Dry Mouth by Report or by Measure? Comparison of Fox and Xerostomia Inventory Scores with Sialometry Authors: Magdalena Ostrowska, Layla Settaf-Cherif, Jan Kiersztyn, Adam Ostrowski, Tomasz A Drewa, Jan Adamowicz, Maciej J Wróbel Published: 2026 ABSTRACT AbstractBackgroundDespite its high prevalence, xerostomia remains frequently misunderstood and often incorrectly assumed to be synonymous with reduced salivary flow. In reality, these two terms- xerostomia and hyposalivation - are not interchangeable.We aimed to quantify how Fox questionnaire and the Xerostomia Inventory (XI) relate to unstimulated salivary flow, identify the most informative items, and evaluate whether total scores and item‑based models discriminate clinically relevant hyposalivation thresholds and treatment‑related deterioration.MethodsWe analyzed 101 adults with overactive bladder contributing 202 paired assessments (fixed order: off‑treatment baseline, on‑treatment follow‑up). Xerostomia was assessed using the Fox questionnaire (0–10) and Xerostomia Inventory (XI; 11–55). Unstimulated salivary flow (mL/min) was measured with the 5‑min spitting method. Associations between questionnaire scores and salivary flow were summarized using Pearson correlations. Discrimination for hyposalivation thresholds (<0.10, <0.066, <0.033 mL/min) was evaluated using ROC/AUC with cluster bootstrap 95% CIs; item‑based screening was explored using LASSO with patient‑level cross‑validation.ResultsContinuous associations: Fox scores were more strongly associated with salivary flow than XI (r = −0.356 vs −0.160), while Fox and XI totals were strongly correlated (r = 0.71). Threshold discrimination: for hyposalivation <0.10 mL/min, AUC was 0.704 for Fox and 0.586 for XI. A sparse XI item model (six items) improved discrimination compared with the XI total score (CV‑AUC 0.709 vs 0.580). During treatment, 29/101 participants met the deterioration endpoint (≥50% reduction in flow and/or incident hyposalivation).ConclusionsQuestionnaire scores showed limited criterion validity for objective hypofunction. Fox performed better than XI for identifying hyposalivation, and item‑focused models were promising but exploratory. We propose a pragmatic diagnostic approach in which questionnaires quantify symptom burden and triage risk, while sialometry confirms hyposalivation when clinically relevant.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12903-026-08450-z.Keywords: Xerostomia, Fox, Hyposalivation, Medication-induced, OAB Background Dryness of the oral cavity (xerostomia) is a widespread complaint, affecting approximately 22% of the general population [1]. At some point, every physician, regardless of specialty, is likely to encounter a patient experiencing xerostomia This is largely due to the fact that various groups of medications are among the most common causes of dry mouth [2]. Despite its high prevalence, xerostomia remains frequently misunderstood and often incorrectly assumed to be synonymous with reduced salivary flow [3]. In reality, these two terms - xerostomia and hyposalivation - are not interchangeable [4–6]. Xerostomia, by definition, is a subjective perception of dry mouth, whereas hyposalivation refers to objectively reduced salivary secretion, typically quantified by sialometry [7]. Xerostomia can be assessed using single screening questions or self-reported instruments (e.g., Fox, Xerostomia Inventory, RODI, Bother Index) [8–12]. Among available patient‑reported tools, we focused on the Fox questionnaire and the Xerostomia Inventory (XI) because they are among the most widely used xerostomia instruments and represent two contrasting formats: a short dichotomous symptom checklist (Fox) versus a multi‑item Likert scale capturing the frequency of dryness‑related experiences, including extra‑oral symptoms (XI). Importantly, xerostomia symptom reports and these instruments (or closely related xerostomia questionnaires) have been examined against objective salivary measures in prior work, providing a foundation for evaluating criterion validity against sialometry [6, 8, 13, 14]. These measures are intended to capture the patient’s symptom experience and functional impact; however, symptom severity is sometimes used in practice as a surrogate for salivary gland hypofunction, despite the constructs being distinct [3]. Although diagnostic thresholds vary with the measurement protocol, hyposalivation is commonly defined as salivary flow ≤ 0.1 mL/min for unstimulated methods, and ≤ 0.5–0.7 mL/min. for