Development of Definitive and Reliable Grading

Scales for Meibomian Gland Dysfunction

REIKO ARITA, IZUMI MINOURA, NAOYUKI MORISHIGE, RIKA SHIRAKAWA, SHIMA FUKUOKA, KEI ASAI,
TATEKI GOTO, TAKAHIRO IMANAKA, AND MASATSUGU NAKAMURA

PURPOSE: To develop and validate grading scales for Author(s). Published by Elsevier Inc. This is an
meibomian gland dysfunction (MGD) that allow consis- open access article under the CC BY-NC-ND license
tent diagnosis of MGD and are suitable for clinical (http://creativecommons.org/licenses/by-nc-nd/4.0/).)
studies.

DESIGN: Development and validation study of grading
scales.

T

METHODS: Lid margin and meibomian gland photo- HE TEAR FILM THAT COVERS THE OCULAR SURFACE
graphs were taken in the multicenter, prospective is protected from evaporation by a thin layer of lipid
cross-sectional study for MGD and control subjects. secreted by the meibomian glands.1 Meibomian
New grading scales for MGD signs (abnormal lid gland dysfunction (MGD) is one of the most common dis-
margin findings of vascularity, plugging of gland ori- orders encountered in ophthalmic clinics and is now recog-
fices, lid margin irregularity, lid margin thickening, nized as a major cause of dry eye syndrome.2–5 It can result
partial glands, and gland dropout) in both upper and in tear film instability, damage to the ocular surface
lower eyelids were developed. Three MGD experts, 3 epithelium, chronic blepharitis, and contact lens
general ophthalmologists, and 3 non-physicians inde- intolerance.6–10 MGD is commonly characterized by a
pendently tested the scales by evaluating photographs. chronic, diffuse abnormality of meibomian glands,
The levels of interrater and intrarater agreement for terminal duct obstruction, and qualitative or quantitative
each grading scale were estimated with the use of kappa changes in the glandular secretion.11 The key signs of
statistics. MGD are meibomian gland dropout, altered meibomian

RESULTS: Thirty-eight patients with MGD and 20 gland secretion, and changes in morphology of the lid
control subjects were enrolled and photographed. New margin. The lid margin abnormalities may become detect-
grading scales were developed using a total of 226 pho- able with a slit-lamp microscope as the disease progresses.12
tographs. The interrater kappa values for MGD experts Lid margin findings and meibomian gland morphology are
and for general ophthalmologists and non-physicians therefore important for diagnosis of MGD. Diagnosis and
with reference to an MGD expert ranged from 0.36 to quantification of MGD thus require assessment of symp-
0.87 (median of 0.66), 0.41 to 0.73 (0.60), and 0.30 toms, altered meibomian gland secretion, changes in lid
to 0.77 (0.59), respectively. Those for intrarater reli- morphology, and meibomian gland dropout.12 Evaluation
ability for 2 MGD experts ranged from 0.49 to 0.93 of meibomian gland expressibility as a dynamic process is
(0.82). also important. Assessment of the efficacy of treatment

CONCLUSIONS: New grading scales for MGD signs for MGD requires precise evaluation of changes in lid
were developed and found to have appropriate inter- morphology, meibomian gland dropout, and meibomian
and intrarater reliabilities for grading MGD. gland expressibility. The ability to perform an objective
These grading scales are suitable for MGD diagnosis evaluation of MGD based on photographs would be useful
and application to multicenter trials. (Am J as a standardized procedure for multicenter clinical trials.
Ophthalmol 2016;169:125–137. Ó 2016 The Grading scales for diagnosis of MGD have been proposed
and adopted in clinical practice.13–18 These scales are based
on assessment of lid margin findings13–15 or of meibomian
Supplemental Material available at AJO.com. glands, with the latter scales being based on the
Accepted for publication Jun 13, 2016. proportion of meibomian glands showing dropout in only
From the Department of Ophthalmology, Graduate School of
Medicine, The University of Tokyo, Tokyo, Japan (R.A., R.S.); the lower tarsal plate16,17 or on the number of whole
Department of Ophthalmology, Itoh Clinic, Saitama, Japan (R.A.); glands and proportion of partial glands.18 The application
Research and Development Division, Santen Pharmaceutical Co Ltd, of these existing scales to the clinic is difficult, however,
Osaka, Japan (I.M., K.A., T.G., T.I., M.N.); Department of
Ophthalmology, Graduate School of Medicine, Yamaguchi University, because there are many grading subdivisions for each sign
Yamaguchi, Japan (N.M.); and Tokyo Kyosai Hospital, Tokyo, Japan or their targets are limited to the lower eyelids. Moreover,
(S.F.). information on grading reliability has been available for
Inquiries to Reiko Arita, Department of Ophthalmology, Graduate
School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo- only a few scales. There is thus still an unmet need for reli-
ku, Tokyo 113-0033, Japan; e-mail: [email protected] able and widely adoptable grading scales based on

0002-9394/$36.00 Ó 2016 THE AUTHOR(S). PUBLISHED BY ELSEVIER INC. 125

http://dx.doi.org/10.1016/j.ajo.2016.06.025
 TABLE 1. Background and Clinical Characteristics of Subjects Evaluated for Establishment of Grading Scales for Meibomian Gland
Dysfunction

Characteristic Controls (N ¼ 20) Initial MGD Patients (N ¼ 38) Additional MGD Patients (N ¼ 18)

