This study was approved by the Ethics Committee of Xiamen Eye Center, an affiliate of Xiamen University. All methods below were performed in accordance with the relevant guidelines and regulations. Written informed consent was obtained from participants in this study in accordance with the Declaration of Helsinki and its subsequent revision. Informed consent for online open-access publication of images or information from participants was also obtained.

Ninety-nine patients who were diagnosed with type 2 diabetes [according to the 1999 WHO diagnostic criteria (10) for diabetes] were recruited as the diabetes mellitus group (DM group). Thirty-three patients without diabetes who were diagnosed with dry eye (11–13) [Ocular Surface Disease Index (OSDI)] ≥ 13 and tear film breakup time (BUT) < 5 s were recruited as the dry eye group (DE group). Meanwhile, 40 normal individuals were recruited as normal control (NC group). The subjects of the NC group have no history of ocular disease or systemic pathology that could affect the test results, have no history of topical medication within the last 6 months, and have no signs or symptoms of ocular surface disease. All the examinations were performed during the period from 1 October, 2016 to 31 December, 2016 in the Affiliated Xiamen Eye Center of Xiamen University, Xiamen, China. Patients with blepharitis, allergic conjunctivitis, infectious keratoconjunctivitis, pterygium, eye trauma, eye operation, eye continuous medication history, or corneal contact lens wearing history were excluded from this study. Meanwhile, patients with systemic diseases, such as Sjogren’s syndrome, Stevens–Johnson syndrome, hyperthyroidism, and pemphigoid, were excluded.

The degree of ocular discomfort was evaluated by an OSDI questionnaire in all subjects. OSDI (14) evaluates 12 symptoms, and every score was calculated according to the duration of the symptoms with 100 points in total. The scale is graded as follows: normal: 10–12 points, mild: 13–22 points, moderate: 23–33 points, and severe: 33–100 points. Subjects with dry eye have at least mild ocular surface symptomology, which means OSDI ≥ 13 points (15).

A series of dry eye parameters were performed in the following order: lipid layer thickness (LLT), tear breakup time (BUT), corneal fluorescein staining scoring (FL), and Schirmer I test (SIT). MGD parameters were evaluated. All the tests were carried out by the same masked investigator. Only the values of one eye (right or left eye, randomly selected by simple randomization using a table of random digit) were included in this study. The HbA_1c level was also tested in all subjects.

The tear LLT test was performed using an interferometer (Lipiview™, TearScience Inc., Morrisville, NC, United States). The patient was asked to blink naturally when tested, and tear film images were captured for 30–60 s (16).

For the BUT test, the sodium fluorescein filter paper was applied to the center of the lower eyelid causing fluorescein sodium to flow into the conjunctival sac. After blinking for three times, the test was conducted by inspecting with a slit lamp under low brightness of a cobalt blue light (BQ900IM9900, Haag-Streit, Switzerland). The average tear BUT of three repeated measurement was considered the BUT of one eye (seconds). Corneal FL was scored under cobalt blue light by the fluorescent paper after a corneal surface of patient was stained for 90 s as mentioned in the BUT test. Cornea was graded as follows: 0: no staining, 1: slight scatted staining, 2: moderate staining between 1 and 3, and 3: severe staining (17).

For SIT, the tear filter paper strip was placed in the conjunctival sac about 1/3 beneath the eyelid center, and the length of filter strip wetting (mm) was read after 5 min.

Lid margin abnormalities (lid margin score) were assessed with a slit lamp as follows (18): irregular lid margins, score 1; vascular engorgement, score 2; glandular orifice obstruction, score 3; and anterior or posterior displacement of the mucocutaneous junction, score 4. No presence of above 4 signs was indicated as score 0.

The meibomian gland yielding liquid secretion score (MGYLS score) was graded under the diffusing light of a slit lamp. A constant pressure about 20 kPa was applied to the lower eyelid using a meibomian gland evaluator (MGE, TearScience Inc., Morrisville, NC, United States) for 10 s, and then the orifice opening of five consequent glands where pressure was applied were evaluated. The number of obstructed openings of the total 15 glands in three different locations (nasal, central, and temporal sides) was inspected and recorded. The MGYLS score of each eye was the average number of obstructed glands of the upper and lower eyelids (15 points in total). The level of MGYLS was graded as follows: grade 0, 0–3 points; grade 1, 4–7 points; grade 2, 8–11 points; and grade 3, 12–15 points.

The qualitative appearance of lipid secretion of each meibomian gland (meibum score) was graded as follows: 0 point (clear or slight yellow liquid secretion); 1 point (creamy yellow or cloudy liquid secretion); 2 points (granular in liquid with white and/or yellow color); and 3 points (toothpaste shape) (19). The total 15 glands in three different locations (nasal, central, and temporal sides) were inspected and recorded. The total score of each eye was 0–90, including the upper and lower eyelids. The level of the meibum score was graded as follows: grade 0, 0–3 point; grade 1, 4–10 points; grade 2, 11–30 points; grade 3, 31–60 points; and grade 4, 61–90 points.

