Integrating basal and reserve salivary function defines clinically anchored functional phenotypes and contextualises salivary end points in Sjögren’s disease

Abstract

Background

Salivary gland dysfunction in Sjögren’s disease (SjD) is clinically heterogeneous. Unstimulated whole salivary flow (UWSF) reflects basal secretion, while stimulated tests such as the Saxon test assess secretory reserve. Whether integrating these complementary domains defines clinically meaningful glandular functional states remains unclear.

Methods

We conducted a retrospective cohort study of 193 anti-Sjögren’s syndrome antigen A (anti-SSA)-positive patients with SjD. Using established clinical thresholds for UWSF and the Saxon test, we defined three functional phenotypes: preserved basal and reserve (PP), partial impairment (PL/LP) and combined low function (LL). Clinical features, patient-reported outcomes, submandibular gland (SMG) gland ratio and longitudinal salivary changes were evaluated.

Results

Dryness and swallowing difficulty differed significantly across phenotypes. Compared with PP, the LL phenotype exhibited higher European Alliance of Associations for Rheumatology (EULAR) Sjögren’s Syndrome Patient-Reported Index dryness (β=1.67, 95% CI 0.64 to 2.70) and increased odds of swallowing difficulty (OR 6.19, 95% CI 2.71 to 14.9) after multivariable adjustment. Histological analyses showed markedly reduced SMG preservation in LL. Longitudinally, PP showed measurable declines in UWSF and Saxon test, whereas PL/LP remained relatively stable. LL had minimal longitudinal change in UWSF and a directionally positive Saxon slope, best interpreted in the context of floor-constrained end point behaviour rather than biological recovery. Baseline-to-last transition analyses showed substantial same-phenotype retention, and most longitudinally evaluable LL patients remained below the impaired Saxon threshold throughout follow-up.

Conclusion

Integrated basal-reserve functional phenotyping defines clinically and structurally anchored salivary gland functional states in SjD. This framework may support clinical stratification and state-dependent interpretation of longitudinal salivary end points in studies targeting glandular outcomes.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Salivary end points in Sjögren’s disease (SjD) clinical trials have shown inconsistent and often modest improvements despite effective systemic immunomodulation.

• Baseline salivary flow has been suggested to influence therapeutic responsiveness, yet unstimulated and stimulated measures are typically interpreted separately, without an integrated framework to define glandular functional states.

• This limits the ability to stratify patients by residual salivary function or to interpret heterogeneous salivary responses.

WHAT THIS STUDY ADDS

• We introduce a dual-domain functional phenotyping framework integrating basal and reserve salivary function to define three clinically and structurally anchored glandular states.

• These phenotypes were associated with dryness burden, swallowing difficulty and submandibular glandular preservation.

• Longitudinal analyses further showed that salivary end point changes should be interpreted in relation to baseline functional state, floor constraints and proximity to phenotype-defining thresholds.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Functional phenotyping may provide a pragmatic strategy for stratifying baseline salivary functional states in SjD studies targeting glandular outcomes.

• By identifying baseline states with different measurable ranges, floor constraints and susceptibility to threshold-related reclassification, this framework may improve the interpretation of salivary end points and reduce misclassification of apparent non-response.

Introduction

Sjögren’s disease (SjD) is a systemic autoimmune disorder with heterogeneous clinical manifestations.^{1 2} Although recent biologic therapies reduce systemic disease activity and clinical trials emphasise extraglandular end points, patient-prioritised glandular symptoms, particularly xerostomia, remain largely unmet.^{1 3}

Recent post hoc analyses from the ianalumab (VAY736) trials have demonstrated that patients with better-preserved baseline salivary gland function were more likely to achieve functional improvement following B-cell-targeted therapy.⁴ These findings suggest that salivary gland function is not merely an outcome but a state variable that may constrain the therapeutic window and ceiling of response for salivary end points. However, despite recommendations incorporating unstimulated and stimulated salivary flow in therapeutic decision-making,⁵ no widely adopted framework conceptualises salivary gland function as a multidimensional state integrating basal and reserve domains.^{5 6} Consequently, clinical assessment lacks a structured approach to stratify patients according to residual glandular function.

To address this gap, we propose a pragmatic dual-domain framework integrating unstimulated whole salivary flow (UWSF) and stimulated reserve measured by the Saxon test to define clinically applicable functional phenotypes. We examined the clinical characteristics, patient-reported symptom burden, histological structural integrity of the submandibular gland (SMG) and longitudinal functional trajectories associated with these phenotypes.

Methods

Study design and participants

This retrospective cohort study was conducted using longitudinal salivary gland functional data from a tertiary medical centre in Taiwan.

We consecutively enrolled anti-Sjögren’s syndrome antigen A (anti-SSA)-positive patients with SjD who fulfilled either the 2002 American-European Consensus Group criteria⁷ or the 2016 American College of Rheumatology/European Alliance of Associations for Rheumatology (EULAR) classification criteria⁸ and had undergone both UWSF measurement and the Saxon test at our institution between October 2022 and December 2025. Patients lacking baseline salivary flow assessment were excluded. Two complementary measures of disease chronicity were ascertained for each patient. Disease duration from institutional diagnosis was calculated as the time from documented institutional diagnosis of SjD to baseline salivary assessment. Symptom duration was calculated from patient-reported symptom onset (extracted from clinical records) to baseline assessment. For analysis of functional trajectory, all available repeated UWSF and Saxon measurements were included and patients could contribute one or more assessments. No formal sample size calculation was performed, given the retrospective design; all eligible patients during the study period were included.

