Abstract
Purpose:
To develop a screening questionnaire to identify dry eye patients with a high likelihood of having underlying Sjögren’s Syndrome (SS).
Methods:
Cross-sectional study of participants with dry eye complaints who were self-referred or referred by an ophthalmologist to the Sjögren’s International Collaborative Clinical Alliance (SICCA) study. Symptoms and ocular surface exam findings were candidate predictors. Univariable and multivariable logistic regression analyses were performed to estimate odds ratios (OR) and 95% confidence intervals (95% CI) for the association of a symptom and/or ocular sign with SS. Area-under-the-receiver-operating-characteristic curve (AUC) was used to summarize the predictive ability of different regression models and the derived likelihood score.
Results:
Four questions were statistically significant in the final multivariable model including: 1) Is your mouth dry when eating a meal? [yes = OR 1.63 (1.18–2.26)]; 2) Can you eat a cracker without drinking a fluid or liquid? [no = OR 1.46 (1.06–2.01)]; 3) How often do you have excessive tearing? [none of the time = OR 4.06 (1.81–9.10); and 4) Are you able to produce tears? [no = OR 2.24 (1.62–3.09)]. The SS likelihood score had an AUC of 0.70 (95% CI 0.66–0.73), while including TBUT and conjunctival staining yielded an AUC of 0.79 (95% CI: 0.77–0.82).
Conclusions:
This questionnaire can be used to identify dry eye patients with a high likelihood of having SS. With future refinement and validation, this screening tool could be used alone, or in combination with exam findings to identify SS patients earlier, thereby facilitating better clinical outcomes.
Keywords: Sjogren’s syndrome, dry eye, screening questionnaire
INTRODUCTION
Sjögren’s syndrome (SS) is a serious, chronic autoimmune disorder that targets the lacrimal and salivary glands, leading to the classic symptoms of dry eye and dry mouth.1 The disease is also associated with autoantibody production, systemic complications, and a higher risk of non-Hodgkin lymphoma (NHL) that increases with disease duration.2–5 Sjögren’s syndrome may affect up to 4 million Americans, and the diagnosis is frequently delayed by an average of up to 4–7 years from the onset of symptoms.6 Currently, SS patients whose primary complaint is dry eye are diagnosed on average 10 years after the onset of dry eye.7 Early diagnosis of SS is critical in order to prevent complications from ocular morbidities such as corneal melting and scleritis7, as well as life threatening manifestations such as vasculitis, pneumonitis, neuropathies, and lymphoma.2 Since dry eye is one of the most common presenting symptoms of SS, an encounter with an eye care provider is often the first point of care for this patient population. Therefore, the development of an effective screening tool to facilitate the timely diagnosis of SS could potentially not only lessen overall medical morbidity but also reduce healthcare costs and improve patient quality of life.
The diagnosis of SS is complex due to the fact that it is a multi-system disease requiring collaboration among multiple specialists and is based on a constellation of signs, symptoms, histopathology, and laboratory results. There are three sets of classification criteria that have historically been used for SS which include the 2002 American-European Consensus Group (AECG) criteria8, the 2012 SICCA/American College of Rheumatology (ACR) criteria9, and, most recently, the 2016 ACR/European League Against Rheumatism (EULAR) criteria.10 All include presence of clinically significant dry eye as one of the main clinical criterion.
While the majority of SS patients experience dry eye symptoms11, the early detection of SS is challenging against the background of highly prevalent dry eye disease (DED).12 Approximately 1 in 10 patients with dry eye symptoms presenting to an ophthalmologist has underlying SS.1, 13, 14 However, on average ophthalmologists are significantly under-referring dry eye patients for systemic evaluations for SS. For example, in a recent survey, among cornea specialists who specialize in the management of ocular surface disease, 44% reported that they refer less than 5% of dry eye patients for SS work-ups, with even lower referral rates reported by non-cornea specialists.15 These low referral rates may be attributable to an underappreciation of the severity of SS, the absence of any evidence-based screening algorithms to distinguish SS from non-SS dry eye patients, and a reluctance to refer a high proportion of patients who do not have the disease. Therefore, developing new instruments to help identify patients with a high likelihood of having SS is of critical importance, as not referring these patients delays treatment, while referring all dry eye patients for unnecessary work-ups is prohibitively expensive and impractical.
The Sjögren’s International Collaborative Clinical Alliance (SICCA) study was an international, multi-center NIH-sponsored registry that was designed to develop new classification criteria for SS, to improve the definition of the SS phenotype, and to collect and store clinical data as well as biospecimens for future studies.11 Due to the large sample size of the SICCA study, detailed symptom data, along with the well-verified SS status, the SICCA registry provides a unique opportunity to develop a screening tool for the clinician to discriminate dry eye patients with SS from patients without SS. Therefore, the purpose of this project was to develop a screening questionnaire based on symptoms alone (ocular and systemic), or in combination with ocular surface exam signs that could be used to identify dry eye patients with a high likelihood of having SS.
