Risk factors for positive minor salivary gland biopsy findings in Sjögren's syndrome and dry mouth patients

Authors


Abstract

Objective

To investigate risk factors for positive minor salivary gland biopsy results in Sjögren's syndrome (SS) and dry mouth patients.

Methods

A total of 289 patients with dry mouth symptoms were evaluated. Potential risk factors for positive minor salivary gland biopsy results (>1 focus of lymphocytes) were studied in 2 phases. In phase 1, predictor variable candidates were identified for the test study (phase 2). Odds ratios were calculated for predictor variables.

Results

IgG, IgA, keratoconjunctivitis sicca, and sex, identified as the best predictor variables from phase 1 data, were included in a logistic regression model using phase 2 data. Only IgG demonstrated association with biopsy results (χ2 = 20.4, P = 0.0001). An elevated IgG level (>1,482 mg/dl) had a high specificity (97% and 97%), high positive predictive value (PPV) (97% and 97%), but poor sensitivity (40% and 45%) in predicting positive biopsy results and SS, respectively.

Conclusion

Elevated serum IgG levels best predicted a positive biopsy result and SS with high PPV and specificities.

INTRODUCTION

Sjögren's syndrome (SS) is an autoimmune exocrinopathy mainly affecting salivary and lacrimal glands. SS may develop in the absence (primary SS) or the presence (secondary SS) of other connective tissue diseases such as rheumatoid arthritis and systemic lupus erythematosus (1). In patients with SS, evidence suggests that inflammatory-mediated events diminish salivation and lacrimation, leading to symptoms of dry mouth and eyes.

All classification criteria require evidence of salivary dysfunction and keratoconjunctivitis sicca (KCS) for primary SS (2). KCS is best established with an ophthalmologic examination to demonstrate signs of lacrimal dysfunction and/or corneal and conjunctival epithelial damage. Currently, the presence of serum autoantibodies is part of the inclusion criteria in the European and San Diego criteria (3). Antibodies commonly found in primary SS include elevated antinuclear antibodies (ANA), anti-SSA/Ro, anti-SSB/La, and rheumatoid factors (RF). Patients with primary SS also frequently have an increased erythrocyte sedimentation rate (ESR) and polyclonal and monoclonal hypergammaglobulinemia (4, 5). Specifically, IgG is more commonly elevated in primary SS compared with IgA and IgM, and has been found to correlate with focus scores of minor salivary glands and van Bijsterveld scores, a measure of KCS severity (6–8).

The minor salivary gland biopsy has been established as a valuable test for the evaluation of the salivary component in SS (3). Characteristic histopathologic involvement of the exocrine tissues includes an intense, focal, mononuclear cell infiltration in a periductal or perivascular distribution, loss of acinar structures, and relative preservation of ductal anatomy (2). A disadvantage of the minor salivary gland biopsy is that some patients may experience postoperative numbness or pain. A patient's refusal of the procedure may result in an inconclusive diagnosis. In addition, the invasive nature of the minor salivary gland biopsy limits its usefulness to follow patients longitudinally, restricting our ability to document disease progression in populations.

Risk factors for positive minor salivary gland biopsy results in SS patients are not clearly delineated. If the presence of one or a combination of several risk factors could predict the presence of lymphocytic foci in salivary glands, a minor gland biopsy may not be needed to diagnose SS. Thus, it is important to seek candidate risk factors for a positive minor salivary gland biopsy result. The objective in the current study was to answer the question: Do patients with dry mouth symptoms present with risk factors for positive minor salivary gland biopsy results? We hypothesized that objective measures of ocular and oral dryness and the presence of autoantibodies are risk factors for a positive minor salivary gland biopsy result in dry mouth patients.

PATIENTS AND METHODS

Study population

Participants were patients with reported dry mouth evaluated at the National Institutes of Health (NIH) Salivary Gland Dysfunction Clinic between 1984 and 2000. Patients with a history of head and neck radiation and patients without dry mouth symptoms were excluded. We evaluated 289 patients by performing a minor salivary gland biopsy.