stimulated [8, 9]. Consequently, it remains important to clarify how patient‑reported xerostomia relates to measured salivary output. Clinically, symptom management targets xerostomia, whereas prevention of oral complications depends on identifying and addressing true salivary hypofunction. Therefore, our study aimed to compare two widely used xerostomia instruments - the Fox questionnaire and the Xerostomia Inventory against objectively measured unstimulated salivary flow to quantify criterion validity. We further examined which individual questions were most closely associated with objective flow and evaluated the ability of total scores and item-based models to discriminate clinically relevant hyposalivation thresholds. We believe that accurate assessment and management of xerostomia are critical for patient comfort and prevention of complications. Methods Study design and participants The analysis was based on a paired repeated-measures dataset that included patient visits occurring both during periods with and without pharmacological treatment for overactive bladder (OAB); the analytic aim was to quantify agreement between patient-reported xerostomia and objective salivary output, and to assess the ability of questionnaire scores (and their individual items) to identify hyposalivation. We performed complete-case analyses for each comparison. Each participant contributed two assessments in a fixed order (off‑treatment baseline followed by on‑treatment follow‑up). Follow-up assessments were performed after 2 weeks of pharmacotherapy initiation. Participants were recruited from the Department of Urology, University Hospital No 1 in Bydgoszcz, Poland. Demographic data, including age, sex, weight, and height, were recorded, along with medical history details such as comorbid conditions, concurrent medications, and smoking status. Enrollment criteria were: adults (≥ 18 years) with overactive bladder diagnosed by a specialist urologist according to contemporary definitions [10, 11]. Symptoms were considered to have a clinically meaningful impact based on patient-reported bother and the treating clinician’s decision to initiate pharmacotherapy (no separate QoL instrument was administered). All participants provided written informed consent; the study was approved by the Local Bioethics Committee (KB/205/2023). Measures I Questionnaires Xerostomia was assessed with the Fox questionnaire (ten yes/no items; total 0–10, higher scores indicate worse dryness) [15] and the Xerostomia Inventory (XI; eleven items rated on a 5‑point Likert scale; total 11–55, higher scores indicate worse dryness) [8]. Questionnaires were administered in Polish using a forward‑translation and clinical‑review process; however, formal cross‑cultural validation of the Polish versions was not performed. II Objective salivary flow Unstimulated saliva was measured using the 5-minute spitting method. The collected mass (grams) over 5 min was recorded and converted to flow (mL/min) assuming 1 g ≈ 1 mL Due to logistical constraints, all collections were scheduled in the afternoon (approximately 14:00–16:00) at both time points; participants were instructed to refrain from eating, drinking, smoking, or oral rinsing for at least two hours before testing. Hyposalivation was evaluated using three pre-specified unstimulated salivary flow rates Hyposalivation was evaluated using three pre-specified unstimulated salivary flow thresholds: Threshold I: <0.10 mL/min; Threshold II: <0.066 mL/min; Threshold III: <0.033 mL/min [9]. These cutoffs reflect a graded severity scheme (mild/moderate/severe hypofunction) used in prior salivary dysfunction literature; we acknowledge that other unstimulated cutoffs (e.g., < 0.20 mL/min and < 0.30 mL/min) are also reported depending on protocol and population [15–17]. For exploratory analyses, a deterioration endpoint was defined as either (I) a ≥ 50% reduction in unstimulated salivary flow from the patient’s baseline measurement, or (II) incident hyposalivation (≤ 0.10 mL/min) after treatment initiation. Baseline was always the off‑treatment assessment, and follow‑up was always the on‑treatment assessment for the same participant (fixed order). Statistical analysis Analyses were performed using TIBCO Statistica 13.3. Penalized regression (LASSO) was implemented in Python. Analyses were conducted as complete cases for each comparison. All tests were two‑sided. Because each participant contributed two assessments, analyses using both visits accounted for within-person