Abstract Background Mammography is the current diagnostic standard for breast cancer screening and monitoring. However, accessibility challenges, accuracy issues and patient discomfort all contribute to reduced patient compliance and utilization, resulting in a need for more effective diagnostic tools. A blood test to detect early-stage breast cancer is sought to increase the screening detection rate, provide more accurate monitoring tools, and improve outcomes for patients. We have previously reported a series of lipidomic studies and derived a lipid signature from plasma enriched extracellular vesicles (EVs) that effectively distinguished people with localized breast cancer from cancer-free controls. Here we report on a significant refinement to the test methodology allowing the assessment of the lipid signature directly from plasma samples and its performance, with the aim of advancing the commercial viability of the test as we move towards clinical application. Methods Lipids in EVs from fasted breast cancer and control blood samples (4 separate cohorts (n = 793)) were extracted and analysed by liquid chromatography-high resolution mass spectrometry (LC-HRAM-MS). Over 400 manually curated lipids were quantified. From these, an independent review and retrospective analysis of the data established a lipid signature. The lipid signature was modelled on each of the cohorts using leave-one-out internal cross-validation. Following variable selection, a lipid signature capable of distinguishing breast cancer samples from controls was derived. Furthermore, we investigated if the lipid signature can be assessed from plasma instead of EVs with the aim of further streamlining the process. We analysed the lipids in cancer and control plasma samples (n = 256) previously used for EV preparations, corresponding to patients from Cohorts 3 and 4, and applied the signature derived using EVs on plasma lipidomic data. NIST Standard Reference Material (SRM) 1950-Metabolites in Frozen Human Plasma was used for inter-laboratory quality control. Results Breast cancer subjects were differentiated from controls by the lipid signature with an area under the curve (AUC) of 0.77–0.89 across four cohorts. Assessing the signature directly from plasma, the test achieved a comparable AUC of 0.84. This improvement would make the test more clinically viable and easier to perform. Conclusion We derived a lipid signature, which shows high potential for distinguishing breast cancer samples from controls directly from plasma instead of EVs, reducing the test complexity. Ongoing studies will optimize the plasma lipidomic signature and prospectively compare the test against mammographic and pathological diagnoses.