METHODS
STUDY PARTICIPANTS
The details regarding the SICCA study and characteristics of its participants have been summarized previously.9, 10, 16, 17 Briefly, participants with a history of suspected or confirmed SS were recruited from 9 international sites from various sources, which included the following: SS specialty clinics, dentists, rheumatologists, ophthalmologists, other doctors, website/internet, advertisements, friends/relatives, and other miscellaneous sources. At the baseline visit, participants answered questionnaires regarding a review of systems and medical history. In addition, a full ocular surface exam was performed in the following order as previously described16: Schirmer testing without anesthesia, tear break-up time (TBUT) and, ocular surface staining of the cornea with fluorescein and of the conjunctiva with lissamine green. In addition, all participants underwent dental and rheumatologic exams, serological testing and a lip biopsy (if not performed within one year of study entry), in order to determine their SS status. All preexisting lip biopsies were re-reviewed and independently scored by an oral pathologist.
For the purpose of this study, the 2016 American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) SS classification criteria set10 was used to define SS status. This classification criteria set is based on the weighted sum of five items which include the following: 1) Anti-SSA (Ro) antibody positivity (3 points); 2) focal lymphocytic sialadenitis with a focus score of ≥1 focus/ 4 mm2 (3 points); 3) abnormal ocular staining score of ≥5 (or a van Bijsterveld score ≥4) in at least one eye (1 point); 4) unanesthetized Schirmer test of ≤5 mm/5 minutes in at least one eye (1 point); and 5) unstimulated whole mouth salivary flow rate of ≤0.1 mL/minute (1 point).10 Patients who met the eligibility criteria and who had a total score of ≥4 for the items above met the criteria for SS. SS study participants were defined as those who met the 2016 ACR/EULAR criteria, while those who did not comprised the non-SS participants.
Of the 3,514 participants enrolled in the SICCA study, 3,389 were able to be classified regarding their SS status and completed the baseline questionnaire, with 2,908 responding “yes” to the question “Do your eyes feel dry?”. Of that subset, 848 participants were classified as “self-referred” (via friends/relatives, website/internet, or seeing an advertisement) or referred by an ophthalmologist (Figure 1). The analyses for this paper were based on the data from these 848 participants as those referred from other sources likely had a different pre-test probability of having SS than those who were self-referred or referred by an ophthalmologist. The Institutional Review Board of the University of Pennsylvania ruled that approval was not required for this study.
Figure 1:
Selection of participants for analysis from Sjögren’s International Collaborative Clinical Alliance (SICCA) study cohort
DATA ANALYSIS
We utilized the baseline questionnaire and SS status at baseline to assess the cross-sectional association between SS status and the responses to various questions regarding symptoms of organ systems that may be affected by SS. Eighty-eight potential predictors of SS (Supplemental Table 1) were selected from the baseline questionnaire and included the following categories: 1) General/Physical/Emotional Health (21 questions); 2) reproductive and hormonal history (2 questions); 3) symptoms affecting the mouth (10 questions); 4) symptoms affecting the eyes (18 questions); 5) medical history (17 questions), and 6) systems review (20 questions)(Supplemental Table 1). The chi-squared test was used to compare the proportion with a specific symptom between those with or without SS in order to identify which symptoms were useful in distinguishing the two groups. The linear trend p-value was used for assessing the association between ordinal variables and SS. Univariable and multivariable logistic regression models were used to estimate the odds ratio (OR) and its 95% confidence interval (95% CI) for the association between SS and the presence of a specific symptom or ocular sign.
We first used the univariable logistic regression model to identify the questions that may be useful for distinguishing those with or without SS (p<0.15), which were then included in the multivariable logistic regression model. The multivariable model then went through a stepwise variable selection process, to reach the final multivariable model, keeping only the statistically significant symptoms (p<0.05) for likelihood score development. Similar analyses were performed by considering the ocular signs: TBUT, corneal staining, conjunctival staining with lissamine green, and Schirmer without anesthesia.
Based on the regression coefficient associated with the presence of each symptom in the final multivariable logistic regression model, we then assigned a point value to the presence of each symptom and ocular sign, following the approach developed by Sullivan.18 The “SS likelihood score” of a participant was calculated as the sum of the number of points corresponding to symptoms and ocular surface signs present in that individual.
We calculated the area under the ROC curve (AUC) to assess the discriminative power of the likelihood score for distinguishing participants with or without SS. Using various cut points of the likelihood score, we calculated the sensitivity, specificity, positive and negative predictive values corresponding to 5% and 10% prevalence rates of SS.
RESULTS
Of the 848 participants in the SICCA study who met the eligibility criteria for this study, the majority (92% SS; 87% controls) were female, with the demographics and clinical characteristics of those included in this study summarized in Table 1. Those in the SS group were more likely to be female than those in the non-SS group (91.9% vs. 87.3%, p= 0.03). In addition, SS participants had a significantly lower median TBUT (2 vs. 4, p<0.0001) and Schirmer score (5 vs. 8, p<0.0001), and a significantly higher median corneal staining score (4 vs. 2, p<0.0001) and conjunctival staining score (6 vs. 2, p<0.0001) than those without SS (Table 1).