clustering. Confidence intervals for correlations and ROC/AUCs were obtained using patient-level (cluster) bootstrap resampling; key estimates were repeated using the baseline assessment only as a sensitivity analysis (Supplementary Table S3). Concordance and correlations. Associations between Fox total score, XI total score, and unstimulated salivary flow (mL/min) were summarized using Pearson correlation coefficients. Spearman correlations were computed as a sensitivity analysis. Within-person baseline vs. follow-up changes in salivary flow and questionnaire totals were summarized descriptively and tested using paired Wilcoxon signed-rank tests (Table 5). Concordance and correlations. Associations between Fox total score, XI total score, and unstimulated salivary flow (mL/min) were summarized using Pearson correlation coefficients. Spearman correlations were computed as a sensitivity analysis. Within-person baseline vs. follow-up changes in salivary flow and questionnaire totals were summarized descriptively and tested using paired Wilcoxon signed-rank tests (Table 5). Item-level signals. For each item (Fox: F1–F10; XI: X1–X11), we correlated item scores with saliva (Pearson’s r). Item-level signals. For each item (Fox: F1–F10; XI: X1–X11), we correlated item scores with saliva (Pearson’s r). Discrimination for hyposalivation. For each unstimulated salivary flow threshold (I: <0.10, II: <0.066, III: <0.033 mL/min), we computed ROC curves and AUC with bootstrap 95% CIs (1,000 resamples). Optimal cut-offs were derived by maximizing Youden’s J; at these cut-offs, we report sensitivity, specificity, PPV, NPV with bootstrap Cis. Discrimination for hyposalivation. For each unstimulated salivary flow threshold (I: <0.10, II: <0.066, III: <0.033 mL/min), we computed ROC curves and AUC with bootstrap 95% CIs (1,000 resamples). Optimal cut-offs were derived by maximizing Youden’s J; at these cut-offs, we report sensitivity, specificity, PPV, NPV with bootstrap Cis. Sparse item models. To compare “sum-scores” with item-level screening, we fit L1-penalized logistic regression (LASSO) for threshold I using (I) Fox items, (II) XI items, (III) all items, each with 5-fold stratified cross-validation over a log-spaced C grid. Performance was summarized as mean CV-AUC. No clinical covariates were included by design. Internal consistency was summarized with Cronbach’s α for Fox and XI item sets. All tests were two-sided. Predictors were standardized within each training fold. Fox items were treated as binary (0/1) and XI items as ordinal (1–5). Models were fit as complete cases for the relevant item set. Hyperparameter C was tuned via cross-validation on the training data; To assess potential optimism, we additionally ran a nested cross validation sensitivity analysis (outer 5-fold for performance estimation with inner tuning), reported in Supplementary Table S4. Predictors were standardized within training folds; Fox items were modeled as binary and XI items as ordinal scores. No resampling or class-weighting was applied; model performance was summarized using AUC, which is prevalence-invariant. For item-level models, we used complete cases for the respective item set (participants with any missing item in the set were excluded for that model). Sparse item models. To compare “sum-scores” with item-level screening, we fit L1-penalized logistic regression (LASSO) for threshold I using (I) Fox items, (II) XI items, (III) all items, each with 5-fold stratified cross-validation over a log-spaced C grid. Performance was summarized as mean CV-AUC. No clinical covariates were included by design. Internal consistency was summarized with Cronbach’s α for Fox and XI item sets. All tests were two-sided. Predictors were standardized within each training fold. Fox items were treated as binary (0/1) and XI items as ordinal (1–5). Models were fit as complete cases for the relevant item set. Hyperparameter C was tuned via cross-validation on the training data; To assess potential optimism, we additionally ran a nested cross validation sensitivity analysis (outer 5-fold for performance estimation with inner tuning), reported in Supplementary Table S4. Predictors were standardized within training folds; Fox items were modeled as binary and XI items as ordinal scores. No resampling or class-weighting was applied; model performance was summarized using AUC, which is prevalence-invariant. For item-level models, we used complete cases for the respective item set (participants with any