Abstract Purpose Development of ipsilateral breast carcinoma following a diagnosis of breast ductal carcinoma in situ (DCIS) has been assumed to represent recurrence of the primary tumour. However, this may not be the case and it is important to know how often recurrences are new primary tumours to ensure appropriate individualised therapy. Experimental Design Ipsilateral primary-recurrence pairs (n=78) were sequenced to test their clonal relatedness. Shared genetic events were identified from whole exome sequencing (n=54 pairs) using haplotype-specific copy number and phylogenetic analysis. The remaining pairs were sequenced by a targeted panel or low-coverage whole genome sequencing. We included 32 non-recurrent DCIS to compare the genetic profiles between recurrent and non-recurrent disease to develop a predictive biomarker. Results We found that 14% of DCIS recurrences were non-clonal, indicative of a new breast carcinoma. Four chromosomal changes (5q, 11q, 17q and 20q) and TP53 mutation were enriched in clonal primaries compared with non-recurrent DCIS (p<0.05, Fisher’s exact test). The prognostic value of TP53 was validated in an independent cohort using immunohistochemistry (HR=3.1; 95% CI 1.3-7.8). Non-clonal DCIS primaries had a very similar genetic profile to non-recurrent DCIS, suggesting this subset of cases would be identified as being at “low risk” of recurrence using tumour-intrinsic markers. Conclusions We have identified a substantial rate of new ipsilateral primary carcinomas after a diagnosis of DCIS. Our results suggest, as with invasive breast cancer, that if a recurrent tumour with an independent origin occurs, then the patient is at a high risk through the environment, the breast microenvironment and/or has a genetic predisposition. Importantly, the frequency of new primaries will influence the interpretation of findings in DCIS biomarker discovery studies as the true recurrence rate will be incorrect, affecting these efforts. Translational relevance Our finding that >10% of recurrent tumours are new primaries provides genetic evidence that the presence of DCIS confers a risk of a de novo breast cancer as well as recurrence. Identifying a biomarker of such risks might allow preventive actions, such as genetic testing, chemoprevention with tamoxifen or aromatase inhibitors, or bilateral mastectomy. The corollary of these findings is that de novo primaries in DCIS biomarker studies may have undermined efforts to find a biomarker of recurrence by reducing statistical power, since a tumour cell-intrinsic marker is unlikely to be predictive for a new primary. Even if a tumour molecular biomarker could stratify between non-recurrent and recurrent patients, it will under-detect patients at risk of new primaries. This issue raises concerns about utilising only a tumour cell-intrinsic biomarker in the clinical setting.