Sex (male/female) 8/12 13/25 8/10

Age (y) 64.5 6 6.7 66.9 6 15.0 69.1 6 13.2
BUT (s) 6.5 6 1.6 3.4 6 2.1 2.7 6 1.1
Schirmer test value (mm) 16.4 6 9.9 10.6 6 7.2 7.9 6 3.7
Corneal staining scorea 0.0 6 0.0 0.5 6 0.6 0.8 6 0.4
Conjunctival staining scoreb 0.0 6 0.0 0.3 6 0.6 0.4 6 0.5
Telangiectasiac
Upper 0.1 6 0.3 1.7 6 0.9 1.8 6 0.9
Lower 0.1 6 0.2 1.3 6 0.8 1.3 6 0.6
Pluggingd
Upper 0.1 6 0.2 1.3 6 0.7 1.5 6 0.5
Lower 0.0 6 0.0 1.0 6 0.7 1.1 6 0.4
Meiboscoree
Upper 0.4 6 0.5 1.9 6 0.8 2.0 6 0.8
Lower 0.2 6 0.4 1.9 6 0.9 1.6 6 0.8
Meibum gradef
Upper 0.1 6 0.4 1.5 6 0.9 1.6 6 0.5
Lower 0.2 6 0.5 1.6 6 0.7 1.4 6 0.6
Gland dropoutg
Upper 0.1 6 0.3 1.1 6 0.8 2.0 6 0.8
Lower 0.0 6 0.0 1.1 6 0.8 1.7 6 0.8

MGD ¼ meibomian gland dysfunction.

Data are means 6 SD.
a
Corneal staining was scored 0-3 for the entire cornea.
b
Conjunctival staining was scored 0-3 for each of the nasal and temporal conjunctiva and then summed.
c
Telangiectasia was assessed on a scale of 0-3.
d
Plugging was assessed on a scale of 0-2.
e
Partial or complete loss of meibomian glands was scored 0-3 (meiboscore).
f
Meibum grade was scored 0-3.
g
Complete loss of meibomian glands was scored 0-2.

evaluation of both upper and lower eyelids for the consis- study adhered to the tenets of the Declaration of Helsinki
tent diagnosis of MGD. and was prospectively approved by the Institutional Re-
We have now developed new grading scales for MGD view Board of Tokyo University School of Medicine. All
that can be used by ophthalmologists without special expe- subjects provided written informed consent before entry
rience, and we have performed a validation study to into the study.
confirm the robustness of these scales. We propose that The MGD patients were aged > _20 years and were diag-
these grading scales are suitable for the diagnosis of MGD nosed on the basis of previously described criteria19,20:
or for its evaluation in clinical studies. (1) at least 1 symptom, such as ocular fatigue, discharge,
foreign body sensation, dryness, uncomfortable sensation,
sticky sensation, pain, epiphora, itching, redness, heavy
sensation, glare, excessive blinking, burning sensation,
METHODS and ocular discomfort on arising; (2) at least 1 abnormal
lid margin finding, such as vascular engorgement, anterior

STUDY DESIGN AND TARGET POPULATION: This study or posterior replacement of the mucocutaneous junction,
was conducted at The University of Tokyo Hospital, Itoh and irregular lid margin; and (3) plugged meibomian
Clinic (Saitama City, Saitama, Japan), and Maeda gland orifices and poor meibum expressibility in the
Ophthalmic Clinic (Aizuwakamatsu City, Fukushima, target eye. The control subjects had never been
Japan). MGD patients and control subjects were randomly diagnosed with blepharitis or MGD, were aged > _20 years,
enrolled from outpatients who visited the 3 medical facil- and had no history of contact lens wear or eye surgery.
ities from December 4, 2012 to December 7, 2013. The Individuals with severe systemic illness or with squamous

126 AMERICAN JOURNAL OF OPHTHALMOLOGY SEPTEMBER 2016

cell debris (collarette) around the base of the eyelashes
were excluded. TABLE 2. Distribution of Severity for Lid Margin and
Meibomian Gland Findings in Subjects Evaluated for