missing item in the set were excluded for that model). Exploratory deterioration models. For the deterioration endpoint, discrimination of item-based and total-score models was summarized using ROC/AUC. Deterioration analyses were conducted at the participant level using one baseline–follow‑up pair per participant. Operating characteristics were reported at pre-specified total-score cut-offs. Exploratory deterioration models. For the deterioration endpoint, discrimination of item-based and total-score models was summarized using ROC/AUC. Deterioration analyses were conducted at the participant level using one baseline–follow‑up pair per participant. Operating characteristics were reported at pre-specified total-score cut-offs. Table 5Baseline vs. on-treatment changes in salivary flow and questionnaire totalsMeasureBaselineOn-treatmentPaired p-value (Wilcoxon)Unstimulated flow (mL/min), mean (SD)0.34 (0.25)0.27 (0.22)< 0.001Fox total (0–10), median (IQR)4 (2–6)5 (2–6)0.059XI total (11–55), median (IQR)28 (21–34)30 (22–34)0.178Hyposalivation < 0.10 mL/min, n (%)9 (8.9%)20 (19.8%)—Hyposalivation < 0.066 mL/min, n (%)7 (6.9%)17 (16.8%)—Hyposalivation < 0.033 mL/min, n (%)3 (3.0%)8 (7.9%)— Baseline vs. on-treatment changes in salivary flow and questionnaire totals Results We enrolled 101 patients undergoing 202 assessments. Out of this group, 51 (50.5%) patients were on solifenacin 5 mg, 29 (28.7%) patients on solifenacin 10 mg and 21 (20.8%) patients on mirabegron 50 mg. Detailed group characteristics are presented in Table 1. Unstimulated salivary flow is summarized in Table 1 (mean ± SD at baseline and on-treatment) together with the distribution across hyposalivation thresholds. Distribution of concomitant medications is presented in Table 2. Table 1Group characteristicsNumber of patientsEntire study population101Median age (years);(IQR)69.0(IQR 63.0–76.0)Women/Men n; (%)61 (60.4%) / 40 (39.6%)Median BMI (IQR)28.0 (IQR 25.3–32.3)Number of people with hypertension n; (%)67 (66.3%)Number of people with diabetes n; (%)23 (22.7%)Number of people with rheumatological conditions n; (%)17 (16.8%)Neurological disorders n; (%)8 (7.9%)LS degeneration n (%)60 (59.4%)Post-RT/CHT head and neck n; (%)3 (3.0%)Smoking n; (%)17 (16.8%)Polypharmacy n; (%)41 (40.6%)Unstimulated salivary flow (mL/min), mean (SD) – baseline0.34 (0.25)Unstimulated salivary flow (mL/min), mean (SD) – on-treatment0.27 (0.22)Hyposalivation I (< 0.10 mL/min), n (%) – baseline9 (8.9%)Hyposalivation I (< 0.10 mL/min), n (%) – on-treatment20 (19.8%)Hyposalivation II (< 0.066 mL/min), n (%) – baseline7 (6.9%)Hyposalivation II (< 0.066 mL/min), n (%) – on-treatment17 (16.8%)Hyposalivation III (< 0.033 mL/min), n (%) – baseline3 (3.0%)Hyposalivation III (< 0.033 mL/min), n (%) – on-treatment8 (7.9%)Abbreviations used are LS Lumbosacral, RT Radiotherapy, and CHT Chemotherapy, IQR Interquartile range Group characteristics Median age (years); (IQR) 69.0 (IQR 63.0–76.0) Abbreviations used are LS Lumbosacral, RT Radiotherapy, and CHT Chemotherapy, IQR Interquartile range Table 2Concomitant medications’ distributionDrug Groupsolifenacin 5mg(n, %)solifenacin 10mg(n, %)mirabegron 50 mg(n, %)All (n, %)Antiplatelets8 (15.7%)7 (22.6%)5 (26.3%)20 (19.8%)Oral Anticoagulants6 (11.8%)2 (6.5%)2 (10.5%)10 (9.9%)Statins and Fibrates21 (41.2%)16 (51.6%)11 (57.9%)48 (47.5%)Alpha Blockers7 (13.7%)1 (3.2%)1 (5.3%)9 (8.9%)Beta Blockers13 (25.5%)11(35.5%)8 (42.1%)32 (31.7%)Diuretics9 (17.6%)7 (22.6%)4 (21.1%)20 (19.8%)Metformin7 (13.7%)8 (25.8%)3 (15.8%)18 (17.8%)Other Antidiabetic Drugs2 (3.9%)5 (16.1%)0 (0.0%)7 (6.9%)ACE Inhibitors or Angiotensin Receptor Blockers26 (51.0%)18 (58.1%)7 (36.8%)51 (50.5%)Painkillers and Anti-Inflammatory12 (23.5%)4 (12.9%)4 (21.1%)20 (19.8%)Allopurinol3 (5.9%)4 (12.9%)1 (5.3%)8 (7.9%)Calcium Channel Blockers11 (21.6%)7 (22.6%)3 (15.8%)21 (20.8%)Psychiatric Drugs9 (17.6%)5 (16.1%)4 (21.1%)18 (17.8%)L-Thyroxine4 (7.8%)6 (19.4%)4 (21.1%)14 (13.9%)Vitamin D310 (19.6%)5 (16.1%)0 (0.0%)15 (14.9%)Proton Pump Inhibitors9 (17.6%)5 (16.1%)5 (26.3%)19 (18.8%)Other Cardiovascular Drugs9 (17.6%)10 (32.3%)6 (31.6%)25 (24.8%) Concomitant medications’ distribution Questionnaire responses vs. unstimulated salivary flow. Questionnaire responses vs. unstimulated salivary flow. All correlation coefficients for individual items of both questionnaires are presented in Table 3. The Fox questionnaire demonstrated a moderate negative correlation (r = -0.356) with saliva production. The items most strongly correlated were question 9 (Do you perceive the volume of saliva in your mouth as too small? ), question 2 (Do you experience mouth dryness during the day? ), and question 10 (Do you need to moisten your mouth frequently? ) with correlation coefficients of r = -0.398, r = -0.385, and r = -0.325, respectively. Table 3Summary of the correlation coefficients between the individual questions of the Fox questionnaire and Xerostomia Inventory with the measured unstimulated salivary flow rateIndividualQuestionCorrelation coefficient rFox 1. Do you experience mouth dryness during night or upon waking up?