Purpose: To compare the efficiency of different cytogenetic tools in estimating the doses received by four people involved in the Lilo accident and to monitor the dose estimate over 4.5 years. Materials and methods: Several young Georgian frontier guards handled at least one of the 12 Caesium sources found in a former Russian military camp. Overexposure lasted from July 1996 to May 1997. The Institute for Radiological Protection and Nuclear Safety (IRSN) obtained blood samples taken at several intervals post-exposure from the four most highly-exposed people. Dose estimation was performed using dicentric and translocation scoring. Results: The first dose estimations performed by dicentric scoring gave whole-body doses ranging from 0.4 to 1.3 Gy. Overexposure was complex and several mathematical models were used to take this complexity into account. This could provide information concerning the circumstances of overexposure. Concerning follow-up, the yield of dicentrics decreased by about 50% in the first 4 months following the end of overexposure whereas translocations were stable over the period of analysis. Conclusion: It has been useful to compare cytogenetic results with clinical results. The results presented here reveal good stability of translocations. However the first dose estimation was not attempted until 6 months after the last exposure.