Abstract
This study collected and analyzed clinical data of psoriasis patients to develop and validate a psoriasis relapse risk prediction model. It aims to support early relapse risk assessment in clinical practice and inform the design of preventive interventions. To develop and validate a risk prediction model for psoriasis relapse. A convenience sampling method was used to select 504 psoriasis patients admitted to a tertiary hospital in China between January 2022 and December 2024, including 353 cases in the training set and 151 cases in the testing set. Independent risk factors for psoriasis relapse were identified through univariate analysis and logistic regression analysis to develop a prediction model. A nomogram and SHAP summary plot were generated for model visualization, and the model’s goodness of fit and discriminative ability were evaluated. The 1-year relapse rate of psoriasis patients after treatment was 66.67%. Logistic regression identified six independent risk factors for psoriasis relapse: BMI, diabetes, biologic use, smoking, upper respiratory tract infection (URTI), and non-standard medication, all of which were incorporated into the model. The area under the ROC curve (AUC) values for the training and testing sets were 0.767 [95% CI 0.715–0.818] and 0.704 [95% CI 0.620–0.789], respectively. The model showed moderate discrimination and good calibration. Decision curve analysis (DCA) confirmed clinically meaningful net benefit in both training and test sets. The predictive model for psoriasis relapse risk established in this study demonstrated only moderate predictive performance. This model can serve as a preliminary exploratory tool, providing a certain degree of quantitative reference for assessing the risk of psoriasis relapse; however, rigorous external validation in independent multicenter cohorts is still required before clinical application.
Keywords: Psoriasis, Relapse, Prediction model
Subject terms: Computational biology and bioinformatics, Diseases, Health care, Medical research, Risk factors
Introduction
Psoriasis is a chronic, relapsing, inflammatory, systemic skin disease mediated by the immune system, triggered by a combination of genetic and environmental triggers. Its typical clinical manifestation is well-demarcated red plaques covered with silvery-white scales, which can be localized or widely distributed1. Patients with psoriasis often experience varying degrees of itching and pain, and some also have comorbidities such as cardiovascular and metabolic diseases2–4. Psoriasis is widely distributed worldwide. Epidemiological surveys show that there are approximately 125 million psoriasis patients globally, with a worldwide prevalence ranging from 0.51 to 11.43%5. The latest large-scale international study indicates that the global average prevalence of adult psoriasis is approximately 4.4%, with the prevalence in the East Asian region reaching as high as 5.7%6. Psoriasis lacks a curative therapy and carries a high long‑term relapse rate. Frequent relapses substantially impair patients’ physical and mental health and quality of life7.
Although BMI, upper respiratory tract infection, smoking, and other risk factors are well documented, individual factors cannot directly quantify a patient’s overall relapse probability in routine clinical care. Currently, there is a lack of efficient early warning and multi-factor quantitative assessment tools for the management of psoriasis relapse. Although there are numerous assessment scales for psoriasis, when using these scales for screening, patients often already exhibit typical psoriasis characteristics and have entered the relapse stage. Therefore, this study aims to integrate these known risk factors and construct a multifactorial comprehensive quantitative prediction model to explore rapid screening risk warning systems. Strengthening the early detection and warning of relevant risk factors before the onset of symptoms is of great significance for controlling psoriasis relapse.
In recent years, with the growing emphasis on precision medicine, the use of predictive models for early identification of disease risk and individualized intervention has become a research focus. Clinical prediction models are developed by utilizing existing clinical data to construct appropriate statistical models that summarize the regularity of the probability of a specific outcome occurring in particular clinical scenarios8. These models can provide patients and physicians with more accurate and scientific evidence to support earlier and better-informed decision-making.
Current research on predictive models for psoriasis relapse risk primarily focuses on the specific clinical decision point of “post-discontinuation of biologic agents”9,10. Although biologic agents have demonstrated remarkable efficacy in the treatment of psoriasis, their actual global utilization remains restricted by multiple factors, including economic constraints, policies, and medical accessibility, resulting in limited access to biologic therapies for a significant proportion of patients11,12. However, there is a lack of effective quantitative prediction tools for the relapse risk in a large number of patients who do not use biologic agents, such as those receiving only conventional systemic drugs, phototherapy, or topical treatments. Therefore, it is necessary to develop a relapse risk prediction model applicable to all psoriasis patients, covering the entire population including both biologic users and those receiving conventional treatments, with the aim of enhancing the model’s clinical applicability and generalizability.
To address this limitation, the present study designed, developed, and validated a relapse risk prediction model based on psoriasis risk factors, applicable to all psoriasis patients. The model is expected to provide a reference for the early risk assessment of psoriasis relapse in clinical practice and to offer insights for the implementation of preventive measures.
Materials and methods
Study design
This was a single-center, retrospective study conducted in the Department of Dermatology at a tertiary general hospital in China from January 2022 to December 2024. Patient data were retrieved from electronic medical records (EMRs) and follow‑up. Written informed consent was obtained from all patients at admission. The study adhered to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement as the reporting guideline13. It was approved by the Ethics Committee of the Second Affiliated Hospital of Hunan University of Traditional Chinese Medicine (Approval Number: 2025-KY-024-01). All methods were performed in accordance with the relevant guidelines and regulations.
Study population
This study employed a convenience sampling method to select psoriasis patients admitted to a tertiary hospital in China between January 2022 and December 2024. Their relapse status within 1 year was obtained through follow-up.
Patients meeting the inclusion criteria are required to fulfill the following conditions: (1) Diagnosis in accordance with the Guideline for the diagnosis and treatment of psoriasis in China (2023 edition)14; (2) Aged between 18 and 80 years at admission; (3) Clear consciousness, absence of cognitive impairment, and possession of basic communication and comprehension abilities; (4) Voluntary participation in this study and provide written informed consent; (5) Availability of complete clinical data and accessibility for follow-up.
The exclusion criteria included the following: (1) History of severe psychiatric disorders or current use of psychotropic medications; (2) Presence of severe primary diseases affecting the cardiovascular, cerebrovascular, hepatic, renal, or hematopoietic systems with unstable conditions; (3) Presence of other autoimmune diseases, malignant tumors, or active infections; (4) Severe missing clinical data that affect the extraction of study variables and outcome assessment; (5) No improvement or progressive deterioration of psoriasis following standardized treatment.
