Detection of disseminated tumour cells (DTCs) in bone marrow by immunocytochemistry (ICC) includes morphological evaluation of cytokeratin immunopositive cells. The aim of this study was to disclose the prognostic significance of different morphological categories of ICC-positive cells according to treatment status and tumour subtype. Bone marrow samples (at surgery) were analysed for the presence of cytokeratin-positive DTCs by a standard immunocytochemical method. The immunopositive cells were classified into the following categories, prior to any analysis of the association between DTCs and clinical outcome: tumour cells (TC), uninterpretable cells (UIC), hematopoietic cells (HC), and questionable HC (QHC). The analysis included 747 early breast cancer patients. Median follow-up was 84 months for relapse, and 99 months for death. The categorisation of the ICC positive cells revealed TC in 13.3 % of the patients, whereas 13.1, 17.8, and 21.4 % of the cases were positive for UIC, QHC, and HC, respectively. Analysing all patients, only TC and UIC predicted systemic relapse. Separate analysis of all patients not receiving adjuvant systemic treatment (No-Adj; n = 389) showed that only QHC were associated with reduced survival (DDFS: p = 0.008; BCSS: p = 0.004, log rank) and the presence of QHC also remained significant in multivariate analysis. Primary tumour subgroup analysis (of all patients) by hormone receptors (HR) and HER2, demonstrated that only TC/UIC had prognostic impact in the HR+/HER2- patients, whereas presence of QHC was associated with unfavourable outcome only in triple negative patients (DDFS: p = 0.004; BCSS: p = 0.024). Patients with ≥3HC had improved outcome compared to those with fewer/no HC (DDFS: p = 0.005; BCSS: p = 0.009). Hence, morphological DTC subgroups may differ in clinical significance according to primary tumour subtype and treatment status. This emphasises the importance of DTC characterisation, and separate analyses of DTC categories according to tumour subtype. Hematopoietic ("false positive") cells might predict an immune-related favorable clinical outcome.