Identifying Risk Factors for Sialolithiasis

Abstract

Objective:

To investigate potential risk factors associated with the development of sialolithiasis in a database-driven case-control study.

Methods:

The All of Us database was queried for participants with sialolithiasis. These participants were matched to controls by age, race, and gender. Demographics and medical history data, such as obesity, hypertension, tonsillitis, gout, and other conditions associated with salivary stones, were extracted and compared among participants. A logistic regression model using optimized parameters was used to identify variables associated with sialolithiasis.

Results:

In total, 2160 participants were included in the analysis, with 540 diagnosed with sialolithiasis and 1620 matched controls. Sjogren’s syndrome (OR=2.057, 95% CI: 1.106-3.824, p=0.023) and obesity (OR=1.419, 95% CI: 1.118-1.802, p=0.004) were significantly associated with increased odds of sialolithiasis, while essential hypertension (OR=0.279, 95% CI: 0.23-0.339, p=0.0) and type 2 diabetes without complication (OR =0.771, 95% CI: 0.601-0.989, p=0.041) were inversely associated. No significant associations were observed for dehydration, hyperparathyroidism, acute tonsillitis, alcohol abuse, smoking, hypercalcemia, gout, nephrolithiasis, gallstones, systemic lupus erythematosus, osteoporosis, or bipolar disorder.

Conclusions:

This study, using a large-scale database, aimed to identify potential risk factors associated with sialolithiasis, thereby improving our understanding of its pathogenesis and paving the way for early diagnosis and intervention. The commonality among the identified risk factors appears to be their association with reduced salivary flow or alterations in fluid dynamics and composition, which are known contributors to the formation of salivary stones.

Level of Evidence:

3

Introduction

Sialolithiasis is a multifactorial condition arising from the interplay of altered salivary composition, ductal epithelial injury, immune and inflammatory activity, and microbial biofilm formation. Alterations in salivary composition, characterized by increased calcium and phosphate levels and reduced crystallization inhibitors, promote mineral precipitation and hydroxyapatite deposition on an organic nidus formed by desquamated epithelial cells, mucins, bacteria, or inflammatory debris.^1-3 The retrograde theory further proposes that oral microorganisms or food particles may migrate into the salivary ducts and serve as a nidus for calcification, thereby contributing to stone initiation.⁴ Biofilm formation within the salivary ducts appears to be a critical initiating event, inducing local epithelial injury and immune activation.^5-7 It triggers neutrophil recruitment and the formation of extracellular traps that aggregate calcium crystals and proteins.⁸ Repeated cycles of inflammation, mineral deposition, and ductal obstruction lead to the concentric, onion-skin architecture of mature sialoliths, linking biochemical imbalance with innate immune responses in salivary stone formation.⁹

Imaging plays a central role in diagnosis. Ultrasound is the preferred first-line modality due to its high sensitivity for stones >2mm, cost-effectiveness, and absence of radiation exposure, though its sensitivity decreases for small or anterior submandibular stones.^10-14 Non-contrast computed tomography is the most sensitive modality for detecting radiopaque sialoliths and is considered the gold standard for identifying calcified stones.¹⁵ Traditional sialography, performed via ductal cannulation and iodinated contrast injection, provides detailed visualization of ductal anatomy and strictures but can be associated with patient discomfort and requires specialized skill and expertise to perform, limiting its use.¹⁰ Magnetic resonance (MR) sialography offers a noninvasive, radiation-free alternative with high sensitivity for detecting non-calcified stones, ductal stenosis, and chronic inflammatory changes, and can be safely performed during acute infections.¹⁵ Compared to conventional sialography, MR sialography demonstrates similar diagnostic accuracy for ductal pathology while avoiding procedural risks, making it particularly valuable when ductal cannulation is difficult or when evaluating chronic sialadenitis and sicca syndromes.¹⁵

Management of sialolithiasis has evolved from gland excision to a range of minimally invasive and gland-preserving techniques.¹⁶ Extracorporeal shock wave lithotripsy facilitates stone fragmentation for natural expulsion but is less effective for large calculi.^17,18 Interventional sialendoscopy now represents the cornerstone of management, allowing stone retrieval, duct dilation, and laser or mechanical fragmentation, with success rates of ~80% for stones ≤5mm. For larger or impacted stones, combined endoscopic-transoral or -transfacial/-transcervical approaches are used while preserving gland function. Gland excision remains reserved for deeply intraparenchymal, multiple, or recurrent stones refractory to conservative methods.¹⁸ It should be noted that while laser lithotripsy is used in certain centers, its use in salivary gland surgery remains off-label in the U.S.