The criteria for withdrawal and dropout included: (1) Patients voluntarily requesting to withdraw or being lost to follow-up during the study period; (2) Major discrepancies between provided clinical data and actual conditions that could not be corrected after verification; (3) Occurrence of serious adverse events, complications, or death during the follow-up period, rendering the patient unable to continue participating in the study; (4) Use of non-protocol medications or therapeutic interventions during the study period that may affect disease relapse; (5) Missing data affecting the assessment of the primary outcome (relapse).
Sample size calculation
Sample size was calculated using the Events Per Variable (EPV) rule, a standard approach for clinical prediction model development, which requires at least 10 outcome events per predictor variable. In other words, a minimum of 10 outcome events per predictor variable is required to ensure the stability and reliability of the model parameters15. This study plans to include 35 independent variables. Research indicates that the 1-year relapse rate of psoriasis is as high as 86.1%16. Therefore, the minimum sample size is calculated as 35 × 10 ÷ 0.861 ≈ 407. A total of 504 patients with psoriasis were finally enrolled in this study and randomly divided into a training set (n = 353) and a testing set (n = 151) at a ratio of 7:3 using a computer-generated random number sequence.
Outcome variables
The outcome variable in this study was relapse. In the routine clinical follow-up practice of this study, relapse was primarily defined based on clinical skin lesion assessment: patients with a confirmed history of psoriasis who, after achieving lesion clearance with clinical treatment, developed either the relapse of original lesions or new lesions involving more than 30% of the body surface area. After discharge, patients were followed up monthly by nursing staff, and the follow-up period lasted for 1 year.
Candidate variables
Based on an analysis of previous relevant literature17–23, integration with clinical experience, and group discussions, we identified potential risk factors for psoriasis relapse. Accordingly, the following data were collected in this study: (1) Demographic and baseline characteristics: gender, age, body mass index (BMI), sleep disorders, smoking history, history of alcohol consumption, and anxiety; (2) Disease-related factors: duration of illness, severity of the condition, type of psoriasis, season of onset, family history, allergy history, history of diabetes, coronary heart disease, hypertension, hyperlipidemia, upper respiratory tract infection, irregular medication use, trauma/surgery, exposure to chemical agents (e.g., hair dyes, pesticides), and exposure to biologic agents; (3) Laboratory parameters: presence of hypoproteinemia and whether biochemical indicators were normal (erythrocyte sedimentation rate, C-reactive protein, immunological indicators).
Biologic exposure was defined as either current use at baseline or previous use of biologics. Immunological indicators included serum levels of IgA, IgG, IgM, and complement C3 and C4, with their normal or abnormal status determined based on the hospital’s laboratory reference ranges.
Data collection
This study retrospectively collected data from the hospital Electronic Medical Record (EMR) using a structured questionnaire. The questionnaire covered three specific domains, and patient relapse within 1 year was obtained through follow-up. The data collection process strictly adhered to the predefined inclusion and exclusion criteria. Two trained researchers used a standardized questionnaire to ensure consistency in data collection and cross-verified the completeness, authenticity, and accuracy of the data. Concurrently, a specialized dermatology nurse conducted monthly follow-ups for a duration of 1 year. Strict quality control measures were implemented throughout the follow-up process to ensure the accuracy and completeness of the information. Missing data were addressed via multiple imputation for variables with a missing rate of less than 5%; variables or samples with a missing rate greater than 5% were excluded. The data entry process strictly followed a double-entry system. After data entry, the data were cross-checked against the original medical records to identify and resolve any discrepancies or missing information, thus ensuring high data quality and completeness.
Statistical analysis
Statistical analyses were performed using R Studio (version 4.2.2). All hypothesis tests were two-sided with a significance level set at 0.05. Continuous data with a normal distribution are presented as the mean ± standard deviation, and comparisons between groups are analyzed using the independent samples t-test. Measurement data that do not follow a normal distribution are presented as median (M) with interquartile range (P25, P75), and comparisons between groups are analyzed using the Mann–Whitney U-test. Categorical data were analyzed using the chi-square test. This study employed univariate pre-screening combined with backward stepwise logistic regression for variable selection. This approach is a commonly used strategy in the development of clinical prediction models, facilitating model interpretability and aligning with the exploratory nature of this initial investigation. Univariate analysis was used to screen potential prognostic factors (P < 0.05). Before inclusion in the Logistic regression model, he linear relationship between continuous variables and logit-transformed outcome values was assessed via the Box-Tidwell test to confirm that the linearity assumption was met. Subsequently, multivariate logistic regression analysis was performed to identify independent risk factors (P < 0.05), and a predictive model was constructed. A nomogram and SHAP summary plot were constructed for model visualization. The Receiver Operating Characteristic (ROC) curve was plotted to evaluate the diagnostic performance of the model. The Hosmer–Lemeshow test was used to assess the goodness of fit. Calibration curves were plotted to evaluate the calibration of the model, and Decision Curve Analysis (DCA) was performed to assess the clinical utility of the model.
Results
Characteristics of the participants
This study collected clinical data from a total of 504 patients, among whom 336 cases (66.67%) experienced relapse within 1 year after treatment. Based on the stratification of the dependent variable, all samples were randomly divided into a training set (n = 353) and a testing set (n = 151) at a 7:3 ratio. The training set included 233 relapsed and 120 non-relapsed cases, with a relapse rate of 66.0%, while the testing set had 103 relapsed and 48 non-relapsed cases, with a relapse rate of 68.2%. The analysis of baseline characteristic differences showed that, in the comparison of clinical data between the training and testing sets, baseline features of C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) exhibited statistically significant differences. However, these differences were due to random allocation and did not affect model development. No statistically significant differences were observed between the two groups in terms of demographics and disease-related factors. Detailed data are provided in Table 1.
Table 1.