The loss of meibomian glands (meiboscore) was examined using a non-contact infrared meibography (20, 21) (Keratograph 5M, OCULUS, Wetzlar, Germany). Meiboscore was graded as follows: 0 point (no absence); 1 point (absence of less than 1/3 of total glands); 2 points (absence of more than 1/3 but less than 2/3 of total glands); and 3 points (absence of more than 2/3 of total glands). Upper and lower eyelids are 0–6 points in sum. Examinations were performed by the same masked ophthalmologist.

The subject was diagnosed as MGD according to the criteria briefly as follows (18, 22): the presence of lid margin abnormality (lid margin score ≥ 1), and/or an altered quality of expressed secretions (meibum grade ≥ 1), and/or a decreased or absent expression (MGYLS grade ≥ 1). The subject with MGD could be further diagnosed as “asymptomatic MGD” if he/she had no symptoms (OSDI < 13) (23, 24).

All data were analyzed using software SPSS 26.0 (SPSS Inc., Chicago, IL, United States). Measurement data were presented as mean and standard deviation. Difference of sex ratio was evaluated by the chi-square test. Differences in other parameters among groups were analyzed by using the one-way ANOVA test. The Dunn–Bonferroni test was used to make a pairwise comparison between two groups. For non-parametric data (corneal FL score and meibomian gland parameters), a Kruskal–Wallis ANOVA test along with the post hoc test was used for multiple comparisons. The correlation between the HbA_1c level and various factors was estimated by the Spearman rank correlation analysis. For the secondary analysis, patients with diabetes (DM group) were further subgrouped according to the dry eye condition, the HbA_1c level, and the duration of diabetes. A multivariate analysis was performed to identify the variables that most correlated with any of the meibomian gland parameters in the DM group. The differences were regarded statistically significant when the value of p was <0.05.

The tear film parameters and the meibomian gland parameters of subjects are shown in Table 2. The OSDI score of the NC group was significantly lower than those of both the DE group and the DM group (p < 0.001). Meanwhile, the OSDI score of the DM group was also significantly lower than that of the DE group (p = 0.016). Furthermore, the average SIT, BUT, and tear LLT values in both DE and DM groups were all significantly lower (p < 0.001) than those in the NC group, whereas the values of meibomian gland parameters (except lid margin score) were all higher (p < 0.001) than those in the NC group (Table 2).

The average SIT value in the DM group (9.59 ± 3.17 mm) was significantly higher than that in the DE group (5.14 ± 2.12 mm).

The four MGD parameters in the DM group were all significantly higher (p < 0.001) than those of the other two groups (Table 2). Although the average SIT value of the DM group is higher than that of the DE group, the severity of MGD is more apparent in the DM group. Interestingly, all the subjects in DE and DM groups could be eventually diagnosed as MGD according to the diagnostic criteria.

The detailed distribution of meibomian gland parameters in different groups is shown in Figure 1. Twenty-six patients (26.3%) in the DM group had a high lid margin score (score ≥ 3), whereas two patients (6.1%) in the DE group and only one patient (2.5%) in the NC group had high score. A high level of the MGYLS grade (grade ≥ 2) was observed in 35 patients (35.4%) in the DM group, whereas only two patients (6.1%) in the DE group and one patient (2.5%) in the NC group were observed. High meibum grade (grade ≥ 3) was seen in 38 patients (38.4%) in the DM group, whereas only four patients (12.1%) in the DE group and only one patient (2.5%) in the NC group were seen. A high meiboscore (score ≥ 3) was seen in 82 patients (82.8%) in the DM group and 24 patients (72.7%) in the DE group, whereas only 17 patients (42.5%) in the NC group were seen.

According to the recently published article (11), the existence of ocular symptoms and BUT for less than 5 s was proposed as the diagnostic criteria for dry eye. Based on the values of OSDI (≥13) and BUT (<5 s), 47 patients with diabetes (47.5%) in the DM group that were diagnosed with dry eye were marked as the DM + DE group, whereas 40 patients (40.4%) without dry eye served as the DM − DE group (OSDI < 13 and BUT > 5 s). Interestingly, the SIT and LLT in the DM − DE group were significantly higher (p < 0.01) than those of the DM + DE group. The values of four meibomian gland parameters, the OSDI score, and the FL score were significantly lower (p < 0.01) in the DM − DE group than those of the DM + DE group, although no significant difference of BUT was found. More importantly, the values of four meibomian gland parameters were significantly higher in the DM − DE group than those of the NC group, although the OSDI were below 13 in both groups. In fact, the values of DM − DE group met the diagnostic criteria of asymptomatic MGD. This suggested that asymptomatic MGD might emerge before the ocular discomfort develops in T2D (Table 3).

