Clinical Implications of Auricular Keloid Subsite for Multisite Keloid Involvement: An Exploratory Retrospective Cohort Study

Background: Auricular keloids are common, but whether anatomic subsite identifies patients with broader keloid susceptibility remains uncertain. Because ear piercing can involve repeated local trauma and chronic inflammation, the anatomic subsite and inciting cause may be closely related. The authors explored whether the auricular subsite was associated with multisite keloid involvement. Methods: The authors reviewed 182 consecutive patients who underwent excision and postoperative radiotherapy for auricular keloids at a single tertiary center. Multisite involvement was defined as keloids at one or more additional body sites. Logistic regression estimated associations between subsite and multisite involvement; additional analyses examined collinearity, model fit, discrimination, trend across a prespecified subsite order, and sensitivity after excluding the small retroauricular subgroup. Results: Multisite involvement differed across all 4 subsites ( P =0.020): earlobe 72.0% (54/75), helix 59.0% (49/83), multiple auricular 50.0% (9/18), and retroauricular 16.7% (1/6). The fully adjusted earlobe-versus-helix estimate was directional but did not reach statistical significance (OR: 1.99, 95% CI: 0.94–4.18, P =0.070). Piercing was independently associated with multisite involvement (OR: 7.10, 95% CI: 2.10–23.95, P =0.002). After excluding retroauricular cases, the three-group subsite comparison was not statistically significant ( P =0.107). The Cochran-Armitage trend test was also not significant ( P =0.823). Collinearity diagnostics did not show problematic multicollinearity (maximum VIF 1.42); model discrimination was modest (C-statistic 0.682), with no evidence of poor calibration by Hosmer-Lemeshow testing ( P =0.337). Conclusions: Auricular subsite showed descriptive differences in multisite keloid involvement, but the primary earlobe-versus-helix comparison and prespecified trend test were not statistically significant. These findings should be interpreted as exploratory and hypothesis-generating rather than confirmatory evidence of a constitutional-to-reactive spectrum.