determined changes in outcomes across time. RESULTS: Analysis of 2006 regular questions and 1728 retrieval practice questions revealed a significantly higher percentage of correct and false answers, and a significantly lower percentage of non-answers, in retrieval practice questions versus regular questions (all P < 0.05). Scores did not change across time. RiT-values were slightly lower in retrieval practice questions, with a small inverse trend across time. CONCLUSION: Our data indicate preservation of knowledge, possibly related to retrieval practice and/or spaced learning. Although the RiT-values of retrieval practice questions were slightly lower than those of regular questions, the discriminative capacity was well within acceptable range. These data highlight the potency of retrieval practice questions to prevent knowledge decrement, without altering exam quality.