Comparison of baseline characteristics between the training and testing sets.
| Variables | Total (n = 504) | Training set (n = 353) | Testing set (n = 151) | Statistic | P-value |
|---|---|---|---|---|---|
| Age [M(P25, P75)] | 54 (39, 65) | 54 (39, 65) | 56 (40, 67) | 24,268.5 | 0.112 |
| Gender (%) | 0.448 | 0.503 | |||
| Female | 149 (29.56) | 108 (30.59) | 41 (27.15) | ||
| Male | 355 (70.44) | 245 (69.41) | 110 (72.85) | ||
| BMI [M(P25, P75)] | 24.61 (22.86, 26.84) | 24.61 (22.72, 26.85) | 24.61 (22.92, 26.37) | 26,713.5 | 0.967 |
| Disease duration [M(P25, P75)] | 9 (5, 14) | 9 (5, 13) | 10 (5, 20) | 24,147 | 0.093 |
| Type (%) | 3.935 | 0.269 | |||
| Arthropathic type | 20 (3.97) | 16 (4.53) | 4 (2.65) | ||
| Erythrodermic type | 38 (7.54) | 30 (8.5) | 8 (5.3) | ||
| Plaque type | 426 (84.52) | 291 (82.44) | 135 (89.4) | ||
| Pustular type | 20 (3.97) | 16 (4.53) | 4 (2.65) | ||
| Disease severity (%) | Fisher | 0.696 | |||
| Mild | 10 (1.98) | 6 (1.7) | 4 (2.65) | ||
| Moderate | 387 (76.79) | 273 (77.34) | 114 (75.5) | ||
| Severe | 107 (21.23) | 74 (20.96) | 33 (21.85) | ||
| Onset season (%) | 2.804 | 0.423 | |||
| Autumn | 88 (17.46) | 67 (18.98) | 21 (13.91) | ||
| Spring | 187 (37.1) | 133 (37.68) | 54 (35.76) | ||
| Summer | 138 (27.38) | 92 (26.06) | 46 (30.46) | ||
| Winter | 91 (18.06) | 61 (17.28) | 30 (19.87) | ||
| Family history (%) | 0.336 | 0.562 | |||
| No | 460 (91.27) | 320 (90.65) | 140 (92.72) | ||
| Yes | 44 (8.73) | 33 (9.35) | 11 (7.28) | ||
| Diabetes (%) | 0.391 | 0.532 | |||
| No | 414 (82.14) | 287 (81.3) | 127 (84.11) | ||
| Yes | 90 (17.86) | 66 (18.7) | 24 (15.89) | ||
| Hyperlipidemia (%) | 0.381 | 0.537 | |||
| No | 378 (75) | 268 (75.92) | 110 (72.85) | ||
| Yes | 126 (25) | 85 (24.08) | 41 (27.15) | ||
| Hypertension (%) | 1.102 | 0.294 | |||
| No | 339 (67.26) | 243 (68.84) | 96 (63.58) | ||
| Yes | 165 (32.74) | 110 (31.16) | 55 (36.42) | ||
| CHD (%) | 0.155 | 0.694 | |||
| No | 453 (89.88) | 319 (90.37) | 134 (88.74) | ||
| Yes | 51 (10.12) | 34 (9.63) | 17 (11.26) | ||
| Allergy (%) | 1.581 | 0.209 | |||
| No | 434 (86.11) | 299 (84.7) | 135 (89.4) | ||
| Yes | 70 (13.89) | 54 (15.3) | 16 (10.6) | ||
| Traumatic surgery (%) | 0.915 | 0.339 | |||
| No | 420 (83.33) | 290 (82.15) | 130 (86.09) | ||
| Yes | 84 (16.67) | 63 (17.85) | 21 (13.91) | ||
| Chemical agent (%) | 1.609 | 0.205 | |||
| No | 460 (91.27) | 318 (90.08) | 142 (94.04) | ||
| Yes | 44 (8.73) | 35 (9.92) | 9 (5.96) | ||
| Biologic agent (%) | 0.003 | 0.957 | |||
| No | 368 (73.02) | 257 (72.8) | 111 (73.51) | ||
| Yes | 136 (26.98) | 96 (27.2) | 40 (26.49) | ||
| Anxiety (%) | 0.153 | 0.696 | |||
| No | 424 (84.13) | 295 (83.57) | 129 (85.43) | ||
| Yes | 80 (15.87) | 58 (16.43) | 22 (14.57) | ||
| Sleep disorder (%) | 1.014 | 0.314 | |||
| No | 175 (34.72) | 128 (36.26) | 47 (31.13) | ||
| Yes | 329 (65.28) | 225 (63.74) | 104 (68.87) | ||
| Smoking (%) | 0.788 | 0.375 | |||
| No | 356 (70.63) | 254 (71.95) | 102 (67.55) | ||
| Yes | 148 (29.37) | 99 (28.05) | 49 (32.45) | ||
| Drinking (%) | 0.46 | 0.498 | |||
| No | 408 (80.95) | 289 (81.87) | 119 (78.81) | ||
| Yes | 96 (19.05) | 64 (18.13) | 32 (21.19) | ||
| URTI (%) | 0.325 | 0.569 | |||
| No | 442 (87.7) | 312 (88.39) | 130 (86.09) | ||
| Yes | 62 (12.3) | 41 (11.61) | 21 (13.91) | ||
| Non-standard medication (%) | 1.116 | 0.291 | |||
| No | 397 (78.77) | 283 (80.17) | 114 (75.5) | ||
| Yes | 107 (21.23) | 70 (19.83) | 37 (24.5) | ||
| Hypoproteinemia (%) | 0.137 | 0.711 | |||
| No | 456 (90.48) | 321 (90.93) | 135 (89.4) | ||
| Yes | 48 (9.52) | 32 (9.07) | 16 (10.6) | ||
| CRP (%) | 3.888 | 0.049 | |||
| Abnormal | 57 (11.31) | 33 (9.35) | 24 (15.89) | ||
| Normal | 447 (88.69) | 320 (90.65) | 127 (84.11) | ||
| ESR (%) | 5.097 | 0.024 | |||
| Abnormal | 260 (51.59) | 170 (48.16) | 90 (59.6) | ||
| Normal | 244 (48.41) | 183 (51.84) | 61 (40.4) | ||
| Immunological indicators (%) | 0.437 | 0.508 | |||
| Abnormal | 250 (49.6) | 179 (50.71) | 71 (47.02) | ||
| Normal | 254 (50.4) | 174 (49.29) | 80 (52.98) | ||
| Relapse (%) | 0.143 | 0.705 | |||
| No | 168 (33.33) | 120 (33.99) | 48 (31.79) | ||
| Yes | 336 (66.67) | 233 (66.01) | 103 (68.21) | ||
BMI, Body mass index; CHD, Coronary heart disease; URTI, Upper respiratory tract infection; CRP, C-reactive protein; ESR, Erythrocyte sedimentation rate.
Univariate analysis for relapse in psoriasis
Table 2 summarizes the results of the univariate analysis. the training set was further stratified into non-relapse and relapse groups based on relapse status. The results of this study showed that BMI, disease duration, diabetes, use of biologic agents, smoking, URTI, and non-standard medication use exhibited significant differences (all P < 0.05).
Table 2.