Subjective symptoms of dry mouth were documented when patients presented with a chief complaint of oral dryness, or if the patient answered “yes” to any of the following questions: 1) “Does your mouth usually feel dry when eating a meal?” 2) “Do you have difficulties swallowing foods if you eat without additional fluids?” and 3) “Do you usually drink fluids with meals to ease swallowing dry foods?” Patients answering “too little” to the question “Does the amount of saliva in your mouth most of the time seem to be: too little, too much, or don't notice it” also were included in the group with dry mouth complaints (9).

Study design

This retrospective cohort study was completed in 2 phases. Between 1984 and September 1997, many patients with subjective dry mouth had only clinical and laboratory examinations. A minor salivary gland biopsy was done only when objective signs of eye dryness and/or autoimmune serology were present. From October 1997 to May 2000, all patients presenting with a dry mouth complaint had a more complete diagnostic evaluation, including a minor salivary gland biopsy. The potential selection bias was greater for the first of these 2 time periods. The data from 178 patients between 1984 and September 1997 were therefore first analyzed separately and used only to identify predictor variable candidates for the test study (1997–2000). One hundred eleven patients from 1997 to 2000 served as the cohort for the test study.

Diagnostic tests and clinical information

Patients routinely underwent a structured interview; a complete head, neck, and oral examination; laboratory studies; eye examination; and labial minor salivary gland biopsy. They also completed a standardized questionnaire. The following information was collected:

  • 1Demographic information (age, race, and sex).
  • 2Degree of lymphocytic infiltration (focus score) in the minor salivary gland biopsy. A minimum of 5 lobules were evaluated following the technique proposed by Daniels (10). Lobules with mucus plug inspissation were excluded. A focus score ≤1 was considered a negative minor salivary gland biopsy and a focus score >1 was considered positive.
  • 3Duration of dry mouth symptoms.
  • 4Objective eye findings: presence of KCS by staining with ocular dyes (van Bijsterveld score ≥4) and/or Schirmer's I test without anesthesia ≤5 mm/5 min (10).
  • 5History of and/or presence of salivary gland swelling.
  • 6Serum IgG, IgA, IgM (mg/dl).
  • 7Presence or absence of anti-SSA and anti-SSB determined by Ouchterlony (1984–1991), hemagglutination (1991–1999), and enzyme-linked immunosorbent assay kit (Hemagen Diagnostics, Columbia, MD) (1999–2000).
  • 8ANA titer by slide test using mouse liver touch preparations (1984), Hep2 (1985–1993), and Hep2 enhanced (1994–2000) (Immunoconcepts, Mt. Laurel, NJ).
  • 9RF (IU/ml) by latex agglutination (1984–1991) and nephelometry (1991–2000); ESR by Westergren (1984–1991) and Vesmatic (1992–2000); and white blood cell count (WBC).
  • 10Salivary flow rates (unstimulated and 2% citric acid stimulated parotid and submandibular flow), which were collected individually from the major salivary glands using established techniques (11).

All diagnostic tests were completed within 1 year of the minor salivary gland biopsy. Data collected from clinical charts were verified after entry in a spreadsheet, and 20 charts were randomly selected to ensure accuracy.

A diagnosis of primary SS and secondary SS utilized the European Classification Criteria for SS (12). Because data were incomplete for the specific questions of subjective ocular symptoms, only objective diagnostic tests were utilized for a diagnosis. A modified unstimulated whole saliva was determined by the sum of unstimulated salivary flow from the right and left parotids and the submandibular/sublingual gland collection.

Statistical analysis

The associations and the magnitude of associations between risk factors and a positive minor salivary gland biopsy result were studied in 2 phases. Potential predictor variables were identified from the 1984 to September 1997 data. Clinical and diagnostic test results for negative and positive minor salivary gland biopsy groups were compared using Fisher's exact test or Wilcoxon's rank sum test, as appropriate. Because of the number of multiple comparisons, statistical significance was considered at an alpha level of <0.002. Clinical and laboratory variables thought likely to have prognostic value from former studies and variables identified with univariate analysis were evaluated with a multivariable logistic regression analysis. The dichotomous response variable for the present study was a negative (≤1 focus of lymphocytes) or positive (>1 focus of lymphocytes) minor salivary gland biopsy result. The potential set of predictor variables that remained in the regression model at P < 0.05 with backward selection were used for a logistic regression analysis of the dry mouth cohort from 1997 to 2000. Odds ratios were calculated to represent the relative risk of the predictor variables. The SAS statistical program (SAS Institute, Cary, NC) was used for analyses.