-0.059 2. Do you experience mouth dryness during the day?-0.385 3. Do you keep a glass of water next to your bed?-0.172 4. Do you drink fluids while swallowing dry foods?-0.166 5. Do you experience mouth dryness during meals?-0.041 6. Do you experience problems with swallowing foods?-0.157 7. Do you use chewing gum on a daily basis to eliminate a feeling of mouth dryness?-0.140 8. Do you use hard fruit or mint candies on a daily basis to eliminate a feeling of mouth dryness?-0.117 9. Do you perceive the volume of saliva in your mouth as too small?-0.398 10. Do you need to moisten your mouth frequently?-0.325Xerostomia Inventory 1. I sip liquids to aid in swallowing food-0.143 2. My mouth feels dry when eating a meal-0.138 3. I get up at night to drink-0.088 4. My mouth feels dry-0.326 5. I have difficulty in eating dry foods-0.081 6. I suck sweets or cough lollies to relieve dry mouth-0.063 7. I have difficulties swallowing certain foods-0.094 8. The skin of my face feels dry-0.079 9. My eyes feel dry-0.059 10. My lips feel dry-0.209 11. The inside of my nose feels dry.-0.002 Summary of the correlation coefficients between the individual questions of the Fox questionnaire and Xerostomia Inventory with the measured unstimulated salivary flow rate The Xerostomia Inventory demonstrated a weak negative correlation with salivary flow (r = -0.160). Among its items, the strongest negative correlations with actual saliva production were observed for XI 4 (My mouth feels dry), and XI10 (My lips feel dry), with correlation coefficients of r = -0.326, and r = -0.209, respectively. Overall the Fox questionnaire showed a moderate negative correlation with actual saliva production exceeding the Xerostomia Inventory. 2.Inter-questionnaire consistency Inter-questionnaire consistency A strong positive correlation was observed between the total scores of the Fox Questionnaire and the Xerostomia Inventory, with a coefficient of r = 0.71. The strongest cross-questionnaire item correlations are summarized in Table 4. The full item-by-item correlation matrix is provided in Supplementary Table S2. Table 4Strongest item-to-item correlations between Fox and Xerostomia InventoryFoxXerostomia InventoryPearson rF8(Do you use hard fruit or mint candies on a daily basis to eliminate a feeling of mouth dryness? )XI6(I suck sweets or cough lollies to relieve dry mouth)0.71F6(I suck sweets or cough lollies to relieve dry mouth)XI7(I have difficulties swallowing certain foods)0.57F3(Do you keep a glass of water next to your bed? )XI3(I get up at night to drink)0.52F4(Do you drink fluids while swallowing dry foods? )XI1(I sip liquids to aid in swallowing food)0.49F5(Do you experience mouth dryness during meals? )XI7(I have difficulties swallowing certain foods)0.44F6(Do you experience problems with swallowing foods? )XI5(I have difficulty in eating dry foods)0.40 Strongest item-to-item correlations between Fox and Xerostomia Inventory F8 (Do you use hard fruit or mint candies on a daily basis to eliminate a feeling of mouth dryness? ) XI6 (I suck sweets or cough lollies to relieve dry mouth) F6 (I suck sweets or cough lollies to relieve dry mouth) XI7 (I have difficulties swallowing certain foods) F3 (Do you keep a glass of water next to your bed? ) XI3 (I get up at night to drink) F4 (Do you drink fluids while swallowing dry foods? ) XI1 (I sip liquids to aid in swallowing food) F5 (Do you experience mouth dryness during meals? ) XI7 (I have difficulties swallowing certain foods) F6 (Do you experience problems with swallowing foods? ) XI5 (I have difficulty in eating dry foods) Within-person changes from baseline (off treatment) to follow-up (on treatment) are summarized in Table 5. Mean (SD) unstimulated flow decreased from 0.34 (0.25) to 0.27 (0.22) mL/min. The proportion of assessments below the predefined hyposalivation thresholds increased at follow-up (< 0.10: 9/101 to 20/101; <0.066: 7/101 to 17/101; <0.033: 