Results

Patient characteristics and adherence classification

A total of 54 patients with newly diagnosed pSjD were enrolled. The mean age was 49.7 ± 9.5 years, and 52 patients (96.3%) were female. At follow-up, 40 patients (74.1%) were available for reassessment. The mean follow-up duration was 21.36 ± 4.79 months.

Among reassessed patients, the CQR-5 identified 16 (40.0%) as having low adherence and 24 (60.0%) as having high adherence.

Baseline characteristics of patients who completed follow-up and those lost to follow-up were comparable, with no significant differences in age, sex, body mass index, or symptom duration (all p > 0.05).

Sociodemographic characteristics

Baseline characteristics according to adherence status are presented in Table 1. Age (p = 0.971), body mass index (p = 0.380), and symptom duration (p = 0.279) were similar between the low- and high-adherence groups. Sex distribution, employment status, smoking habits, and education level also did not differ significantly between groups.

 | Low adherence (n:16) | High adherence (n:24) | p-value
Age, years, mean (SD) | 49.94 (8.44) | 50.04 (9.27) | 0.971a
BMI, kg/m2, mean (SD) | 30.29 (5.99) | 28.68 (5.33) | 0.380a
Symptom duration, years, median (25th–75th) | 3.5 (2.0–8.75) | 2.0 (1.0–5.0) | 0.279b
Sex, n (%) female | 15 (93.8%) | 24 (100.0%) | 0.400c
Employment status, n (%) not active | 13 (81.2%) | 21 (87.5%) | 0.668c
Smoking, n (%) non-smoker | 14 (87.5%) | 22 (91.7%) | 0.999c
Education, n (%) high school graduate or above | 4 (25.0%) | 6 (25.0%) | 0.999c

Disease activity and Sjögren-specific symptom burden

Cross-sectional comparisons

Baseline and follow-up ESSDAI and ESSPRI values are presented in Table 2, and categorical distributions are shown in Table 3. Systemic disease activity was low in both adherence groups at baseline and remained low at follow-up, with no significant between-group differences (baseline p = 0.621; follow-up p = 0.736).

 | Time point | Low adherence (n:16) | High adherence (n:24) | p-value
ESSDAI score, median (25th–75th percentile) | Baseline | 1.00 (0.00–8.75) | 1.00 (0.00–3.75) | 0.733a
 | Follow-up | 1.50 (0.00–8.75) | 1.00 (0.00–4.00) | 0.452a
 | Change (follow-up–baseline) | 0.0 (0.0–0.0) | 0.0 (− 0.75–0.0) | 0.469a
ESSPRI total score | Baseline, mean (SD) | 5.97 (2.41) | 5.34 (2.08) | 0.389b
 | Follow-up, median (25th–75th percentile) | 7.32 (6.08–8.17) | 6.33 (5.41–7.58) | 0.134a
 | Change (follow-up–baseline), mean (SD) | 0.85 (2.52) | 0.92 (1.95) | 0.918b

 | Category | Low adherence n (%) | High adherence n (%) | p-value
ESSDAI baseline | Low (0–4) | 11 (68.8) | 20 (83.3) | 0.621a
 | Moderate (5–13) | 3 (18.8) | 3 (12.5) | 
 | High (≥ 14) | 2 (12.5) | 1 (4.2) | 
ESSDAI Follow-up | Low (0–4) | 11 (68.8) | 19 (79.2) | 0.736a
 | Moderate (5–13) | 3 (18.8) | 4 (16.7) | 
 | High (≥ 14) | 2 (12.5) | 1 (4.2) | 
ESSPRI baseline | Low (0–4) | 4 (25.0) | 10 (41.7) | 0.279b
 | High (≥ 5) | 12 (75.0) | 14 (58.3) | 
ESSPRI Follow-up | Low (0–4) | 2 (12.5) | 2 (8.3) | 0.999c
 | High (≥ 5) | 14 (87.5) | 22 (91.7) | 

ESSPRI scores were also comparable between groups. At baseline, total scores and the proportion of patients with high symptom burden did not differ significantly (p = 0.389 and p = 0.279, respectively). Similar findings were observed at follow-up (p = 0.134 and p = 0.999, respectively).

Longitudinal analyses

ESSDAI remained stable over time in both groups. In the low-adherence group, median values changed from 1.00 (0.00–8.75) to 1.50 (0.00–8.75) (p = 0.999). In the high-adherence group, ESSDAI remained at 1.00 (p = 0.366).

ESSPRI increased in both groups. In low-adherence patients, scores rose from 5.97 ± 2.41 to 6.82 ± 1.81 (mean change + 0.85, p = 0.197). In high-adherence patients, ESSPRI increased from 5.34 ± 2.08 to 6.24 ± 1.64 (mean change + 0.90, p = 0.032).

However, between-group comparisons of change scores did not show statistically significant differences. Median ESSDAI change was 0.0 (0.0–0.0) versus 0.0 (− 0.75 to 0.0) (p = 0.469), and mean ESSPRI change was 0.85 ± 2.52 versus 0.92 ± 1.95 (p = 0.918).