CLINICAL ASSESSMENT AND IMAGE COLLECTION: Establishment of Grading Scales for Meibomian Gland
One
Dysfunction
eye was selected as the target eye in each subject. Compli-
cations, history of contact lens wear or eye surgery, the MGD Patients (N ¼ 38) Controls (N ¼ 20)
presence of ocular allergy, and concomitant medications
Parameter Upper Lower Upper Lower
were noted as background information. The subjects were
assessed for lid margin and meibomian gland findings, as Telangiectasia
0 3 (7.9) 4 (10.8) 17 (89.5) 18 (94.7)
well as for their experience of subjective symptoms.
1 14 (36.8) 21 (56.8) 2 (10.5) 1 (5.3)
Lid margin findings were evaluated for the upper and
2 14 (36.8) 8 (21.6) 0 (0) 0 (0)
lower eyelids with the use of a slit-lamp microscope. Tel-
3 7 (18.4) 4 (10.8) 0 (0) 0 (0)
angiectasia was assessed on a scale from 0 to 3: 0 ¼ no Missing data 0 1 1 1
findings; 1 ¼ mild telangiectasia; 2 ¼ moderate telangiec- Mucocutaneous junction
tasia or redness; 3 ¼ severe telangiectasia or redness. 0 3 (8.3) 4 (11.8) 18 (94.7) 18 (94.7)
Mucocutaneous junction was assessed on a scale from 1 12 (33.3) 12 (35.3) 1 (5.3) 1 (5.3)
0 to 3: 0 ¼ Marx line (ML) courses on the skin side of 2 11 (30.6) 12 (35.3) 0 (0) 0 (0)
the meibomian orifice (MO) line and does not touch 3 10 (27.8) 6 (17.6) 0 (0) 0 (0)
MOs at all; 1 ¼ parts of ML touch MOs; 2 ¼ ML courses Missing data 2 4 1 1
through MOs; 3 ¼ ML courses along the eyelid margin Irregularity
side of MOs.21 Irregularity, plugging, foaming, and thick- 0 22 (57.9) 19 (50.0) 18 (94.7) 18 (94.7)
1 14 (36.8) 14 (36.8) 1 (5.3) 1 (5.3)
ness were assessed on a scale from 0 to 2: 0 ¼ no findings;
2 2 (5.3) 5 (13.2) 0 (0) 0 (0)
1 ¼ mild findings; 2 ¼ severe findings.
Missing data 0 0 1 1
Corneal and conjunctival staining were scored from Plugging
0 to 9.22 The tear film breakup time (BUT) was 0 5 (13.2) 9 (23.7) 18 (94.7) 19 (100)
measured 3 times consecutively after the instillation of 1 18 (47.4) 21 (55.3) 1 (5.3) 0 (0)
fluorescein, and the mean value was adopted. Tear fluid 2 15 (39.5) 8 (21.1) 0 (0) 0 (0)
production was evaluated with the Schirmer test without Missing data 0 0 1 1
anesthesia. Foaming
Meibomian glands were evaluated for the upper and 0 37 (97.4) 34 (91.9) 19 (100) 19 (100)
lower lids with the use of a noncontact meibography system 1 1 (2.6) 2 (5.4) 0 (0) 0 (0)
2 0 (0) 1 (2.7) 0 (0) 0 (0)
attached to a slit-lamp microscope. Partial or complete loss
Missing data 0 1 1 1
of meibomian glands was scored on a scale from 0 to 3
Thickness
(meiboscore), as described previously.23 The extent of
0 29 (76.3) 30 (78.9) 19 (100) 19 (100)
meibomian gland dropout was determined on a scale 1 7 (18.4) 7 (18.4) 0 (0) 0 (0)
from 0 to 2 based on the number of affected glands: 0 ¼ 2 2 (5.3) 1 (2.6) 0 (0) 0 (0)
none; 1 ¼ small number; 2 ¼ large number. Missing data 0 0 1 1
The degree of ease with which meibum could be Meiboscore
expressed at the central area of both upper and lower eyelid 0 1 (2.7) 0 (0) 12 (60.0) 16 (80.0)
was evaluated semiquantitatively on a scale from 0 to 3: 0 ¼ 1 11 (29.7) 15 (40.5) 8 (40.0) 4 (20.0)
clear meibum readily expressed; 1 ¼ cloudy meibum 2 17 (45.9) 10 (27.0) 0 (0) 0 (0)
expressed with mild pressure; 2 ¼ cloudy meibum expressed 3 8 (21.6) 12 (32.4) 0 (0) 0 (0)
with more than moderate pressure; 3 ¼ meibum could not Missing data 1 1 0 0

be expressed even with strong pressure.16 MGD ¼ meibomian gland dysfunction.

Images of lid margins and orifices of meibomian glands at Data are n (%).
the upper and lower eyelids of 1 eye were obtained with a
digital camera and meibography system attached to a slit-
lamp microscope (SL-D701 DC4 BG-5; Topcon Japan,
Tokyo, Japan).23 The images of the lid margins and those
of meibomian glands were both acquired at 103 magnifica- were acquired by an ophthalmologist (R.A. or R.S.)
tion for the full length of each eyelid within a single photo- with a specialty in MGD. Four signs for lid margin find-
graph. Four images (2 of each eyelid) were collected for ings detected with a slit-lamp microscope and 2 signs of
each subject. meibomian glands detected by meibography were selected
for development of the grading scales. Three MGD ex-

NEW GRADING SYSTEM: We developed grading scales perts (R.A., R.S., and S.F.), each participating at a
for MGD with the use of printed images. All images different institution, developed draft grading scales for

VOL. 169 RELIABLE GRADING SCALES FOR MEIBOMIAN GLAND DYSFUNCTION 127
FIGURE 1. Distribution of the severity of meibum grade (Left) and meibomian gland dropout (Right) for the enrolled meibomian
gland dysfunction patients (n [ 38). Meibum grade was scored on a scale of 0-3 for each of the upper and lower eyelids and then
summed. Meibomian gland dropout was scored on a scale of 0-2 for each of the upper and lower eyelids and then summed.

TABLE 3. Proposed Grading Scales for Meibomian Gland Dysfunction

Abnormal Lid Margin Findings of Vascularity

0 ¼ No or slight redness in lid margin conjunctiva and no telangiectasia crossing meibomian gland orifices
1 ¼ Redness in lid margin conjunctiva and no telangiectasia crossing meibomian gland orifices
2 ¼ Redness in lid margin conjunctiva and telangiectasia crossing meibomian gland orifices with a distribution of less than half of the full
length of the lid
3 ¼ Redness in lid margin conjunctiva and telangiectasia crossing meibomian gland orifices with a distribution of half or more of the full length
of the lid
Plugging of Gland Orifices
0 ¼ No plugging of gland orifices
1 ¼ Fewer than 3 pluggings of gland orifices
2 ¼ Three or more pluggings of gland orifices with a distribution of less than half of the full length of the lid
3 ¼ Three or more pluggings of gland orifices with a distribution of half or more of the full length of the lid
Lid Margin Irregularity
0 ¼ No lid margin irregularity
1 ¼ Fewer than 3 lid margin irregularities with shallow notching
2 ¼ Three or more lid margin irregularities or deep notching
Lid Margin Thickening
0 ¼ No lid margin thickening
1 ¼ Lid margin thickening with or without localized rounding
2 ¼ Lid margin thickening with diffuse rounding
Partial Glands
0 ¼ No partial glands
1 ¼ Fewer than 3 partial glands
2 ¼ Three or more partial glands and fewer than 3 partial glands with loss of half or more of the full length
3 ¼ Three or more partial glands with loss of half or more of the full length
Gland Dropout
0 ¼ No gland dropout
1 ¼ Fewer than 3 gland dropouts
2 ¼ Three or more gland dropouts

