Purpose
To evaluate the use of tear osmolarity in the diagnosis of dry eye disease.
Design
A prospective, observational case series to determine the clinical usefulness of tear osmolarity and commonly used objective tests to diagnose dry eye disease.
Methods
A multicenter, 10-site study consisting of 314 consecutive subjects between 18 and 82 years of age. Bilateral tear osmolarity, tear film break-up time (TBUT), corneal staining, conjunctival staining, Schirmer test, and meibomian gland grading were performed. Diagnostic performance was measured against a composite index of objective measurements that classified subjects as having normal, mild or moderate, or severe dry eye. The main outcome measures were sensitivity, specificity, area under the receiver operating characteristic curve, and intereye variability.
Results
Of the 6 tests, tear osmolarity was found to have superior diagnostic performance. The most sensitive threshold between normal and mild or moderate subjects was found to be 308 mOsms/L, whereas the most specific was found at 315 mOsms/L. At a cutoff of 312 mOsms/L, tear hyperosmolarity exhibited 73% sensitivity and 92% specificity. By contrast, the other common tests exhibited either poor sensitivity (corneal staining, 54%; conjunctival staining, 60%; meibomian gland grading, 61%) or poor specificity (tear film break-up time, 45%; Schirmer test, 51%). Tear osmolarity also had the highest area under the receiver operating characteristic curve (0.89). Intereye differences in osmolarity were found to correlate with increasing disease severity ( r 2 = 0.32).
Conclusions
Tear osmolarity is the best single metric both to diagnose and classify dry eye disease. Intereye variability is a characteristic of dry eye not seen in normal subjects.
Dry eye disease is a commonly encountered condition in clinical practice and affects up to 20% of the population in North America. The knowledge base concerning its pathogenesis, classification, and characteristics has grown considerably over the last 15 years, but its diagnosis, particularly in the early or mild stages, has been hampered by the lack of objective tests with sufficient sensitivity and specificity, adequate repeatability, ease of performance, and suitability for the clinical practice setting. In addition, although symptoms of ocular irritation are common, there is a lack of correlation between signs and symptoms, particularly in mild dry eye disease, rendering symptoms alone unreliable for diagnosis and determination of disease severity. Moreover, there is a lack of consensus on the clinical usefulness of individual objective tests in the diagnosis of dry eye disease.
An increase in tear osmolarity is a hallmark of dry eye disease and is thought to be the central mechanism in the pathogenesis of ocular surface damage in the disease, as noted in the Dry Eye Workshop Report. Tear osmolarity has been reported to be the single best marker for dry eye disease, but measurement has been limited to laboratory instruments requiring large microliter volumes; collection and manipulation of the tear specimens induce reflex tearing in most subjects, and collected specimens can be concentrated by evaporative loss during handling and collection. Further, microliter volumes are not available in many dry eye patients. The current study was undertaken to determine the clinical usefulness of tear osmolarity in the diagnosis and management of dry eye disease using a new tear osmometer that minimizes the previous limitations compared with the other most commonly used tests.
Methods
A prospective, exploratory, multicenter study was undertaken at 10 sites in the European Union and the United States. The subject population consisted of randomly presenting subjects between the ages of 18 and 82 years of both sexes, including those with and without a history of dry eye disease. Investigators were instructed to recruit roughly a 2:1 ratio of presumed dry eye patients to normals. This report documents the results of the analysis of the initial 314 subjects, 15 of whom were removed for incomplete data reporting (n = 218 female, n = 81 male; average age, 46.3 ± 16.9 years).
Subjects were excluded from the study if they exhibited any active infection of the eye, active ocular allergy, evidence of lid deformity or abnormal lid movement disorder, refractive surgery within one year of the study visit, pregnancy or lactation, abnormal nasolacrimal drainage, punctal plug placement within 30 days of testing, or evidence of a systemic disease (except Sjögren syndrome) known to affect tear production, such as thyroid eye disease or graft-versus-host disease. Moreover, patients were excluded if they initiated or altered the dose of chronic systemic medication known to affect tear production within 30 days of testing (for instance, initiation or dosage change of antihistamines, antidepressants, diuretics, corticosteroids, or immunomodulators were listed as exclusion criteria) or had a known hypersensitivity to any of the agents used in testing (eg, sodium fluorescein or lissamine green). Subjects were required to remove contact lenses at least 8 hours before examination and not to use artificial tears within two hours of screening. Patients were excluded from the study if they did not wish to participate in the study or could not cooperate with the collection of tear samples.