Despite advances in treatment, the pathogenesis and risk factors of sialolithiasis remain incompletely understood. Several theories have emerged based on analyses of extracted stones.^1,19 While reduced salivary flow and ductal stasis are recognized as key factors,⁴ their mechanistic links to systemic and lifestyle influences and medications that reduce salivary output are not fully elucidated, with studies reporting variable associations with these conditions.^20-23 Emerging computational approaches, including machine learning models, offer new opportunities to analyze large datasets and identify potential predictive factors for sialolithiasis. By leveraging advanced computational models, this study aims to investigate risk factors associated with sialolithiasis and to provide new insights into its etiology, to facilitate prevention, early diagnosis, and personalized treatment strategies.

Materials and Methods

In this retrospective, case-control study, we selected 540 individuals diagnosed with sialolithiasis from the 409,420 participants in the All of Us Research Program between 1980 and 2022. This group was then matched to a control group at a 3:1 ratio based on age, race, and gender. In total, 2160 participants were included, of whom 540 had a diagnosis of sialolithiasis. To examine the association between sialolithiasis and a set of comorbidities, we collected data on each participant’s history of the following diagnoses: acute tonsilitis, alcohol abuse, essential hypertension, type 2 diabetes mellitus without complication, dehydration, kidney stones, cholelithiasis without obstruction, gout, bipolar disorder, hypercalcemia, gallstone, Sjogren’s syndrome, systemic lupus erythematous, hyperparathyroidism, obesity, osteoporosis, and smoking. A participant was classified as having a specific condition if a prior diagnosis was documented in their electronic health record in the “condition” category. The time of diagnosis for each condition was extracted and encoded into one-hot vectors, where 1 indicated the presence of the condition and 0 indicated its absence. Smoking status was defined as a survey response indicating the participant had smoked at least 100 cigarettes in their lifetime and currently smoked “every day” or “some days.” This variable was also encoded into a one-hot vector. The outcome variable, a diagnosis of sialolithiasis, was similarly converted into a one-hot vector.

A total of 16 binary variables were used as inputs for the logistic regression model using Python’s Scikit-learn (sklearn) library. The dataset was split into training (80%) and testing (20%) subsets. An L1-regularized logistic regression model was trained with a learning rate (alpha) of 0.001 to generate predictions. Model performance was assessed using accuracy, sensitivity, specificity, F1 score, and the area under the receiver operating characteristic curve (ROC-AUC). Feature importance was assessed based on odds ratios and p-values obtained from the logistic regression model, with variables considered significant if p<0.05. Additionally, a standalone SHAP (Shapley Additive exPlanations) explainer analysis was performed using an XGBoost model to gain further insights into the features contributing to a diagnosis of sialolithiasis. SHAP is a model-agnostic framework that quantifies the impact of each variable on the model’s predictions. The computed SHAP values allow for individual-level interpretation of how each variable influences the probability of a sialolithiasis diagnosis.

Results

Among the 2160 participants included in the analysis, the mean age was 67.7 ±13.4 years, and 540 (25%) had a diagnosis of sialolithiasis. The logistic regression model identified several variables as significant predictors of sialolithiasis (Figure 1). Sjögren’s Syndrome (OR = 2.057, 95% CI: 1.106-3.824, p = 0.023) was associated with over double the odds of sialolithiasis. Obesity was also a significant contributor (OR=1.419, 95% CI: 1.118-1.802, p=0.004). Interestingly, essential hypertension was negatively associated with sialolithiasis (OR=0.279, 95% CI: 0.23-0.339, p=0.0), suggesting a potential protective effect. Having type 2 diabetes (OR=0.771, 95% CI: 0.601-0.989, p=0.041) was also modestly negatively associated with sialolithiasis. Other examined comorbidities did not reach statistical significance. All other variables did not show significant associations with outcome (Table 1).

