In Vivo Confocal Microscopy in Dry Eye Disease Associated With Chronic Graft-Versus-Host Disease

Abstract

Purpose: To compare densities of corneal epithelial dendritic cells (DCs), corneal subbasal nerves, and conjunctival epithelial immune cells (EICs) between patients with dry eye disease (DED) with and without graft-versus-host disease (GVHD) using in vivo confocal microscopy (IVCM).

Methods: This study included 54 patients who had moderate to severe DED either associated with (n = 33) or without (n = 21) chronic GVHD. In addition to evaluating clinical parameters of DED, images obtained by laser-scanning IVCM of the central cornea and superior tarsal conjunctiva were analyzed to measure densities of corneal epithelial DCs, corneal subbasal nerves, and conjunctival EICs.

Results: Although there were no significant differences between GVHD and non-GVHD groups in symptom scores, the GVHD group had significantly worse corneal fluorescein staining, tear break-up time, and Schirmer's scores than the non-GVHD group. Corneal epithelial DC density, corneal subbasal nerve density, and conjunctival EIC density were 148 ± 135 cells/mm², 16.3 ± 6.1 mm/mm², and 670 ± 267 cells/mm², respectively, in the GVHD group; and 122 ± 99 cells/mm², 18.3 ± 5.1 mm/mm², and 572 ± 271 cells/mm², respectively, in the non-GVHD group. After adjusting for clinical parameters, including the DED severity, none of the IVCM parameters was significantly different between the GVHD versus non-GVHD groups (P = 0.82, P = 0.21, and P = 0.60, respectively).

Conclusions: In GVHD-associated DED, cellular changes in the cornea and conjunctiva observed by IVCM were similar to those seen in patients who have non-GVHD dry eye with the same level of disease severity. Therefore, corneal and conjunctival IVCM findings in GVHD-associated DED are possibly reflective of the local disease (DED) severity rather than the underlying systemic disease process.

Graft-versus-host disease (GVHD) is a major complication of allogeneic hematopoietic stem cell transplantation (HSCT), in which donor cells recognize recipient antigens as foreign and target them. Graft-versus-host disease develops in 30% to 70% of patients after HSCT and affects many organ systems with variable clinical features resembling autoimmune disorders.^1–3 One of the organs most commonly involved in GVHD is the eye. In fact, ocular involvement occurs in approximately 60% to 90% of patients with GVHD.^4–6 Ocular GVHD may manifest as corneal epitheliopathy, ocular surface inflammation, conjunctival cicatricial disease, cataract, eyelid inflammation and scarring, uveitis, scleritis, retinal microvasculopathy, or infectious retinitis. However, dry eye disease (DED) is the most common form of ocular involvement in GVHD, occurring in 40% to 76% of patients.^4–7 

The exact pathogenesis of DED in GVHD has not been fully determined. Although DED may be partly caused by lacrimal gland dysfunction due to the disease process in GVHD, the underlying hematologic disease and its treatment modalities, such as radiation and chemotherapy, may also contribute to the development of DED.^4–7 Evaluation of cellular changes in the ocular surface of patients with GVHD can be useful in unraveling the disease pathogenesis. Clinical manifestations of GVHD-associated DED have been well described, and reported to be similar to DED due to non-GVHD causes, albeit often more severe.^8–10 However, there are limited data on cellular changes of the ocular surface in these patients. More importantly, it remains unknown whether cellular changes in the ocular surface of patients with GVHD-associated DED are different from those in DED due to non-GVHD conditions. 

In vivo confocal microscopy (IVCM) is a relatively novel technology for evaluating cellular changes in the cornea and conjunctiva. This imaging modality is a powerful, noninvasive diagnostic tool to study corneal and conjunctival structures at the cellular level.^11,12 Using IVCM, numerous studies^13–15 have demonstrated ocular surface changes in various conditions including DED. 

This study was designed to use IVCM for evaluation of cellular changes in the cornea and conjunctiva of patients with GVHD-associated DED and to compare these changes with those in patients with DED due to non-GVHD causes. 

