Anti-PD-(L)1 plus BRAF/MEK inhibitors (triplet therapy) after failure of immune checkpoint inhibition and targeted therapy in patients with advanced melanoma.

BACKGROUND: Effective treatment options are limited for patients with advanced melanoma who have progressed on immune checkpoint inhibitors (ICI) and targeted therapies (TT). Preclinical models support the combination of ICI with TT; however, clinical trials evaluating the efficacy of triplet combinations in first-line setting showed limited advantage compared to TT only. METHODS: We conducted a retrospective, multicenter study, that included patients with advanced melanoma who were treated with BRAF/MEK inhibitors in combination with an anti-PD-(L)1 antibody (triplet therapy) after failure of at least one anti-PD-(L)1-based therapy and one TT in seven major melanoma centers between February 2016 and July 2022. RESULTS: A total of 48 patients were included, of which 32 patients, 66.7% had brain metastases, 37 patients (77.1%) had three or more metastatic organs and 21 patients (43.8%) had three or more treatment lines. The median follow-up time was 31.4 months (IQR, 22.27-40.45 months). The treatment with triplet therapy resulted in an ORR of 35.4% (n = 17) and a DCR of 47.9% (n = 23). The median DOR was 5.9 months (range, 3.39-14.27 months). Patients treated with BRAF/MEK inhibitors as the last treatment line showed a slightly lower ORR (29.6%) compared to patients who received ICI or chemotherapy last (ORR: 42.9%). Grade 3-4 treatment-related adverse events occurred in 25% of patients (n = 12), with seven patients (14.6%) requiring discontinuation of treatment with both or either drug. CONCLUSIONS: Triplet therapy has shown activity in heavily pretreated patients with advanced melanoma and may represent a potential treatment regimen after failure of ICI and TT.

Federated Learning for Decentralized Artificial Intelligence in Melanoma Diagnostics.

Importance: The development of artificial intelligence (AI)-based melanoma classifiers typically calls for large, centralized datasets, requiring hospitals to give away their patient data, which raises serious privacy concerns. To address this concern, decentralized federated learning has been proposed, where classifier development is distributed across hospitals. Objective: To investigate whether a more privacy-preserving federated learning approach can achieve comparable diagnostic performance to a classical centralized (ie, single-model) and ensemble learning approach for AI-based melanoma diagnostics. Design, Setting, and Participants: This multicentric, single-arm diagnostic study developed a federated model for melanoma-nevus classification using histopathological whole-slide images prospectively acquired at 6 German university hospitals between April 2021 and February 2023 and benchmarked it using both a holdout and an external test dataset. Data analysis was performed from February to April 2023. Exposures: All whole-slide images were retrospectively analyzed by an AI-based classifier without influencing routine clinical care. Main Outcomes and Measures: The area under the receiver operating characteristic curve (AUROC) served as the primary end point for evaluating the diagnostic performance. Secondary end points included balanced accuracy, sensitivity, and specificity. Results: The study included 1025 whole-slide images of clinically melanoma-suspicious skin lesions from 923 patients, consisting of 388 histopathologically confirmed invasive melanomas and 637 nevi. The median (range) age at diagnosis was 58 (18-95) years for the training set, 57 (18-93) years for the holdout test dataset, and 61 (18-95) years for the external test dataset; the median (range) Breslow thickness was 0.70 (0.10-34.00) mm, 0.70 (0.20-14.40) mm, and 0.80 (0.30-20.00) mm, respectively. The federated approach (0.8579; 95% CI, 0.7693-0.9299) performed significantly worse than the classical centralized approach (0.9024; 95% CI, 0.8379-0.9565) in terms of AUROC on a holdout test dataset (pairwise Wilcoxon signed-rank, P < .001) but performed significantly better (0.9126; 95% CI, 0.8810-0.9412) than the classical centralized approach (0.9045; 95% CI, 0.8701-0.9331) on an external test dataset (pairwise Wilcoxon signed-rank, P < .001). Notably, the federated approach performed significantly worse than the ensemble approach on both the holdout (0.8867; 95% CI, 0.8103-0.9481) and external test dataset (0.9227; 95% CI, 0.8941-0.9479). Conclusions and Relevance: The findings of this diagnostic study suggest that federated learning is a viable approach for the binary classification of invasive melanomas and nevi on a clinically representative distributed dataset. Federated learning can improve privacy protection in AI-based melanoma diagnostics while simultaneously promoting collaboration across institutions and countries. Moreover, it may have the potential to be extended to other image classification tasks in digital cancer histopathology and beyond.