RESULTS

Phase 1: Data analysis of patients seen between 1984 and September 1997

Between 1984 and 1997, 178 patients with dry mouth symptoms were evaluated at the NIH Salivary Gland Dysfunction Clinic. Demographic, clinical, and diagnostic test results for all patients combined, as well as patients dichotomized by minor salivary gland biopsy results (negative or positive), are given in Table 1. Of the 178 patients, 123 had a positive minor salivary gland biopsy result, and the minor gland biopsy results of 55 were negative. Numerous diagnostic tests were different between the negative and positive minor salivary gland biopsy groups. Those with positive minor gland biopsy results had a greater number of objective tests of eye dryness (KCS and Schirmer's I test), decreased salivary flows, elevated immune parameters (IgG, IgA, anti-SSA, anti-SSB, ANA, RF, and ESR) and lower WBCs (Table 1).

Table 1. Clinical and diagnostic test results of dry mouth patients with a negative and positive minor salivary gland biopsy result, 1984 to September 1997*
All patientsNegative biopsy resultPositive biopsy resultP value
  • *

    (Q1–Q3) = 25th to 75th percentile; KCS = keratoconjunctivitis sicca; ANA = antinuclear antibodies; RF = rheumatoid factor; WBC = white blood cell count; ESR = erythrocyte sedimentation rate.

  • All values reported as <20 IU/ml were given the value of 0.

Total number17855123
Median age (Q1–Q3)54 (42–64)52 (43–62)55 (42–64)0.80
Sex
 Male1596
 Female163461170.017
Duration of dry mouth complaint; median years (Q1–Q3)4 (1.5–10)3 (1–6.5)4 (2–10)0.18
Positive KCS, 136 total (%)113 (83)17 (57)96 (91)<0.001
Schirmer's I test; median mm/5 min (Q1–Q3)4.3 (1–10)7.25 (3.5–15)3.75 (1–8.5)0.0015
Unstimulated parotid saliva; median ml/min/gland (Q1–Q3)0 (0–0.02)0 (0–0.046)0 (0–0)0.026
Stimulated parotid saliva; median ml/min/gland (Q1–Q3)0.19 (0.07–0.40)0.31 (0.16–0.43)0.13 (0.03–0.37)<0.001
Unstimulated submandibular saliva; median ml/min/gland (Q1–Q3)0 (0–0.03)0.02 (0–0.06)0 (0–0.01)<0.001
Stimulated submandibular saliva; median ml/min/gland (Q1–Q3)0.08 (0–0.27)0.22 (0.12–0.39)0.02 (0–0.16)<0.001
History of gland swelling, 174 total (%)76 (44)17 (31)59 (59)0.022
Presents with gland swelling, 176 total (%)46 (26)12 (22)34 (28)0.460
IgG; median mg/dl (Q1–Q3)1,500 (1,130–2,050)1,090 (901–1,330)1,800 (1,280–2,400)<0.001
IgA; median mg/dl (Q1–Q3)264 (187–375)186 (151–251)318 (224–411)<0.001
IgM; median mg/dl (Q1–Q3)146 (99–211)138 (94–158)165 (100–231)0.06
Positive anti-SSA, 173 total (%)73 (42)5 (9)68 (57)<0.001
Positive anti-SSB, 173 total (%)40 (23)2 (4)38 (47)<0.001
ANA; median titer (Q1–Q3)320 (0–640)80 (0–320)640 (160–640)<0.001
Positive ANA, 177 total (%)127 (72)31 (60)96 (79)0.015
Rheumatoid factor; median IU/ml (Q1–Q3)0 (0–97)0 (0–0)34 (0–174)<0.001
Positive RF, 172 total (%)80 (47)8 (15)72 (61)<0.001
WBC; median × 1,000/μl (Q1–Q3)5.5 (4.4–6.9)6.7 (5.4–7.9)5.2 (4.2–6.2)<0.001
ESR; median mm/hour (Q1–Q3)30 (17–53)36.5 (20–59)18 (9–29)<0.001

Univariate analysis by this dichotomous response variable and findings from past studies identified 15 candidate covariates for a logistic regression model. IgG, IgA, KCS, and sex were the best explanatory variables in the above model (χ2 = 44.6, P = 0.0001; Table 2).

