Meibomian gland dysfunction is highly prevalent among first-time visitors at a Norwegian dry eye specialist clinic

To investigate the prevalence of meibomian gland dysfunction (MGD) in patients presenting with subjective dry eye-related symptoms at their first-time consultation in a Norwegian specialized ocular surface clinic. Additionally, to explore the accuracy of the ocular surface disease index score (OSDI) as an extensively applied tool to assess the severity of dry eye symptoms and MGD diagnosis. Patients with subjective dry eye-related complaints (n = 900) attending the clinic for the first time, from 2012 to 2016, were included in the study. At the baseline, patients completed the OSDI questionnaire. Subsequently, objective clinical tests, including fluorescein break-up time (FBUT), Schirmer-I test, ocular surface staining (OSS), and meibomian gland function assessment using gland expressibility and meibum quality were performed. The association between MGD and its severity in relation to symptom severity defined by OSDI-score was examined. MGD was found in 93.8% of the study group. MGD prevalence was not significantly different between groups based on age (p = 0.302) or sex (p = 0.079). There was a significant association between severity of MGD and dry eye-related symptoms (p = 0.014). OSS was significantly higher in patients with severe symptoms (p = 0.031). Sensitivity and specificity of positive symptom-score (OSDI ≥ 13) for disclosing MGD were 85.5% and 30.4%, respectively. MGD was highly prevalent, not associated with age and sex. OSDI ≥ 13 had high sensitivity and high positive predictive value (PPV), but low specificity and negative predictive value (NPV) for disclosing MGD. This underscores the importance of meibomian gland assessment in patients with dry eye-related symptoms.

www.nature.com/scientificreports/ prevalence worldwide, characterized by using different diagnostic definitions to diagnose MGD. In the US study of Salisbury, prevalence of dry eye was 3.5%, while among those with dry eye symptoms and signs 20.7% had MGD mentioned in the study as meibomititis 25 .
The prevalence of MGD is high in Asian populations [26][27][28] . In a Chinese study, the prevalence of subjective dry eye symptoms was 21%, however, using the presence of telangiectasia at the lid margin as definition of MGD resulted in a prevalence of 69% 26 . An Indonesian population-based study of subjects of Malay ethnicity found MGD prevalence to be 56.3% 27 . In a Japanese clinic-based, study MGD prevalence among patients scheduled for cataract surgery was 74.5% 28 . Reported MGD prevalence in European (21.9%) 29 and U.S. (38.9%) populations is markedly lower 20 . Notably, the diagnostic criteria used to define and diagnose MGD across these studies varies considerably 16 . In this study, MGD diagnosis was made according to the report of the International Workshop on Meibomian Gland Dysfunction from 2011 11 .
The specialized ocular surface clinic in Oslo is unique in the Nordic region by being dedicated to diagnosis and treatment of patients with dry eye-related symptoms. To our knowledge, no studies have explored how prevalent MGD is in Norwegian patients with dry eye symptoms. The aim of this study was to explore how prevalent MGD is among the first-time visitors to the clinic and to explore the accuracy of symptoms in identifying MGD.