Measurement of salivary flows and functional phenotyping

UWSF was collected by the spitting method and expressed as mL/min, serving as an index of basal salivary secretion. Stimulated salivary reserve was assessed using the Saxon test and expressed as g/2 min. Established clinical thresholds were used to define preserved basal secretion (UWSF >0.1 mL/min) and preserved stimulated reserve (Saxon test >2.0 g/2 min).⁸,¹⁰ Detailed testing procedures are provided in the online supplemental methods. Based on the combination of basal and reserve function, patients were initially categorised into four quadrants and then classified into three functional phenotypes: preserved-preserved (PP), preserved-low/low-preserved (PL/LP) and low-low (LL). PL and LP were combined as a pragmatic discordant-function category because preservation of either domain represents partial functional impairment and the PL subgroup was small. Salivary flow assessments were routinely performed at baseline, approximately 6 months after baseline and annually thereafter. Additional measurements were obtained when clinically indicated.

Biopsy procedure, histological assessment and gland ratio quantification

SMG biopsies were available in a subset of patients (n=51) and were used to provide histological structural anchoring for functional phenotypes. SMG biopsies were performed as previously described.¹¹ Briefly, bilateral SMG ultrasonographic evaluation was conducted using a 12 MHz linear-array transducer, and core needle biopsy was performed on the gland with more pronounced ultrasonographic abnormalities, using an 18-gauge biopsy needle via a free-hand technique by an experienced otolaryngologist (T-CC).

Histological evaluation focused on quantification of the SMG gland ratio, defined as the proportion of preserved glandular parenchyma relative to the total biopsy area, by an experienced head and neck pathologist (M-SH).¹² Glandular parenchyma was defined as acinar and ductal components, while lymphocytic foci, adipose tissue, fibrotic areas and haemorrhage were excluded from glandular area calculations.

Medication exposure

Medication exposure was evaluated at baseline and during follow-up. Baseline medication-adjusted sensitivity analyses included hydroxychloroquine, glucocorticoids, immunosuppressants, anticholinergics and secretagogues. Longitudinal drug-accounting generalised estimating equations (GEE) models additionally adjusted for baseline immunomodulatory treatment and visit-level secretagogue use. Definitions of medication exposure are provided in the online supplemental methods.

Statistical analysis

Continuous variables were summarised using medians and IQRs, and categorical variables as frequencies and percentages. Comparisons across functional phenotypes were performed using the Kruskal-Wallis test or Fisher’s exact test, as appropriate. Trend tests were applied, treating the three phenotypes as ordinal categories (PP, PL/LP, LL) using the Jonckheere-Terpstra test and the Cochran-Armitage trend test, as appropriate. Missing data were assessed at the variable level. Baseline missingness was <5% for all variables and did not differ across functional phenotypes; therefore, complete-case analyses were performed without imputation.

Multivariable linear regression and logistic regression models were used to assess the independent associations between functional phenotypes, patient-reported outcomes and SMG gland ratio. Covariates included age, sex, comorbid autoimmune disease, rheumatoid factor, EULAR Sjögren’s Syndrome Disease Activity Index (ESSDAI) and disease duration from institutional diagnosis.¹³ Sensitivity analyses substituting ESSDAI with serum IgG were performed as a complementary marker of immune burden.

To assess whether the categorical functional phenotype provided information beyond the component salivary measurements, we compared four regression model structures for key outcomes: functional phenotype alone, continuous UWSF and Saxon values, UWSF-by-Saxon interaction and continuous UWSF/Saxon values plus functional phenotype. Model fit was compared using Akaike information criterion, adjusted R² or pseudo-R² as appropriate and likelihood ratio tests for nested models.

Longitudinal salivary flow trajectories were visualised using spaghetti plots combined with locally estimated scatter plot smoothing (LOESS) curves with 95% CIs. For visualisation purposes, LOESS curves were truncated at 600 days to reduce instability arising from sparse observations at extended follow-up durations. Longitudinal trajectories were modelled using GEE with a Gaussian distribution, identity link and exchangeable working correlation structure. Phenotype-by-time interaction terms were used to estimate phenotype-specific slopes, which were converted to annualised rates of change. Models were adjusted for age and sex, and inference was based on robust SEs. Cluster bootstrap resampling at the patient level with 1000 replicates was used to assess slope stability. Sensitivity analyses examined follow-up frequency, marked within-subject variability, alternative Saxon thresholds, working correlation structures and medication exposure.

To complement continuous trajectory analyses, patient-level baseline-to-last-visit and visit-level consecutive transition matrices were constructed among patients with assignable phenotypes at baseline and follow-up. Boundary proximity was quantified as the standardised minimum distance from baseline UWSF or Saxon value to the nearest phenotype-defining threshold, and its association with phenotype reclassification was assessed using logistic regression adjusted for follow-up duration, number of visits and baseline phenotype. Detailed methods for statistical analyses are provided in the online supplemental methods.

