OBJECTIVES: Sample quality in prospective long-term drug safety studies can be impaired by selective lost-to-follow-up. Attrition will especially bias the sample, when patients with relevant risk factors selectively drop out. In this case, effects in endpoints cannot be related to study-relevant independent variables. The present contribution will demonstrate how careful follow-up procedures can prevent disease-related drop-out bias of the sample.  METHODS: For a long-term prospective safety study of new Oral Contraceptive (OC) 25,213 women aged 20 to 40 years were enrolled from gynecological practices in Germany. The women filled-in a baseline questionnaire and were followed-up over two years with four follow-up questionnaires in total. Whenever safety-relevant signs were reported in the questionnaires, physicians validated the report. After two years 12,823 women were still in the sample and completed the fourth questionnaire. Disease differences between the “Retained” and the “Lost” group, which could indicate sample bias, were analysed using multivariate methods. RESULTS: The “retained” and the “lost-to-follow-up” group did not differ in initial disease status or in risk factor at study start: High blood pressure:  2.9% in “Retained Group”; 2.7% in “Lost Group” (phi=.006; n.s.), diabetes : 0.6% vs. 0.6% (phi=.001; n.s.), high cholesterol:  2.6% vs. 2.4% (phi=.007; n.s.),venous thrombosis : .8% vs. .8% (phi=.002; n.s.), smoker-rate : 34.9% vs. 42.3% (phi=.127), BMI>30: r=.004 (n.s.), age r=.048 (n.s.). The results show that drop-out of the initial sample is not related to study relevant morbidity and that sample bias cannot be concluded. CONCLUSIONS: Careful follow-up methods guarantee low lost-to-follow-up in longterm prospective studies of drug safety. Since drop-out cannot be attributed to study-relevant confounders, attrition does not lead to sample bias.