Material and methods
Study population, inclusion and exclusion criteria. This study had a cross-sectional design. Patients with dry eye-related symptoms at their baseline visit to the Norwegian specialized ocular surface clinic between January 2012 to January 2016, either self-referred, referred by optometrists, ophthalmologists, or general practitioners, were consecutively recruited to the study group. Patients who attended the clinic having followed their habitual dry eye treatment, either self-initiated or optometrists, ophthalmologists or general practitioners. Follow-up visitors or patients with a history of any ocular surgery in the past 12 months, conditions including ocular infections, or ocular allergy were not included.
Prevalence. Prevalence is a frequency measure of morbidity that is defined as the proportion or rate of individuals who have a particular disease or condition at or during a particular time period. The frequency measure for the prevalence used here was proportion 30,31 . In this study, we defined the prevalence of MGD as the proportion of patients with MGD of the total number of patients in the study group, in the group as a whole or in age-and sex-stratified groups (Table 1) and in each symptom-category (Table 2).
Tear film stability was assessed by measuring FBUT after instillation of 5 µL of 2.0% fluorescein dye into the conjunctival sac with a micro-pipette, with values ≤ 10 s defined as unstable tear film 32,36 . The staining of interpalpebral cornea and conjunctiva with fluorescein constituted ocular surface staining (OSS) was used to evaluate ocular surface damage, which was graded according to the Oxford grading scheme, ranging on a scale from 0 to 15 32,37 . Tear production was assessed by performing the Schirmer-I test (without anesthesia).Wetting of the Schirmer strip after 5 min was measured, and values < 10 mm were defined as abnormal 32,36 .
Meibomian gland expressibility (ME) and meibum quality (MQ) were evaluated under a slit-lamp microscope by applying firm pressure using a cotton-tipped applicator onto the lower lid margin. ME was graded on a 4-point scale based on the number of expressible glands in the central five glands: grade 0, all five glands expressible; grade 1, 3-4 glands expressible; grade 2, 1-2 glands expressible; and grade 3, 0 glands expressible 11 . MQ was graded on a scale from 0 to 3 by examining the secretion of central eight meibomian glands in the lower lid: grade 0: clear meibum; grade 1: cloudy meibum; grade 2: cloudy with particles; and grade 3: inspissated or toothpastelike meibum. A sum score for the central eight glands (range 0-24) was then calculated 11 . The severity of MGD was graded according to MQ and ME scores: grade 0, no MGD; grade 1, ME score = 1 and/or MQ ≥ 2-4; grade 2, ME score = 1 and/or MQ ≥ 4-8; grade 3, ME score = 2 and/or MQ ≥ 8 but < 13; and grade 4, ME score = 3 and/ or MQ ≥ 13 38 . The criteria for MGD diagnosis in patients aged ≤ 20 years was score > 1 for either MQ or ME; and in patients aged > 20 years, a score of 1 for both MQ and ME or a score > 1 for either MQ or ME 11,38 . OSDI-score as described above, was used as an instrument to assess the severity of dry eye-related symptoms 34,35 . Thus, we examined the accuracy of OSDI-score (≥ 13) by measuring its sensitivity and specificity in relation to disclose MGD that was the "disease" in focus for this study. It is worth underscoring that OSDI is not a tool to diagnose MGD rather to assess symptoms. OSDI measures dry eye-related symptoms. On this basis, the accuracy of OSDI measuring symptoms vis-à-vis presence or absence of MGD diagnosis was investigated, using binary classification of patients according to absence or presence of dry eye symptoms (normal symptom load OSDI < 13 and symptomatic OSDI ≥ 13) versus absence or presence of MGD. On this basis, using a contingency table reporting binary classification of patients according to the OSDI score (normal symptom load/OSDI < 13, and symptomatic category: OSDI ≥ 13) relative to absence or presence of MGD (Table 3). The sensitivity and specificity of OSDI score ≥ 13 in revealing MGD were determined. Subsequently, positive and negative predictive values (PPV, and NPV, respectively) were calculated based on the MGD prevalence 39 . PPV is defined as the probability of the subject/patient having MGD given that the OSDI score ≥ 13, that is having mild, moderate and severe dry eye symptoms. NVP is defined as the probability of the patient/subject not having MGD given that the OSDI score < 13, that is the patient having a normal symptom load. The formula and calculations of PPV and NPV are detailed in Eqs. (1) and (2), respectively. Statistical analysis. Statistical analysis was performed using the Statistical Package for the Social Sciences (IBM SPSS Statistics, 24.0). The data are reported as frequencies, percentages with 95% confidence intervals (CIs), and means with standard deviation (SD). Pearson´s chi-square (χ 2 ) tests were used for testing associations between categorical groups; MGD, MGD severity, sex, age, symptomatic and asymptomatic OSDI, and sever-

Results
Demographics. In total, 1027 patients were eligible for the study; complete data, including OSDI, ME, and MQ was available in 900 patients (87.6%) and were included in the analyses. The mean age of patients was 52 ± 16.7 years, 74.2% of patients (n = 668) were female. Figure 1 shows the age distribution in the study group and Table 1 presents the distribution of study subjects based on age and sex.
Overall MGD prevalence and laterality of MGD. In

Relationship between OSDI-symptom severity categories and clinical signs; FBUT, OSS, and
Schirmer. The mean values of FBUT, OSS, and Schirmer-I test for both eyes of all patients by the four OSDI categories are presented in Table 6. There was a statistically significant difference for OSS between the groups (p = 0.031). However, pairwise post hoc analysis using Dunn's test with Bonferroni correction failed to identify      www.nature.com/scientificreports/