MGD with key conceptual components based on morpho- evaluated with a preliminary validation test in which
logic and anatomic criteria described in previous re- each of the 3 MGD experts classified the printed images
views.13,14 These draft grading scales were then independently according to the draft scales, and the

128 AMERICAN JOURNAL OF OPHTHALMOLOGY SEPTEMBER 2016

FIGURE 2. Representative images of lid margin findings of vascularity. Grade 0 (Upper left), grade 1 (Upper right), grade 2 (Middle
left), and grade 3 (Middle right) are based on the degree of redness in the lid margin and the distribution of telangiectasia crossing
meibomian gland orifices. Note that telangiectasia is observed but does not cross the gland orifices in half of the eyelid in grade 2 (Bot-
tom left), whereas it crosses the gland orifices in half or more of the eyelid in grade 3 (Bottom right).

consistency for each scale was determined. The 3 experts upper and lower eyelids of the original 58 subjects together
then reviewed the results and adjusted the grading scales with printed grading scales and representative images. Each
accordingly. The 3 experts then reviewed the results and printed image had a randomly assigned number for analysis.
adjusted the grading scales accordingly with the advice of Raters individually classified each image according to the
another expert. The final versions proceeded to the grading scales. The test was performed by each rater at a
validation test as the proposed grading scales. separate site. After the test, the assigned numbers of the
classified images for each grading scale were recorded for

VALIDATION TESTING: Validation testing was statistical analysis.
performed to evaluate the robustness of the proposed
grading scales after an interval of >2 weeks since their Interrater and intrarater reliability. Three MGD experts
development. The raters received printed images of the (R.A., R.S., and S.F.) performed the validation test, and

VOL. 169 RELIABLE GRADING SCALES FOR MEIBOMIAN GLAND DYSFUNCTION 129
FIGURE 3. Representative images of plugging of gland orifices. Grade 0 (Upper left), grade 1 (Upper right), grade 2 (Middle left),
and grade 3 (Middle right) are based on the number and distribution of abnormal findings for meibomian gland orifices such as capping,
pouting, and ridge. Arrows indicate plugging (Upper right). Plugging is apparent but shows a distribution of less than half of the lid
length in grade 2 (Bottom left), whereas it is distributed along half or more of the lid length in grade 3 (Bottom right).

consistency among their results for each grading scale was validation test. An introduction and explanation of the
evaluated. grading scales were given before the test by an MGD
Two of the MGD experts who participated in the deter- expert (R.A.). Consistency between the results for each
mination of interrater reliability (R.A. and S.F.) performed group of raters and those for an MGD expert (first test
the test a second time after an interval of >2 weeks. Con- performance by R.A.) was evaluated as interrater
sistency between the results of the first and second tests for reliability.
each grading scale was evaluated.
Additional validation. For confirmation of the intrarater
Effect of clinical experience. Three general ophthalmolo- and interrater reliability of the grading scales, 2 MGD ex-
gists (non–MGD experts) who had been certified for 3, 5, perts (N.M. and Tohru Sakimoto [Nihon University])
or 8 years, as well as 3 non-physicians, also performed the who did not contribute to their development performed

130 AMERICAN JOURNAL OF OPHTHALMOLOGY SEPTEMBER 2016

FIGURE 4. Representative images of lid margin irregularity. Grade 0 (Top left), grade 1 (Middle left), and grade 2 (Bottom left) are
based on the number of irregularities and the form of notching. Notching is not observed in grade 0 (Top right), is shallow in grade 1
(Middle right, arrow), and is deep in grade 2 (Bottom right, arrowheads).

the validation test as described above but with images of used for the calculation. Receiver operating characteristic
the upper and lower eyelids of 18 additional MGD pa- (ROC) curves and the area under the curve (AUC) were
tients that were not used for scale development. Those calculated for each grading scale.
images of additional MGD patients were obtained at
Itoh Clinic.
STATISTICAL ANALYSIS: Data were collected into an
Excel 2010 worksheet and were analyzed with the use of

DIAGNOSTIC ABILITY OF THE PROPOSED GRADING SAS version 9.2 software (SAS Institute, Cary, North Car-
SCALES: For evaluation of the diagnostic ability of the pro- olina, USA). Weighted kappa values and 95% confidence
posed grading scales, the MGD patients and control sub- intervals (CIs) for each grading scale were calculated to
jects were classified on the basis of the validation test evaluate consistency in test performance. Median values
results for an MGD expert (first test performance by were used to summarize the results for weighted kappa
R.A.). The total score of the upper and lower eyelids was values for each test.

VOL. 169 RELIABLE GRADING SCALES FOR MEIBOMIAN GLAND DYSFUNCTION 131
FIGURE 5. Representative images of lid margin thickening. Grade 0 (Upper left), grade 1 (Upper right), and grade 2 (Bottom left)
are based on the presence of thickening and rounding.