The univariate analysis for relapse in psoriasis (n = 353).
| Variables | Total (n = 353) | Non-Relapse (n = 120) | Relapse (n = 233) | statistic | P-value |
|---|---|---|---|---|---|
| Age [M(P25, P75)] | 54 (39, 65) | 52 (36, 65) | 55 (40, 64) | 13,231 | 0.41 |
| Gender (%) | 0.462 | 0.497 | |||
| Female | 108 (30.59) | 40 (33.33) | 68 (29.18) | ||
| Male | 245 (69.41) | 80 (66.67) | 165 (70.82) | ||
| BMI [M(P25, P75)] | 24.61 (22.72, 26.85) | 23.72 (21.15, 25.5) | 25.31 (23.38, 27.66) | 9052 | < 0.001 |
| Disease duration [M(P25, P75)] | 9 (5, 13) | 7 (4, 11) | 10 (5, 16) | 11,684 | 0.011 |
| Type (%) | 5.105 | 0.164 | |||
| Arthropathic-type | 16 (4.53) | 5 (4.17) | 11 (4.72) | ||
| Erythrodermic-type | 30 (8.5) | 7 (5.83) | 23 (9.87) | ||
| Plaque-type | 291 (82.44) | 99 (82.5) | 192 (82.4) | ||
| Pustular-type | 16 (4.53) | 9 (7.5) | 7 (3) | ||
| Disease severity (%) | Fisher | 0.893 | |||
| Mild | 6 (1.7) | 2 (1.67) | 4 (1.72) | ||
| Moderate | 273 (77.34) | 91 (75.83) | 182 (78.11) | ||
| Severe | 74 (20.96) | 27 (22.5) | 47 (20.17) | ||
| Onset season (%) | 0.419 | 0.936 | |||
| Autumn | 67 (18.98) | 22 (18.33) | 45 (19.31) | ||
| Spring | 133 (37.68) | 48 (40) | 85 (36.48) | ||
| Summer | 92 (26.06) | 30 (25) | 62 (26.61) | ||
| Winter | 61 (17.28) | 20 (16.67) | 41 (17.6) | ||
| Family history (%) | 0 | 1 | |||
| No | 320 (90.65) | 109 (90.83) | 211 (90.56) | ||
| Yes | 33 (9.35) | 11 (9.17) | 22 (9.44) | ||
| Diabetes (%) | 3.996 | 0.046 | |||
| No | 287 (81.3) | 105 (87.5) | 182 (78.11) | ||
| Yes | 66 (18.7) | 15 (12.5) | 51 (21.89) | ||
| Hyperlipidemia (%) | 0.396 | 0.529 | |||
| No | 268 (75.92) | 94 (78.33) | 174 (74.68) | ||
| Yes | 85 (24.08) | 26 (21.67) | 59 (25.32) | ||
| Hypertension (%) | 3.667 | 0.055 | |||
| No | 243 (68.84) | 91 (75.83) | 152 (65.24) | ||
| Yes | 110 (31.16) | 29 (24.17) | 81 (34.76) | ||
| CHD (%) | 0 | 0.982 | |||
| No | 319 (90.37) | 109 (90.83) | 210 (90.13) | ||
| Yes | 34 (9.63) | 11 (9.17) | 23 (9.87) | ||
| Allergy (%) | 0.336 | 0.562 | |||
| No | 299 (84.7) | 104 (86.67) | 195 (83.69) | ||
| Yes | 54 (15.3) | 16 (13.33) | 38 (16.31) | ||
| Traumatic surgery (%) | 0.316 | 0.574 | |||
| No | 290 (82.15) | 101 (84.17) | 189 (81.12) | ||
| Yes | 63 (17.85) | 19 (15.83) | 44 (18.88) | ||
| Chemical agent (%) | 0.363 | 0.547 | |||
| No | 318 (90.08) | 106 (88.33) | 212 (90.99) | ||
| Yes | 35 (9.92) | 14 (11.67) | 21 (9.01) | ||
| Biologic agent (%) | 11.001 | < 0.001 | |||
| No | 257 (72.8) | 101 (84.17) | 156 (66.95) | ||
| Yes | 96 (27.2) | 19 (15.83) | 77 (33.05) | ||
| Anxiety (%) | 0.452 | 0.501 | |||
| No | 295 (83.57) | 103 (85.83) | 192 (82.4) | ||
| Yes | 58 (16.43) | 17 (14.17) | 41 (17.6) | ||
| Sleep disorder (%) | 0.487 | 0.485 | |||
| No | 128 (36.26) | 47 (39.17) | 81 (34.76) | ||
| Yes | 225 (63.74) | 73 (60.83) | 152 (65.24) | ||
| Smoking (%) | 12.534 | < 0.001 | |||
| No | 254 (71.95) | 101 (84.17) | 153 (65.67) | ||
| Yes | 99 (28.05) | 19 (15.83) | 80 (34.33) | ||
| Drinking (%) | 2.35 | 0.125 | |||
| No | 289 (81.87) | 104 (86.67) | 185 (79.4) | ||
| Yes | 64 (18.13) | 16 (13.33) | 48 (20.6) | ||
| URTI (%) | 6.803 | 0.009 | |||
| No | 312 (88.39) | 114 (95) | 198 (84.98) | ||
| Yes | 41 (11.61) | 6 (5) | 35 (15.02) | ||
| Non-standard medication (%) | 12.007 | < 0.001 | |||
| No | 283 (80.17) | 109 (90.83) | 174 (74.68) | ||
| Yes | 70 (19.83) | 11 (9.17) | 59 (25.32) | ||
| Hypoproteinemia (%) | 2.009 | 0.156 | |||
| No | 321 (90.93) | 105 (87.5) | 216 (92.7) | ||
| Yes | 32 (9.07) | 15 (12.5) | 17 (7.3) | ||
| CRP (%) | 0.012 | 0.913 | |||
| Abnormal | 33 (9.35) | 12 (10) | 21 (9.01) | ||
| Normal | 320 (90.65) | 108 (90) | 212 (90.99) | ||
| ESR (%) | 0.025 | 0.873 | |||
| Abnormal | 170 (48.16) | 59 (49.17) | 111 (47.64) | ||
| Normal | 183 (51.84) | 61 (50.83) | 122 (52.36) | ||
| Immunological indicators (%) | 0.673 | 0.412 | |||
| Abnormal | 179 (50.71) | 65 (54.17) | 114 (48.93) | ||
| Normal | 174 (49.29) | 55 (45.83) | 119 (51.07) | ||
Logistic regression analysis for relapse in psoriasis
Table 3 summarizes the results of the logistic regression analysis. A logistic regression model was constructed using the training set data. Backward stepwise logistic regression was performed for feature selection, with 1-year relapse as the dependent variable and variables with statistical significance in the univariate analysis as independent variables. The logistic regression results indicated that BMI, diabetes, use of biologic agents, smoking, URTI, and non-standard medication were independent risk factors for relapse (all P < 0.05).