Table 1:
Demographics and Clinical Characteristics among participants with or without Sjögren’s Syndrome in the Sjögren’s International Collaborative Clinical Alliance (SICCA) Study
| Referral eligible (n=848)* | |||
|---|---|---|---|
| Sjogren’s Syndrome | |||
| Characteristics | No (n=464) | Yes (n=384) | P-value┼ |
| Gender | 0.03 | ||
| Female | 405 (87.3%) | 353 (91.9%) | |
| Male | 59 (12.7%) | 31 (8.1%) | |
| Age | |||
| Mean (SD) | 51.4 (12.6) | 50.5 (12.7) | 0.30 |
| Race/ethnicity | <0.0001 | ||
| Caucasian | 266 (57.3%) | 155 (40.4%) | |
| Asian or pacific islander | 115 (24.8%) | 169 (44.0%) | |
| Hispanic/Latino | 40 (8.6%) | 37 (9.6%) | |
| African American | 9 (1.9%) | 5 (1.3%) | |
| Native American | 3 (0.7%) | 1 (0.3%) | |
| More than 1 race category reported | 31 (6.7%) | 17 (4.4%) | |
| Tearing break-up time (seconds) | |||
| Median (1st quartile, 3rd quartile)* | 4 (2, 7) | 2 (1, 4) | <0.0001 |
| <10 seconds in either eye | 77 (16.7%) | 16 (4.2%) | <0.0001 |
| Schirmer test (mm/5 minutes) | |||
| Median (1st quartile, 3rd quartile)* | 8 (5, 14) | 5 (2, 8) | <0.0001 |
| Lissamine green conjunctiva staining | |||
| Median (1st quartile, 3rd quartile)* | 2 (1, 4) | 6 (4, 6) | <0.0001 |
| Fluorescein cornea staining | |||
| Median (1st quartile, 3rd quartile)* | 2 (1, 3) | 4 (2, 5) | <0.0001 |
Based on worse eye.
From chi-squared test for comparison of proportions, t-test for comparison of means, and Wilcoxon rank sum test for comparison of medians.
Eleven out of 88 questions were identified from the univariable analysis (Table 2) that were able to distinguish the SS group from the control group. From this set, four questions were statistically significant in the final multivariable logistic regression model (Table 3) including: 1) Does your mouth feel dry when eating a meal? [yes = OR 1.63(1.18–2.26)]; 2) Can you eat a cracker without drinking a fluid or liquid? [no = OR 1.46 (1.06–2.01)]; 3) How often do you have excessive tearing? [none of the time = OR 4.06 (1.81–9.10)]; and 4) Are you able to produce tears? [no = OR 2.24 (1.62–3.09)]. The area under the ROC curve (AUC) for the multivariable logistic regression model that included these four questions and gender was 0.70 (95% CI 0.66–0.73) (Figure 2). When the ocular signs TBUT and conjunctival lissamine green staining were added into the model, the AUC increased to 0.79 (95% CI: 0.77–0.82)(Table 3; Figure 2). Adding corneal staining with fluorescein and Schirmer test scores did not improve the performance of the model over using symptoms, TBUT and conjunctival lissamine green staining alone (data not shown).
Table 2:
Results of the univariable analysis among participants with or without Sjögren’s Syndrome in the Sjögren’s International Collaborative Clinical Alliance (SICCA) Study
| Sjögren’s Syndrome | |||
|---|---|---|---|
| Characteristic | No (n=464) | Yes (n=384) | P-value |
| Gender | 0.03 | ||
| Female | 405 (87.3%) | 353 (91.9%) | |
| Male | 59 (12.7%) | 31 (8.1%) | |
| Does your mouth feel dry? | 0.06 | ||
| Yes | 420 (90.5%) | 361 (94.0%) | |
| No | 44 (9.5%) | 23 (6.0%) | |
| Is your mouth dry when eating a meal? | <0.0001 | ||
| Yes | 253 (54.6%) | 265 (69.0%) | |
| No | 210 (45.4%) | 119 (31.0%) | |
| Do you have difficulty swallowing any foods? | <0.0001 | ||
| Yes | 246 (53.0%) | 264 (68.8%) | |
| No | 218 (47.0%) | 120 (31.3%) | |
| Do you need to sip liquids to swallow dry foods | 0.001 | ||
| Yes | 291 (62.9%) | 281 (73.2%) | |
| No | 172 (37.2%) | 103 (26.8%) | |
| Is the amount of saliva in your mouth: | 0.003 | ||
| Too little | 288 (62.2%) | 281 (73.2%) | |
| Too much | 19 (4.1%) | 11 (2.9%) | |