Discussion
The vast majority of patients in this clinic-based study population had MGD. MGD was not associated with either age or sex. However, there was a significant association between the presence of MGD and subjective dry eye symptoms. Moreover, MGD severity and dry eye symptom severity also showed significant association. Population-based studies of MGD prevalence have reported conflicting results with respect to age and sex. Whereas some studies have found MGD to be more prevalent in men 27,29 , others have not supported the association with sex 25 . Our results are in agreement with the latter study, as we did not find a relationship between MGD prevalence and sex. This result conflicts with findings in an Austrian dry eye clinic population, reporting MDG more frequently in females 40 .
The differences in prevalence can reflect population differences, and partly be because the MGD criteria used in different prevalence studies vary significantly. MGD prevalence studies, conducted as population-based studies, have shown substantial variation in MGD prevalence 20,[25][26][27]29,41 . MGD is often clinically defined based on evaluation of meibomian gland obstruction, gland dropout, and abnormal gland secretions. In 26 studies reviewed by the International Workshop on MGD clinical trials subcommittee, 53.8% contained meibomian gland secretion assessment, 50% included the symptoms associated with DED, as entry, diagnostic or outcome criteria, and lid abnormalities such as telangiectasia was reported in 38.5% 42 .
The high prevalence of MGD in the present study likely reflects the sample population and diagnostic criteria. The study was conducted in a clinic-based population of first-time visitors with primarily subjective dry eyerelated complaints and not in a general population; hence, a higher prevalence of MGD is expected. Nevertheless, the prevalence was higher than found in an Austrian dry eye clinic population 40 , a result that likely reflects variations in diagnostic criteria. In the present study, the diagnostic criteria for MGD were based on meibum expressibility and quality according to the international workshop on meibomian gland dysfunction: report of the diagnosis subcommittee 11,38 , whereas the Austrian study also included morphological changes as part of the diagnostic criteria.
Age is a risk factor of MGD 16 . In our study, age was not associated with MGD, possibly because the population were patients seeking help because of dry eye symptoms. However, the age distribution in our study was skewed, with the majority of patients with MGD (75%) being older than 40 years, corresponding with the age distribution found in the Austrian dry eye clinic study 40 .
In this research, MGD was significantly associated with the presence of dry eye symptoms (OSDI score ≥ 13) and symptom severity. MGD is the most common cause of evaporative dry eye disease 16,43,44 , resulting in dry eye-related symptoms 29,[45][46][47] , and symptomatic MGD is characterized by symptoms of ocular discomfort such as irritation, soreness, redness of the eyes and eyelids, irritation, burning, itching of the eye, dryness, heavy or/ puffy eyelids, and watery eyes 2,10,11,21,48-52 . The prevalence of MGD among first-time visitors to the Norwegian dry eye clinic reflects this pattern. However, the MGD prevalence was also high in the subgroup with normal symptom load. Of the total study subjects, 15.4% had a normal symptom load; nevertheless, from these patients' perspectives, the symptoms were sufficiently severe to seek ophthalmological help. Consequently, high MGD prevalence even in the OSDI subgroup with normal symptom range (87.8%) indicates the substantial role of MGD in patients with dry eye-related symptoms of all symptom severities and multiple etiologies.
Some studies have indicated that asymptomatic MGD is more common than symptomatic MGD 28,29 and most individuals with anatomical features of MGD are asymptomatic 25 . In Amano and colleagues' research conducted in a Japanese population, the overall MGD prevalence was 74.5%, whereas the prevalence of symptomatic vs asymptomatic MGD was 11.2% vs 63.3% 28 , respectively. These observations corresponds with studies that have indicated a large proportion of asymptomatic patients have some degree of MGD 28,29 . Patients with undetected MGD are at risk of gradual and continuous deterioration and progression to a chronic phase, with potentially irreversible changes in the anatomy and function of the meibomian glands. Consequently, deterioration of tear composition and quality, with aggravation of DED symptoms will follow 11,45,[53][54][55] . Chronic and irreversible MGD is a major etiological factor of DED, which imposes substantial economic burdens on both patients and society 56 . The high percentage of patients with normal symptom load diagnosed with MGD in our study supports the premise that a substantial percentage of patients with MGD are asymptomatic, and may remain undetected, undiagnosed, and untreated. Dry eye-specific work-up should include assessment of meibomian gland function to detect MGD regardless of the symptoms, specifically to detect MGD in patients in the normal symptom load category.
DED is a multifactorial disorder of tears and the ocular surface 8 , and the two major etiological causes of dry eye disease are aqueous-deficient and evaporative dry eye, the former characterized by reduced tear volume measured by Schirmer-I test, and the latter by decreased tear film stability measured by FBUT 8,9 . The mean value of the Schirmer-I test in all OSDI groups was over the pathological cut-off value of 10 mm/5 min. However, a broad range of values was detected, indicating that aqueous deficiency was not the predominant objective clinical sign in our study population. Furthermore, FBUT, which indicates the stability of the tear film, was below the pathological cut-off of < 10 s in all OSDI groups. It is important to note that mild and moderate dry eye patients may show a broad range of FBUT values 57 . FBUT decreased with increasing OSDI severity, with the mean FBUT being lowest in the severe OSDI symptom group, although the intergroup difference did not reach statistical significance. Moreover, OSS was not significantly different between groups of patients with varying dry eye symptom severity. Considering that MGD is the most common cause of evaporative dry eye 2,14,15 , the very high MGD prevalence in the study supports the evaporative etiology as the more prominent cause of DED in the study population.
The present study highlights the poor diagnostics of patient self-reported symptoms. Dry eye symptoms (OSDI-score ≥ 13) correctly identified 85% of the patients with MGD. The predictive values of a test are dependent on the prevalence of the condition. The increased prevalence of MGD causes a high positive predictive value