Table 3.
The logistic regression analysis for relapse in psoriasis.
| Variables | β | SE | Z | P-value | OR | 95% CI |
|---|---|---|---|---|---|---|
| (Intercept) | − 5.127 | 1.094 | − 4.687 | < 0.001 | 0.006 | 0.001–0.051 |
| BMI | 0.201 | 0.045 | 4.511 | < 0.001 | 1.223 | 1.121–1.335 |
| Diabetes | 0.987 | 0.352 | 2.802 | 0.005 | 2.683 | 1.345–5.353 |
| Biologic agent | 0.992 | 0.310 | 3.197 | 0.001 | 2.696 | 1.468–4.952 |
| Smoking | 0.843 | 0.317 | 2.657 | 0.008 | 2.324 | 1.248–4.330 |
| URTI | 1.160 | 0.497 | 2.332 | 0.020 | 3.191 | 1.204–8.459 |
| Non-standard medication | 0.835 | 0.376 | 2.222 | 0.026 | 2.305 | 1.104–4.813 |
Development of model
Nomogram for clinical risk stratification
The partial regression coefficients of the independent predictors identified via logistic regression analysis were used to develop the model. The fitted regression equation of the prediction model for the risk of relapse within 1 year in patients with psoriasis is as follows: Logit(P) = − 5.127 + 0.201 × BMI + 0.987 × Diabetes + 0.992 × Biologic agent + 0.843 × Smoking + 1.160 × URTI + 0.835 × Non-standard medication. The results are shown in Fig. 1.
Fig. 1.
Static nomogram for predicting relapse in psoriasis.
SHAP-based interpretability analysis
This study evaluated the relative importance of various risk factors influencing relapse in patients with psoriasis, as shown in Fig. 2. The SHAP summary plot quantifies each variable’s contribution to relapse risk. The horizontal axis shows SHAP values: positive values indicate higher relapse risk, negative values lower risk. Colors represent the magnitude of the feature value (blue for high, red for low). Binary features such as URTI, smoking, non-standard medication, diabetes, and biological agents exhibited a consistent pattern of influence: when present, they increased the risk of relapse, and when absent, they decreased the risk. As a continuous variable, BMI exhibited the widest distribution of SHAP values, suggesting its greatest impact on the risk of relapse, with higher BMI being associated with an increased risk of relapse. Overall, BMI is an important factor influencing relapse, highlighting the significance of weight management in controlling psoriasis relapse.
Fig. 2.
SHAP-based interpretability analysis of the nomogram model.
Prediction model performance
Model calibration
The model demonstrated excellent calibration performance in both the training and testing sets. The Hosmer–Lemeshow goodness-of-fit test yielded a χ2 value of 7.577 with a P-value of 0.476 for the training set, and a chi-square value of 10.391 with a P-value of 0.239 for the testing set. The results demonstrated that the model exhibited good fitting performance, and the calibration curves showed a high degree of concordance between the predicted and observed outcomes. The bootstrap-corrected calibration results further confirmed good agreement, with the calibration curves (green and blue solid lines) showing good alignment with the ideal reference line (black dashed line), as shown in Fig. 3.
Fig. 3.
Calibration curve of the relapse risk prediction model.
Model discrimination
The ROC curves for the predictive model indicate that it has moderate discriminative ability in predicting relapse risk in psoriasis patients. The area under the curve (AUC) was 0.767 [95% CI 0.715–0.818] for the training set (red solid line) and 0.704 [95% CI 0.620–0.789] for the testing set (blue dashed line). The results are presented in Fig. 4 and Table 4.
Fig. 4.
Receiver operating characteristic (ROC) curve of the model.
Table 4.
Model diagnostic performance indicators.
| Datasets | Sensitivity | Specificity | Accuracy | AUC | AUC [95% CI] |
|---|---|---|---|---|---|
| Training data | 0.682 | 0.742 | 0.703 | 0.767 | 0.767 [0.715–0.818] |
| Testing data | 0.466 | 0.917 | 0.609 | 0.704 | 0.704 [0.620–0.789] |
Clinical utility
Figure 5 presents the clinical Decision Curve Analysis (DCA). This method circumvents the need to explicitly consider false positives and false negatives by directly calculating the net benefit (NB) and maximizing its value. The DCA plot displays the threshold probability on the x-axis and the Net Benefit (NB) on the y-axis. the horizontal reference line represents the “no intervention” strategy (i.e., no clinical intervention for any patients), for which the net benefit is zero. The vertical reference line represents the “treat all” strategy (treating all samples as positive and providing intervention to all patients), which is characterized by a negative slope. The DCA curves for the training set (red line) and the testing set (green line) both exhibited net benefits exceeding those of the “treat-none” (black horizontal line) and “treat-all” (light gray diagonal line) reference strategies within a threshold probability range of approximately 0 to 0.92. This indicates that the model demonstrates a certain level of net benefit within this threshold range.
Fig. 5.
Decision curve analysis (DCA) of the model.
Discussion
Psoriasis is a chronic, recurrent, inflammatory skin disease, and its high relapse rate remains a major challenge and key focus in clinical management. Therefore, developing a prediction model for early identification of patients at high risk of relapse has significant clinical value for exploring individualized intervention strategies and optimizing treatment regimens. This study constructed and internally validated a 1-year relapse risk prediction model for psoriasis patients, identifying URTI, smoking, non-standard medication, diabetes, BMI, and biologic agent as independent predictive factors. Although these risk factors have been extensively reported in previous literature, the clinical significance of this study lies in integrating these multidimensional routine clinical indicators into an intuitive quantitative assessment tool. Moreover, the study population encompasses both biologic and non-biologic treatment recipients, addressing the limitation of existing models that focus primarily on relapse following biologic agent discontinuation.