| You don’t notice it | 156 (33.7%) | 92 (24.0%) | |
| Can you eat a cracker without drinking a fluid/liquid? | <0.0001 | ||
| Yes | 238 (51.3%) | 127 (33.1%) | |
| No | 226 (48.7%) | 257 (66.9%) | |
| How would you describe your oral and dental health? | 0.003 | ||
| Excellent | 57 (12.3%) | 29 (7.6%) | |
| Good | 189 (40.7%) | 128 (33.3%) | |
| Fair | 137 (29.5%) | 134 (34.9%) | |
| Poor | 81 (17.5%) | 93 (24.2%) | |
| How often do you have excessive tears? | <0.0001 | ||
| None of the time | 297 (64.0%) | 321 (83.6%) | |
| Some of the time | 133 (28.7%) | 55 (14.3%) | |
| Half of the time | 19 (4.1%) | 2 (0.5%) | |
| Most of the time | 10 (2.2%) | 6 (1.6%) | |
| All of the time | 5 (1.1%) | 0 (0.0%) | |
| Are you able to produce tears? | <0.0001 | ||
| Yes | 367 (79.1%) | 215 (56.3%) | |
| No | 97 (20.9%) | 167 (43.7%) | |
| How often do you use artificial tears? | 0.001 | ||
| 10 times a day or more | 36 (7.8%) | 29 (7.6%) | |
| 4 to 9 times a day | 165 (35.7%) | 174 (45.3%) | |
| 1 to 3 times a day | 167 (36.2%) | 138 (35.9%) | |
| Never | 94 (20.4%) | 43 (11.2%) | |
| During last week, have you experienced tearing? | <0.0001 | ||
| None of the time | 268 (57.8%) | 290 (75.5%) | |
| Some of the time | 161 (34.7%) | 81 (21.1%) | |
| Half of the time | 20 (4.3%) | 6 (1.6%) | |
| Most of the time | 6 (1.3%) | 4 (1.0%) | |
| All of the time | 9 (1.9%) | 3 (0.8%) | |
Certain question responses do not add up to a total of 848 participants due to the fact that a small number of participants did not answer all baseline questions.
Table 3:
Multivariable analysis for the prediction of Sjögren’s Syndrome using symptoms with or without ocular signs
| Multivariable model: Gender + Symptoms only | Multivariable model: Gender + Symptoms + Tear breakup time + conjunctival staining | |||||
|---|---|---|---|---|---|---|
| Predictors | N* | Sjögren’s Syndrome (%) | OR (95% CI) | P-value | OR (95% CI) | P-value |
| Gender | ||||||
| Female | 755 | 351 (46.5%) | 1.66 (1.03 – 2.70) | 0.04 | 1.63 (0.97 – 2.75) | 0.07 |
| Male | 90 | 31 (34.4%) | 1.00 | 1.00 | ||
| Mouth dry when eating meal | ||||||
| Yes | 516 | 263 (51.0%) | 1.63 (1.18 – 2.26) | 0.003 | 1.47 (1.03 – 2.08) | 0.03 |
| No | 329 | 119 (36.2%) | 1.00 | 1.00 | ||
| Eat a cracker without drinking a fluid/liquid | ||||||
| Yes | 364 | 127 (34.9%) | 1.00 | 1.00 | ||
| No | 481 | 255 (53.0%) | 1.46 (1.06 – 2.01) | 0.02 | 1.55 (1.09 – 2.21) | 0.01 |
| How often do you have excessive tears? | <0.0001 | <0.0001 | ||||
| None of the time | 616 | 319 (51.8%) | 4.06 (1.81 – 9.10) | 0.0007 | 3.75 (1.59 – 8.87) | 0.003 |
| Some of the time | 187 | 55 (29.4%) | 1.85 (0.79 – 4.33) | 0.15 | 1.66 (0.67 – 4.10) | 0.27 |
| More than some of the time | 42 | 8 (19.1%) | 1.00 | 1.00 | ||
| Are you able to produce tears? | ||||||
| Yes | 581 | 215 (37.0%) | 1.00 | 1.00 | ||
| No | 264 | 167 (63.3%) | 2.24 (1.62 – 3.09) | <0.0001 | 1.42 (1.01 – 2.03) | 0.049 |
| Tear breakup time less than10 seconds in either eye | ||||||
| No | 93 | 16 (17.2%) | 1.00 | |||
| Yes | 752 | 367 (48.8%) | 3.04 (1.65 – 5.60) | 0.0004 | ||
| Lissamine green staining of conjunctiva (nasal or temporal region of either eye) | ||||||
| Negative (<2) | 280 | 48 (17.1%) | 1.00 | |||
| Positive (≥2) | 565 | 335 (59.3%) | 5.68 (3.91 – 8.24) | <0.0001 | ||
| AUC (95% CI) | 0.70 (0.66 – 0.73) | 0.79 (0.77 – 0.82) | ||||
3 participants with missing data in any of these variables were excluded.
Figure 2:
ROC curves from the multivariable logistic regression model with gender and symptoms only, and from the model with gender, symptoms, tear break up time (TBUT) and conjunctival staining.
