Axillary evaluation in ductal cancer in situ of the breast: challenging the diagnostic accuracy of clinical practice guidelines

Abstract

Abstract Background Staging of the axilla is not routine in ductal cancer in situ (DCIS) although invasive cancer is observed in 20–25 per cent of patients at final pathology. Upfront sentinel lymph node dissection (SLND) is advocated in clinical practice guidelines in certain situations. These include expected challenges in subsequent SLN detection and when the risk for invasion is high. Clinical practice guidelines are, however, inconsistent and lead to considerable practice variability. Methods Clinical practice guidelines for upfront SLND in DCIS were identified and applied to patients included in the prospective SentiNot study. These patients were evaluated by six independent, blinded raters. Agreement statistics were performed to assess agreement and concordance. Receiver operating characteristic curves were constructed, to assess guideline accuracy in identifying patients with underlying invasion. Results Eight guidelines with relevant recommendations were identified. Interobserver agreement varied greatly (kappa: 0.23–0.9) and the interpretation as to whether SLND should be performed ranged from 40–90 per cent and with varying concordance (32–88 per cent). The diagnostic accuracy was low with area under the curve ranging from 0.45 to 0.55. Fifty to 90 per cent of patients with pure DCIS would undergo unnecessary SLNB, whereas 10–50 per cent of patients with invasion were not identified as ‘high risk’. Agreement across guidelines was low (kappa = 0.24), meaning that different patients had a similar risk of being treated inaccurately. Conclusion Available guidelines are inaccurate in identifying patients with DCIS who would benefit from upfront SLNB. Guideline refinement with detailed preoperative work-up and novel techniques for SLND identification could address this challenge and avoid overtreatment.