Table 2. Variables in the phase 1 logistic regression model*
VariableBeta valueStandard errorOdds ratioP
  • *

    KCS = keratoconjunctivitis sicca; IgG in mg/dl × 103; IgA in mg/dl × 102.

Intercept−5.471.38
Sex1.920.826.800.019
KCS1.410.584.100.015
IgG0.1630.071.180.021
IgA0.0680.031.070.036

Phase 2: Data analysis of patients seen between October 1997 and 2000

A cohort of 111 dry mouth patients was evaluable in this group. Minor salivary gland biopsy results were negative for 33 patients and positive for 78 patients (Table 3). Only the predictor variables identified in the first phase of the study (IgG, IgA, KCS, and sex) were included in a logistic regression model with the data from 1997 to 2000 (Table 4). Of the 4 variables placed in the model, only IgG demonstrated an association with biopsy results (χ2 = 20.4, P = 0.0001).

Table 3. October 1997–2000 clinical and diagnostic test results of dry mouth patients with negative and positive minor salivary gland biopsy results*
All patientsNegative biopsy resultsPositive biopsy results
  • *

    (Q1–Q3) = 25th to 75th percentile; KCS = keratoconjunctivitis sicca; Unstim. = unstimulated; ANA = antinuclear antibodies; RF = rheumatoid factor; WBC = white blood cell count; ESR = erythrocyte sedimentation rate.

Total number (%)11133 (30)78 (70)
Median age (Q1–Q3)54 (44–62)49 (43–57)54 (46–62)
Sex
 Male18 (16)8 (24)10 (13)
 Female93 (84)25 (76)68 (87)
Duration of dry mouth complaint; median years (Q1–Q3)3.3 (1.4–6.5)3 (1–6)4 (2–7)
Positive KCS, 108 total (%)81 (73)21 (64)60 (77)
Schirmer's I test; median mm/5 min (Q1–Q3)5 (2–11)7 (3.5–13)4.5 (1.5–9.5)
Unstim. parotid saliva; median ml/min/gland (Q1–Q3)0 (0–0)0 (0–0)0 (0–0)
Stimulated parotid saliva; median ml/min/gland (Q1–Q3)0.23 (0–0.44)0.24 (0.04–0.37)0.18 (0–0.44)
Unstim. submandibular saliva; median ml/min/gland (Q1–Q3)0 (0–0.03)0.03 (0–0.07)0 (0–0.0005)
Stimulated submandibular saliva; median ml/min/gland (Q1–Q3)0.13 (0.005–0.21)0.18 (0.12–0.21)0.08 (0–0.20)
History of gland swelling (%)52 (47)18 (55)34 (44)
Presents with gland swelling (%)34 (31)10 (30)24 (31)
IgG; median mg/dl (Q1–Q3)1,130 (938–1,610)1,030 (805–1,140)1,250 (1,010–1,780)
IgA; median mg/dl (Q1–Q3)250 (170–349)224 (162–274)284 (182–378)
IgM; median mg/dl (Q1–Q3)123 (92–174)120 (87–145)133 (98–191)
Positive anti-SSA, 111 total (%)34 (31)0 (0)34 (44)
Positive anti-SSB, 111 total (%)18 (16)0 (0)18 (23)
ANA median titer (Q1–Q3)80 (0–640)0 (0–0)320 (0–1,280)
Positive ANA (%)59 (53)7 (21)52 (67)
RF; median IU/ml (Q1–Q3)0 (0–39)0 (0–0)0 (0–92)
WBC; median × 1000/μL (Q1–Q3)5.8 (4.5–7.5)6.05 (5.5–7.5)5.4 (4.4–7.1)
ESR; median mm/hour (Q1–Q3)26 (16–50)18 (11–24)36 (21–54)
Table 4. Variables in the phase 2 logistic regression model
VariableBeta valueStandard errorOdds ratio (95% CI)P value
  1. CI = confidence interval; IgG in mg/dl × 103.