The findings of this study indicate that URTI is an independent risk factor for psoriasis relapse. Streptococcal upper respiratory tract infection is recognized as a key trigger for psoriasis relapse24. Certain components of Streptococcus spp., such as the M protein, exhibit molecular mimicry with human keratinocytes, potentially initiating cross-immune reactions, activating T cells, and leading to the exacerbation of psoriatic skin lesions25. Additionally, upper respiratory tract infections caused by Staphylococcus aureus may also be associated with psoriasis relapse, as its superantigens can non-specifically activate a large number of T cells, triggering a strong inflammatory response and exacerbating psoriatic skin lesions26. Overall, infections can trigger abnormal immune responses that induce psoriasis relapse. Therefore, preventing and promptly treating infections is of great significance in reducing the risk of relapse. Relevant studies have also found that targeted vaccination can effectively reduce the risk of infection-induced relapse in patients with psoriasis27.
This study found that smoking is an independent risk factor for psoriasis relapse, consistent with previous research28,29. The mechanism by which smoking induces relapse is relatively complex20,30. Components in tobacco, such as nicotine, can directly stimulate immune cells to release inflammatory factors (e.g., IL-6, TNF-α), thereby promoting the chemotaxis and activation of neutrophils. Meanwhile, oxidants and free radicals in smoke can damage the skin barrier, trigger oxidative stress responses, and activate inflammatory pathways such as NF-κB. Furthermore, smoking may also predispose the body to inflammatory responses by influencing gene expression (such as upregulating susceptibility genes like HLA-C*06:02) and epigenetic modifications. In clinical management, smoking cessation should be incorporated as an integral component of the comprehensive therapeutic strategy for psoriasis. Patients should be explicitly educated on the role of smoking in precipitating disease relapse and the underlying inflammatory mechanisms, thereby fostering awareness of the necessity to quit. Furthermore, comprehensive smoking cessation support should be provided to address the challenges of smoking cessation, including encompassing behavioral interventions, psychological counseling, and pharmacological aids where indicated.
Non-standard medication practices, including unplanned discontinuation, dose reduction, or drug misuse, constitute a critical factor for psoriasis relapse. Topical treatment requires patients to maintain long-term adherence. Sudden discontinuation or improper withdrawal may trigger disease rebound. For example, abrupt discontinuation following prolonged high-dose glucocorticoid use can lead to rapid lesion spread and even induce severe phenotypes (e.g., erythrodermic or pustular psoriasis)31. Additionally, drug resistance or reduced efficacy during treatment may also contribute to disease relapse. For example, long-term administration of the same biological agent in certain patients may induce the formation of anti-drug antibodies, resulting in diminished efficacy and the relapse of skin lesions. Non-standard medication is also a modifiable factor contributing to psoriasis relapse. Adhering to standardized and continuous treatment is essential for preventing relapse. Studies have shown that implementing mobile health reminder systems, combined with smart pillboxes and SMS notifications, significantly improves patient medication adherence, which in turn helps reduce the risk of relapse32.
The results of this study indicate that diabetes is an independent risk factor for psoriasis relapse. A population-based cohort study has shown that psoriasis patients with diabetes exhibit higher disease activity and an increased risk of relapse33. Diabetes, as a common comorbidity of psoriasis, promotes relapse through the IL-23/IL-17 axis: a hyperglycemic environment activates macrophages through advanced glycation end products (AGEs), leading to the upregulation of IL-23 secretion and subsequently exacerbating cutaneous inflammation34. In clinical practice, diabetes screening for psoriasis patients can be strengthened, especially by conducting regular blood glucose monitoring for high-risk groups such as individuals with obesity, to enable early identification and intervention. Furthermore, for psoriasis patients with established diabetes, glycemic control should be intensified through dietary regulation, regular exercise, and rational pharmacotherapy to maintain blood glucose levels within the target range, thereby reducing the hyperglycemia-induced release of inflammatory factors.
This study found that increased BMI is an independent risk factor for psoriasis relapse. BMI, as a core indicator for assessing obesity, is defined by the World Health Organization as a BMI of 30 or higher in adults35. Research has shown that obesity can promote psoriasis relapse: adipose tissue secretes various pro-inflammatory cytokines (such as IL-6, TNF-α, leptin), which overlap with the inflammatory pathways involved in psoriasis and can lead to disease worsening and an increased risk of relapse36,37. For patients with elevated BMI, the prevention of psoriasis relapse can be achieved by prioritizing weight management and lifestyle interventions.
This study found that the prior biologic agent exposure was associated with an increased risk of psoriasis relapse. A study from Taiwan found that each additional prior exposure to biologics increases the risk of psoriasis relapse by 23%38. This phenomenon of “exposure-dependent relapse acceleration” may be related to the production of anti-drug antibodies (ADAs), which can accelerate drug clearance and reduce treatment efficacy39. For patients with a history of multiple exposures to biologic agents, greater caution should be exercised when selecting future treatment regimens. If necessary, rotating drugs with different mechanisms of action may be considered to reduce the risk of relapse. However, this finding may contradict conventional clinical intuition. We believe this may not reflect a direct pathogenic effect of biologic agents themselves, but rather indicates potential indication confounding and differences in disease severity. Typically, patients who meet the criteria for biologic agent use often have more severe disease, longer disease duration, or resistance to conventional treatments. These patients inherently have a higher natural risk of relapse. Additionally, some patients may have irregular discontinuation of medication during follow-up, which could also lead to an increased relapse rate.
Notably, in the selection of candidate variables, this study only included past biologic agent exposure history as a treatment-related candidate variable in the model, without separately listing conventional systemic medications, phototherapy, topical treatments, and other different treatment regimens as individual candidate variables for analysis. The primary reason is that the core research objective of this study was to construct a universal relapse risk prediction model applicable to all psoriasis patients, focusing on exploring the impact of generalizable risk factors such as demographics, comorbidities, lifestyle, infections, and medication compliance on relapse, rather than comparing the differences in relapse risk among different treatment regimens. Furthermore, due to the overlapping use of different treatment modalities among patients in this study, separately analyzing individual treatment regimens would have led to ambiguous variable definitions. Additionally, the uneven distribution of sample sizes across different treatment modalities in this single-center study made it difficult to conduct a valid statistical analysis. Therefore, this study did not include different treatment regimens as independent candidate variables, but rather reflected the potential impact of treatment-related factors on relapse through biologic agent exposure history.