Methods

This study included a post hoc analysis of the baseline data of a previously conducted randomized clinical trial (clinicaltrials.gov, NCT01456780). The study included 54 patients with moderate to severe DED including 33 patients with chronic GVHD and 21 patients without GVHD. The protocol of the study was approved by the Institutional Review Board of Massachusetts Eye and Ear Infirmary, Boston, Massachusetts, and the study was in compliance with the Health Insurance Portability and Accountability Act and in adherence to the tenets of the Declaration of Helsinki. 

Chronic GVHD was diagnosed based on the history of allogeneic HSCT and the presence of systemic GVHD in other organs than eyes. In the GVHD group, DED symptoms started after the development of systemic GVHD. In both groups, inclusion criteria included the presence of typical symptoms of DED, an Ocular Surface Disease Index (OSDI) score of >22, and a corneal fluorescein staining score ≥ 4 (National Eye Institute [NEI] grading scale, 0–15). Exclusion criteria consisted of the following: use of steroids, antibiotics, or optical soft contact lenses within 2 weeks before enrollment, and any change in the dosage of topical anti-inflammatory medications, other than steroids, or oral tetracyclines within 2 weeks before enrollment. Additional exclusion criteria included a history of Stevens-Johnson syndrome or mucous membrane pemphigoid, a history of herpetic keratitis, and active ocular allergies. 

The following clinical data were obtained from patients: demographics, ocular and systemic medications, scores for DED symptom questionnaires including OSDI and Symptom Assessment iN Dry Eye (SANDE), corneal fluorescein staining (NEI scale, 0–15), conjunctival staining with lissamine green (NEI scale, 0–18), tear break-up time (TBUT), and Schirmer's test with anesthesia. 

In Vivo Confocal Microscopy

All patients in both GVHD and non-GVHD groups had undergone imaging with a laser IVCM for the central cornea and superior tarsal conjunctiva of both eyes, using the Heidelberg Retina Tomograph 3 with the Rostock Cornea Module (Heidelberg Engineering GmbH, Heidelberg, Germany) as previously described.^16,17 This microscope uses a 670-nm red wavelength diode laser source and obtains images that each represents a coronal section of 400 × 400 μm (horizontal × vertical). Digital images were recorded with the sequence mode at a rate of 3 frames per second, including 100 images per sequence. There was a separation of 1 μm between adjacent images; the lateral resolution was 1 μm. 

For the cornea, a total of six to eight sequence scans of nonoverlapping areas were recorded from full thickness of the central cornea, with three to four sequence scans focusing on the subbasal layer, resulting in 300 to 400 images of the subbasal layer alone. For IVCM of the superior tarsal conjunctiva, the upper eyelid was first everted exposing the conjunctiva. Then, a total of four to six sequence scans of nonoverlapping areas were obtained from full thickness of the conjunctiva, with two to three sequence scans focusing on the conjunctival epithelium. 

Image Analysis

Three most representative images of corneal epithelial immune dendritic cells (DCs), corneal subbasal nerves, and conjunctival epithelium were chosen for each eye. All measurements were performed with ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA). 

To measure corneal epithelial DC density, the Cell Count tool of ImageJ was used in the manual mode as described previously.¹⁸ Briefly, epithelial DCs were morphologically identified as bright individual dendritiform structures with cell bodies. All complete DCs present in each image, as well as partial cells on the top and right borders of each frame, were counted. The average density of epithelial DCs for three images was then calculated and expressed as cells/mm². 

To measure subbasal nerve density, subbasal nerve fibers were traced with NeuronJ (http://www.imagescience.org/meijering/software/neuronj/; provided in the public domain by Erik Meijering), which is a semiautomated nerve analysis plug-in program of ImageJ.¹⁸ The subbasal nerve density was defined as the total length of all nerve fibers traced per image, which was then expressed as mm/mm². The average subbasal nerve density for three images was then calculated and recorded. 