Dermatologist-like explainable AI enhances trust and confidence in diagnosing melanoma.

Artificial intelligence (AI) systems have been shown to help dermatologists diagnose melanoma more accurately, however they lack transparency, hindering user acceptance. Explainable AI (XAI) methods can help to increase transparency, yet often lack precise, domain-specific explanations. Moreover, the impact of XAI methods on dermatologists' decisions has not yet been evaluated. Building upon previous research, we introduce an XAI system that provides precise and domain-specific explanations alongside its differential diagnoses of melanomas and nevi. Through a three-phase study, we assess its impact on dermatologists' diagnostic accuracy, diagnostic confidence, and trust in the XAI-support. Our results show strong alignment between XAI and dermatologist explanations. We also show that dermatologists' confidence in their diagnoses, and their trust in the support system significantly increase with XAI compared to conventional AI. This study highlights dermatologists' willingness to adopt such XAI systems, promoting future use in the clinic.

Standardized Computer-Assisted Analysis of 5-hmC Immunoreactivity in Dysplastic Nevi and Superficial Spreading Melanomas.

5-Hydroxymethylcytosine (5-hmC) is an important intermediate of DNA demethylation. Hypomethylation of DNA is frequent in cancer, resulting in deregulation of 5-hmC levels in melanoma. However, the interpretation of the intensity and distribution of 5-hmC immunoreactivity is not very standardized, which makes its interpretation difficult. In this study, 5-hmC-stained histological slides of superficial spreading melanomas (SSM) and dysplastic compound nevi (DN) were digitized and analyzed using the digital pathology and image platform QuPath. Receiver operating characteristic/area under the curve (ROCAUC) and t-tests were performed. A p-value of <0.05 was used for statistical significance, and a ROCAUC score of >0.8 was considered a "good" result. In total, 92 5-hmC-stained specimens were analyzed, including 42 SSM (45.7%) and 50 DN (54.3%). The mean of 5-hmC-positive cells/mm2 for the epidermis and dermo-epidermal junction and the entire lesion differed significantly between DN and SSM (p = 0.002 and p = 0.006, respectively) and showed a trend towards higher immunoreactivity in the dermal component (p = 0.069). The ROCAUC of 5-hmC-positive cells of the epidermis and dermo-epidermal junction was 0.79, for the dermis 0.74, and for the entire lesion 0.76. These results show that the assessment of the epidermal with junctional expression of 5-hmC is slightly superior to dermal immunoreactivity in distinguishing between DN and SSM.

Threshold Optimization for Tumor Markers S100b and MIA in Uveal Melanoma - A Single Center Analysis.

BACKGROUND/AIM: Uveal melanoma (UM) is the most common malignant tumor of the eye in adults. Metastases develop in 50% of the patients, predominantly in the liver. In UM, the cut-off concentrations of the blood-based tumor markers S100b and MIA are inconclusive. PATIENTS AND METHODS: In this retrospective monocenter study, we statistically evaluated 1,878 S100b and 1,768 MIA measurements in 244 patients with UM from 2011-2020. Threshold optimization was performed using receiver operating characteristic (ROC) curves. RESULTS: A total of 171 patients with non-metastatic UM (nmUM) and 73 patients with metastatic UM (mUM) showed no differences in sex, age at diagnosis or the affected eye. In mUM, 80% of the patients developed metastases to the liver at a median of 46 months after initial diagnosis. The sensitivity and specificity of S100b was 16.10% and 94.52%, and that of MIA was 31.86% and 81.42%, respectively. ROC curves revealed poor values for the area under the curve of 0.57 for S100b and 0.55 for MIA. The optimal cut-off concentration to detect metastases was 0.14 µg/l for S100b and 17.4 ng/ml for MIA. With at least one tumor marker elevated, optimized sensitivity was 20.40% and specificity 96.76%. CONCLUSION: Current thresholds for S100b and MIA in UM are not able to detect early metastatic disease and require additional diagnostics to clarify false positive results. Threshold optimization considering both S100b and MIA results in a better diagnostic validity with an acceptable specificity and a poor sensitivity. Highly sensitive blood-based and imaging methods to detect metastases early in UM are urgently needed.

Tirbanibulin 1% Ointment Significantly Reduces the Actinic Keratosis Area and Severity Index in Patients with Actinic Keratosis: Results from a Real-World Study.