A SS likelihood scoring system was developed based on the regression coefficients corresponding to each symptom and ocular sign from the multivariable logistic regression model (Table 4). This scoring system included the assignment of points for the following items: gender, four symptom questions, and two ocular signs (TBUT and conjunctival staining with lissamine green). The question “How often do you have excessive tearing?” was assigned 1 point for the response of “Some of the time”, and 3 points for a response of “None of the time” (“Half of the time”, “Most of the time”, or “All of the time” were assigned 0 points). Having a TBUT value of less than 10 seconds was assigned 2 points, while positive lissamine green staining of the conjunctiva (≥2 on a scale of 0–3 for the temporal or nasal conjunctiva in one or both eyes16) was assigned 4 points. The maximum score using this system was 13 points.
Table 4:
Likelihood Scoring system for screening dry eye patients for Sjögren’s Syndrome
| Items | Response | Likelihood score points |
|---|---|---|
| Gender | Female | 1 |
| Mouth dry when eating meal | Yes | 1 |
| Eat a cracker without drinking a fluid/liquid | No | 1 |
| How often do you have excessive tears? | None of the time | 3 |
| Some of the time | 1 | |
| Are you able to produce tears? | No | 1 |
| Tear breakup time less than 10 seconds in either eye | Yes | 2 |
| Lissamine green staining of conjunctiva in nasal or temporal region of either eye (≥ 2) | Positive (≥2) | 4 |
Table 5 shows that the likelihood of having SS is associated with the number of likelihood score points in a dose-response manner regardless of whether or not the likelihood score includes dry eye signs.
Table 5:
Likelihood scores for predicting Sjögren’s Syndrome in dry eye patients
| Likelihood score from Gender + Symptoms only | Likelihood score from Gender + Symptoms + tear breaking time + conjunctival staining | ||||
|---|---|---|---|---|---|
| Likelihood Score* | N | Sjögren’s Syndrome n (%) | Likelihood Score | N | Sjögren’s Syndrome n (%) |
| 1 | 13 | 3 (23.1%) | 1 | 1 | 0 (0.0%) |
| 2 | 61 | 14 (23.0%) | 2 | 8 | 0 (0.0%) |
| 3 | 98 | 25 (25.5%) | 3 | 13 | 0 (0.0%) |
| 4 | 193 | 67 (34.7%) | 4 | 28 | 1 (3.6%) |
| 5 | 155 | 58 (37.4%) | 5 | 51 | 9 (17.7%) |
| 6 | 207 | 123 (59.4%) | 6 | 94 | 17 (18.1%) |
| 7 | 121 | 94 (77.7%) | 7 | 54 | 14 (25.9%) |
| 8 | 82 | 27 (32.9%) | |||
| 9 | 59 | 17 (28.8%) | |||
| 10 | 109 | 57 (52.3%) | |||
| 11 | 105 | 49 (46.7%) | |||
| 12 | 140 | 102 (72.9%) | |||
| 13 | 104 | 91 (87.5%) | |||
| AUC from likelihood score (95% CI) | 0.70 (0.66 – 0.73) | 0.79 (0.76 – 0.82) | |||
Calculated using the likelihood scoring algorithm in Table 4.
Table 6 summarizes the sensitivity, specificity, and predictive values for 2 models: 1) using gender and symptoms only, or 2) gender, symptoms, TBUT and lissamine green staining of the conjunctiva. For model 1 (gender + symptoms only), when a cut point of 6 points was used for classifying the screening result as positive for SS, the likelihood score had a sensitivity of 56% (95% CI: 51.5% – 61.4%) and a specificity of 76% (95% CI: 72.0% – 79.7%). In a population where the prevalence rate is 5%, this would have a positive predictive value (PPV) of 11% and negative predictive value (NPV) of 97.1%, whereas if the prevalence rate were 10%, the PPV would increase to 21% and the NPV would remain high (94%). For model 2 (gender + symptoms + TBUT + conjunctival staining), a likelihood score using a cut point of 10 points had a sensitivity of 78% (95% CI: 73.5% – 81.7%) with a specificity of 66% (61.5% – 70.1%). If the prevalence rate were 5%, this would have a PPV of 11% and a NPV of 98%, whereas if the prevalence rate were 10%, the PPV would increase to 20% and the NPV would remain high (96%).