Psychological symptoms, fatigue, functional status, and health-related quality of life

Baseline and follow-up data are summarized in Table 4. At baseline, SF-12 physical, mental, and total scores did not differ significantly between low- and high-adherence patients (p = 0.484, 0.405, and 0.415, respectively). HADS anxiety and depression scores were also similar (p = 0.821 and p = 0.580, respectively). Fatigue severity and HAQ scores showed no significant differences (p = 0.902 and p = 0.090, respectively).

 | Low adherence (n:16) | High adherence (n:24) | p-value
SF-12 physical component, mean (SD) | 33.44 (18.92) | 38.08 (21.24) | 0.484a
SF-12 mental component, mean (SD) | 44.00 (17.78) | 49.58 (22.18) | 0.405a
SF-12 total score, mean (SD) | 39.81 (17.21) | 44.96 (20.62) | 0.415a
Fatigue severity, median (25th–75th percentile) | 5.44 (3.08–6.08) | 5.74 (3.49–6.45) | 0.902b
HAQ, median (25th–75th percentile) | 0.63 (0.16–1.09) | 0.22 (0.00–0.63) | 0.090b
HADS-Anxiety, mean (SD) | 8.13 (4.96) | 8.50 (5.20) | 0.821a
HADS-Depression, mean (SD) | 7.69 (5.03) | 8.54 (4.55) | 0.580a
SF-12 physical component (follow-up), mean (SD) | 39.81 (18.13) | 42.63 (23.89) | 0.692a
SF-12 mental component (follow-up), mean (SD) | 53.31 (16.97) | 51.67 (14.59) | 0.745a
SF-12 total score (follow-up), mean (SD) | 48.38 (17.46) | 47.50 (17.57) | 0.878a
Fatigue severity (follow-up), median (25th–75th percentile) | 6.30 (4.13–7.00) | 6.10 (4.91–6.95) | 0.774b
HAQ (follow-up), median (25th–75th percentile) | 0.63 (0.00–1.00) | 0.75 (0.28–1.50) | 0.113b
HADS-Anxiety (follow-up), mean (SD) | 7.38 (3.05) | 7.96 (4.59) | 0.658a
HADS-Depression (follow-up), mean (SD) | 8.31 (3.30) | 8.17 (4.28) | 0.909a

At follow-up, results remained comparable across groups. SF-12 scores (p = 0.692, 0.745, and 0.878), HADS anxiety and depression scores (p = 0.658 and p = 0.909), fatigue (p = 0.774), and HAQ (p = 0.113) were not significantly different between groups.

Longitudinal changes in patient-reported outcomes

SF-12 scores showed numerical improvement in both groups but did not reach statistical significance. In the low-adherence group, changes were + 6.38, + 9.31, and + 8.56 (p = 0.272, p = 0.102, and p = 0.127), while in the high-adherence group, changes were + 4.54, + 2.08, and + 2.54 (p = 0.362, p = 0.621, and p = 0.521).

HADS anxiety decreased slightly in both groups (− 0.75 vs − 0.54; p = 0.509 and p = 0.712), and depression scores showed minimal variation (+ 0.63 vs − 0.38; p = 0.703 and p = 0.785). Fatigue increased modestly (+ 0.68 vs + 0.63; p = 0.169 and p = 0.064). HAQ scores changed by − 0.15 (p = 0.495) and + 0.43 (p = 0.062) in low- and high-adherence groups, respectively.

Between-group comparisons of change scores did not reveal statistically significant differences for any of these outcomes.

Beliefs about medicines

Specific necessity scores were numerically higher in the high-adherence group (3.37 ± 0.79) compared with the low-adherence group (3.04 ± 1.06), but this difference did not reach statistical significance (p = 0.267). Specific concerns tended to be higher in low-adherence patients (3.46 ± 0.92 vs 3.07 ± 0.63, p = 0.115). General harm beliefs were similar between groups (p = 0.530), and general overuse beliefs did not differ significantly (p = 0.062).

In contrast, the necessity-concern differential (NCD) differed significantly between groups, with lower values observed in low-adherence patients (− 0.43 ± 1.16) compared with high-adherence patients (0.30 ± 0.96; p = 0.038). Although the absolute difference in the necessity-concern differential was modest, its direction was consistent with the prespecified conceptual framework: patients with lower adherence had a less favorable balance between perceived necessity and medication-related concerns.

Logistic regression

In the exploratory binary logistic regression model, the necessity-concern differential was entered as the main explanatory variable for high medication adherence. The model showed statistically significant overall fit compared with the null model (Omnibus _χ_2 = 4.46, p = 0.035). Each one-point increase in the necessity-concern differential was associated with higher odds of being classified in the high-adherence group (OR = 1.95, 95% CI 1.01–3.76). The model explained a modest proportion of variance, with Cox and Snell R2 of 10.6% and Nagelkerke R2 of 14.3%, and had an overall classification accuracy of 65%. Given the small sample size and the cross-sectional assessment of adherence and medication beliefs at follow-up, this model was interpreted as hypothesis-generating rather than confirmatory.