A protocol for developing a multivariable predictive model to assess clinical response to topical pharmacological treatment in patients with atopic dermatitis. A cohort study.

<ns3:p> Atopic dermatitis (AD) is a skin disease, but its clinical impact is systemic with more than 30% of patients suffering from other allergic diseases and mental health disorders. Early disease control could prevent the development of comorbidities or reduce the severity of the condition. Topical treatments, such as skin moisturizers, steroids, calcineurin inhibitors, and phosphodiesterase inhibitors, have proven to be effective in achieving clinical control in AD; However, sustained improvement is observed in approximately 50–70% of patients, depending on disease severity, treatment adherence, and individual response. Early detection of patients with a low probability of response to topical treatment would help start with other therapies and improve the quality of life of patients. Several clinical and laboratory biomarkers have been proposed to predict the prognosis of AD. In this study, we present a protocol for developing a model to predict the clinical response of AD patients to topical treatment. To achieve this goal, a cohort study will be performed for the prognostic model using the following steps: <ns3:list list-type="bullet"> <ns3:list-item> <ns3:p>1. Selection of predictive variables for the model</ns3:p> </ns3:list-item> <ns3:list-item> <ns3:p>2. Evaluation of the quality of the collected data and control of lost data</ns3:p> </ns3:list-item> <ns3:list-item> <ns3:p>3. Data statistical management</ns3:p> </ns3:list-item> <ns3:list-item> <ns3:p>4. Strategies to select the variables to include at the end of the model</ns3:p> </ns3:list-item> <ns3:list-item> <ns3:p>5. Evaluation of the performance of different possible models (predictive accuracy)</ns3:p> </ns3:list-item> <ns3:list-item> <ns3:p>6. selection of the best model.</ns3:p> </ns3:list-item> <ns3:list-item> <ns3:p>7. Validation</ns3:p> </ns3:list-item> </ns3:list> The performance and internal validation of the model will be assessed, and if high diagnostic performance is confirmed, external validation will be justified to assess its clinical impact ( Figure 1). </ns3:p>