The model achieved AUC values of 0.767 and 0.704 on the training and test sets, respectively, indicating moderate predictive performance. The low sensitivity (0.466) of the test set indicates that the model may fail to effectively identify some high-risk patients. Therefore, clinical application should incorporate comprehensive clinical judgment and the model should not be used as an independent decision-making tool. The goodness-of-fit tests showed χ2 = 7.577 (P = 0.476) for the training set and χ2 = 10.391 (P = 0.239) for the testing set, indicating that the model has good overall consistency. Decision Curve Analysis (DCA) demonstrated a certain net benefit in both the modeling cohort and the validation cohort, supporting the reference value of this model as a preliminary exploratory tool for assessing the risk of psoriasis relapse. Clinical practitioners may use this model solely as a preliminary exploratory tool for assessing the risk of psoriasis relapse. It should be integrated with patients’ clinical characteristics, medical history, and examination results to comprehensively identify individuals at higher risk of relapse and to facilitate the exploration of targeted preventive interventions. These measures include but are not limited to: the prevention and prompt treatment of upper respiratory tract infections; smoking cessation via interventions such as nicotine replacement therapy; strict adherence to prescribed medication regimens; enhanced blood glucose monitoring and control for patients with comorbid diabetes; and scientific weight management through nutritional counseling and exercise prescriptions. In summary, early identification of high-risk patients and the implementation of timely and effective interventions are key steps in reducing the risk of psoriasis relapse.
Conclusion
This study found that upper respiratory tract infection (URTI), smoking, non-standard medication use, diabetes, BMI, and biologic agent exposure are independent risk factors influencing psoriasis relapse. The psoriasis relapse risk prediction model developed in this study demonstrates moderate predictive performance and is intended as a preliminary exploratory tool only. It should not be used as an independent clinical decision-making instrument. This model can only provide limited quantitative reference for clinicians in the preliminary assessment of psoriasis relapse risk in patients. Given that this was a single-center retrospective study predominantly involving hospitalized patients with more severe conditions, the generalizability of the model to the general outpatient population is limited. Prior to its widespread clinical application, rigorous external validation in independent, multicenter, large-scale cohorts is warranted.
Limitations
This study has several limitations that need to be addressed in future research. First, the observed annual relapse rate of 66.67% in this study was higher than that reported in some previous cohort studies. This discrepancy may be attributed to the single-center retrospective design of our study, where all data were derived from one medical center. The study population consisted of hospitalized patients from a tertiary hospital, who generally had more severe and refractory disease, representing a relatively complex psoriasis population, thus introducing significant selection bias. This selection bias may render the applicability of the model in outpatient populations with milder disease uncertain, thereby compromising its generalizability and highlighting the need for further multicenter validation of the findings. Second, in defining relapse, this study relied primarily on clinical assessment of affected body surface area rather than systematic histopathological confirmation for all relapse cases, which may compromise the rigor of outcome classification to some extent. Third, this study did not include traditional systemic medications, phototherapy, and topical treatments as independent candidate variables in the model. Only prior biologic exposure history was incorporated as a treatment-related factor. The differential impact of various treatment regimens and their combinations on psoriasis relapse risk was not thoroughly analyzed, limiting the model’s ability to provide targeted references for clinical treatment selection. Future studies could expand sample sizes, clearly define grouping criteria for different treatment regimens, and incorporate these factors for in-depth analysis. Fourth, the predictive variables included in this study were all routine clinical indicators, and no novel biomarkers (e.g., genetic factors, specific cytokines, or microbiomic profiles) were incorporated. Future research could integrate multi-omics data to enhance the model’s predictive accuracy and scientific significance. Fifth, this study employed a variable selection strategy combining univariate pre-screening with backward stepwise logistic regression. Although this approach is a common method for constructing clinical prediction models, offering advantages such as model simplicity and strong clinical interpretability, it also has drawbacks, including instability in variable selection and a risk of overfitting, which could affect the model’s stability and reproducibility. Future research plans to adopt more robust variable selection methods, such as penalized regression (e.g., LASSO), to optimize and validate the model. Sixth, this study only focused on relapse within 1 year and did not evaluate the longer-term risk of psoriasis relapse or the patterns of disease progression. Future research could extend the follow-up period to develop a more comprehensive disease trajectory prediction model. Finally, this study only conducted internal validation, and the AUC values indicated that the model has moderate predictive performance only. Therefore, rigorous external validation in independent, multicenter large-cohort studies is necessary before applying this model in clinical practice.
Author contributions
XZ: Conceptualization, Methodology, Investigation, Data curation, Writing—original draft. CZ: Investigation, Data curation, Writing Original Draft. YL: Coordination, Formal analysis, Project administration, Writing—review and editing. HH: Coordination, Project administration, Writing—review and editing. JG: Project administration, Writing—review and editing. XT: Project administration, Writing—review and editing. GJ: Funding acquisition, Project administration, Writing—review and editing. All the authors have read and approved the final version of the manuscript.
Funding
This study was supported by the 2025 Hunan Provincial Natural Science Foundation Project (2025JJ80911).
Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Hua He, Email: 827179673@qq.com.
Gufen Jiang, Email: ylyyw118@163.com.
References
- 1.Boehncke, W.-H. & Schön, M. P. Psoriasis. Lancet386, 983–994 (2015). [DOI] [PubMed] [Google Scholar]
- 2.Dubertret, L. et al. European patient perspectives on the impact of psoriasis: The EUROPSO patient membership survey. Br. J. Dermatol.155, 729–736 (2006). [DOI] [PubMed] [Google Scholar]
- 3.Peng, C. et al. Epidemiological variations in the global burden of psoriasis, an analysis with trends from 1990 to 2017. Front. Med.8, 585634 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Takeshita, J. et al. Psoriasis and comorbid diseases: Epidemiology. J. Am. Acad. Dermatol.76, 377–390 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.World Health Organization. Global Report on Psoriasis. (2016).