Figure 1.

Logistic regression model showing statistically significant comorbid predictors of sialolithiasis.

Table 1.

Adjusted Odds Ratios for the Association Between Comorbid Conditions and Sialolithiasis

Condition	Odds Ratio (OR)	95% CI	p-value
Sjögren's syndrome	2.057	1.106–3.824	0.023
Obesity	1.419	1.118–1.802	0.004
Alcohol abuse	1.337	0.897–1.991	0.154
Bipolar disorder	1.114	0.719–1.726	0.63
Gout	1.067	0.702–1.62	0.762
Hypercalcemia	1.019	0.617–1.685	0.94
Dehydration	0.967	0.701–1.335	0.84
Osteoporosis	0.964	0.709–1.312	0.817
Systemic lupus erythematosus	0.97	0.487–1.727	0.788
Acute tonsillitis	0.881	0.426–1.822	0.732
Gallstones	0.863	0.594–1.253	0.438
Kidney stone	0.81	0.574–1.142	0.228
Hyperparathyroidism	0.776	0.394–1.529	0.463
Type 2 diabetes mellitus without complication	0.771	0.601–0.989	0.041
Smoking	0.762	0.552–1.052	0.098
Essential hypertension	0.279	0.23–0.339	0.0

The logistic regression model achieved an ROC-AUC value of 0.568, with a prediction accuracy of 69.9%. The sensitivity was 8% and the specificity was 74% (Figure 2). A SHAP analysis identified obesity (SHAP=0.282), essential hypertension (SHAP=0.237), type 2 diabetes without complication (SHAP=0.228), and dehydration (SHAP=0.127) as the most influential features (Figure 3). Additional contributors included osteoporosis (SHAP=0.099), kidney stone (SHAP=0.095), smoking (SHAP=0.090), and alcohol abuse (SHAP=0.084). These SHAP values represent the average magnitude by which each variable impacted the model’s prediction of sialolithiasis, providing a complementary perspective to the logistic regression findings.

Figure 2.

Receiver operating characteristic (ROC) curve for the multivariate logistic regression model predicting sialolithiasis (AUC=0.57), indicating modest discriminative performance.

Figure 3.

Machine learning model (XGBoost) showing the relative contribution of each variable to the prediction of sialolithiasis.

Discussion

The present study identified several comorbid conditions significantly associated with sialolithiasis, suggesting a multifactorial etiology. Sjogren’s syndrome and obesity emerged as statistically significant predictors, while essential hypertension and type 2 diabetes were inversely associated. These findings align partially with prior literature and provide new insights into the systemic factors potentially contributing to salivary stone formation.

Sjögren’s syndrome has historically been considered a risk factor for sialolithiasis due to its association with reduced salivary flow, and our findings show a strong association (OR=2.057, p=0.023). Chronic lymphocytic ductal infiltration in Sjögren’s syndrome is associated with salivary gland stenosis and chronic obstructive sialadenitis, which have strong associations with sialolithiasis.^24,25 In a retrospective study of 333 patients, it was found that Sjögren’s syndrome is significantly associated with duct stricture, which was the most common cause of chronic obstructive sialadenitis.²⁴ Further, several case report studies have documented sialolithiasis occurring in multiplicity in patients with Sjögren’s syndrome, an uncommon finding which may suggest that microlith formation occurs with greater frequency in patients with Sjögren’s syndrome.²⁶ While sialolithiasis and Sjögren’s syndrome share a pathophysiologic mechanism, some prior studies have failed to find an association.²⁷ There is currently no body of epidemiological evidence linking the two conditions.²⁸ This underscores the need for more controlled studies to further explore the relationship between autoimmune salivary disorders and sialolithiasis.²⁸

