Search for a command to run...
Keloids are benign fibro-proliferative tumours resulting from dysregulated wound healing with chronic inflammation and excessive collagen deposition.1 Multiple therapeutic approaches exist, yet recurrence rates remain high. Surgical excision followed by tailored post-operative radiotherapy can reduce recurrence from 50%–80% to <10%, particularly when doses are adjusted for high-risk sites such as the chest.2 Cryosurgery remains a minimally invasive option that induces controlled lesional necrosis and subsequent lesion regression.3 Intralesional cryosurgery allows uniform deep freezing of fibrotic tissue while sparing surrounding skin, overcoming the limitations of contact cryosurgery.4-6 The CryoShapeTM device standardized this approach,7 but predicted hold times derived from earlier algorithmic models8 have not yet been validated prospectively. We, therefore, evaluated the accuracy of algorithmic predictions for intralesional cryosurgery hold times and examined anatomical factors influencing treatment outcomes. Twenty-five patients (51 keloids; median age 28 years, 52% male) underwent intralesional cryosurgery using the CryoShape™ device. Lesions were located on the ears (n = 10), shoulders (n = 13), sternum (n = 16), neck (n = 6), inguinal (n = 4), pectoral (n = 1) and chin (n = 1). The patients have given written informed consent to the publication of their case details. Both treatment-naïve and previously treated keloids were analysed together to reflect real-world practice. Predicted hold times were obtained from previously published thermal models based on keloid volume and probe geometry,8 where linear regression functions (y = a x + b) were derived for auricular, sternal and shoulder sites (Figure 1). Actual hold times corresponded to the freezing durations recorded by the CryoShape™ device during treatment until the entire keloid was completely frozen. Differences between predicted and actual values were analysed using the Wilcoxon signed-rank test for paired non-parametric data (two-tailed, significance set at p < 0.05). No correction for multiple comparisons was applied given the exploratory nature and limited sample size of this pilot study. Significant discrepancies were found between predicted and actual hold times (p = 0.00004). The closest alignment occurred in ear keloids (p = 0.008), whereas shoulder (p = 0.0005) and sternal (p = 0.0002) lesions showed greater deviations. Sternal keloids displayed a bimodal distribution, implying distinct thermal responses depending on proximity to bone or soft tissue—evidence that tissue composition and heat capacity influence freezing dynamics. Long-term volume reduction was assessed in eight representative patients (11 keloids) 6 months post-treatment, revealing a median 72.4% reduction (p = 0.002). Overall, 9 out of 11 keloids were reduced after 6 months (Figure 2). However, volume data were available only for this subset, which may introduce selection bias. No adverse events, including infection or unwanted tissue necrosis, occurred. Necrosis within the lesion represents the intended therapeutic mechanism enabling regression, while surrounding skin remained intact. These findings confirm intralesional cryosurgery as an effective and safe therapy for keloids while demonstrating that algorithmic hold-time predictions require refinement for clinical use. Observed discrepancies likely reflect anatomical variations in vascularity, tissue density and underlying structures as well as minor device modifications in the current CryoShape™ model. This single-centre pilot design, limited sample size, absence of a direct comparison with other cryosurgical techniques and lack of skin-type stratification restrict generalisability. Moreover, ear lesions were not differentiated between auricular and earlobe sites, which may show distinct responses and recurrence risks. Additionally, long-term follow-up data were available only for a subset of patients, which may introduce selection bias. Future predictive models should incorporate parameters such as bone proximity, vascularity and tissue density to improve accuracy and better account for the anatomical variability observed across different keloid sites. Based on our findings, actual freezing durations in thick sternal and shoulder keloids were approximately 15%–20% longer than algorithmic predictions, suggesting that clinicians should interpret these models with this margin in mind, while ear lesions generally conformed to existing estimates. Larger multicentre cohorts and integration of device-based thermal feedback could further standardize and optimize outcomes. Intralesional cryosurgery remains a reproducible, minimally invasive procedure providing substantial keloid regression when guided by refined algorithmic models. This article has no funding source. All authors have no relevant conflict of interest to declare. Since the data evaluated the results of medical treatments not included in a clinical study, no Ethics Committee approval was required. The patients in this manuscript have given written informed consent to the publication of their case details. The data that support the findings of this study are available from the corresponding author upon reasonable request.