- 6.Skayem, C., Taieb, C., Halioua, B., Baissac, C. & Saint Aroman, M. Epidemiology of psoriasis: A worldwide global study. Acta Derm Venereol105, adv42945 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bu, J., Ding, R., Zhou, L., Chen, X. & Shen, E. Epidemiology of psoriasis and comorbid diseases: A narrative review. Front. Immunol.13, 880201 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cadrin-Tourigny, J. et al. A new prediction model for ventricular arrhythmias in arrhythmogenic right ventricular cardiomyopathy. Eur. Heart J.43, e1–e9 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Huang, S., Bai, Y., Qi, R., Yu, H. & Duan, X. Personalized prediction of psoriasis relapse post-biologic discontinuation: A machine learning-driven population cohort study. J. Dermatolog. Treat.36, 2480743 (2025). [DOI] [PubMed] [Google Scholar]
- 10.Huang, S., Chen, B., Qi, Y., Duan, X. & Bai, Y. Development and external validation of a prediction model for the risk of relapse in psoriasis after discontinuation of biologics. Front. Med.11, 1488096 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Girolomoni, G. et al. Increasing access to effective systemic treatments in patients with moderate-to-severe psoriasis: Narrative review. Dermatol. Ther.13, 2171–2185 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wan, V. et al. Disparities and barriers to the access of biologics in moderate-to-severe adult psoriasis. Int. J. Dermatol.63, 1293–1301 (2024). [DOI] [PubMed] [Google Scholar]
- 13.Collins, G. S., Reitsma, J. B., Altman, D. G. & Moons, K. G. M. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): The TRIPOD statement. BMJ350, g7594 (2015). [DOI] [PubMed] [Google Scholar]
- 14.Committee on Psoriasis, Chinese Society of Dermatology. Guideline for the diagnosis and treatment of psoriasis in China (2023 edition). Chin. J. Dermatol.56, 573–625 (2023). [Google Scholar]
- 15.Riley, R. D. et al. Calculating the sample size required for developing a clinical prediction model. BMJ368, m441 (2020). [DOI] [PubMed] [Google Scholar]
- 16.Huang, Y.-W. & Tsai, T.-F. Remission duration and long-term outcomes in patients with moderate-to-severe psoriasis treated by biologics or tofacitinib in controlled clinical trials: A 15-year single-center experience. Dermatol. Ther.9, 553–569 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Francis, L., Capon, F., Smith, C. H., Haniffa, M. & Mahil, S. K. Inflammatory memory in psoriasis: From remission to relapse. J. Allergy Clin. Immunol.154, 42–50 (2024). [DOI] [PubMed] [Google Scholar]
- 18.Lee, E. B., Wu, K. K., Lee, M. P., Bhutani, T. & Wu, J. J. Psoriasis risk factors and triggers. Cutis102, 18–20 (2018). [PubMed] [Google Scholar]
- 19.Zhang, L.-J. et al. Antimicrobial peptide LL37 and MAVS signaling drive interferon-β production by epidermal keratinocytes during skin injury. Immunity45, 119–130 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Liu, S. et al. Triggers for the onset and relapse of psoriasis: a review and update. Cell Commun. Signal22, 108 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Stewart, T. J., Tong, W. & Whitfeld, M. J. The associations between psychological stress and psoriasis: A systematic review. Int. J. Dermatol.57, 1275–1282 (2018). [DOI] [PubMed] [Google Scholar]
- 22.Potestio, L. et al. Risk factors for psoriasis flares: A narrative review. Psoriasis (Auckl)14, 39–50 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Yang, H., Li, X., Zhang, L., Xue, F. & Zheng, J. Immunomodulatory effects of sleep deprivation at different timing of psoriasiform process on skin inflammation. Biochem. Biophys. Res. Commun.513, 452–459 (2019). [DOI] [PubMed] [Google Scholar]
- 24.Thorleifsdottir, R. H. et al. Throat Infections are associated with exacerbation in a substantial proportion of patients with chronic plaque psoriasis. Acta Derm. Venereol.96, 788–791 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Johnston, A., Gudjonsson, J. E., Sigmundsdottir, H., Love, T. J. & Valdimarsson, H. Peripheral blood T cell responses to keratin peptides that share sequences with streptococcal M proteins are largely restricted to skin-homing CD8(+) T cells. Clin. Exp. Immunol.138, 83–93 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Travers, J. B. et al. Epidermal HLA-DR and the enhancement of cutaneous reactivity to superantigenic toxins in psoriasis. J. Clin. Investig.104, 1181–1189 (1999). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Chiricozzi, A., Gisondi, P., Bellinato, F. & Girolomoni, G. Immune response to vaccination in patients with psoriasis treated with systemic therapies. Vaccines10.3390/vaccines8040769 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Shen, F. et al. Tobacco smoking was positively associated with disease relapse at week 24 and 48 among patients with psoriasis vulgaris in Shanghai: A prospective study. Psoriasis (Auckl)15, 261–272 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lee, E. J., Han, K. D., Han, J. H. & Lee, J. H. Smoking and risk of psoriasis: A nationwide cohort study. J. Am. Acad. Dermatol.77, 573–575 (2017). [DOI] [PubMed] [Google Scholar]
- 30.Armstrong, A. W., Armstrong, E. J., Fuller, E. N., Sockolov, M. E. & Voyles, S. V. Smoking and pathogenesis of psoriasis: A review of oxidative, inflammatory and genetic mechanisms. Br. J. Dermatol.165, 1162–1168 (2011). [DOI] [PubMed] [Google Scholar]
- 31.Carey, W. et al. Relapse, rebound, and psoriasis adverse events: An advisory group report. J. Am. Acad. Dermatol.54, S171-181 (2006). [DOI] [PubMed] [Google Scholar]
- 32.Svendsen, M. T. et al. A smartphone application supporting patients with psoriasis improves adherence to topical treatment: A randomized controlled trial. Br. J. Dermatol.179, 1062–1071 (2018). [DOI] [PubMed] [Google Scholar]
- 33.Armstrong, A. W., Harskamp, C. T. & Armstrong, E. J. Psoriasis and the risk of diabetes mellitus: A systematic review and meta-analysis. JAMA Dermatol.149, 84–91 (2013). [DOI] [PubMed] [Google Scholar]
- 34.De Luca, D. A., Papara, C., Hawro, T. & Thaçi, D. Psoriasis and diabetes: A review of the pathophysiological and therapeutic interconnections. Minerva Med.116, 195–222 (2025). [DOI] [PubMed] [Google Scholar]
- 35.World Health Organization. Obesity and overweight. https://www.who.int/news-room/fa-ct-sheets/detail/obesity-and-overweight (2025).
- 36.Musumeci, M. L., Nasca, M. R., Boscaglia, S. & Micali, G. The role of lifestyle and nutrition in psoriasis: Current status of knowledge and interventions. Dermatol. Ther.35, e15685 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Kunz, M., Simon, J. C. & Saalbach, A. Psoriasis: Obesity and fatty acids. Front. Immunol.10, 1807 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Huang, Y.-H. et al. Predicting the time to relapse following withdrawal from different biologics in patients with psoriasis who responded to therapy: A 12-year multicenter cohort study. Am. J. Clin. Dermatol.25, 997–1008 (2024). [DOI] [PubMed] [Google Scholar]
- 39.Sun, X. et al. Formation and clinical effects of anti-drug antibodies against biologics in psoriasis treatment: An analysis of current evidence. Autoimmun. Rev.23, 103530 (2024). [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.