BACKGROUND: Actinic keratosis (AK) is a cutaneous lesion resulting from the proliferation of atypical epidermal keratinocytes caused by long-term exposure to ultraviolet radiation. AK may progress to cutaneous squamous cell carcinoma (cSCC) and therefore is often treated with topical agents such as 5-fluorouracil, diclofenac, imiquimod, and photodynamic therapy. Tirbanibulin has been approved based on two phase III trials in the USA. However, real-world evidence for tirbanibulin is absent. METHODS: This was a single-centre study of adult patients with clinically typical, visible AK on the face or scalp treated with tirbanibulin 1% ointment. Treatment was administered as per label once daily for 5 consecutive days on the same lesions or field. Treatment outcomes were assessed 4 weeks after treatment, with additional optional assessments conducted at later time points. Efficacy was measured using the actinic keratosis area and severity index (AKASI) and digital dermoscopy. RESULTS: A total of 33 patients were treated of whom 30 were analysed. The median AKASI score was 5.6 (1.4-11) pre-treatment and 1.2 (0-7.4) post-treatment (p < 0.0001). Complete clearance as defined by AKASI scores less than 1 was achieved in 47% (n = 14) and 57% (n = 13) at the first and second follow-up, respectively. All local reactions resolved spontaneously and without sequelae. The most common local reactions were erythema (80%, n = 26) and flaking or scaling (43%, n = 13). CONCLUSIONS: Tirbanibulin 1% ointment significantly and rapidly reduced the AKASI score in a real-world setting. The complete clearance rates were in line with those observed in the two pivotal trials.

Adjuvant treatment and outcome of stage III melanoma patients: Results of a multicenter real-world German Dermatologic Cooperative Oncology Group (DeCOG) study.

PURPOSE: Clinical trials demonstrated significantly improved recurrence-free survival (RFS) of melanoma patients receiving adjuvant treatment. As data from controlled trials are based on selected populations, we investigated adjuvantly treated stage III melanoma patients under real-world conditions. PATIENTS AND METHODS: In a prior multicenter cohort study, stage III-IV melanoma patients were analysed for their choice of adjuvant therapy. In this follow-up study, we examined RFS, overall and melanoma-specific survival (MSS) and response to the subsequent treatment of 589 stage III patients (232 BRAF-mutated) receiving adjuvant PD-1 inhibitors (PD1; n = 479) or targeted therapy (TT; n = 110). RESULTS: The median follow-up of the total cohort was 25.7 months. The main reason for premature discontinuation of adjuvant therapy was disease progression in PD1- (28.8%, n = 138/479) and adverse events in TT-treated patients (28.2%, n = 31/110). Among BRAF-mutated patients, RFS at 24 months was 49% (95% CI 40.6-59.0%) for PD1- and 67% (95% CI 58-77%) for TT-treated patients. The risk of recurrence was higher for BRAF-mutated PD1 than TT (hazard ratio 1.99; 95% CI 1.34-2.96; hazard ratio adjusted for age, sex and tumour stage, 2.21; 95% CI 1.48-3.30). Twenty-four months MSS was 87% (95% CI 81.0-94.1) for PD1 and 92% (95% CI 86.6-97.0) for TT. Response to subsequent systemic treatment for unresectable disease was 22% for all PD1- and 16% for TT-treated patients. CONCLUSIONS: PD1-treated patients had more and earlier recurrences than TT patients. In BRAF-mutated patients, adjuvant TT might prevent early recurrences more effectively than PD1 treatment. Management of recurrence despite adjuvant treatment is challenging, with low response to current therapeutic options.

Opportunities to inform German residents about the possibility of skin cancer screening and to inform stakeholders to take appropriate actions: A qualitative approach.

BACKGROUND: The national skin cancer screening (SCS) was introduced in Germany in 2008. However, public awareness and participation rates remain low. There are no campaigns or target group-specific invitation strategies for SCS yet. Thus, our aim was to derive potential suggestions on how to best inform German residents about the possibility of SCS. METHODS: Semi-structured, individual interviews with male and female German residents aged ≥35 years were conducted in Erlangen (Germany) to explore opportunities on raising awareness of SCS. Interviews were audiotaped, transcribed verbatim, and analyzed using qualitative content analysis. RESULTS: Overall, 39 persons were interviewed. About 79.5% (31/39) had already undergone at least one SCS. Numerous opportunities to raise awareness of the possibility of SCS were suggested which were categorized into three main topics: the role of public promotion, health-related caregivers, and health insurance. Similar themes were identified for inviting entitled persons to undergo SCS after 2 years. Furthermore, age-dependent communication approaches were proposed, that is, younger persons should be approached electronically, while the older generation should be targeted with traditional media like mail. CONCLUSIONS: The results of this project will inform stakeholders to take appropriate actions. The findings may contribute to increase participation rates in SCS and thus lead to earlier detection of skin cancer.