RESULTS were evaluated in the full-length images of each eyelid ob-

tained with a slit-lamp microscope. Partial glands
THIRTY-EIGHT MGD PATIENTS (13 MEN AND 25 WOMEN; (Figure 6) and gland dropout (Figure 7) were evaluated
mean age 6 SD, 66.9 6 15.0 years) and 20 control subjects by noncontact meibography for meibomian glands in the
(8 men and 12 women; 64.5 6 6.7 years) were eventually middle two-thirds of each eyelid, given that it is difficult
enrolled. All subjects underwent clinical assessment and to capture and examine glands at the ends of each eyelid
image collection, and their clinical characteristics are in a single photograph (Figure 6).
shown in Table 1. Each grading scale is based on specific features.
For the MGD patients, lid margin and meibomian gland Abnormal lid margin findings of vascularity are based on
findings encompassed the range of severity levels and 2 key components: the degree of redness at the lid margin
appeared to reflect the distribution of findings encountered and the distribution of telangiectasia crossing meibomian
in the clinic (Table 2, Figure 1). For the control subjects, gland orifices. Plugging of gland orifices includes abnormal
almost all findings were normal (Table 2). With the excep- findings for meibomian gland orifices such as capping, pout-
tion of 6 missing images, 4 images acquired from each sub- ing, and ridge and is evaluated on the basis of the number
ject (total of 226 images) were used to develop the new and distribution of these abnormal findings. Lid margin ir-
grading system. regularity is evaluated on the basis of the number of lid
margin irregularities and form of notching. Lid margin

NEW GRADING SYSTEM: We developed 6 grading scales thickening is evaluated based on the presence of thick-
for MGD based on images of the upper and lower eyelids ening and rounding. Partial glands are defined as meibo-
of 38 MGD patients and 20 control subjects (Table 3, mian glands showing partial loss from the orifice or fornix
Supplemental Figure; Supplemental Material available at and are evaluated on the basis of their number and length
AJO.com). Abnormal lid margin findings of vascularity (Figure 6). Gland dropout is evaluated on the basis of the
(Figure 2), plugging of gland orifices (Figure 3), lid margin number of meibomian glands with complete loss from
irregularity (Figure 4), and lid margin thickening (Figure 5) orifice to fornix (Figure 7).

132 AMERICAN JOURNAL OF OPHTHALMOLOGY SEPTEMBER 2016

FIGURE 6. Representative images of partial glands. Grade 0 (Upper left), grade 1 (Upper right), grade 2 (Middle left), and grade 3
(Middle right) are based on the number and extent of partial meibomian glands, which are defined as glands with partial loss between
the orifice and fornix. The scheme illustrates the definition of partial glands and gland dropout (Bottom), both of which are assessed
over the middle two-thirds of each eyelid.

VALIDATION TESTING: Evaluation of interrater reli- (median) of weighted kappa values of 0.36-0.81 (0.68)
ability for performance of a validation test for the 6 for the upper eyelid and 0.57-0.87 (0.65) for the lower
grading scales by 3 MGD experts yielded a range eyelid, as shown in Table 4, referred with agreement.24

VOL. 169 RELIABLE GRADING SCALES FOR MEIBOMIAN GLAND DYSFUNCTION 133
FIGURE 7. Representative images of gland dropout. Grade 0 (Upper left), grade 1 (Upper right), and grade 2 (Bottom left) are based
on the number of meibomian glands showing complete loss from orifice to fornix.

The range (median) of weighted kappa values for eval-
DIAGNOSTIC ABILITY OF THE PROPOSED GRADING
uation of intrarater reliability for 2 MGD experts who SCALES: Finally, with regard to the diagnostic ability
performed the test twice was 0.49-0.92 (0.79) for the of the proposed grading scales, generation of ROC
upper eyelid and 0.67-0.93 (0.85) for the lower eyelid curves revealed that gland dropout showed the greatest
(Table 4). With regard to the effect of clinical experi- AUC with a value of 0.78 (95% CI, 0.66-0.90),
ence, evaluation of interrater reliability yielded a range followed by partial glands (AUC ¼ 0.72; 95% CI,
(median) of weighted kappa values of 0.42-0.66 (0.59) 0.58-0.86), plugging of gland orifices (AUC ¼ 0.70;
for the upper eyelid and 0.41-0.73 (0.65) for the lower 95% CI, 0.55-0.85), lid margin thickening (AUC ¼
eyelid for 3 general ophthalmologists and of 0.35-0.68 0.69; 95% CI, 0.54-0.84), lid margin irregularity
(0.53) for the upper eyelid and 0.30-0.77 (0.63) for (AUC ¼ 0.59; 95% CI, 0.46-0.73), and abnormal lid
the lower eyelid for 3 non-physicians (Table 4). For margin findings of vascularity (AUC ¼ 0.54; 95% CI,
confirmation of the reliability of the grading scales, 0.37-0.71) (Figure 8).
the validation test was performed by 2 additional
MGD experts who did not contribute to their develop-
ment and with images obtained from a different group
of 18 MGD patients that were also not used for scale DISCUSSION
development. The range (median) of weighted kappa
values for evaluation of intrarater reliability for the 2 WE HAVE HEREIN DEVELOPED 6 NEW GRADING SCALES FOR
additional MGD experts who performed the test twice typical clinical findings in MGD and have evaluated the
was 0.47-0.92 (0.65) for the upper eyelid and 0.49- reliability of these scales with the use of 226 images ob-
0.91 (0.80) for the lower eyelid (Table 5). The evalua- tained from 58 subjects. The new grading scales were based
tion of interrater reliability yielded a range (median) of on real clinical findings and were found to be robust and to
weighted kappa values of 0.49-0.92 (0.53) for the upper show better than moderate agreement in most instances on
eyelid and 0.30-0.74 (0.51) for the lower eyelid validation testing. Whereas evaluation of dynamic features
(Table 5). of meibomian gland function such as meibum expressibility

134 AMERICAN JOURNAL OF OPHTHALMOLOGY SEPTEMBER 2016

 and color is informative with regard to assessment of MGD,
such evaluation is difficult to record as dynamic results. The