Intercept−2.180.83
IgG0.25210.07311.287 (1.13–1.51)0.0006

To assess whether the assumption that a 1-year window was appropriate, we controlled for length of time between minor salivary gland biopsy and the laboratory studies. We found that 71% of patients in the test study had laboratory evaluations completed within 2 weeks of the minor salivary gland biopsy. Adding length of time to the logistic regression analysis had essentially no effect on the model (not shown).

To further analyze the predictive value of an elevated serum IgG for a positive minor salivary gland biopsy result and for a diagnosis of SS, we considered the data in light of the upper limit of normal for IgG at our hospital. A serum IgG level ≤1,482 mg/dl will account for 97.5% of the normal population at the NIH Clinical Center. It is of interest to know whether high serum IgG levels determined by a routine laboratory test are associated with a positive minor salivary gland biopsy result and a diagnosis of SS. The results of this analysis are summarized in Table 5 and Table 6, respectively. The use of an elevated IgG level had poor sensitivity, high specificity, and a high positive predictive value (PPV) in predicting a positive minor salivary gland biopsy result and a SS diagnosis. An elevated IgG level had a low negative predictive value for a positive minor salivary gland biopsy result and a SS diagnosis. In addition, 45% of SS patients had an elevated IgG level (>1,482 mg/dl), whereas 55% had an IgG level in the normal range.

Table 5. Diagnostic performance of an elevated IgG (>1,482 mg/dl) for a positive minor salivary gland biopsy result*
Positive biopsy resultNegative biopsy result
  • *

    PPV = positive predictive value; NPV = negative predictive value.

Elevated IgG311PPV (31/32)
97%
Not elevated IgG4732NPV (32/79)
41%
SensitivitySpecificity
31/78 = 40%32/33 = 97%
Table 6. Diagnostic performance of an elevated IgG (>1,482 mg/dl) for a diagnosis of SS*
SS diagnosisNo SS
  • *

    SS = Sjögren's syndrome; PPV = positive predictive value; NPV = negative predictive value.

Elevated IgG311PPV (31/32)
97%
Not elevated IgG3837NPV (37/75)
49%
SensitivitySpecificity
31/69 = 45%37/38 = 97%

DISCUSSION

The minor salivary gland biopsy is an important procedure for the diagnosis of Sjögren's syndrome (3, 13), but the invasiveness of the procedure can discourage its use as a diagnostic test or as a measure with which to follow disease progression. The identification of strong, less invasive predictors for a positive biopsy result would provide an alternative and important measure of salivary gland disease activity. The findings from the present study identified numerous differences in diagnostic tests for dry mouth patients. However, the logistic model employed in phase 2 data utilizing predictor variables identified from earlier data demonstrated limited value to predict biopsy results. Serum IgG was identified as the best predictor of a positive or negative minor salivary gland biopsy result.

The strongest utility of an elevated serum IgG in the test study (phase 2) was in specificity and PPV. In the present study, the strong PPV demonstrates that 97% of dry mouth patients with an IgG level above the normal range (>1,482 mg/dl) had a positive minor salivary gland biopsy result. In addition, an elevated IgG level was present in patients with dry mouth who had been diagnosed with primary or secondary SS in 97% of cases. On the other hand, a low sensitivity indicates the use of elevated IgG will miss a number of cases of positive minor salivary gland biopsy results and SS. In the present population, patients with a positive minor salivary gland biopsy result and/or a diagnosis of primary or secondary SS had a wide range of IgG levels. This variation in IgG levels likely accounts for the low sensitivity of elevated IgG levels in identifying positive biopsy results and cases of SS. In addition, the 45% of SS patients with elevated IgG may represent a group with higher disease activity.