Standardized Computer-Assisted Analysis of PRAME Immunoreactivity in Dysplastic Nevi and Superficial Spreading Melanomas.

PRAME (PReferentially expressed Antigen in MElanoma) is a cancer testis antigen that is frequently expressed in melanoma compared to benign melanocytic proliferations and nevi. However, the interpretation of the intensity and distribution of PRAME immunostaining is not standardized a lot, which makes interpretation difficult. PRAME-stained histological slides of superficial spreading melanomas (SSM) and dysplastic nevi (DN) were digitized and analyzed using the digital pathology and image platform QuPath. t-tests and ROC AUCs were performed with SPSS. A p-value of <0.05 was used for statistical significance, and a ROC AUC score of >0.8 was considered a good result. A cut-off score was defined in an evaluation cohort and subsequently analyzed in an independent validation cohort. In total, 81 PRAME-stained specimens were included. The evaluation cohort included 32 (50%) SSM and 32 (50%) DN, and the mean of PRAME-positive cells/mm2 for the entire lesion was 455.3 (SD 428.2) in SSM and 60.5 (SD 130.1; p < 0.001) in DN. The ROC AUC of PRAME-positive cells of the entire lesion was 0.866, and in the epidermis it was 0.901. The defined cut-off score to distinguish between DN and SSM was 97.67 cells/mm2. In the validation cohort, 16 out of 17 cases (94.1%) were correctly classified by the cut-off score. The computer-aided assessment of PRAME immunostaining is a useful tool in dermatopathology to distinguish between DN and SSM. Lesions with a moderate expression and indifferent morphologic features will remain a challenge for dermatopathologists.

S1-guideline cutaneous and subcutaneous leiomyosarcoma.

Superficial leiomyosarcomas (LMS) are rare skin cancers (2-3% of cutaneous sarcomas) that originate from dermally located hair follicle muscles, dartos or areolar muscles (cutaneous/dermal LMS), or from vascular muscle cells of the subcutaneous adipose tissue (subcutaneous LMS). These superficial LMS are distinct from LMS of the deep soft tissues. Leiomyosarcomas are typically localized at the lower extremities, trunk or capillitium, and present as painful, erythematous to brownish nodules. Diagnosis is made by histopathology. The treatment of choice for primary LMS is complete (R0) microscopically controlled excision, with safety margins of 1 cm in dermal LMS, and 2 cm in subcutaneous LMS, if possible. Non-resectable or metastatic LMS require individual treatment decisions. After R0 resection with 1 cm safety margins, the local recurrence rate of dermal LMS is very low, and metastasis is very rare. Subcutaneous LMS, very large, or incompletely excised LMS recur and metastasize more frequently. For this reason, clinical follow-up examinations are recommended every six months for cutaneous LMS, and every three months for subcutaneous LMS within the first two years (in subcutaneous LMS including locoregional lymph node sonography). Imaging such as CT/MRI is indicated only in primary tumors with special features, recurrences, or already metastasized tumors.

Conventional and three-dimensional photography as a tool to map distribution patterns of in-transit melanoma metastases on the lower extremity.

Background: In melanoma, in-transit metastases characteristically occur at the lower extremity along lymphatic vessels. Objectives: The objective of this study was to evaluate conventional or three-dimensional photography as a tool to analyze in-transit metastasis pattern of melanoma of the lower extremity. In addition, we assessed risk factors for the development of in-transit metastases in cutaneous melanoma. Methods: In this retrospective, monocentric study first we compared the clinical data of all evaluable patients with in-transit metastases of melanoma on the lower extremity (n = 94) with melanoma patients without recurrence of disease (n = 288). In addition, based on conventional (n = 24) and three-dimensional photography (n = 22), we defined the specific distribution patterns of the in-transit metastases on the lower extremity. Results: Using a multivariate analysis we identified nodular melanoma, tumor thickness, and ulceration as independent risk factors to develop in-transit metastases ITM (n = 94). In patients with melanoma on the lower leg (n = 31), in-transit metastases preferentially developed along anatomically predefined lymphatic pathways. In contrast when analyzing in-transit metastases of melanoma on the foot (n = 15) no clear pattern could be visualized. In addition, no difference in distance between in-transit metastases and primary melanoma on the foot compared to the lower leg was observed using three-dimensional photography (n = 22). Conclusion: A risk-adapted follow-up of melanoma patients to detect in-transit metastases can be applied by knowledge of the specific lymphatic drainage of the lower extremity. Our current analysis suggests a more complex lymphatic drainage of the foot.

