This was a cross-sectional observational study that included 331 SLE patients recruited from a tertiary hospital’s rheumatology and immunology department between 2021 and 2022. Sleep and self-reported outcomes in were assessed. In previous studies we have used baseline data from this program to construct a nomogram for poor sleep quality in patients with systemic lupus erythematosus (SLE) (10). However, the current submission aims to identify distinct sleep disturbance profiles in SLE patients and examine their associations with demographic, disease-related, and psychosocial factors.

Participants were enrolled based on predefined inclusion and exclusion criteria. The inclusion criteria were as follows: (a) patients with a history of ANA positivity who met the 2019 EULAR/ACR, 2012 Systemic Lupus International Collaborating Clinics (SLICC), or 1997 ACR criteria for SLE classification (11–13); (b) age between 18 and 70 years; and (c) ability to understand the study procedures and provide written informed consent.

Patients were excluded if they met any of the following criteria: (a) presence of other autoimmune rheumatic diseases (e.g., rheumatoid arthritis, Sjögren’s syndrome, ankylosing spondylitis, scleroderma, dermatomyositis, or fibromyalgia); (b) concomitant vital organ failure or malignancy; (c) severe cognitive impairment, dementia, psychiatric disorders, or other neurodegenerative diseases; (d) suspected or confirmed pregnancy; or (e) inability to complete the study questionnaires independently or with assistance.

This study was conducted in accordance with the Declaration of Helsinki and was reviewed and approved by the Ethics Committee of West China Hospital, Sichuan University. All eligible participants were informed about the study’s purpose, procedures, potential risks and benefits, and their right to withdraw at any time.

SLE Activity: Disease activity was assessed using the Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2000), which includes 24 descriptors across nine organ systems. The recall period for disease activity assessment was the previous 10 days. Scores range from 0 to 105, with higher scores indicating greater disease activity (16).

Anxiety and Depression: Anxiety and depression were evaluated using the HADS, which consists of 14 items equally divided into two subscales: HADS-Anxiety (HADS-A) and HADS-Depression (HADS-D) (17). HADS is widely used to assess anxiety and depression in physically ill patients, as it excludes symptoms such as insomnia, loss of appetite, and fatigue that may be attributed to physical illness. A cutoff score of ≥8 is recommended to identify potential cases of anxiety and depression (18–21).

Fatigue Severity: Fatigue severity was measured using the Fatigue Severity Scale (FSS), which comprises nine items. The fatigue severity score is calculated as the mean of all items, ranging from 1 (no fatigue) to 7 (maximum fatigue). Alternatively, a total score of ≥36 was used as the cutoff for significant fatigue (22). The FSS is commonly used to assess fatigue in SLE patients (23).

Sleep Quality: Sleep quality was assessed using the PSQI, a 19-item questionnaire comprising seven components: subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction over the past month. Each component is scored from 0 (no difficulty) to 3 (severe difficulty), with a total score ranging from 0 to 21. Higher scores indicate poorer sleep quality. Since its development, PSQI has been widely used in both clinical and non-clinical populations (24). In this study, a PSQI score ≥7 was considered indicative of poor sleep quality, while a score <7 was classified as acceptable sleep quality (25).

A database was established using Excel. Data entry was performed independently by two researchers, and 12 questionnaires with missing values were excluded. The final analysis was conducted on 331 valid questionnaires.

For descriptive analysis, the Shapiro-Wilk normality test was used to assess data distribution. Normally distributed variables were presented as mean ± standard deviation, while skewed data were reported as medians with interquartile ranges (25th and 75th percentiles). Categorical variables were expressed as n (%).

LPA was performed using the tidyLPA package. Model fit was evaluated using the Akaike information criterion, Bayesian information criterion, adjusted Bayesian information criterion, and entropy. The bootstrap likelihood ratio test and likelihood ratio test were applied for model comparison.

The following components of the PSQI were included in the cluster analysis: subjective sleep quality, sleep latency, sleep duration, sleep efficiency, and sleep disturbances. The use of sleep medication and daytime dysfunction were excluded due to their susceptibility to the effects of fatigue and mood disorders. Including unnecessary variables may compromise model identifiability and lead to overparameterization (26).

Variables with significant differences between clusters (p < 0.05) were incorporated into a logistic regression model to predict cluster membership probability. A two-tailed p < 0.05 was considered statistically significant.

There is no universally accepted guideline for the minimum sample size required in cluster analysis (27). However, some studies have suggested a minimum sample size of 2k, where k represents the number of variables (28). Based on this recommendation, a minimum sample size of 128 (k = 7, 27 = 128) was considered necessary to ensure the reliability and interpretability of the results in this study.