3/101 to 8/101). Overall, 29/101 participants met the deterioration endpoint; 26/101 had a ≥ 50% reduction in flow and 14/101 developed incident hyposalivation (≤ 0.10 mL/min), with overlap in 11 participants. Across paired assessments, reductions in salivary flow did not translate into proportional changes in questionnaire totals. Within-person changes in flow were weakly associated with changes in FOX and XI totals (Δflow vs. ΔFOX: Pearson r = − 0.00; Δflow vs. ΔXI: r = 0.11). Changes in FOX and XI totals were moderately correlated with each other (Pearson r = 0.44), indicating that both instruments captured a similar direction of symptom change despite limited correspondence with measured flow. 3.Discrimination for hyposalivation Discrimination for hyposalivation Subanalysis of ROC curves was performed in the division by the hyposalivation threshold. AUCs (95% CI) for Fox vs. XI totals are presented in Table 6. Across all thresholds, the Fox total outperformed XI. Table 6Discrimination of Fox and Xerostomia Inventory total scores for hyposalivation defined by unstimulated salivary flow thresholdsOutcome definitionScaleAUC (95% CI)Cut-off (Youden)Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)I < 0.10 mL/minFOX0.705 (0.590–0.814)≥ 40.897 (0.786–1.000)0.453 (0.360–0.547)0.217 (0.140–0.301)0.963 (0.915–1.000)I < 0.10 mL/minXI0.586 (0.460–0.724)≥ 310.621 (0.437–0.833)0.613 (0.521–0.705)0.212 (0.116–0.321)0.906 (0.845–0.965)II < 0.066 mL/minFOX0.693 (0.570–0.810)≥ 50.750 (0.520–0.966)0.576 (0.489–0.667)0.194 (0.111–0.283)0.944 (0.886–0.992)II < 0.066 mL/minXI0.561 (0.424–0.709)≥ 310.625 (0.423–0.833)0.607 (0.517–0.694)0.176 (0.098–0.269)0.923 (0.872–0.973)III < 0.033 mL/minFOX0.732 (0.580–0.861)≥ 41.000 (1.000–1.000)0.426 (0.342–0.518)0.092 (0.043–0.146)1.000 (1.000–1.000)III < 0.033 mL/minXI0.625 (0.413–0.818)≥ 310.727 (0.400–1.000)0.597 (0.510–0.683)0.094 (0.037–0.167)0.974 (0.943–1.000)AUC indicates the area under the receiver operating characteristic curve. 95% confidence intervals were obtained by nonparametric bootstrap resampling (1,000 resamples). Optimal cut-offs were derived by maximizing Youden’s J. PPV, and NPV depend on the outcome prevalence in the sample. Confidence intervals were obtained using patient-level (cluster) bootstrap resampling (1,000 resamples)XI Xerostomia Inventory, PPV Positive predictive value, NPV Negative predictive value Discrimination of Fox and Xerostomia Inventory total scores for hyposalivation defined by unstimulated salivary flow thresholds I < 0.10 mL/min I < 0.10 mL/min II < 0.066 mL/min II < 0.066 mL/min III < 0.033 mL/min III < 0.033 mL/min AUC indicates the area under the receiver operating characteristic curve. 95% confidence intervals were obtained by nonparametric bootstrap resampling (1,000 resamples). Optimal cut-offs were derived by maximizing Youden’s J. PPV, and NPV depend on the outcome prevalence in the sample. Confidence intervals were obtained using patient-level (cluster) bootstrap resampling (1,000 resamples) XI Xerostomia Inventory, PPV Positive predictive value, NPV Negative predictive value 4.Prediction of hyposalivation LASSO versus Sum Scores. Prediction of hyposalivation LASSO versus Sum Scores. In patient-level cross-validated logistic models, the Fox sum-score outperformed the XI sum-score (CV-AUC 0.697 vs. 0.578) for hyposalivation < 0.10 mL/min. At the item level, LASSO applied to XI items achieved CV-AUC 0.689 using 7 items (X1, X2, X3, X4, X5, X6, X8), whereas LASSO on Fox items showed limited shrinkage (CV-AUC 0.638; all 10 items retained). A combined LASSO over all items achieved the highest CV-AUC (0.718) but selected 12 predictors (F1, F4, F7, F8, F9, F10, X1, X2, X4, X5, X8, X10). Nested cross-validation yielded similar, slightly more conservative estimates (Table S4). Internal consistency was acceptable for the Fox questionnaire (Cronbach’s α = 0.693; n = 201 assessments with complete Fox items) and good for the Xerostomia Inventory (α = 0.812; n = 202). 