Immune signatures predict development of autoimmune toxicity in patients with cancer treated with immune checkpoint inhibitors.

BACKGROUND: Immune checkpoint inhibitors (ICIs) are among the most promising treatment options for melanoma and non-small cell lung cancer (NSCLC). While ICIs can induce effective anti-tumor responses, they may also drive serious immune-related adverse events (irAEs). Identifying biomarkers to predict which patients will suffer from irAEs would enable more accurate clinical risk-benefit analysis for ICI treatment and may also shed light on common or distinct mechanisms underpinning treatment success and irAEs. METHODS: In this prospective multi-center study, we combined a multi-omics approach including unbiased single-cell profiling of over 300 peripheral blood mononuclear cell (PBMC) samples and high-throughput proteomics analysis of over 500 serum samples to characterize the systemic immune compartment of patients with melanoma or NSCLC before and during treatment with ICIs. FINDINGS: When we combined the parameters obtained from the multi-omics profiling of patient blood and serum, we identified potential predictive biomarkers for ICI-induced irAEs. Specifically, an early increase in CXCL9/CXCL10/CXCL11 and interferon-γ (IFN-γ) 1 to 2 weeks after the start of therapy are likely indicators of heightened risk of developing irAEs. In addition, an early expansion of Ki-67+ regulatory T cells (Tregs) and Ki-67+ CD8+ T cells is also likely to be associated with increased risk of irAEs. CONCLUSIONS: We suggest that the combination of these cellular and proteomic biomarkers may help to predict which patients are likely to benefit most from ICI therapy and those requiring intensive monitoring for irAEs. FUNDING: This work was primarily funded by the European Research Council, the Swiss National Science Foundation, the Swiss Cancer League, and the Forschungsförderung of the Kantonsspital St. Gallen.

One Website to Gather them All: Usability Testing of the New German SKin Cancer INFOrmation (SKINFO) Website-A Mixed-methods Approach.

Skin cancer patients increasingly search the internet to acquire disease-related information. However, information on the internet may be misleading. Recently, SKINFO has been launched, a website exclusively created for German-speaking skin cancer patients providing information as well as additional resources of verified quality. Here, we describe the results of the first usability test of SKINFO using a mixed-methods approach. Ten adult patients with skin cancer were recruited for usability testing in the skin cancer units of the University Hospitals of Erlangen and Dresden, Germany. Testing consisted of three different scenarios where patients were asked to find specific information on the SKINFO website guided by the think-aloud method. Descriptive analysis and content analyses were performed. All patients would recommend SKINFO and appreciated its content, design, and structure. Think-aloud analysis identified the topics layout, navigation, and content and structure which would benefit from refinement. Major criticism included the navigation through the website, and the desire for more specific information addressing patients' relatives and the latest, up-to-date information. Overall, usability testing showed that the unique web-based information platform has the potential to support patients coping with skin cancer and thus strengthen informed decision-making.

Characterization of the impact of immune checkpoint inhibitors on platelet activation and aggregation.

Immune checkpoint inhibitors (ICIs) are effective oncological drugs which block cellular check-point receptors typically targeted by tumor immune evasion strategies. Despite their benefits, clinicians have reported treatment-associated thromboembolism during ICI therapy in recent years. Though several theories on this ICI-associated pathogenesis exist, the direct effects of ICIs on platelets remains unknown. We therefore investigated the potential direct and indirect effect of PD-1, PD-L1 and CTLA-4-targeting ICIs on platelet functionality in multifaceted in vitro experiments. Interestingly, we could not observe a clear effect of ICI on platelet aggregation and primary hemostasis in whole blood and platelet concentrate-based assays. Furthermore, the presence of ICIs in toll-like receptor stimulation had no significant impact on platelet surface marker expression. In a second approach, we investigated the indirect immunological impact of ICIs on platelet activation by exposing platelets to supernatants from ICI- and Staphylococcal enterotoxin B-exposed PBMCs. Whereas ICIs affected IL-2 levels in supernatants, we could not detect clear differences in the secretion of pro-thrombogenic factors and platelet responses. The obtained data suggest that the direct influence of ICIs on platelet activation or the influence of altered T cell function on platelet activation cannot be considered a major factor in the development of thrombotic events.