Other studies have shown a relationship between IgG and minor salivary gland lymphocytic infiltrate (8, 10, 11, 13, 14). One important limitation in many of these studies is selection bias. Patients frequently received a minor salivary gland biopsy with evidence of other positive diagnostic tests consistent with SS. This was also the case in phase 1 of the present study. Selection bias can lead to an overestimation of the relationship between the minor salivary gland biopsy and other objective diagnostic tests. In the present study, after September 1997, all patients presenting with a complaint of dry mouth had a full SS diagnostic evaluation as described herein. Analysis of this cohort should theoretically approximate these relationships more accurately.

Another important limitation of previous studies has been a variation in the time interval between the minor salivary gland biopsy and other diagnostic tests, which was not addressed thoroughly in past studies. To assess these relationships, the amount of time between the minor salivary gland biopsy and diagnostic tests was limited to 1 year in the present study. The 1-year period was acceptable, as demonstrated by further data analysis. Controlling for length of time between the biopsy and laboratory tests had minimal effect on the logistic regression model when time was less than 1 year.

Due to the length of data collection for the present study (i.e., 17 years) different laboratory techniques were used over time. The use of new laboratory techniques can lead to bias with inconsistency in findings from one time period to another. Additionally, due to the length of time of data collection, two pathologists read biopsy slides for the duration of the study; one pathologist from 1984 to 1995 and the other from 1996 to 2000. Each biopsy was scored by one of the pathologists and was not reread. We have limited data for inter- and intraobserver reliability due to the retrospective nature of the present study.

The only exclusion factors utilized in the present study included no report of dry mouth and a history of radiotherapy to the head and neck. Medications were not an exclusion in the present study. However, data on specific medications were not abstracted from the chart. It is very possible that some dry mouth symptoms were side effects of medications.

We measured salivary flow from the individual major salivary glands. A previous study suggests that flow rates from the submandibular/sublingual glands could better predict a diagnosis of primary SS (15). Thus, we have utilized the salivary flow from the individual major salivary glands as explanatory variables in the present study. In the present study, no measure of salivary flow was demonstrated to predict biopsy results with the modeling used. The unstimulated parotid flow median values for the negative biopsy group in Tables 1 and 3 were not detectable. Considering 22% of a group of healthy individuals with no systemic medications also had no measurable unstimulated parotid flow (16), the flow rate of the negative biopsy result group may fall within the range of normal. Similarly, the median values for stimulated parotid, unstimulated submandibular/sublingual, and stimulated submandibular/sublingual flow may also fall within the range of normal. Although all patients reported dry mouth, subjective symptoms are not always related to objective measures of salivary flow (17, 18).

A main limitation of the current logistic model is the low odds ratio (1.3). In the test study, an increase of 1,000 mg/dl in IgG was associated with only a 30% increase in the odds of having a positive minor salivary gland biopsy result. However, the predictive value of IgG for a positive minor salivary gland biopsy result was much better when examined as a dichotomous variable (i.e., elevated IgG versus not elevated IgG). The limit of IgG as a continuous variable to predict a positive minor salivary gland biopsy result likely reflects the inability of a low IgG to consistently predict a negative minor salivary gland biopsy.

The limited value of modeling prediction variables in the present study may also be related to the nature of the variables chosen. An elevated IgG was associated with biopsy results as well as numerous diagnostic tests. A high level of colinearity existed between IgG and the following: IgA (r = 0.34, P < 0.0002), anti-SSA (r = 0.56, P < 0.0001), anti-SSB (r = 0.52, P < 0.0001), ANA (r = 0.49, P < 0.0001), ESR (r = 0.55, P < 0.0001), and RF (r = 0.45, P < 0.0001). Independent variables with a high degree of colinearity have limited usefulness as predictors in a logistic regression model (19).

An increased prevalence of patients with a positive lip biopsy result or primary SS in the study population will increase the estimate of PPV, thus affecting the generalizability of the current results. Because of referral bias, the generalizability of the predictive utility of an elevated IgG in other dry mouth populations is only speculative.

In summary, although elevated IgG and dry mouth complaints support the presence of minor salivary gland biopsy results with focal lymphocytic infiltrates, a normal serum IgG does not exclude it.

Acknowledgements

We thank Joanne Brennan and Sunday Akintoye for assistance in data entry.

Ancillary