To measure conjunctival epithelial immune cells (EICs), the Cell Count tool of ImageJ was used in the manual mode as described previously.^19,20 These cells were identified as hyperreflective round, oval, or polymorphous structures within the epithelium, ranging in size from 5 to 20 μm. These cells are morphologically different from goblet cells and epithelial cells and have higher reflectivity.²⁰ All complete immune cells present in each image, as well as partial cells on the top and right borders of each frame, were counted. For oblique images, in which the epithelium was present only in a part of the frame, the area of the epithelium was measured by the Polygon tool of ImageJ and was taken into account in calculation of the cell density per mm². The average value for conjunctival EIC density of three images, expressed as cells/mm², was then recorded. 

All IVCM measurements were performed by two masked observers and the averaged values of these two observers were used for the analysis. 

Statistical Analysis

Statistical analysis was performed by SPSS for Windows version 21 (SPSS, Inc., Chicago, IL, USA). For each variable, except for symptom scores, the patient's data were calculated by averaging the values from both eyes. The lack of normal distribution of the data was first verified by using Shapiro-Wilk test. To compare clinical parameters between the GVHD and non-GVHD groups, as well as between men and women, Mann-Whitney test and χ² test were used for quantitative and qualitative variables, respectively. As there were significant differences in clinical parameters between the GVHD and non-GVHD groups, to compare the IVCM data between the two groups, stepwise linear regression models were used to adjust for potential confounding factors including age, sex, clinical severity of the disease, and the use of systemic immunosuppression. Moreover, to correlate the clinical and IVCM parameters, Spearman's rank-order correlation test was carried out. Two-sided P values of less than 0.05 were considered as statistically significant for all comparisons. 

Results

This study included 54 patients (32 women and 22 men) with a mean age of 54.4 ± 12.2 years (range, 23–79 years). There were 33 patients with chronic GVHD and 21 patients without GVHD. In addition to lubricating eye drops, patients were being treated with topical anti-inflammatory medications other than steroids in 18 (33.3%) and autologous serum drops in 20 (37.0%). Moreover, 26 (78.8%) patients in the GVHD group were using systemic immunosuppressive medications. 

Overall, the mean symptom scores were 65.2 ± 17.7 for SANDE and 56.3 ± 20.6 for OSDI, and 44 patients (81.5%) had an OSDI score more than 32. The average values for ocular surface parameters included 6.5 ± 2.3 for corneal fluorescein staining, 2.5 ± 2.6 for conjunctival lissamine green staining, 3.1 ± 2.0 seconds for TBUT, and 5.4 ± 5.6 mm for Schirmer's test. In IVCM, the average values for corneal epithelial DC density, corneal subbasal nerve density, and conjunctival EIC density were 137.9 ± 122.1 cells/mm², 17.1 ± 5.8 mm/mm², and 631.6 ± 270.2 cells/mm², respectively. 

The clinical and imaging data of both groups are shown in Table 1. There were no significant differences in OSDI and SANDE symptom scores between the GVHD group (57.9 ± 20.5 and 62.7 ± 18.2, respectively) and the non-GVHD group (53.8 ± 21.0 and 69.1 ± 16.5, respectively). However, compared with the non-GVHD group, the GVHD group had a significantly higher corneal fluorescein staining score (5.4 ± 1.8 vs. 7.1 ± 2.4, respectively, P = 0.005), lower TBUT (4.3 ± 1.9 vs. 2.4 ± 1.8 seconds, respectively, P = 0.001) and lower Schirmer's score (7.6 ± 6.8 vs. 3.9 ± 4.3 mm, respectively, P = 0.008). Conjunctival lissamine green scores were similar in both groups (Table 1). 

Table 1

 

Clinical and In Vivo Confocal Microscopic Parameters in Patients Who Had Dry Eye Disease With or Without Ocular GVHD

 

In vivo confocal microscopy showed that corneal epithelial DC density, corneal subbasal nerve density, and conjunctival EIC density were 148 ± 135 cells/mm², 16.3 ± 6.1 mm/mm², and 670 ± 267 cells/mm², respectively, in the GVHD group; and 122 ± 99 cells/mm², 18.3 ± 5.1 mm/mm², and 572 ± 271 cells/mm², respectively, in the non-GVHD group. After adjusting for age, sex, DED severity, and systemic immunosuppression, using regression models, all these IVCM parameters did not show any significant difference between the two groups (P = 0.82, P = 0.21, and P = 0.60, respectively; Fig.). 

