To evaluate the clinical impact of minimal residual disease in multiple myeloma, apheretic products from 51 autotransplanted patients were tested by fluorescent (GeneScan) polymerase chain reaction (PCR). Sixty‐nine per cent of harvests were contaminated when evaluated for IgH rearrangement. Forty‐six patients responded to transplant, with 52·9% achieving complete response (CR). The clinical response of patients was significantly influenced by the number of re‐infused CD34⁺ cells. Positive PCR results of re‐infused harvests were not significantly related to patient outcome. Median overall survival (OS) was 33 months, and a significant advantage for patients transplanted by 12 months from diagnosis was observed. Moreover, OS was longer for patients receiving PCR‐negative stem cells, with 72% of patients surviving to 70 months in the group receiving PCR‐negative harvests vs 48% in the group transplanted with contaminated precursors (not statistically significant). Ex vivo purging caused a reduction of contamination of up to 3 logs; nevertheless, 80% of purged harvests remained PCR‐positive and the purging procedure did not alter response or survival rates. Thus, the failure of a predictive role for this highly sensitive molecular method could be explained by the assumption that in vivo persisting malignant cells are the true source of relapse in MM.