This single visit study included the following common objective tests for dry eye disease, performed on both eyes: tear osmolarity, tear film breakup time (TBUT), corneal staining (National Eye Institute/Industry scale), conjunctival staining, Schirmer test without anesthesia, and meibomian gland grading (Foulks/Bron scoring ). Data including demographic information were recorded on case report forms and sent to a central data collecting center, where they were entered in digital form for analysis. The more severe of the bilateral measurements was used in analysis because of the asymmetrical effects of transient compensatory mechanisms attempting to drive down tear osmolarity in response to environmental stress. Intereye variability in osmolarity was calculated as the absolute difference in the two eye measurements (|OD−OS|).
Optimal cutoff values for each sign were determined post hoc, assuming equal risk for false-positive and false-negative results. In particular, Gaussian distributions were generated based on the mean and standard deviation of normal and dry eye disease populations. Diagnostic cutoff thresholds were located at the intersection between these curves. Sensitivity was determined as the percentage of true positives, whereas specificity was calculated as the percentage of true negatives.
In certain clinical situations, the acceptable risks of false-positive or false-negative diagnosis are unequal. Although the exact weighting for these risks has not been quantified in the literature, more sensitive or specific thresholds were established at the intersection between normal and mild or moderate or between normal and severe subsets of the patient population. For more sensitive detection, cutoff values were located at the intersection between normal and mild or moderate subjects, whereas more specific detection thresholds were located at the intersection between normal and severe dry eye patient distributions.
Classification of mild or moderate and severe patients was based on a composite disease severity index, derived from the Dry Eye Workshop severity scale. To convert the various clinical measurements into common units, an expert panel of clinicians, using a consensus approach, agreed to the severity distribution for each diagnostic test, based on the Dry Eye Workshop categorical severity scale. For each test performed, the more severe of the bilateral measurements for each clinical sign was mapped onto a 0 (least evidence of disease) to 1 (greatest evidence of disease) scale, normalized by an independent component analysis to remove overlap in mutual information, and then combined into a single Euclidean distance from the origin to produce a final composite severity score for each subject. In this way, no single test outweighed the others. In terms of classification, the composite severity score does not reveal etiologic information; rather, it provides an unbiased, objective quantification of dry eye disease severity that does not vary according to the cause of the disease.
Normal, mild, moderate, and severe patients were assigned according to the 4 quartiles of composite disease severity, but because of the lack of clinical differentiation between the mild and moderate quartiles, they were combined into a single group, mild or moderate. The independent component analysis methodology is discussed in detail in a recently published report.
Results
Tear osmolarity was found to have a 72.8% sensitivity and 92.0% specificity at a cutoff value of 312 mOsms/L (i.e., values > 311 mOsms/L; Table 1 ). No other clinical sign exhibited more than 62% performance in both categories. Corneal staining, conjunctival staining, and meibomian grading lacked sensitivity (54.0%, 60.3%, and 61.2% respectively), whereas TBUT and Schirmer results lacked specificity (45.3% and 50.7%, respectively). The performance of osmolarity was consistent with earlier studies. In particular, the meta-analysis performed by Tomlinson and associates reported a 69% sensitivity and 92% specificity at a referent value of 316 mOsm/L. Citing the 15% prevalence used in Tomlinson and associates, tear osmolarity was found to have an 88.6% accuracy in the current study.
| Test | Cutoff | Sensitivity (n = 224) | Specificity (n = 75) |
|---|---|---|---|
| Osmolarity | >311 mOsms/L | 72.8% | 92.0% |
| TBUT | <10 secs | 84.4% | 45.3% |
| Schirmer | <18 mm | 79.5% | 50.7% |
| Corneal stain | >Grade 1 | 54.0% | 89.3% |
| Conjunctival stain | >Grade 2 | 60.3% | 90.7% |
| Meibomian grade | >Grade 5 | 61.2% | 78.7% |
a Cutoff values were located at the intersection between normal subjects and the entire subset of dry eye patients.
The average osmolarity values across normal, mild or moderate, and severe severity grades were 300.8 ± 7.8 mOsms/L, 315.5 ± 10.4 mOsms/L, and 336.7 ± 22.2 mOsms/L, respectively. Within each severity subset, no significant differences were found across age or sex. Specifically, there was no difference in mean for subjects younger than 30 years, 30 to 50 years, 50 to 70 years, and older than 70 years (all P > .01, with a Bonferroni correction of n = 6 on 2-tailed t tests). Similarly, there was no difference between males or females across different levels of disease severity (all P > .20). Therefore, the primary determinant of osmolarity was disease severity.
Calculated cutoff thresholds for each test generally were in agreement with published values, although the Schirmer results were much higher than the clinically accepted range of 5 to 10 mm/5 minutes. When placed at a cutoff of 5 mm or less, the Schirmer test results improved the classification of normal subjects to 84.0%, but suffered in diagnosis of dry eye subjects, correctly reporting 24.2% of the mild to moderate patients and 57.3% of the severe patients.