All statistical analyses were conducted using R software (V.4.4.3) and a two-sided p value <0.05 was considered statistically significant.

Results

Cohort characteristics and functional phenotyping

A total of 193 anti-SSA-positive patients with SjD were included in the study cohort (table 1, online supplemental figure S1). Based on combined assessment of basal salivary secretion (UWSF) and reserve capacity (Saxon test), patients were classified into three functional phenotypes according to established thresholds: preserved basal and reserve function (PP), partial impairment of either basal or reserve function (PL/LP) and combined low basal and reserve function (LL) (figure 1).

Table 1. Total characteristics for enrolled anti-SSA-positive patients with SjD.

Characteristic	N=193
Age, year, median (IQR)	58.3 (49.2–69.5)
Female, n (%)	170 (88)
Comorbid autoimmune disease, n (%)	49 (25)
Disease duration from institutional diagnosis, year, median (IQR)	0 (0.0–0.0)
Symptom duration from patient-reported onset, year, median (IQR)	1.7 (0.4–5.0)
UWSF, mL/min, median (IQR)	0.1 (0.0–0.2)
Saxon, g/2 min, median (IQR)	3.0 (1.5–4.6)
Schirmer’s test ≤5 mm/5 min in at least one eye, n (%)	149 (78)§
Rheumatoid factor, IU/mL, median (IQR)	0.0 (0.0–29.7)†
C3, mg/dL, median (IQR)	115.3 (102.2–130.5)‡
C4, mg/dL, median (IQR)	27.0 (22.6–34.1)‡
IgG, mg/dL, median (IQR)	1510.7 (1271.2–1798.2)‡
ESSDAI, median (IQR)	1 (0–4)
clinESSDAI, median (IQR)	0 (0–5)
Ocular symptoms: dry eye, n (%)	128 (66)
Ocular symptoms: foreign body sensation, n (%)	112 (58)*
Ocular symptoms: tear substitutes, n (%)	69 (36)
Oral symptoms: dry mouth, n (%)	141 (73)
Oral symptoms: swollen salivary glands, n (%)	28 (15)
Oral symptoms: drink liquids to aid in swallowing dry food, n (%)	89 (46)*
Patient global assessment, median (IQR)	5 (3–7)
ESSPRI—dryness, median (IQR)	6 (3–8)
ESSPRI—fatigue, median (IQR)	6 (3–7)
ESSPRI—pain, median (IQR)	4 (1–7)
Use of hydroxychloroquine in baseline, n (%)	63 (33)
Use of glucocorticoids in baseline, n (%)	25 (13)
Use of immunosuppressants in baseline, n (%)	25 (13)
Use of pilocarpine or cevimeline in baseline, n (%)	9 (4.7)
Use of anticholinergics in baseline, n (%)	10 (5.2)

Data presented as median (IQR) or n (%).

^*Data were missing in one patient.

^†Data were missing in four patients.

^‡Data were missing in three patients.

^§Data were missing in two patients.

clinESSDAI, clinical ESSDAI; ESSDAI, EULAR Sjögren’s Syndrome Disease Activity Index; ESSPRI, EULAR Sjögren’s Syndrome Patient Reported Index; EULAR, European Alliance of Associations for Rheumatology; IQR, interquartile range; SjD, Sjögren’s disease; UWSF, unstimulated whole salivary flow.

Figure 1. Functional phenotyping based on basal and reserve salivary function. Scatter plot of UWSF (mL/min) versus Saxon test (g/2 min) in 193 patients with Sjögren’s disease. Dashed lines indicate predefined cut-offs (UWSF 0.1 mL/min; Saxon 2 g/2 min), generating four quadrants (PP, PL, LP, LL). PL and LP were combined for downstream analyses to define three phenotypes: PP, PL/LP and LL. LL, low basal/low reserve; LP, low basal/preserved reserve; PL, preserved basal/low reserve; PP, preserved basal/preserved reserve; UWSF, unstimulated whole salivary flow.

At baseline, the PP phenotype exhibited the highest UWSF and Saxon test values, whereas the LL phenotype showed uniformly low values across both measures. The PL/LP group demonstrated intermediate functional profiles (table 2). Serum rheumatoid factor and IgG levels were higher in the LL phenotype compared with the other groups, while ESSDAI and clinical ESSDAI did not differ significantly across phenotypes. Baseline use of hydroxychloroquine, glucocorticoid and immunosuppressants was comparable across groups. Secretagogue use showed a numerical increase across more impaired phenotypes but did not reach statistical significance. Disease duration from institutional diagnosis was tightly clustered around zero in all phenotypes, reflecting the cohort’s recently diagnosed predominance, while a minority of LL patients had documented institutional follow-up prior to baseline assessment. Symptom duration did not differ significantly across phenotypes. Detailed characteristics of the original four quadrants are provided in online supplemental table S1.

Table 2. Characteristics of anti-SSA-positive patients with SjD among different phenotypes of glandular function.