0.74 (0.60-0.88) 0.86 (0.77-0.95) 0.80 (0.67-0.93) 0.90 (0.80-1.00) 0.81 (0.71-0.91) 0.75 (0.65-0.86) 0.59 (0.43-0.74) 0.69 (0.59-0.80) 0.68 (0.54-0.82) 0.68 (0.55-0.81)

0.49 (0.27-0.71) 0.77 (0.58-0.95) 0.72 (0.56-0.88) 0.67 (0.48-0.85) 0.58 (0.42-0.74) 0.63 (0.41-0.86) 0.59 (0.44-0.74) 0.41 (0.17-0.65) 0.35 (0.18-0.52) 0.30 (0.14-0.45)

0.58 (0.25-0.92)
0.44 (0.32-0.57)
0.68 (0.56-0.81)
0.77 (0.65-0.88)
TABLE 4. Weighted Kappa Values and 95% Confidence Intervals for Intrarater and Interrater Variability in Initial Evaluation of Grading Scales for Meibomian Gland Dysfunction
new grading scales with representative images should thus

Lower
prove helpful for the diagnosis of MGD in a consistent

Kappa values of <0.01, 0.01-0.20, 0.21-0.40, 0.41-0.60, 0.61-0.80, and 0.81-1.00 correspond to poor, slight, fair, moderate, substantial, and almost-perfect agreement, respectively.24
Non-physicians manner and be particularly useful for the evaluation of
MGD in multicenter clinical studies.

0.46 (0.25-0.67)
0.60 (0.46-0.74)
0.44 (0.28-0.59)
0.60 (0.45-0.75)
The reliability of the new grading scales is likely attribut-
able at least in part to the clear definitions and descriptions
Upper

on which they are based, with avoidance of imprecise terms

such as mild, moderate, or severe. The scales thus allow an
0.62 (0.31-0.92)
objective classification of morphologic and anatomic find-
0.46 (0.32-0.61)
0.72 (0.60-0.83)
0.73 (0.59-0.88)
ings. Moreover, the grading scales are applicable to both
Lower

upper and lower eyelids. Given that the degree of morpho-

General Ophthalmologists

logic change of meibomian glands has been found to differ

Interrater

between the upper and lower eyelids in MGD patient and

dry eye patient populations, examination of both eyelids
0.42 (0.19-0.65)
0.60 (0.46-0.74)
0.47 (0.35-0.60)
0.66 (0.51-0.81)

is essential for proper evaluation of the pathologic condi-

Upper

tion of such individuals.25,26

The new grading scales showed a high level of consis-
tency among MGD experts, general ophthalmologists,
0.57 (0.29-0.86)
0.66 (0.50-0.81)
0.64 (0.46-0.82)
0.87 (0.80-0.94)

and non-physicians. These findings indicate that any

ophthalmologist should be able to diagnose MGD and
Lower

classify its severity in a consistent manner with the

MGD Experts

grading scales. This ability is likely to prove especially

useful in clinical studies or trials, with the grading scales
0.36 (0.15-0.57)
0.66 (0.51-0.81)
0.71 (0.50-0.91)
0.73 (0.58-0.88)

allowing investigators at different sites to enroll appro-

Upper

priate subjects and to evaluate the efficacy of medica-

tion in a consistent manner and thus allowing the
results obtained at the different sites to be compared
directly.
0.76 (0.58-0.95)
0.75 (0.62-0.87)
0.87 (0.77-0.98)
0.83 (0.74-0.93)

Our evaluation of the new grading scales revealed that

Lower

consistency among raters was lowest for plugging of gland

MGD Expert 2 (S.F.)

orifices and lid margin irregularity. Plugging of gland ori-

fices is usually assessed by determining the expressibility
0.87 (0.77-0.97)
0.82 (0.70-0.95)
0.86 (0.74-0.97)
0.90 (0.82-0.98)

of meibum, and lid margin irregularity is usually assessed

with a slit-lamp microscope by changing the angle of the
Upper

light relative to the patient’s eyelids. The consistent

grading of these clinical findings based only on images
Intrarater

might thus be expected to be more difficult than that for

0.87 (0.70-1.00)
0.82 (0.70-0.94)
0.86 (0.75-0.97)
0.93 (0.87-1.00)

the other assessed signs.

We also investigated the diagnostic ability of the new
Lower

grading scales for the enrolled MGD patients and con-

trol subjects. Construction of ROC curves and calcula-
tion of AUC values revealed that the scales for both
MGD Expert 1 (R.A.)

0.67 (0.40-0.94)
0.75 (0.62-0.89)
0.77 (0.58-0.95)
0.92 (0.86-0.99)

MGD ¼ meibomian gland dysfunction.

partial glands and gland dropout showed sufficient diag-

Upper

nostic ability, suggesting that both signs are MGD spe-

cific and should be evaluated separately. These
findings are consistent with the results of a previous
study.20 On the other hand, the AUC for abnormal
findings of vascularity

Lid margin irregularity

Lid margin thickening

lid margin findings of vascularity was lowest among

Abnormal lid margin

Plugging of gland

the grading scales. The grading scale for this sign ap-
Gland dropout
Partial glands

pears to be affected not only by MGD but also by aging

Grading Scale

orifices

or other factors, such as workplace exposure to dust par-

ticles, urban living, and cosmetics.15 As far as we are
aware, this is the first study to evaluate diagnostic ability
of partial glands and gland dropout separately on the

VOL. 169 RELIABLE GRADING SCALES FOR MEIBOMIAN GLAND DYSFUNCTION 135
 TABLE 5. Weighted Kappa Values and 95% Confidence Intervals for Intrarater and Interrater Variability in Additional Evaluation of
Grading Scales for Meibomian Gland Dysfunction