Correlations between clinical and IVCM parameters are shown in Table 2. There was a statistically significant correlation between the score of corneal fluorescein staining and the conjunctival EIC density (P = 0.01, Rs = 0.34). In addition, there was a marginally significant correlation between the score of corneal fluorescein staining and the density of corneal epithelial DCs (P = 0.07, Rs = 0.25). No other significant correlations were found. 

Table 2

 

Correlations Between Clinical and In Vivo Confocal Microscopic Parameters in Patients Who Had Dry Eye Disease With or Without GVHD

 

Comparisons of clinical and IVCM parameters between women and men in patients who had DED with or without ocular GVHD are shown in Table 3. 

Table 3

 

Clinical and In Vivo Confocal Microscopic Parameters Between Women and Men in Patients Who Had Dry Eye Disease With or Without Ocular GVHD

 

Discussion

This study demonstrated significant IVCM changes in the cornea and conjunctiva of patients with chronic GVHD who had moderate to severe DED. However, after adjusting for clinical severity of DED, there were no significant differences in IVCM parameters, including corneal epithelial DC density, corneal subbasal nerve density, and conjunctival EIC density, between patients with GVHD and those without this condition. Thus, ocular surface cellular changes observed by IVCM in GVHD-associated DED are possibly reflective of the local disease severity rather than the underlying systemic disease process. 

As DED in patients with GVHD is often severe,^8–10 for this study we only included patients with moderate to severe DED (OSDI score > 22 and corneal fluorescein staining score ≥ 4 [NEI]) to have a relatively similar degree of severity in both GVHD and non-GVHD groups. In fact, the mean OSDI score was 56.3, and 81.5% of our patients were severely symptomatic. Despite these inclusion criteria, the GVHD group still had a significantly higher score of corneal fluorescein staining and significantly lower values of TBUT and Schirmer's test than the non-GVHD group. Interestingly, even with significantly worse clinical signs in the GVHD group, these patients did not show worse symptom scores. Such a discrepancy, commonly seen in DED,^21,22 may partly be explained by a reduced corneal sensation in GVHD, as has been reported before,¹⁰ as well as a decreased corneal nerve density, as discussed below. 

Compared with corneal epithelial DC density in healthy controls (14.5–34.0 cells/mm²),^23,24 overall our patients had a higher corneal epithelial DC density; however, there was no significant difference between the GVHD group (148 cells/mm²) and the non-GVHD group (122 cells/mm²). Steger and colleagues²⁵ also have compared 12 patients with severe DED due to GVHD to 6 patients without GVHD (5 before and 1 after HSCT) and also found no significant difference in corneal epithelial DC density between these two groups. As epithelial DCs play an important role in corneal immune homeostasis,^26,27 our findings may suggest a similar degree of immune cell changes in DED with and without GVHD as further discussed below. 

In our study, we had a reduced corneal subbasal nerve density in both GVHD and non-GVHD groups (16.3 and 18.3 mm/mm², respectively) as compared with the normal range described before (21.6–24.1 mm/mm²).^28,29 However, no significant difference was noted between our two groups after adjusting for clinical parameters including DED severity. Steger and associates²⁵ also have observed a significantly lower number of subbasal nerves per frame in the GVHD group than in a control group without GVHD. Many previous studies have shown a reduced density of corneal subbasal nerves in DED,^13–15 which correlates with clinical severity of the disease.³⁰ On the other hand, a significant reduction of corneal sensation, which is known to correlate with corneal nerve density,³¹ has also been noted to occur not only in GVHD-associated DED but also in post-HSCT patients without DED.¹⁰ Thus, it may be speculated that such changes in corneal sensation and nerves might be attributed to the associated DED as well as the conditioning regimen before HSCT, such as total body irradiation.¹⁰ 