	Impairment in 0 measure: PP (n=62)	Impairment in one measure: PL/LP (n=74)	Impairment in two measures: LL (n=57)	P value
Age, years, median (IQR)	57.1 (47.1–68.4)	60.7 (50.2–70.2)	58.2 (48.7–69.5)	0.679
Female, n (%)	52 (84)	66 (89)	52 (91)	0.434
Comorbid autoimmune disease, n (%)	15 (24)	14 (19)	20 (35)	0.105
Disease duration from institutional diagnosis, years, median (IQR)	0.0 (0.0–0.0)	0.0 (0.0–0.0)	0.0 (0.0–0.3)	0.006*
Symptom duration from patient-reported onset, years, median (IQR)	0.8 (0.4–4.6)	1.1 (0.4–4.9)	2.2 (0.4–9.5)	0.112
UWSF, mL/min, median (IQR)	0.2 (0.2–0.3)	0.1 (0.0–0.1)	0.0 (0.0–0.1)	<0.001*
Saxon, g/2 min, median (IQR)	4.4 (3.1–5.7)	3.6 (2.6–4.6)	0.9 (0.5–1.3)	<0.001*
Schirmer’s test ≤5 mm/5 min in at least one eye, n (%)	58 (94)	49 (67)†	42 (75)†	<0.001*
Rheumatoid factor, IU/mL, median (IQR)	0.0 (0.0–14.6)‡	0.0 (0.0–19.2)†	13.0 (0.0–47.1)†	0.018*
C3, mg/dL, median (IQR)	115.0 (101.4–129.2)‡	116.8 (104.3–129.5)†	115.0 (99.4–132.3)	0.704
C4, mg/dL, median (IQR)	28.0 (22.3–35.0)§	28.0 (23.7–35.0)	25.6 (21.6–32.2)	0.319
IgG, mg/dL, median (IQR)	1450.0 (1182.4–1584.1)‡	1481.7 (1274.0–1753.1)†	1699.3 (1369.7–2034.3)	0.004*
ESSDAI, median (IQR)	1.0 (0–4)	1.0 (0–6)	2.0 (1–5)	0.151
clinESSDAI, median (IQR)	0.0 (0–3)	0.0 (0–6)	0.0 (0–6)	0.380
Oral symptoms: dry mouth, n (%)	41 (66)	48 (65)	52 (91)	0.001*
Oral symptoms: swollen salivary glands, n (%)	3 (4.8)	11 (15)†	14 (25)	0.010*
Oral symptoms: drink liquids to aid in swallowing dry food, n (%)	16 (26)	34 (47)†	39 (68)	<0.001*
ESSPRI—dryness, median (IQR)	5.0 (3.0–7.0)	6.0 (3.0–8.0)	7.0 (5.0–8.0)	<0.001*
ESSPRI—fatigue, median (IQR)	5.0 (3.0–7.0)	5.0 (3.0–8.0)	6.0 (4.0–7.0)	0.214
ESSPRI—pain, median (IQR)	3.5 (1.0–6.0)	4.0 (1.0–7.0)	4.0 (2.0–7.0)	0.659
Use of hydroxychloroquine in baseline, n (%)	18 (29)	25 (34)	20 (35)	0.753
Use of glucocorticoids in baseline, n (%)	9 (15)	10 (14)	6 (11)	0.797
Use of immunosuppressants in baseline, n (%)	8 (13)	12 (16)	5 (8.8)	0.453
Use of pilocarpine or cevimeline in baseline, n (%)	1 (1.6)	2 (2.7)	6 (11)	0.061
Use of anticholinergics in baseline, n (%)	4 (6.5)	2 (2.7)	4 (7.0)	0.505

Data presented as median (IQR) or n (%).

^*P<0.05.

^†Data were missing in one patient.

^‡Data were missing in two patients.

^§Data were missing in three patients.

clinESSDAI, clinical ESSDAI; ESSDAI, EULAR Sjögren’s Syndrome Disease Activity Index; ESSPRI, EULAR Sjögren’s Syndrome Patient-Reported Index; EULAR, European Alliance of Associations for Rheumatology; IQR, interquartile range; LL, low UWSF/low Saxon; LP, low UWSF/preserved Saxon; PL, preserved UWSF/low Saxon; PP, preserved UWSF/preserved Saxon; SjD, Sjögren’s disease; UWSF, unstimulated whole salivary flow.

Stepwise increases in dryness-related symptom burden and structural impairment across functional phenotypes

Graded differences in dryness-related symptom burden were observed across phenotypes (figure 2). ESSPRI dryness scores increased stepwise from PP (median 5.0) to PL/LP (median 6.0) to LL (median 7.0) (Jonckheere-Terpstra test, p<0.001; figure 2A). A similar gradient was observed for swallowing difficulty, affecting 26%, 47% and 68% of PP, PL/LP and LL patients, respectively (Cochran-Armitage trend test, p<0.001; figure 2B). In contrast, fatigue and pain domains showed no significant monotonic association, indicating that functional phenotyping primarily captures the gland-specific component of patient-reported burden (table 2).