5.Predictive models for hyposalivation or 50% reduction in saliva production during OAB treatment. Predictive models for hyposalivation or 50% reduction in saliva production during OAB treatment. The model based on individual items from the Fox Questionnaire demonstrated superior predictive ability (AUC = 0.79) compared to the model relying on the total Fox score (AUC = 0.57). At a cutoff score of 5 or higher (out of 10), the sensitivity of the Fox Questionnaire in predicting the occurrence of hyposalivation or a 50% reduction in salivary production was 52.6%, with a specificity of 60.5%. ROC curve is presented in Fig. 1. Receiver operating characteristic (ROC) analysis curve for the Fox questionnaire assessing individual items in the prediction of hyposalivation, of a 50% reduction in saliva production after initiating OAB treatment A similar observation was made for the Xerostomia Inventory. In the model incorporating all individual questionnaire items, the AUC was 0.74, while the model based solely on the total score yielded an AUC of 0.59. At a cutoff score of 29 (out of 55) on the XI, the sensitivity of predicting hyposalivation or a 50% reduction in salivary production was 55.0%, with a specificity of 60.5%, the ROC curve is presented in Fig. 2. Receiver operating characteristic (ROC) analysis curve for the Xerostomia Inventory, assessing individual items, in predicting hyposalivation (50% reduction in saliva production) after initiating OAB treatment Discussion Several validated questionnaires have been developed to assess the subjective perception of dry mouth. They are usually low-cost, easy for patients to complete, and can be evaluated by clinicians [18]. One of the earliest tools was introduced by Fox [13], and Thomson et al. later developed the Xerostomia Inventory (XI) [14]. Both aimed to identify a small number of key questions that could reliably indicate true salivary gland dysfunction in patients reporting oral dryness. Accordingly, we compared Fox and XI scores against unstimulated sialometry to quantify their criterion validity. In our cohort, Fox scores were more closely aligned with objective salivary hypofunction than XI. However, the magnitude of both correlations also highlights an important clinical point: subjective xerostomia is not equivalent to hyposalivation. Notably, some individuals experience dryness despite producing normal or even above-normal saliva volumes, whereas others demonstrate markedly reduced salivary flow yet report no subjective discomfort [3]. In Nielander’s study, a hyposalivation was found in 28.6% patients with xerostomia [19]. In a systematic review by Aliko et al. In the majority of the 129 reviewed papers, it was not documented whether xerostomia was accompanied by decreased salivary flow [20]. In a pharmacologically induced dry-mouth experiment involving healthy volunteers, Wolff and Kleinberg noted that oral dryness typically appeared once the total salivary flow rate fell below half its normal level—a threshold consistent with the earlier findings of Dawes [21]. In a study of the effects of anticholinergic agents on salivary flow, patients experienced a sensation of dry mouth when the normal unstimulated salivary flow rate was reduced by 40–50% [22]. Although Fox and XI were designed to capture patient-reported dryness, they differ in item content and focus. Several questions are conceptually overlapping or nearly identical across instruments, particularly those addressing compensatory behaviors (e.g., drinking fluids to aid swallowing, waking at night to drink) and functional difficulties with dry foods or swallowing. This conceptual overlap is reflected in the strong correlation between Fox and XI total scores suggesting that both instruments largely capture the same subjective xerostomia construct. The highest cross‑questionnaire item correspondences clustered around closely matched coping and functional items (Table 3), indicating that agreement between instruments is driven mainly by behaviors and eating/swallowing difficulties rather than by extra‑oral symptoms. In contrast, XI includes a broader set of symptoms that extend beyond the oral cavity, such as dryness of the facial skin, eyes, lips, and the inside of the nose. These items may reflect a wider “mucosal dryness” experience, but they are not expected to map as directly onto salivary gland output as questions that specifically probe oral wetness or perceived saliva volume. Our item-level results suggest that the strongest objective signal is concentrated in questions that directly probe persistent oral dryness and perceived lack of saliva, whereas context-dependent symptoms (e.g., night-time dryness or dryness during meals) may be influenced by factors other than gland output such as mouth breathing, sleep environment, food characteristics, and compensatory drinking behaviors. Sreebny et al., in their study, calculated that a simple question, “Does your mouth usually feel dry?”, demonstrated high sensitivity (93%) for identifying individuals with