Case report: Patient specific combination of surgery and immunotherapy in advanced squamous cell carcinoma of the head and neck - a case series and review of literature.

Background: Prognosis of patients with recurrent or metastatic head and neck cancer is generally poor. Adjuvant immunotherapy (IT) featuring immune checkpoint inhibition (ICI) is standard of care in advanced stage head and neck squamous cell carcinoma (HNSCC) and cutaneous squamous cell carcinoma (CSCC). ICI response rates in CSCC are described as higher than in HNSCC. IT is constantly shifting into earlier disease stages which confronts the surgeon with immunotherapeutically pre-treated patients. It is therefore becoming increasingly difficult to assess which patients with symptomatic tumor disease and a lack of curative surgical option might benefit from salvage surgery. Case presentations: The following 6 cases describe therapeutic decision-making regarding ICI and (salvage) surgery in patients with advanced stage HNSCC or CSCC. Cases A and B focus on neoadjuvant ICI followed by salvage surgery. In Cases C and D salvage surgery was performed after short-term stabilization with partial response to ICI. The last two cases (Cases E and F) address the surgical approach after failure of ICI. All cases are discussed in the context of the current study landscape and with focus on individual decision-making. For better understanding, a timetable of the clinical course is given for each case. Conclusions: ICI is rapidly expanding its frontiers into the neoadjuvant setting, frequently confronting the surgeon with heavily pretreated patients. Salvage surgery is a viable therapeutic concept despite the rise of systemic treatment options. Decision-making on surgical intervention in case of a salvage surgery remains an individual choice. For neoadjuvant ICI monitoring regarding pathological tumor response or tumor necrosis rate, we suggest correlation between the initial biopsy and the definite tumor resectate in order to increase its significance as a surrogate marker. Scheduling of neoadjuvant ICI should be further investigated, as recent studies indicate better outcomes with shorter time frames.

Melanoma 2.0. Skin cancer as a paradigm for emerging diagnostic technologies, computational modelling and artificial intelligence.

We live in an unprecedented time in oncology. We have accumulated samples and cases in cohorts larger and more complex than ever before. New technologies are available for quantifying solid or liquid samples at the molecular level. At the same time, we are now equipped with the computational power necessary to handle this enormous amount of quantitative data. Computational models are widely used helping us to substantiate and interpret data. Under the label of systems and precision medicine, we are putting all these developments together to improve and personalize the therapy of cancer. In this review, we use melanoma as a paradigm to present the successful application of these technologies but also to discuss possible future developments in patient care linked to them. Melanoma is a paradigmatic case for disruptive improvements in therapies, with a considerable number of metastatic melanoma patients benefiting from novel therapies. Nevertheless, a large proportion of patients does not respond to therapy or suffers from adverse events. Melanoma is an ideal case study to deploy advanced technologies not only due to the medical need but also to some intrinsic features of melanoma as a disease and the skin as an organ. From the perspective of data acquisition, the skin is the ideal organ due to its accessibility and suitability for many kinds of advanced imaging techniques. We put special emphasis on the necessity of computational strategies to integrate multiple sources of quantitative data describing the tumour at different scales and levels.

Model soups improve performance of dermoscopic skin cancer classifiers.

BACKGROUND: Image-based cancer classifiers suffer from a variety of problems which negatively affect their performance. For example, variation in image brightness or different cameras can already suffice to diminish performance. Ensemble solutions, where multiple model predictions are combined into one, can improve these problems. However, ensembles are computationally intensive and less transparent to practitioners than single model solutions. Constructing model soups, by averaging the weights of multiple models into a single model, could circumvent these limitations while still improving performance. OBJECTIVE: To investigate the performance of model soups for a dermoscopic melanoma-nevus skin cancer classification task with respect to (1) generalisation to images from other clinics, (2) robustness against small image changes and (3) calibration such that the confidences correspond closely to the actual predictive uncertainties. METHODS: We construct model soups by fine-tuning pre-trained models on seven different image resolutions and subsequently averaging their weights. Performance is evaluated on a multi-source dataset including holdout and external components. RESULTS: We find that model soups improve generalisation and calibration on the external component while maintaining performance on the holdout component. For robustness, we observe performance improvements for pertubated test images, while the performance on corrupted test images remains on par. CONCLUSIONS: Overall, souping for skin cancer classifiers has a positive effect on generalisation, robustness and calibration. It is easy for practitioners to implement and by combining multiple models into a single model, complexity is reduced. This could be an important factor in achieving clinical applicability, as less complexity generally means more transparency.