Figure 2. Clinical and structural anchoring of basal-reserve functional phenotypes in Sjögren’s disease. (A) Distribution of ESSPRI dryness scores across three functional phenotypes defined by combined basal (UWSF) and reserve (Saxon test) salivary function: PP, PL/LP and LL. Boxes indicate IQR with median; points represent individual patients. A significant increasing trend was observed across ordered phenotypes (Jonckheere-Terpstra test). (B) Prevalence of swallowing difficulty across phenotypes. A monotonic increase in dysphagia was observed from PP to LL (Cochran-Armitage trend test). (C) Structural correlates of functional phenotype in SjD. Distribution of SMG gland ratio across three ordered functional phenotype groups defined by combined basal and reserve salivary function: PP (n=17); PL/LP (n=18); LL (n=16). ESSPRI, EULAR Sjögren’s Syndrome Patient-Reported Index; LL, low UWSF/low Saxon; LP, low UWSF/preserved Saxon; PP, preserved UWSF/preserved Saxon; PL, preserved UWSF/low Saxon; SjD, Sjögren's disease; SMG, submandibular gland; UWSF, unstimulated whole salivary flow.

Among patients with available salivary gland biopsy data (n=51), SMG structural integrity, quantified by the SMG gland ratio, differed significantly across phenotypes (p<0.001; figure 2C). The SMG gland ratio was highest in the PP phenotype (n=17) and lowest in the LL phenotype (n=16), with PL/LP (n=18) occupying an intermediate position. Together, these findings show that combined basal-reserve functional classification aligns with both patient-reported dryness burden and objective structural glandular integrity.

Multivariable associations between functional phenotypes and clinical outcomes

Multivariable regression analyses were performed to evaluate the associations between functional phenotypes and key clinical outcomes after adjustment for age, sex, serological markers, comorbid autoimmune disease, disease activity and disease duration from institutional diagnosis (table 3). Functional phenotype remained independently associated with dryness-related outcomes and SMG gland ratio in adjusted models. Compared with the PP phenotype, the LL phenotype was associated with significantly higher ESSPRI dryness scores and increased odds of swallowing difficulty. The adjusted OR for swallowing difficulty was 6.19 (95% CI 2.71 to 14.9) for LL and 2.28 (95% CI 1.08 to 4.98) for PL/LP relative to PP (table 3). A similar graded association was observed between SMG gland ratio and functional phenotype (table 3). Baseline medication-adjusted sensitivity analyses, additionally including baseline hydroxychloroquine, glucocorticoid, immunosuppressant, anticholinergic and secretagogue use, yielded consistent associations between functional phenotype and ESSPRI dryness, swallowing difficulty and SMG gland ratio (online supplemental table S2A).

Table 3. Regression models for ESSPRI—dryness, drink liquids to aid in swallowing food and SMG gland ratio in anti-SSA-positive patients with SjD.

	ESSPRI—dryness (multivariable linear regression)		Drink liquids to aid in swallowing food (multivariable logistic regression)		SMG gland ratio (multivariable linear regression)
	β (95% CI)	P value	OR (95% CI)	P value	β (95% CI)	P value
Age, year	0.01 (−0.02 to 0.04)	0.604	0.99 (0.96 to 1.01)	0.206	−0.01 (−0.01 to −1.69×10−3)	0.008*
Male	0.29 (−0.96 to 1.53)	0.651	0.67 (0.24 to 1.79)	0.433	0.22 (0.03 to 0.41)	0.025*
Comorbid autoimmune disease	−0.36 (−1.28 to 0.57)	0.445	0.75 (0.36 to 1.56)	0.448	0.04 (−0.11 to 0.18)	0.635
Rheumatoid factor	4.32×10−4 (−1.78×10−3 to 2.65×10−3)	0.701	1.00 (1.00 to 1.00)	0.992	−9.80×10−4 (−1.89×10−3 to −7.50×10−5)	0.034*
ESSDAI	−0.01 (−0.10 to 0.09)	0.906	1.04 (0.97 to 1.12)	0.298	−0.01 (−0.02 to 0.01)	0.361
Functional phenotype
PP	Reference		Reference		Reference
PL/LP	0.40 (−0.55 to 1.36)	0.407	2.28 (1.08 to 4.98)	0.034*	−0.20 (−0.33 to −0.06)	0.005*
LL	1.67 (0.64 to 2.70)	0.002*	6.19 (2.71 to 14.9)	<0.001*	−0.43 (−0.59 to −0.28)	<0.001*
Disease duration from institutional diagnosis, year	0.14 (0.01 to 0.28)	0.041*	1.05 (0.94 to 1.19)	0.383	0.00 (−0.02 to 0.02)	0.764

^*P<0.05.

ESSDAI, EULAR Sjögren’s Syndrome Disease Activity Index; ESSPRI, EULAR Sjögren’s Syndrome Patient-Reported Index; EULAR, European Alliance of Associations for Rheumatology; IQR, interquartile range; LL, low UWSF/low Saxon; LP, low UWSF/preserved Saxon; PL, preserved UWSF/low Saxon; PP, preserved UWSF/preserved Saxon; SjD, Sjögren’s disease; SMG, submandibular gland; UWSF, unstimulated whole salivary flow.