Our data showed obesity as a potential risk factor (OR=1.419, p=0.004), which is further supported by a recent meta-analysis showing that obese individuals have significantly reduced stimulated salivary flow rate compared to those with normal weight, and worsening with increasing obesity severity.²⁹ This reduction in salivary flow promotes salivary stagnation and mineral precipitation within the ducts. Regarding alcohol, previous research^27,22 and our findings (OR=1.337, p=0.154) did not demonstrate a significant association with sialolithiasis, although our SHAP analysis ranked alcohol abuse as a moderately influential variable, which warrants further exploration using more complex models. Kraaij et al. found no increased prevalence of diabetes or hypertension in patients with sialolithiasis.²⁷ Our results showed a mild inverse relationship at the lower limit of significance level between type 2 diabetes mellitus without complications (OR= 0.771, 95% CI: 0.601-0.989, p=0.041) and sialolithiasis. Interestingly, we identified a strong inverse association between essential hypertension and sialolithiasis (OR=0.279, p=0.0), which may be related to certain classes of antihypertensives, like angiotensin-converting enzyme inhibitors that modestly increase salivary flow.³⁰ In a controlled study, captopril was associated with increased salivary flow rates without significant changes in salivary composition, potentially reducing stone risk.³⁰

Other systemic factors that reduce salivary secretion, including diuretic use and smoking, have also been explored. Nederfors et al. found that diuretics modestly affect saliva production, possibly due to variable sensitivity in salivary gland ion transporters, though the exact mechanisms remain unclear.³¹ Huoh et al. observed a higher prevalence of diuretic use in sialolithiasis patients, supporting the hypothesis that reduced salivary flow contributes to stone formation.³² They also suggested a potential association between tobacco smoking and sialolithiasis,³² which is possible, given the known effects of tobacco smoke on mucosal inflammation and antimicrobial function.³³ A retrospective study found a significant association between smoking and obstructive salivary disorders,³⁴ while other research showed that current smokers had larger stones than former smokers³⁵ and that smoking was more prevalent among individuals with sialolithiasis (32.4%) compared to the control group (29.1%), with a significantly higher adjusted odds ratio (OR=1.31, 95% CI = 1.08-1.59).²² However, in our study, smoking (OR=0.762, p=0.098) or dehydration (OR=0.967, p=0.840) did not emerge as significant risk factors for sialolithiasis, possibly reflecting differences in population demographics or data limitations regarding diuretic use.

Primary hyperparathyroidism (pHPT) is associated with nephrolithiasis, but its link to sialolithiasis is less clear. One study found a higher prevalence of sialolithiasis (0.6%) in pHPT patients compared to the general population rate of 0.45% (p=0.0028),³⁶ suggesting that 5.4% of symptomatic sialolithiasis patients may have pHPT, sometimes presenting as the first sign. Early detection and treatment of pHPT may prevent recurrent sialolithiasis.³⁶ Osteoporosis has also been linked to sialolithiasis, indicating a possible connection to systemic calcium metabolism disorders.³⁷ However, we found no significant association between sialolithiasis and hyperparathyroidism (OR=0.776, p=0.460), gout (OR=1.067, p=0.762), osteoporosis (OR=0.964, p=0.817), or hypercalcemia (OR=1.019, p=0.94). Gout may still increase the risk of uric-acid-based salivary stones.³⁸ We did not find a significant association between nephrolithiasis and sialolithiasis (OR=0.81, p=0.228). Some researchers suggested both conditions share pathophysiology,^21,39 but results have been mixed,³² often lacking control groups. A population-based study found that patients with sialolithiasis were nearly five times more likely to have a history of nephrolithiasis (adjusted OR=4.74; 95% CI=3.41-6.58) compared to controls after adjusting for various factors.²¹ This may prompt physicians to advise patients with nephrolithiasis on preventive measures, such as smoking cessation. Nevertheless, other studies found no significant correlation, suggesting distinct mechanisms.^23,27

Conclusion

A growing body of research highlights the complex interplay between local factors, such as salivary composition, and systemic conditions, including metabolic conditions and lifestyle habits, in the development of sialolithiasis. Since some associations remain unclear, future prospective studies are needed to clarify these relationships and elucidate the underlying mechanisms, ultimately improving prevention and management strategies.