Table 6:
Screening test characteristics for two algorithms for predicting the presence of Sjögren’s Syndrome
| Presence of Sjögren’s Syndrome (n=384) | Absence of Sjögren’s Syndrome (n=464) | Prevalence rate of Sjögren’s Syndrome =5% | Prevalence rate of Sjögren’s Syndrome =10% | ||||
|---|---|---|---|---|---|---|---|
| Sensitivity | Specificity | Positive predictive value | Negative predictive value | Positive predictive value | Negative predictive value | ||
| Cut point (≥) likelihood score | Likelihood score based on (Gender + Symptoms only) | ||||||
| 2 | 99.2% | 2.2% | 5.1% | 98.1% | 10.1% | 96.1% | |
| 3 | 95.6% | 12.3% | 5.4% | 98.2% | 10.8% | 96.2% | |
| 4 | 89.1% | 28.0% | 6.1% | 98.0% | 12.1% | 95.9% | |
| 5 | 71.6% | 55.2% | 7.8% | 97.4% | 15.1% | 94.6% | |
| 6 | 56.5% | 76.1% | 11.1% | 97.1% | 20.8% | 94.0% | |
| 7 | 24.5% | 94.2% | 18.2% | 96.0% | 31.9% | 91.8% | |
| Likelihood score based on (Gender + Symptoms +tear breakup time + conjunctival staining) | |||||||
| 2 | 100% | 0.2% | 5.0% | 100% | 10.0% | 100% | |
| 3 | 100% | 1.9% | 5.1% | 100% | 10.2% | 100% | |
| 4 | 100% | 4.7% | 5.2% | 100% | 10.4% | 100% | |
| 5 | 99.7% | 10.6% | 5.5% | 99.9% | 11.0% | 99.7% | |
| 6 | 97.4% | 19.6% | 6.0% | 99.3% | 11.9% | 98.5% | |
| 7 | 93.0% | 36.2% | 7.1% | 99.0% | 13.9% | 97.9% | |
| 8 | 89.3% | 44.8% | 7.8% | 98.8% | 15.2% | 97.4% | |
| 9 | 82.3% | 56.7% | 9.1% | 98.4% | 17.4% | 96.6% | |
| 10 | 77.9% | 65.7% | 10.7% | 98.3% | 20.2% | 96.4% | |
| 11 | 63.0% | 76.9% | 12.6% | 97.5% | 23.3% | 94.9% | |
| 12 | 50.3% | 89.0% | 19.4% | 97.1% | 33.7% | 94.2% | |
| 13 | 23.7% | 97.2% | 30.8% | 96.0% | 48.5% | 92.0% | |
DISCUSSION
We have developed an evidence-based screening algorithm to identify dry eye patients with possible SS. Using data from the SICCA study cohort, we identified a set of four screening questions that distinguish dry eye patients with or without SS. The SS likelihood score developed based on these four screening questions and gender has a moderate discriminative ability for SS (AUC=0.70). In addition, the performance of the likelihood score improves (AUC=0.79) if certain ocular signs (TBUT and lissamine green staining of the conjunctiva) are included.
Our finding that female gender was a strong predictor is consistent with the fact that SS, like many autoimmune diseases, is known to have a strong female propensity.19 Interestingly, one of the most common questions used to screen dry eye patients for SS, “Does your mouth feel dry?” was not useful in distinguishing those with or without SS. However, the more specific related questions “Is your mouth dry when eating a meal?” and “Can you eat a cracker without drinking a fluid or liquid?” were helpful for distinguishing the two groups.
No previous studies have assessed the prevalence of multi-organ, systemic symptoms specifically in SS patients with dry eye symptoms. One study conducted by the Sjögren’s Syndrome Foundation (SSF) examined symptom prevalence in SS patients and found that the most prevalent symptoms were dry eye (92%) and dry mouth (92%).11 This study is consistent with our findings in that the four most useful questions to distinguish dry eye patients with or without SS relate to symptoms of dry eye and dry mouth.
The clinical utility of a screening algorithm depends on the sensitivity, specificity, and prevalence of disease in the target population. Ideally, a screening tool for SS should have high sensitivity so that it correctly identifies those with the disease, while also having good specificity to limit the number of false positives.20, 21 However, the prevalence of disease affects the positive and negative predictive values of the test. When the disease prevalence is low, the PPV can be low even for a screening test with good sensitivity and specificity. Therefore, it is important to consider prevalence when assessing the clinical utility of a screening algorithm.
This screening algorithm should also be interpreted in the context of what is occurring in clinical practice. The current low referral rate of dry eye patients for systemic work-ups could be due in part to the lack of a good screening instrument for SS. In a recent survey of ophthalmologists, the majority of those surveyed (83%) felt that there was a need for an evidence-based screening tool that could be used to identify dry eye patients with SS.15 While the algorithm developed in this study was not high in both sensitivity and specificity, it still has clinical utility. For example, if there is a prevalence of SS of 10% in dry eye patients, and all of those with a score of 10 points or above are referred for work-ups (i.e. sensitivity of 78%, specificity 66%), then the positive predictive value increases from 10% to 20%. Therefore, this would double the number of dry eye patients diagnosed with SS. In addition, eye care providers could also choose to use a higher or lower cut-point depending on if they are concerned more about over- or under-referrals for systemic SS work-ups.
It is also important to note that eye care providers do not perform all dry eye tests on every patient at each visit. The choice of which dry eye tests are performed varies greatly based on the examiner. For example, Korb found that there was no single dry eye test that was the dominant choice of ophthalmologists and optometrists evaluating patients for dry eye.22 In that study, the most frequent response was the use of taking a history and/or dry eye questionnaire, followed by TBUT (19%), fluorescein staining (13%), and rose bengal staining (10%). Similarly, Nichols and colleagues also found that symptom assessment was the most common test used in ophthalmic practice (83%), followed by fluorescein staining (56%), and TBUT (41%).23 Schirmer testing and rose bengal staining of the conjunctiva were performed less commonly. Finally, in a recent survey of ophthalmologists, the most common dry eye tests that were performed were corneal staining with fluorescein, TBUT, and anesthetized Schirmer testing.15 Of note, lissamine green staining of the conjunctiva was one of the least commonly performed tests, which could be due to the fact that it requires additional supplies and is not as readily available as fluorescein.23 Interestingly, in our study we found that adding corneal staining or Schirmer test scores did not improve the discriminative ability of the model over using symptoms, TBUT, and conjunctival staining with lissamine green (data not shown).