We also observed a higher conjunctival EIC density in both GVHD and non-GVHD groups (670 and 572 cells/mm², respectively) compared with what has been previously reported in healthy individuals (278 cells/mm²) (Qazi Y, et al. IOVS 2012;53:ARVO E-Abstract 593); however, no significant difference existed between the two groups. Using cytologic techniques, an increased density of inflammatory cells in the tarsal conjunctiva of patients with GVHD-associated DED has also been observed before.^10,32 Moreover, using IVCM similar findings have been seen in DED due to Sjogren's syndrome as well as in atopic keratoconjunctivitis.^33,34 Although conjunctival EICs may play a role in pathogenesis of DED in GVHD, it is unclear whether these are the cause or result of DED. However, as some correlations have been described between the density of conjunctival EICs and the clinical severity of DED,³³ these cells may potentially be used as an objective marker for patients' follow-up. Using IVCM, which is more convenient for patients than cytologic techniques, these changes can thus be monitored over time. 

Interestingly, we did not find any significant differences in IVCM parameters between patients who had DED with or without GVHD. It may be speculated that any IVCM changes caused by GVHD alone may be overshadowed by extensive changes induced by the presence of moderate to severe DED. On the other hand, it is plausible that GVHD alone (without DED), may not be associated with significant changes in corneal and conjunctival immune hemostasis. For example, in patients with autoimmune disease, such as rheumatoid arthritis, systemic lupus erythematosus, and ankylosing spondylitis, increased corneal epithelial DC density and/or decreased corneal nerve density have been reported before.^35–37 However, such diseases are associated with DED with or without Sjogren's syndrome, which may be the cause of corneal changes in IVCM. Notably, Resch and associates³⁶ have demonstrated that in patients with systemic lupus erythematosus with normal tear production, corneal epithelial DC density is similar to normal controls, implicating that corneal IVCM changes in these immune diseases may be due to the associated DED. Moreover, in patients with Sjogren's syndrome, higher densities of corneal epithelial DCs and conjunctival EICs, and a significantly lower density of corneal nerve density, have been reported, compared with non-Sjogren DED.^33,38,39 However, in these studies clinical signs are significantly worse in patients with Sjogren's syndrome, and such differences are not accounted in the comparison of the IVCM data between the two groups. Accordingly, it may be concluded that IVCM changes in systemic immune diseases, including GVHD, may be reflective of the associated DED rather than the underlying disease. Future longitudinal studies on IVCM changes in the cornea and the conjunctiva after HSCT, with or without DED, will shed light on the potential role of individual pathogenetic mechanisms. 

In our study, corneal fluorescein staining had a statistically significant correlation with conjunctival EIC density and a marginally significant correlation with corneal epithelial DC density. Correlations between IVCM and clinical parameters have been reported by some studies, but not others.^11,12 Such conflicting results may be due to associated confounding factors, individual variations, or the degree of disease severity. 

In this study we included patients with GVHD who had stable disease with no recent change in anti-inflammatory medications. In addition, our patients with GVHD were mostly under treatment with systemic immunosuppressive medications, which may dampen immune and inflammatory changes. Inclusion of patients after HSCT with no DED, mild GVHD-associated DED, or without systemic immunosuppressive treatment in future studies may provide a better insight to underlying immune changes in these patients. With these limitations in mind, our study showed that patients with GVHD who have moderate to severe DED demonstrate significant ocular surface cellular changes in IVCM, which are not different from those in patients with non-GVHD dry eye who have comparable disease severity. 

Figure

 

In vivo confocal microscopic images of corneal subbasal layer (upper row) and superior tarsal conjunctiva (lower row) in normal controls (from a preexisting database) and in patients with DED with or without chronic GVHD. Compared with normal individuals, there were significantly higher densities of epithelial dendritic cells and conjunctival epithelial immune cells as well as a significantly lower density of subbasal nerves. However, there were no statistically significant differences in these parameters between GVHD and non-GVHD groups after adjusting for clinical severity of the disease.

Acknowledgments

This study has been partly presented at the Annual Meeting of the Association for Research in Vision and Ophthalmology, Seattle, Washington, United States, May 2016. 

Disclosure: A. Kheirkhah, None; Y. Qazi, None; M.A. Arnoldner, None; K. Suri, None; R. Dana, P