In contrast, ESSDAI and rheumatoid factor showed weaker or non-significant associations with dryness-related outcomes and SMG gland ratio in adjusted models. Sensitivity analyses substituting ESSDAI with serum IgG yielded consistent results, with LL remaining significantly associated with higher dryness scores and increased odds of swallowing difficulty (online supplemental table S2B). Notably, the baseline IgG level was associated with ESSPRI dryness and SMG gland ratio.

Model-comparison analyses further clarified the role of categorical functional phenotype relative to continuous salivary measurements (online supplemental table S3). For ESSPRI dryness, the UWSF×Saxon interaction improved model fit compared with the additive continuous-measure model, supporting joint basal-reserve interpretation (online supplemental table S4). For swallowing difficulty, neither the interaction term nor the addition of phenotype to continuous UWSF and Saxon values clearly improved model fit (online supplemental table S5). In the SMG biopsy subset, however, adding phenotype to continuous UWSF and Saxon values improved model fit for SMG gland ratio, providing structural corroboration for the phenotype framework (online supplemental table S6). These findings indicate that functional phenotype should be viewed as a clinically practical operational summary of basal-reserve configuration, with outcome-dependent incremental value rather than universal superiority over continuous salivary measures.

Empirical longitudinal trajectories of salivary gland function

Among the 193 patients included in the study, 127 (66%) contributed at least two salivary function assessments. Among patients with ≥2 assessments, the median follow-up duration was 504 days (IQR 196–637). The distribution of baseline functional phenotypes differed modestly between patients with longitudinal follow-up and those with only a single visit, with the LL phenotype more frequently represented among patients with repeated assessments (p=0.044). Baseline UWSF, Saxon test values and ESSDAI scores were otherwise comparable between the follow-up group and those with a single visit. The number of follow-up assessments per patient did not differ significantly across phenotypes. Detailed baseline comparisons are provided in online supplemental table S7.

Individual trajectories demonstrated substantial interindividual variability within each phenotype for UWSF (figure 3A; online supplemental figure S2). LOESS curves revealed phenotype-specific patterns. The PP phenotype showed a gradual decline over time, becoming more apparent approximately 1 year after baseline (figure 3A). In contrast, PL/LP remained relatively stable, whereas LL persisted at uniformly low levels with minor fluctuations near the measurement boundary.

Figure 3. Longitudinal salivary end point behaviour and phenotype retention according to baseline functional phenotype. (A) Longitudinal trajectories of UWSF over 0–600 days, stratified by baseline functional phenotype. Thin lines represent individual patient trajectories; bold curves represent LOESS-smoothed trends with 95% confidence bands. (B) Longitudinal trajectories of Saxon test results over 0–600 days, stratified by baseline functional phenotype, displayed using the same format as panel A. (C) Baseline-to-last-visit phenotype retention and reclassification among patients with repeated salivary assessments. Stacked bars show the proportion of patients in each last-visit phenotype according to baseline phenotype. Same-phenotype retention was observed in 76.3% of PP, 59.1% of PL/LP and 69.8% of LL patients. These longitudinal patterns should be interpreted in the context of baseline functional state, floor constraints and proximity to operational UWSF or Saxon thresholds, rather than as a simple natural-history sequence. LL, low UWSF/low Saxon; LP, low UWSF/preserved Saxon; LOESS, locally estimated scatterplot smoothing; PL, preserved UWSF/low Saxon; PP, preserved UWSF/preserved Saxon; UWSF, unstimulated whole salivary flow.

For the Saxon test, the PP phenotype exhibited higher baseline reserve capacity followed by a modest decline (figure 3B, online supplemental figure S2). The PL/LP phenotype again demonstrated relative stability. The LL phenotype showed low baseline values with a modest upward trend. Given the proximity to the measurement floor, this pattern is more consistent with low-range end point variability than with sustained biological recovery.

Model-based estimation of annualised functional change

To quantify longitudinal trends while accounting for within-subject correlation, GEE models were applied to estimate annualised rates of change for UWSF and Saxon test results (online supplemental figure S3; online supplemental tables S8 and S9).

For UWSF, the PP phenotype demonstrated a significant negative annualised slope (−0.026 mL/min/year, 95% CI −0.044 to −0.008), indicating a measurable negative slope in basal secretion within this baseline-defined group. In contrast, neither PL/LP nor LL showed significant longitudinal change. Post hoc comparisons confirmed that the rate of decline in PP differed significantly from both PL/LP and LL (online supplemental table S9). The negative PP slope was robust to cluster bootstrap resampling (1000 replicates; bootstrap 95% CI −0.069 to −0.014), with a negative direction observed in 99.5% of resamples, whereas the slope directions for PL/LP and LL were not directionally stable across resamples (negative in 25.2% and 51.0%, respectively).

For the Saxon test, PP similarly exhibited a negative annualised slope (−0.332 g/2 min/year, 95% CI −0.606 to −0.058), whereas PL/LP remained stable. The LL phenotype showed a directionally positive slope (0.544 g/2 min/year, 95% CI 0.217 to 0.871). These directional patterns were also robust to cluster bootstrap resampling: the PP slope was negative in 99.7% of resamples (bootstrap 95% CI −0.859 to −0.136) and the LL slope positive in 99.2% (bootstrap 95% CI 0.067 to 0.775), whereas PL/LP slope direction was not directionally stable (positive in 47.4%). Among baseline LL patients with follow-up Saxon measurements, 29 of 43 (67.4%) remained below the impaired Saxon threshold throughout follow-up, whereas 14 of 43 (32.6%) crossed the threshold at least once. This supports a cautious interpretation of the directionally positive LL Saxon slope as low-range end point variability rather than uniform biological recovery.