Our study has certain limitations. One limitation is its moderate specificity and sensitivity of this screening tool. While ideally there would be a higher sensitivity and specificity, it is important to remember that currently there are no evidence-based screening tools available and therefore this is a good starting point. In the future, this screening tool could be refined and used with other screening tools to improve its sensitivity or specificity for identifying SS. In addition, this screening tool needs to be externally validated in a new cohort in order to assess how it performs prospectively. Another possible limitation is misclassification bias by outcome. For example, a control subject at baseline could subsequently be diagnosed with SS during the course of the study. However, in the SICCA registry, among the 771 participants who returned for a 2–3 year follow-up visit, only 8–9% of non-SS patients progressed to SS.19 Therefore, delayed diagnoses were unlikely to have had a significant effect on our analyses. In addition, it is likely that the control patients in this study were not completely normal in that there often was some suspicion of SS, which prompted an ophthalmologist or the patient to be referred for the study. However, despite this, our screening questionnaire was able to distinguish these two groups. We would expect the questionnaire to have better specificity with higher positive predictive values when distinguishing dry eye patients without any suspicion of SS from those with a high likelihood of having SS.
Finally, this algorithm was developed using a target population from the SICCA cohort of those who would typically be seen by an ophthalmologist. We chose this population as the pre-test probability of a patient referred from another specialty would likely be different than that of someone referred by ophthalmology. For example, a patient referred to the study from an oral medicine specialist would be more likely to have dry mouth. More studies are needed to further refine and validate the use of this algorithm in other populations, such as patients in a primary care setting or rheumatology clinic. Future studies could also explore the inclusion of other objective test components, such as serological test results, into the model to see if this improves its performance.24 In addition, new methods such as studying the relationship between the ocular surface microbiome and the rest of the body, as well as the use of metagenomics, show promise in improving the diagnosis of SS and could potentially be used in combination with our screening algorithm.24, 25
In summary, eye care providers are often the first practitioners to see dry eye patients with undiagnosed SS. As a result, ophthalmologists and optometrists are in a unique position to identify patients with a high likelihood of SS early in the disease course. Historically, however, dry eye patients are typically not referred until severe, advanced ocular surface disease is present. A paradigm shift in which dry eye patients are evaluated for possible SS sooner would facilitate timely diagnosis. Our algorithm may be helpful for identifying patients with more advanced dry eye disease characteristic of SS.
This new screening algorithm is brief and easy to use in a busy clinical practice. It can be used with symptom data alone, or for better performance, ocular signs (TBUT and lissamine green staining of the conjunctiva) can also be added into the algorithm. In the future, if this screening tool is further refined and validated, it could shorten the time to diagnosis of SS, lead to earlier treatment, and significantly improve clinical outcomes.
Supplementary Material
Eighty-eight potential predictors of Sjögren’s Syndrome were selected from the baseline questionnaire for use in the development of the screening algorithm.
Acknowledgments
Funding:
VYB: National Eye Institute R01 EY026972; Research to Prevent Blindness
GSY: National Eye Institute R01 EY026972; P30 EY001583 45
The sponsors or funding organizations listed above had no role in the design or conduct of this research.
Financial Disclosures
VYB: Bausch & Lomb (funding for research study); Celularity (consultant); Verily (consultant)
EKA: Allergan Inc. (institutional support), Novalique Gmb. (consultant), Regeneron (consultant), Sjogren’s Syndrome Foundation (unpaid board member).
MMG: Celularity (consultant); PRN (personal financial interest); Verily (consultant)
FBV: Novartis (consultant), Biogen-Idec (consultant), Celularity (consultant), Boston Pharmaceuticals (consultant) Bristol-Myers Squibb (consultant); Immco Diagnostics (consultant)
Footnotes
Conflict of Interest
No conflicting relationships exist for any other authors.
Data Availability
The data used in this manuscript were obtained from the Sjögren’s International Collaborative Clinical Alliance (SICCA) Biorepository, funded under contract #HHSN26S201300057C by the National Institute of Dental and Craniofacial Research (NIDCR). This manuscript was prepared using a publicly available SICCA data set and does not necessarily reflect the opinions or views of all SICCA investigators or the NIDCR.