The percentages of correctly diagnosed subjects, segregated by disease severity, are shown in Table 2 . Osmolarity was revealed to be exceptionally good at differentiating normal subjects (92.0% correctly diagnosed) from severe dry eye subjects (89.3% correctly diagnosed), while still correctly identifying approximately two thirds of the difficult-to-diagnose early stage and mild or moderate subjects (64.4%). By contrast, only TBUT and the Schirmer test were more effective identifying the early stage and mild or moderate subjects (76.5% and 75.2%, respectively), but both of those examinations had a high rate of false positives in the normal cohort (only 45.3% and 50.7% correctly diagnosed, respectively).
| Test | Cutoff | % Correctly Diagnosed | ||
|---|---|---|---|---|
| Normal (n = 75) | Mild/Moderate (n = 149) | Severe (n = 75) | ||
| Osmolarity | >311 mOsms/L | 92.0% | 64.4% | 89.3% |
| TBUT | <10 secs | 45.3% | 76.5% | 100.0% |
| Schirmer | <18 mm | 50.7% | 75.2% | 88.0% |
| Corneal stain | >Grade 1 | 89.3% | 43.6% | 74.7% |
| Conjunctival stain | >Grade 2 | 90.7% | 49.7% | 81.3% |
| Meibomian grade | >Grade 5 | 78.7% | 51.7% | 80.0% |
a Cutoff values were located at the intersection between normal subjects and the entire subset of dry eye patients.
Changing the diagnostic cutoff to a more sensitive threshold at the intersection of normal and mild or moderate subjects ( Table 3 ) resulted in the predictable increase in correctly diagnosed dry eye subjects at the expense of an increased false-positive rate. Tear osmolarity, at a cutoff of more than 308 mOsms/L, achieved a 90.7% rate of proper diagnosis of severe dry eye patients, while accurately classifying 81.3% of the normal subjects. Corneal staining seemed to benefit the most from a lower threshold (from grade 2 to grade 1), improving to an 85.3% mark in the severe category, although in practice, distinguishing between a single corneal staining grade is quite challenging.
| Test | Mild Cutoff | % Correctly Diagnosed | ||
|---|---|---|---|---|
| Normal (n = 75) | Mild/Moderate (n = 149) | Severe (n = 75) | ||
| Osmolarity | >308 mOsms/L | 81.3% | 73.2% | 90.7% |
| TBUT | <11 secs | 40.0% | 83.2% | 100.0% |
| Schirmer | <20 mm | 42.7% | 82.6% | 90.7% |
| Corneal stain | >Grade 0 | 82.7% | 60.4% | 85.3% |
| Conjunctival stain | >Grade 1 | 73.3% | 73.2% | 90.7% |
| Meibomian grade | >Grade 4 | 74.7% | 54.4% | 81.3% |
a Cutoff values were located at the intersection between normal subjects and the mild/moderate subset of dry eye patients.
Conversely, changing the diagnostic cutoff to a more specific threshold at the intersection of normal and severe subjects ( Table 4 ) weakened performance in the overall dry eye population. The corresponding improvement in normal classification (4.0% for osmolarity) did not seem to warrant the loss in dry eye performance (8.0% in the mild or moderate group and 2.0% in the severe group).
| Test | Severe Cutoff | % Correctly Diagnosed | ||
|---|---|---|---|---|
| Normal (n = 75) | Mild/Moderate (n = 149) | Severe (n = 75) | ||
| Osmolarity | >314 mOsms/L | 96.0% | 56.4% | 86.7% |
| TBUT | <6 secs | 69.3% | 63.8% | 97.3% |
| Schirmer | <16 mm | 57.3% | 71.8% | 86.7% |
| Corneal stain | >Grade 1 | 89.3% | 43.6% | 74.7% |
| Conjunctival stain | >Grade 2 | 90.7% | 49.7% | 81.3% |
| Meibomian grade | >Grade 5 | 78.7% | 51.7% | 80.0% |
a Cutoff values were located at the intersection between normal subjects and the severe subset of dry eye patients.
The demographics of disease severity broken down by age and sex are shown in Table 5 . Both the relative level of severity and the ratio of people with dry eye disease increased with increasing age. Many more females (2.4 times as many) had severe dry eye than males.
| Test | Normal | Mild/Moderate | Severe |
|---|---|---|---|
| <30 years of age | 33 | 37 | 7 |
| 30 to 50 years of age | 21 | 50 | 14 |
| 50 to 70 years of age | 17 | 45 | 44 |
| >70 years of age | 4 | 17 | 10 |
| Male | 26 | 44 | 10 |
| Female | 49 | 105 | 65 |
Diagnostic results are summarized in Figure 1 , shown as a receiver operating characteristic curve. Osmolarity demonstrated the greatest area under the curve (0.89), followed by conjunctival staining (0.83), TBUT (0.81), meibomian grading (0.78), corneal staining (0.77), and Schirmer test (0.71).
