AI‐based and LC‐OCT‐guided follow‐up of actinic keratoses under treatment with tirbanibulin 1%

Actinic keratosis (AK) is a chronic condition characterized by in-situ proliferation of atypical keratinocytes and basal epidermal proliferation (PRO score). The cumulative lifetime risk for malignant transformation into cutaneous squamous cell carcinoma is 6–10%, while the annual risk for single lesions ranges from 0.03%–20%.1-3 Identification of risk factors and treatment of AK remains challenging as in up to 85% of the cases, recurrence manifests after 1 year.4 Since the clinical appearance of AK is not indicative for grading of atypia and PRO score, treatment should be implemented in all cases upon diagnosis.5 Introducing non-invasive imaging tools such as line-field confocal optical coherence tomography (LC-OCT) to the follow-up (FU) of AK, may guide and facilitate the assessment of therapy response by quantification of objectifiable changes in atypia and PRO score, which are usually only accessible by punch biopsy.6, 7 In this prospective study, a cohort of 10 patients with a total of 24 AK lesions on the head and neck area were treated using tirbanibulin 1% once per day for 5 consecutive days (maximum field of treatment: 25 cm2). For clinical diagnosis, the prevalence of erythema and/or scaling accompanied by a clear strawberry pattern in dermoscopy were obligate. Only well circumscribed lesions were included, based on their accessibility and by being easily identifiable by skin landmarks to provide a comprehensive FU, also under possible clearance under therapy. LC-OCT imaging was performed exemplarily from the centre of each lesion. LC-OCT was used at 6 time points (V0-V5) during 90 days of FU, evaluating stratum corneum (SC)/epidermal thickness, PRO score, dermo-epidermal junction (DEJ) undulation index (DEJ UI) and atypia score by AI-based algorithms previously described.7-10 The initial Olsen score, and the actinic keratosis area and severity index (AKASI) before and after FU were assessed. Due to insufficient imaging quality, 3/24 lesions were excluded from further evaluation. The cohort's mean age was 73.9 ± 7.8 years. Lesions were identified as either Olsen I (14/24) or Olsen II (10/24). The mean AKASI score before treatment (3.7 ± 2.1) was significantly reduced to 2.7 ± 2.1 after treatment (p < 0.006) (Table 1). For 18/24 cases, a clinically apparent inflammatory skin reaction was present on V2. For all cases, a hyper-reflective infiltrate and epidermal spongiosis was observed in LC-OCT. Keratinocyte apoptosis was visualized in 6/24 cases according to the study of Lacarrubba et al.11 The AI-generated metrics were visualized per single AK lesion (Figure 1a). Comparing V0 vs. V5, 16/21 (76%) lesions, showed SC and epidermal thinning by 15% (p = 2.2 × 10−3) and 8% (p = 0.035). DEJ UI did decrease by 15% (p = 5.8 × 10−3) in 14/21 (67%) lesions. PRO score (V0) was PRO I for 19/24 and PRO II for 5/24 AK. For PRO I, 17/19 lesions remained stable, while 2/19 lesions progressed to PRO II (V5). For PRO II, 4/5 lesions improved to PRO I, 1/5 remained stable. Atypia score did significantly improve by 26% (p = 3.3 × 10−4) in 19/21 (91%) lesions. In a case a slight worsening of an AK lesion was clinically noticed on V5. LC-OCT imaging revealed the persistence of atypia, accordingly, leading to the assumption of insufficient therapy response (Figure 1b). Since only PRO I and II lesions were included, assumptions on the possible behaviour of PRO III, so called high-proliferative lesions could not be made. It may be assumed that PRO III lesions are affected the most, under tirbanibulin treatment. Still, a sufficient therapy response was also observed for AK PRO I/II. In conclusion, LC-OCT allows a more comprehensive FU of AK rather than considering clinical lesion aspects only. Moreover, lesions not responding to treatment sufficiently can be identified and can be referred to adequate treatment early on. The research work conducted was financially sponsored by Almirall S.A. as a Phase IV/IIS-RC Study. Janis R. Thamm reports no conflict of interest. Sandra Schuh reports no conflict of interest. Ethical approval was obtained by the votum of the ethics committee of LMU Munich under the Project-Nr. 22-0781 for AI-based training of diagnostic algorithms and Project-Nr. 17-699 for diagnosis of skin lesions using LC-OCT. 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.