Intrarater Interrater

Additional MGD Expert 1 (N.M.) Additional MGD Expert 2 (T.S.) Additional MGD Experts

Grading Scale Upper Lower Upper Lower Upper Lower

Abnormal lid margin 0.92 (0.83-0.99) 0.91 (0.83-0.99) 0.66 (0.52-0.80) 0.84 (0.73-0.95) 0.92 (0.83-0.99) 0.58 (0.43-0.73)
findings of vascularity
Plugging of gland orifices 0.74 (0.61-0.88) 0.84 (0.73-0.95) 0.47 (0.31-0.66) 0.70 (0.57-0.83) 0.50 (0.33-0.66) 0.30 (0.15-0.45)
Lid margin irregularity 0.82 (0.65-0.99) 0.80 (0.67-0.93) 0.64 (0.31-0.97) 0.49 (0.26-0.72) 0.49 (0.30-0.68) 0.72 (0.54-0.91)
Lid margin thickening 0.52 (0.34-0.70) 0.49 (0.32-0.66) 0.50 (0.29-0.71) 0.80 (0.67-0.94) 0.55 (0.39-0.72) 0.41 (0.25-0.56)
Partial glands 0.72 (0.60-0.85) 0.80 (0.66-0.93) 0.50 (0.34-0.66) 0.74 (0.61-0.87) 0.52 (0.37-0.67) 0.43 (0.28-0.57)
Gland dropout 0.82 (0.71-0.93) 0.81 (0.68-0.94) 0.49 (0.33-0.65) 0.82 (0.71-0.94) 0.54 (0.40-0.68) 0.74 (0.61-0.87)

MGD ¼ meibomian gland dysfunction.

Kappa values of <0.01, 0.01-0.20, 0.21-0.40, 0.41-0.60, 0.61-0.80, and 0.81-1.00 correspond to poor, slight, fair, moderate, substantial, and
almost-perfect agreement, respectively.24

of MGD characteristics. MGD is one of the most com-

mon causative conditions of dry eye.27 Application of
the grading scales to both eyes of an individual would
provide information on the bilaterality of MGD or dry
eye.
There is a general consensus regarding the definition of
MGD severity based on meibum score, symptoms, and lid
margin findings.11,14 However, the ability to evaluate
MGD severity more accurately with reliable objective
scales based on standardized criteria is needed. Our
proposed new grading system can be used to define MGD
severity. It should be possible to classify MGD patients
according to disease severity with the use of such a
grading system, including multiple static and dynamic
evaluation procedures.
There are some limitations to the present study. First,
the study was based on static evaluation of photographs
and thus did not take into account dynamic observations
such as meibum expressibility, including meibum quality
FIGURE 8. Receiver operating characteristic curves for the 6 and quantity. Given that evaluation of meibum expressi-
new grading scales. bility is one of the most informative procedures regarding
the pathogenesis of MGD, movie-based assessment of
meibum expressibility and other dynamic features should
be performed in the future. Second, the new grading
basis of AUC values. A previous study examined the scales were developed with the use of images obtained
diagnostic ability of the total score for 4 eyelid signs in a cross-sectional analysis. We were therefore not able
(vascular engorgement, plugged meibomian gland ori- to evaluate whether they are suitable for detection of
fices, irregular lid margin, and anterior or posterior changes in clinical findings. It will be necessary to confirm
replacement of the mucocutaneous junction) by calcula- that the grading scales are appropriate for evaluation of
tion of AUC values.21 We evaluated the AUC for each medical treatment in longitudinal studies. Third, we
of these signs (with the exception of mucocutaneous also evaluated the reliability of the grading scales with
junction) separately. These results can provide useful the same set of images. Given that plugging of gland ori-
information for interpretation of scores obtained with fices and lid margin irregularity can be assessed accurately
the grading scales in clinical practice. only by physical examination with observation in 3 di-
In the present study, we evaluated only 1 eye of each mensions, the assessment of these signs from 2-dimen-
subject; therefore we are unable to address the bilaterality sional images may be problematic. There is thus still

136 AMERICAN JOURNAL OF OPHTHALMOLOGY SEPTEMBER 2016

room for improvement of the proposed clinical scoring for In conclusion, we have developed new grading scales for
MGD, but, despite its basis in static evaluation, it is MGD that are sufficiently reliable for use by any ophthal-
potentially sufficient for clinical application. Its further mologist. Further studies are warranted to determine
improvement will be required for it to become a standard whether the grading scales are suitable for evaluation of
procedure for MGD diagnosis. the efficacy of MGD management.

FUNDING/SUPPORT: THIS STUDY WAS SUPPORTED BY SANTEN PHARMACEUTICAL CO LTD (OSAKA, JAPAN). THE SPONSOR
participated in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article
for publication. Financial disclosures: Reiko Arita has received personal fees from Santen Pharmaceutical Co Ltd (Osaka, Japan), Otsuka Pharmaceutical
Co Ltd (Tokushima, Japan), Kiribai Chemical Co Ltd (Osaka, Japan), Kao Corporation (Tokyo, Japan), Japan Focus Company Ltd (Tokyo, Japan),
TOPCON Japan (Tokyo, Japan), and TearScience (Morrisville, NC). Naoyuki Morishige has received personal fees from Santen Pharmaceutical Co
Ltd (Osaka, Japan), Otsuka Pharmaceutical Co Ltd (Tokushima, Japan), Kao Corporation (Tokyo, Japan), Japan Focus Company Ltd (Tokyo, Japan),
Pfizer Japan Inc (Tokyo, Japan), Wakamoto Co Ltd (Tokyo, Japan), and Astellas Pharma Inc (Tokyo, Japan). Shima Fukuoka has received personal
fees from Santen Pharmaceutical Co Ltd (Osaka, Japan). Izumi Minoura, Kei Asai, Tateki Goto, Takahiro Imanaka, and Masatsugu Nakamura are em-
ployees of Santen Pharmaceutical Co Ltd (Osaka, Japan). The following author has no financial disclosures: Rika Shirakawa. All authors attest that
they meet the current ICMJE criteria for authorship.