Findings remained materially unchanged when restricting analyses to patients with at least two longitudinal assessments (online supplemental table S10). The directionally positive Saxon slope in LL persisted across alternative exclusion strategies and working correlation structures in GEE (online supplemental tables S11 and S12). Under more restrictive Saxon thresholds or after excluding cases with pronounced fluctuations, the magnitude of the positive LL slope attenuated toward zero (online supplemental tables S11 and S13).

Phenotype-specific annualised slopes for UWSF and Saxon remained materially unchanged after accounting for baseline hydroxychloroquine, glucocorticoid and immunosuppressant use, and visit-level secretagogue use (online supplemental tables S8 and S9). During the follow-up period, hydroxychloroquine use differed across phenotypes (78.9% in PP, 93.2% in PL/LP and 95.6% in LL; p=0.043), whereas the use of glucocorticoids, immunosuppressive medications and secretagogues showed no significant differences across the functional phenotypes (online supplemental table S14).

Phenotype stability and transition patterns

To determine whether continuous salivary changes translated into categorical phenotype shifts, we evaluated phenotype transitions among patients with assignable phenotypes at baseline and last follow-up (n=125, online supplemental figure S1). Functional phenotypes showed substantial within-state retention over a median follow-up of 16.6 months (IQR 6.4–21.2) in this population (n=125) (online supplemental table S15). In the baseline-to-last-visit analysis (figure 3C, online supplemental table S15, panel A), same-phenotype retention was 76.3% for PP, 59.1% for PL/LP and 69.8% for LL; transitions spanning the two severity extremes (PP↔LL) were infrequent and symmetric (PP→LL: 5.3%; LL→PP: 7%). Consecutive visit-to-visit transitions showed a concordant pattern (online supplemental table S15, panel B; same-state retention PP 70.3%, PL/LP 60.0%, LL 71.6%). The similar retention patterns in the two matrices suggest that within-phenotype stability was not limited to the baseline-to-last-visit comparison, but was also observed across adjacent clinical assessments. Transitions towards less impaired states were at least as frequent as transitions towards greater impairment (improving: 29 vs worsening: 11; binomial p=0.006), inconsistent with unidirectional monotonic decline within the observed window. Detailed four-quadrant transition patterns are shown in online supplemental table S16.

Boundary proximity correlates with transition probability

To contextualise phenotype transitions, we examined baseline proximity to the operational UWSF and Saxon thresholds.

Greater baseline distance from the nearest UWSF or Saxon threshold was associated with lower odds of phenotype reclassification during follow-up, after adjustment for follow-up duration, number of visits and baseline phenotype (OR 0.018 per one-unit increase in standardised minimum distance, 95% CI 0.002 to 0.130, p<0.001). Conversely, patients closer to either operational threshold were more likely to be reclassified. Online supplemental figure S4A illustrates this association using marginal adjusted transition probabilities, and online supplemental figure S4B shows that patients with phenotype transition clustered near the UWSF or Saxon thresholds. In the LL phenotype, those with transition had higher baseline UWSF and Saxon results than those without transition (online supplemental table S17).

Those in the PP phenotype showed a similar trend: patients with phenotype shift had lower baseline UWSF and Saxon results, which were closer to the threshold, than those with stable PP (online supplemental table S17). The use of hydroxychloroquine, glucocorticoid and immunosuppressants in those with and without phenotype transition were comparable in the stable LL and LL with transition during the follow-up period (online supplemental table S18).

Discussion

This study presents a pragmatic, dual-domain and clinically scalable framework for classifying salivary gland dysfunction in SjD, based on the combined assessment of UWSF and the Saxon test, which reflect basal secretion and secretory reserve, respectively. Three functional phenotypes (PP, PL/LP and LL) were delineated, corresponding to distinct glandular functional states associated with patient-reported dryness, swallowing difficulty and SMG structural integrity. Longitudinal analyses further indicated that salivary changes should be interpreted in light of baseline functional state, floor constraints and proximity to phenotype-defining thresholds. Together, these findings support the value of integrated functional phenotyping as a clinically relevant stratification framework that complements single-domain salivary assessments.

These findings should be interpreted as complementary to recent symptom-based, immune-based and serology-based stratifications of SjD.^{14 15} Using clustering based on symptoms, clinical manifestations and biological features, Nguyen et al identified a B-cell-active, low-symptom-burden subgroup that later showed worsening symptom state, suggesting that immune activity, systemic manifestations and symptom burden may be dissociated at baseline and may evolve differently over time. The present study addresses a distinct glandular functional axis defined by basal secretion and stimulated reserve over a shorter observation window. These findings do not exclude long-term immune-mediated progression from preserved to globally impaired glandular function, but indicate that basal-reserve phenotyping provides a glandular-state readout complementary to immune-endotype stratification when interpreting salivary symptoms, glandular structural integrity and salivary end points.