It was also noted that subjects in the DM + DE and DE groups were represented with similar ODSI, BUT, and FL scores. However, the values of SIT and all of the four MGD parameters were significantly higher (p < 0.05), whereas the LLT was lower (p < 0.05) in the DM + DE group than those of the DE group, indicating the dry eye condition in T2D might mainly originate from the impairment of the meibomian gland function (Table 3). In DM − DE group, the OSDI and FL scores were lower (p < 0.05), whereas the SIT and BUT values and the MGYLS score were higher (p < 0.05) than those in the DE group. No significant difference in LLT, the lid margin score, the meibum score, and meiboscore were found between the DM − DE and DE groups (p > 0.05, Table 3).

In order to gain insight into the importance of the glycemic control level in dry eye in patients with diabetes, the patients were further separated into several subgroups according to HbA_1c levels. The DM − DE group was further separated into two subgroups: the DM − DE-1 subgroup (HbA_1c < 7%) and the DM − DE-2 subgroup (HbA_1c ≥ 7%); the DM + DE group was separated into three subgroups: the DM + DE-1 subgroup (HbA_1c < 7%), the DM + DE-2 subgroup (7% ≤ HbA_1c < 9%), and the DM + DE-3 subgroup (HbA_1c ≥ 9%). Table 4 shows the tear film and meibomian gland parameters of the subgroups and the NC group at different HbA_1c levels.

As the HbA_1c levels increased in the three DM + DE subgroups, the OSDI and FL scores also increased, whereas the SIT, BUT, and tear LLT values decreased (Table 4). Meanwhile, the scores of meibomian gland parameters gradually increased. In comparison with the NC group (13.58 ± 2.92 mm), the SIT value of the DM + DE group with good glycemic control (DM + DE-1 group) was decreased (9.33 ± 2.46 mm, p < 0.01), but was still close to the normal cut-off value (10 mm). In contrast, the SIT value of the DM + DE group with poor glycemic control (DM + DE-3 group) significantly decreased (5.80 ± 2.11 mm, p < 0.001). In addition, the tear BUT, FL, LLT scores, and all meibomian gland parameters of the DM + DE group with good glycemic control (DM + DE-1 group) were significantly abnormal when compared to the normal controls (p < 0.001).

The parameters of tear film and meibomian gland in the two DM − DE subgroups were also different at different HbA_1c levels, but the SIT value were still within the normal range. In the subgroup with good glycemic control (DM − DE-1), the BUT and LLT scores were significantly lower, and meibomian gland parameters were significantly higher when compared with the NC group (p < 0.001). In the subgroup with poor glycemic control (DM − DE-2), the BUT, LLT, and FL scores were significantly altered when compared to those of the NC group (p < 0.001). All meibomian gland parameters of these two subgroups were dramatically higher when compared to those of the NC group (p < 0.001). These results suggested that patients with diabetes might had a state of asymptomatic MGD even though the blood glucose levels were well controlled (HbA_1c < 7%). Asymptomatic MGD might already exist in T2D with good glycemic control but without dry eye. Representative images from patients in each subgroup are shown in Figure 2.

In the DM group, the Spearman correlation analysis showed that the HbA_1c level was positively correlated with the OSDI score (R = 0.644, p < 0.001) and FL score (R = 0.393, p < 0.001) but was inversely correlated with SIT (R = − 0.586, p < 0.001), tear BUT (R = − 0.575, p < 0.001), and LLT (R = − 0.560, p < 0.001) values. The analysis showed that the HbA_1c level was positively correlated with all the four meibomian gland parameters (lid margin score, R = 0.427; MGYLS score, R = 0.407; meibum score, R = 0.452; and meiboscore, R = 0.454; all p < 0.001). Among the relevant variables (i.e., age, blood glucose, HbA_1c, and duration of diabetes), the multiple linear regression analysis revealed that only HbA_1c level was significantly associated with meibum score (β = 0.397, p = 0.036) in the DM group, whereas age was significantly associated with meiboscore (β = 0.362, p = 0.001) and MGYLS score (β = 0.310, p = 0.003). Blood glucose level was not associated with any of the meibomian gland parameters (p > 0.05).

Patients in the DM group were further separated into three subgroups based on the duration of the diabetes. The parameters of tear film and meibomian gland in these subgroups are shown in Table 5. The OSDI score, FL score, the lid margin score, MGYLS score, the meibum score, and meiboscore in the long and moderate duration groups were significantly higher, whereas the SIT value was significantly lower (p < 0.001) than that of the short duration group (p < 0.01). No significant difference was found in these parameters between the moderate duration group and long duration group. Interestingly, as the duration of diabetes increased from moderate to long duration, the value of BUT and LLT also decreased (p < 0.001), but the four MGD parameters did not deteriorate accordingly (p > 0.05). Multiple linear regression analysis also showed that the diabetic duration was significantly associated with meiboscore (β = 0.334, p = 0.002), the lid margin score (β = 0.498, p = 0.000), and the MGYLS score (β = 0.355, p = 0.001) in the DM group.