REFERENCES
- 1.Liew MS, Zhang M, Kim E, et al. Prevalence and predictors of Sjogren’s syndrome in a prospective cohort of patients with aqueous-deficient dry eye. Br J Ophthalmol 2012;96(12):1498–503. [DOI] [PubMed] [Google Scholar]
- 2.Fox RI. Sjogren’s syndrome. Lancet 2005;366(9482):321–31. [DOI] [PubMed] [Google Scholar]
- 3.Nikolov NP, Illei GG. Pathogenesis of Sjogren’s syndrome. Curr Opin Rheumatol 2009;21(5):465–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Baimpa E, Dahabreh IJ, Voulgarelis M, et al. Hematologic manifestations and predictors of lymphoma development in primary Sjogren syndrome: clinical and pathophysiologic aspects. Medicine (Baltimore) 2009;88(5):284–93. [DOI] [PubMed] [Google Scholar]
- 5.Ioannidis JP, Vassiliou VA, Moutsopoulos HM. Long-term risk of mortality and lymphoproliferative disease and predictive classification of primary Sjogren’s syndrome. Arthritis Rheum 2002;46(3):741–7. [DOI] [PubMed] [Google Scholar]
- 6.Fox PC, Bowman SJ, Segal B, et al. Oral involvement in primary Sjogren syndrome. J Am Dent Assoc 2008;139(12):1592–601. [DOI] [PubMed] [Google Scholar]
- 7.Akpek EK, Mathews P, Hahn S, et al. Ocular and systemic morbidity in a longitudinal cohort of Sjogren’s syndrome. Ophthalmology 2015;122(1):56–61. [DOI] [PubMed] [Google Scholar]
- 8.Vitali C, Bombardieri S, Jonsson R, et al. Classification criteria for Sjogren’s syndrome: a revised version of the European criteria proposed by the American-European Consensus Group. Ann Rheum Dis 2002;61(6):554–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Shiboski SC, Shiboski CH, Criswell L, et al. American College of Rheumatology classification criteria for Sjogren’s syndrome: a data-driven, expert consensus approach in the Sjogren’s International Collaborative Clinical Alliance cohort. Arthritis Care Res (Hoboken) 2012;64(4):475–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Shiboski CH, Shiboski SC, Seror R, et al. 2016 American College of Rheumatology/European League Against Rheumatism classification criteria for primary Sjogren’s syndrome: A consensus and data-driven methodology involving three international patient cohorts. Ann Rheum Dis 2017;76(1):9–16. [DOI] [PubMed] [Google Scholar]
- 11.Foundation SS. Moisture Seekers 2006;24:1–3. [Google Scholar]
- 12.Pflugfelder SC. Prevalence, burden, and pharmacoeconomics of dry eye disease. Am J Manag Care 2008;14(3 Suppl):S102–6. [PubMed] [Google Scholar]
- 13.Bunya VY, Ying GS, Maguire MG, et al. Prevalence of Novel Candidate Sjogren Syndrome Autoantibodies in the Dry Eye Assessment and Management (DREAM) Study. Cornea 2018;37(11):1425–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Akpek EK, Klimava A, Thorne JE, et al. Evaluation of patients with dry eye for presence of underlying Sjogren syndrome. Cornea 2009;28(5):493–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Bunya VY, Fernandez KB, Ying GS, et al. Survey of Ophthalmologists Regarding Practice Patterns for Dry Eye and Sjogren Syndrome. Eye Contact Lens 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Whitcher JP, Shiboski CH, Shiboski SC, et al. A simplified quantitative method for assessing keratoconjunctivitis sicca from the Sjogren’s Syndrome International Registry. Am J Ophthalmol 2010;149(3):405–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Shiboski CH, Baer AN, Shiboski SC, et al. Natural History and Predictors of Progression to Sjogren’s Syndrome Among Participants of the Sjogren’s International Collaborative Clinical Alliance Registry. Arthritis Care Res (Hoboken) 2018;70(2):284–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Sullivan LM, Massaro JM, D’Agostino RB Sr., Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med 2004;23(10):1631–60. [DOI] [PubMed] [Google Scholar]
- 19.Patel R, Shahane A. The epidemiology of Sjogren’s syndrome. Clin Epidemiol 2014;6:247–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Trevethan R Sensitivity, Specificity, and Predictive Values: Foundations, Pliabilities, and Pitfalls in Research and Practice. Front Public Health 2017;5:307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Boynton PM, Greenhalgh T. Selecting, designing, and developing your questionnaire. BMJ 2004;328(7451):1312–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Korb DR. Survey of preferred tests for diagnosis of the tear film and dry eye. Cornea 2000;19(4):483–6. [DOI] [PubMed] [Google Scholar]
- 23.Nichols KK, Nichols JJ, Zadnik K. Frequency of dry eye diagnostic test procedures used in various modes of ophthalmic practice. Cornea 2000;19(4):477–82. [DOI] [PubMed] [Google Scholar]
- 24.Trujillo-Vargas CM, Schaefer L, Alam J, et al. The gut-eye-lacrimal gland-microbiome axis in Sjogren Syndrome. Ocul Surf 2020;18(2):335–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Borroni D, Romano V, Kaye SB, et al. Metagenomics in ophthalmology: current findings and future prospectives. BMJ Open Ophthalmol 2019;4(1):e000248. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Eighty-eight potential predictors of Sjögren’s Syndrome were selected from the baseline questionnaire for use in the development of the screening algorithm.