REFERENCES 15. Hykin PG, Bron AJ. Age-related morphological changes in

lid margin and meibomian gland anatomy. Cornea 1992;
1. Mishima S, Maurice DM. The oily layer of the tear film and evap- 11(4):334–342.
oration from the corneal surface. Exp Eye Res 1961;1:39–45. 16. Shimazaki J, Goto E, Ono M, et al. Meibomian gland dysfunc-
2. Lemp MA. Report of the National Eye Institute/Industry tion in patients with Sjogren syndrome. Ophthalmology 1998;
Workshop on Clinical Trials in Dry Eyes. CLAO J 1995; 105(8):1485–1488.
21(4):221–232. 17. Pflugfelder SC, Tseng SC, Sanabria O, et al. Evaluation of
3. Mathers WD. Ocular evaporation in meibomian gland subjective assessments and objective diagnostic tests for diag-
dysfunction and dry eye. Ophthalmology 1993;100(3):347–351. nosing tear-film disorders known to cause ocular irritation.
4. Shimazaki J, Sakata M, Tsubota K. Ocular surface changes Cornea 1998;17(1):38–56.
and discomfort in patients with meibomian gland dysfunc- 18. Nichols JJ, Berntsen DA, Mitchell GL, Nichols KK. An
tion. Arch Ophthalmol 1995;113(10):1266–1270. assessment of grading scales for meibography images. Cornea
5. Lee SH, Tseng SC. Rose bengal staining and cytologic char- 2005;24(4):382–388.
acteristics associated with lipid tear deficiency. Am J Ophthal- 19. Amano S, Arita R, Kinoshita S, et al. The Japan Dry
mol 1997;124(6):736–750. Eye Society Meibomian Gland Dysfunction Working
6. Mathers WD, Shields WJ, Sachdev MS, et al. Meibomian Group. Definition and diagnostic criteria for meibomian
gland dysfunction in chronic blepharitis. Cornea 1991; gland dysfunction. Atarashii Ganka (J Eye) 2010;27(5):
10(4):277–285. 627–631.
7. Korb DR, Henriquez AS. Meibomian gland dysfunction and 20. Arita R, Itoh K, Maeda S, et al. Proposed diagnostic criteria
contact lens intolerance. J Am Optom Assoc 1980;51(3): for obstructive meibomian gland dysfunction. Ophthalmology
243–251. 2009;116(11):2058–2063.
8. Henriquez AS, Korb DR. Meibomian glands and contact lens 21. Yamaguchi M, Kutsuna M, Uno T, et al. Marx line: fluores-
wear. Br J Ophthalmol 1981;65(2):108–111. cein staining line on the inner lid as indicator of meibomian
9. Zengin N, Tol H, Gunduz K, et al. Meibomian gland dysfunc- gland function. Am J Ophthalmol 2006;141(4):669–675.
tion and tear film abnormalities in rosacea. Cornea 1995; 22. Van Bijsterveld OP. Diagnostic tests in the sicca syndrome.
14(2):144–146. Arch Ophthalmol 1969;82(1):10–14.
10. Ong BL. Relation between contact lens wear and meibomian 23. Arita R, Itoh K, Inoue K, Amano S. Noncontact infrared
gland dysfunction. Optom Vis Sci 1996;73(3):208–210. meibography to document age-related changes of the meibo-
11. Nelson JD, Shimazaki J, Benitez-del-Castillo JM, et al. The mian glands in a normal population. Ophthalmology 2008;
International Workshop on Meibomian Gland Dysfunction: 115(5):911–915.
Report of the Definition and Classification Subcommittee. 24. Landis JR, Koch GC. The measurement of observer agree-
Invest Ophthalmol Vis Sci 2011;52(4):1930–1937. ment for categorical data. Biometrics 1977;33:159–174.
12. Tomlinson A, Bron AJ, Korb DR, et al. The International 25. Pult H, Riede-Pult BH, Nichols JJ. Relation between upper
Workshop on Meibomian Gland Dysfunction: Report of and lower lids’ meibomian gland morphology, tear film, and
the Diagnosis Subcommittee. Invest Ophthalmol Vis Sci dry eye. Optom Vis Sci 2012;89(3):E310–E315.
2011;52(4):2006–2049. 26. Koh S, Ikeda C, Fujimoto H, et al. Regional differences in tear
13. Bron AJ, Benjamin L, Snibson GR. Meibomian gland disease. film stability and meibomian glands in patients with aqueous-
Classification and grading of lid changes. Eye 1991;5(4): deficient dry eye. Eye Contact Lens 2016;42(4):250–255.
395–411. 27. Lemp MA, Crews LA, Bron AJ, Foulks GN, Sullivan BD. Dis-
14. Foulks GN, Bron AJ. Meibomian gland dysfunction: a clin- tribution of aqueous-deficient and evaporative dry eye in a
ical scheme for description, diagnosis, classification and clinic-based patient cohort: a retrospective study. Cornea
grading. Ocul Surf 2003;1(3):107–126. 2012;31(5):472–478.

VOL. 169 RELIABLE GRADING SCALES FOR MEIBOMIAN GLAND DYSFUNCTION 137