A total of 331 SLE patients were enrolled in this study, including 300 females (90.63%) and 31 males (9.37%), with a mean age of 35.28 ± 11.37 years and a mean disease duration of 70.79 ± 81.65 months.

On the basis of the PSQI’s five dimensions, it was established that a three-class latent profile model provided the best fit. LPA yielded an entropy value of 0.996 for the three-cluster model. Entropy is a commonly used measure of classification accuracy, ranging from 0 to 1, with values closer to 1 indicating more precise classification. An entropy value above 0.8 is generally considered acceptable, corresponding to a classification accuracy exceeding 90% (29). The three-cluster model demonstrated the lowest Bayesian information criterion and Akaike information criterion values, indicating the best model fit (30). Cluster 1 was characterized by the highest scores in sleep latency, sleep efficiency, and subjective sleep quality. Cluster 2 exhibited relatively high scores in sleep latency and subjective sleep quality. In contrast, three clusters were presented: Cluster 1 (severe sleep disturbance), Cluster 2 (moderate sleep disturbance), and Cluster 3 (mild sleep disturbance), shoeing a gradient of differences ( Table 1 , Figure 1 ).

The mean total PSQI score for the SLE sample was 7.59 ± 3.44, and the mean values for the individual PSQI components were as follows: component 1 (subjective sleep quality), 1.58 ± 0.76; component 2 (sleep latency), 1.92 ± 1.03; component 3 (sleep duration), 0.78 ± 0.91; component 4 (sleep efficiency), 0.45 ± 0.80; component 5 (sleep disturbances), 1.16 ± 0.51; component 6 (use of sleep medication), 0.24 ± 0.76; and component 7 (daytime dysfunction), 1.48 ± 1.19.The P values for all dimensions and total scores were < 0.05, indicating statistically significant intergroup differences, with the most pronounced differences observed in PSQI–sleep latency and PSQI–sleep efficiency (%) ( Table 2 ).

This study enrolled 331 patients who were classified into three clusters based on their characteristics: Cluster 1 (n = 42), Cluster 2 (n = 174), and Cluster 3 (n = 115). Significant differences were observed in the median disease duration among the clusters, with Cluster 2 showing the longest progression (60 months), followed by Cluster 3 (36 months), while Cluster 1 had the shortest duration (17 months). Notably, all divorced patients (two cases) were in Cluster 1. However, no significant differences were found among clusters regarding other variables such as age, gender, BMI, smoking/alcohol history, place of residence, and household income ( Table 3 ).

Significant differences were observed in calcium supplement usage rates among the three groups. Cluster 1 demonstrated the highest calcium supplement utilization rate, while Cluster 2 showed a substantially higher proportion of patients with renal dysfunction compared to the other two groups. Of the 45 patients with renal dysfunction, 30 cases were concentrated in Cluster 2. Regarding psychological symptoms, anxiety and depression incidence rates varied markedly across groups: Cluster 1 had the highest rates whereas Cluster 3 exhibited the lowest. Notably, the rates in Cluster 3 were significantly lower than those in both Cluster 1 and Cluster 2. Cluster 2 also reported the most cases of fatigue cases, while no other differences reached statistical significance ( Table 4 ).

An unordered multinomial logistic regression model was constructed, with the LPA-derived clusters as the dependent variable and significant factors as independent variables, using Cluster 2 as the reference group. The analysis revealed that disease duration, fatigue severity, calcium supplement use, impaired renal function, anxiety, and depression were all significant factors influencing cluster membership. Compared with Cluster 2, Cluster 3 exhibited lower levels of fatigue [OR (95% CI) = 0.522 (0.437, 0.625)], a lower likelihood of calcium consumption [OR (95% CI) = 0.889 (0.748, 1.056)], a higher likelihood of being free from anxiety [OR (95% CI) = 0.380 (0.231, 0.625)], a smaller family size [OR (95% CI) = 0.938 (0.900, 0.978)], and a longer disease duration [OR (95% CI) = 1.003 (1.002, 1.004)]. In comparison with Cluster 2, Cluster 1 was primarily characterized by a shorter disease course [OR (95% CI) = 0.992 (0.990, 0.994)] and a higher proportion of individuals with undergraduate degrees or above [OR (95% CI) = 2.632 (2.062, 3.359)] ( Table 5 ).

The study had several limitations. First, it cross-sectional design, precludes causal inference. Causation should be verified through longitudinal studies. Second the small number of patients using sleep medications prevented a detailed analysis of medication use across clusters. However, all the patients in study were enrolled from our lupus cohort, ensuring comprehensive clinical and treatment data collection. The availability of detailed historical clinical data is a major strength of this research. Despite these limitations, the large sample size provides robust evidence supporting the personalized management of sleep disturbances in SLE patients.