The longitudinal analyses should be interpreted as baseline-state-conditioned salivary end point behaviour rather than as a direct natural-history sequence.^{16 17} Annualised slopes and phenotype transition matrices provide complementary information. Slopes describe group-level continuous changes in UWSF or Saxon values, whereas transition matrices indicate whether these changes cross phenotype-defining thresholds. Because the same salivary measurements were used to define baseline phenotypes and to evaluate subsequent change, group-level slopes are susceptible to regression-to-the-mean effects, particularly in PP and LL phenotypes positioned at the extremes of the baseline distribution. In the present study, PP showed negative UWSF and Saxon slopes, whereas LL showed a directionally positive Saxon slope; these patterns are therefore best interpreted in relation to baseline measurable range and floor constraints rather than as uniform progression or biological recovery. The transition analyses further showed that most patients remained within their baseline phenotype over a median follow-up of 16.6 months, and that transitions across the two extremes were uncommon. Moreover, phenotype reclassification was strongly associated with proximity to the UWSF or Saxon thresholds, indicating that short-term categorical transitions were largely boundary-dependent. Medication-adjusted sensitivity analyses did not materially alter the phenotype-specific slopes, although treatment exposure was observational and non-randomised. Together, these findings support the use of PP, PL/LP and LL as operational baseline functional states for end point interpretation, rather than as fixed disease stages. These findings do not exclude long-term progression from preserved to globally impaired functional states driven by immune-mediated glandular injury.¹⁴ Rather, they indicate that short-term salivary slopes and phenotype transitions should be interpreted in relation to baseline state, floor constraints and threshold proximity.

These findings have implications for the design and interpretation of studies targeting salivary gland outcomes in SjD.⁶¹⁸,²² Although salivary flow can be appropriately analysed as a continuous outcome, a numerically similar change may have different interpretive value depending on baseline functional state, remaining measurable range, proximity to phenotype-defining thresholds and persistence over time. Importantly, the present study does not establish a responder threshold for clinically meaningful salivary improvement. Rather, it suggests that future definitions of meaningful improvement should not rely solely on absolute salivary change or single-visit threshold crossing. Apparent improvement in floor-limited or boundary-adjacent patients may reflect low-range fluctuation or operational reclassification, consistent with prior evidence that stimulated salivary flow measurements show substantial intraindividual variability.²³ Sustained improvement in patients with residual functional reserve may be more interpretable, particularly when accompanied by concordant changes in patient-reported outcomes, imaging, tissue structure or biological markers. Thus, basal-reserve phenotyping may be most useful at present as a pragmatic baseline glandular-state proxy and end point-interpretation layer, rather than as a standalone trial end point. Future prospective studies should evaluate state-conditioned responder definitions that incorporate durability of change, threshold proximity, baseline structural reserve and concordance with patient-reported outcomes, biological markers or imaging.

Several limitations should be acknowledged. First, this was a retrospective single-centre study, and the timing and number of salivary assessments reflected clinical practice rather than a prespecified follow-up protocol. The median follow-up of 16.6 months provides a limited window for characterising the natural salivary trajectory in SjD, and longer prospective observation will be required for direct inference of long-term functional dynamics. Residual selection bias and follow-up heterogeneity therefore cannot be excluded. Second, the phenotypes were defined using established UWSF and Saxon thresholds. This threshold-based approach supports clinical applicability, but longitudinal use of such categories requires attention to patients near the operational cutoffs, who may be reclassified after small fluctuations in salivary measurements. Accordingly, phenotype transitions should be interpreted in the context of boundary proximity rather than assumed to represent biological recovery or progression. Third, PL and LP were combined as a pragmatic discordant-function category because the PL subgroup was small. This grouping does not imply biological equivalence, and larger cohorts are needed to determine whether basal-low/reserve-preserved and basal-preserved/reserve-low states differ in mechanism or prognosis. Fourth, SMG biopsy data were available only in a subset of patients, limiting structural validation. Finally, treatment exposure was heterogeneous and non-randomised. Although sensitivity analyses adjusting for baseline immunomodulators and visit-level secretagogue use yielded similar longitudinal patterns, these models account only for measured medication exposure and should not be interpreted causally; residual confounding by indication remains possible. Long-term progression across functional phenotypes remains biologically plausible, but whether such progression relates to glandular or immune factors could not be addressed within the observed follow-up in this study and should be evaluated in prospective studies. External validation in independent prospective cohorts with standardised salivary testing, imaging and tissue-based biomarkers will be needed.

In conclusion, integrated assessment of basal secretion and stimulated reserve defines clinically meaningful functional phenotypes of salivary gland dysfunction in SjD. These phenotypes are associated with dryness-related symptom burden, swallowing difficulty and SMG structural integrity, supporting their relevance as operational glandular functional states. Longitudinal analyses further indicate that salivary changes should be interpreted in the context of baseline functional state, floor constraints and proximity to phenotype-defining thresholds, rather than as direct evidence of biological progression or recovery. Basal-reserve phenotyping may therefore provide a pragmatic framework for clinical stratification and for the design and interpretation of studies targeting glandular outcomes in SjD.

Data availability statement

Data are available on reasonable request.