objectively low salivary flow rates. However, the specificity was 68% [8]. Within XI, items focused on intra-oral dryness showed the clearest relationship with salivary flow, while extra-oral dryness symptoms (skin/eyes/nose) were weakly related. This pattern is consistent with XI capturing a broader ‘mucosal dryness’ experience, which may not map directly to salivary secretion. Consistent with prior efforts to shorten xerostomia instruments, our exploratory modeling indicates that a small subset of XI items can retain most of the discrimination for objective hypofunction while reducing respondent burden. Importantly, higher internal consistency does not imply better criterion validity; instrument choice should reflect the clinical question (symptom burden vs. salivary hypofunction). The deterioration analysis further underscores that changes in measured flow do not necessarily translate into proportional changes in symptom scores over time. In clinical settings where medication changes or new oral complications occur, objective flow measurement can help distinguish symptom fluctuation from clinically meaningful hyposecretion. A proposed step-by-step diagnostic approach is presented in Fig. 3. Proposed step-by-step diagnostic approach in xerostomia. The algorithm outlines initial clinical screening, questionnaire‑based triage when flow measurement is unavailable, confirmatory unstimulated sialometry when indicated, and management pathways differing for normal flow (symptom‑focused) versus hyposalivation (complication‑risk prevention and reassessment) Limitations Limitations include the afternoon timing of saliva collections. Because salivary flow shows diurnal variation, afternoon sampling may underestimate “morning‑worst” low‑flow cases, although comparisons within a fixed time window should remain internally valid. Moreover, the lack of formal cross-cultural validation of the Polish questionnaire versions, and limited characterization of potential confounders such as obstructive sleep apnea or permanent salivary gland damage after head-and-neck radiotherapy might be potential limitation. Follow-up timing reflected real-world pharmacotherapy exposure and was not standardized to a fixed interval. Finally, detailed medication regimens (agents and dosages beyond OAB therapy and concomitant xerogenic medications) will be added in the revised submission to enable more granular confounding assessment. Conclusions Questionnaire scores showed limited criterion validity for objectively reduced unstimulated salivary flow. Fox performed modestly better than XI for screening, but neither instrument can replace sialometry when confirmation of salivary hypofunction is clinically relevant; item‑focused models may support triage but require external validation. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Abbreviations Overactive bladder Xerostomia Inventory Lumbosacral Radiotherapy Chemotherapy Interquartile range Receiver operating characteristic Authors’ contributions Conceptualization: MO, AO; Methodology: MO, AO, MJW; Investigation: MO, LSC, JK; Writing—original draft preparation: MO; Writing—review and editing: MJW, AO; Supervision: MJW, JA, TAD. Funding No funding was received for conducting this study. Data availability To protect participant privacy and in accordance with European data protection regulations, we did not seek participant consent to deposit the study data in a public repository. However, the underlying data are available from the corresponding author upon reasonable request, subject to ethical and institutional guidelines. Declarations Ethics approval and consent to participate The ethics approval (KB/205/2023), was obtained from the ethics committee of Collegium Medicum University in Bydgoszcz, Nicolaus Copernicus University in Torun. The procedures used in this study adhere to the tenets of the Declaration of Helsinki. Informed consent to participate was obtained from all individual participants included in the study. Consent for publication Not applicable. 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. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Supplementary Material 4. Data Availability Statement To protect participant privacy and in accordance with European data protection regulations, we did not seek participant consent to deposit the study data in a public repository. However, the underlying data are available from the corresponding author upon reasonable request, subject to ethical and institutional guidelines.