Theoretical, Experimental, and Optical Coherence Tomography (OCT) Studies of Graft Apposition and Adhesion in Descemets Stripping Automated Endothelial Keratoplasty (DSAEK)

Abstract

Purpose.: To investigate the effects of adhesion promoting surgical adjuncts in Descemets stripping automated endothelial keratoplasty (DSAEK). The effects of air-fill pressure, duration, use of venting incisions and stromal roughening on fluid dispersion, and donor adhesion strength were examined in theoretical, optical coherence tomography (OCT), and strain gauge models of DSAEK.

Methods.: OCT analysis: DSAEK modeled using a microkeratome prepared lenticule inserted under a “recipient” corneo-scleral rim mounted on an artificial anterior chamber. Pressure of 18 mm Hg (n = 6) or 60 mm Hg (n = 6) was applied. The area of interface fluid was measured sequentially. The area of interface fluid before and after opening of venting incisions was measured (n = 6). Adhesion experiments: Direct measurement of adhesion force using a universal testing machine was performed. Peak adhesion after compression at 60 mm Hg/8 minutes, 60 mm Hg/1 minutes, 18 mm Hg/8 minutes, and 18 mm Hg/1 minutes (n = 8 each group) was measured. Subsequently, adhesion after complete removal of interface fluid and after stromal roughening was measured in separate samples (n = 12).

Results.: Interface fluid diminishes with time during tamponade at both low and high pressures (P < 0.0001). Pressure had no effect on amount or rate of fluid dispersion. Venting incisions eliminated interface fluid in all samples when opened sufficiently. Adhesion is independent of anterior chamber air tamponade pressure (P = 0.38). Complete removal of interface fluid increases average adhesion (16.0 mN vs. 7.8 mN, P = 0.0001). Roughening of the host stroma increased adhesion (13.8 mN vs. 9.8 mN, P = 0.0034).

Conclusions.: Venting incisions and stromal roughening aid adhesion in DSAEK. Sustained high-pressure anterior chamber air tamponade has no demonstrable effect on measured fluid dispersion or adhesion strength.

Introduction

Descemets stripping automated endothelial keratoplasty (DSAEK) has largely replaced penetrating keratoplasty in the treatment of endothelial dysfunction. Graft dislocation is the most common early complication after DSAEK. ^1,2 Graft repositioning after dislocation inevitably involves further surgery and donor manipulation, and the authors have previously found a strong association between graft dislocation and subsequent graft failure. ³ An understanding of graft adhesion and the factors leading to donor dislocation is therefore central to improving surgical results in endothelial keratoplasty (EK). ⁴  

A variety of alterations in DSAEK surgical technique have been utilized with the aim of preventing graft dislocation in the early post-operative period. These include variations in the time and pressure of the anterior chamber air tamponade, mid corneal venting incisions, compressive sweeping across the surface of the host cornea to expel interface fluid, and procedures to roughen the host stroma. ^5,6 To date, no controlled trials have been conducted investigating the benefits of these variations in technique, which may be associated with the risk of additional complications. ^7–9  

To help understand how variations in air tamponade might influence retained interface fluid in DSAEK, the authors developed a mathematical fluid dynamic model to predict the relationship between bubble size, air fill pressure, duration of air compression, and fluid expulsion from the graft/host interface in an eye without venting incisions (see Appendix) (Angunawela RI, et al. IOVS 2010;51:ARVO E-Abstract 785). This model assumed that the graft was thin, with no resistance to deformation, and perfectly smooth. It predicts that interface fluid diminishes over time during air compression, but raising air fill pressure during the air compression phase of EK surgery would have little effect on dispersion of interface fluid. The model also predicts that it is not possible to express all interface fluid using an anterior chamber air bubble alone. 

The authors set out to test these predictions using an optical coherence tomography (OCT) imaging based model of the air compression phase of DSAEK surgery to evaluate the effect of variations in air fill pressure, air fill duration, and venting incisions on graft/host interfacial fluid elimination. In a second in vitro model, the authors measured adhesion strength directly, using a universal testing machine to examine the effects of variations in air fill pressure and air fill duration, interface drying, and stromal roughening. 

Methods

Institutional review board (IRB) approval for the study was obtained from Moorfields Eye Hospital Research Ethics Committee. Pairs of human corneas with consent for research were obtained from Moorfields Lions Eye Bank (Moorfields Eye Hospital, London, UK) the United Kingdom Transplant Service (Bristol, UK), and Ocular Systems (Winston-Salem, NC). Tissue samples were stored in either Optisol (Bausch and Lomb, Irvine, CA) or organ culture media. 

The experimental study consisted of 2 parts: 

Imaging of Interface Fluid on OCT

Measurements of interface fluid were taken sequentially during air compression using the RTVue 4.0 Fourier domain anterior segment optical coherence tomographer (AS-OCT) (Optovue Inc., Fremont, CA). 

A model of DSAEK was developed using an artificial anterior chamber (Moria S.A., Antony, France) to mount the host cornea after desmetorhexis had been performed using a Sinsky hook and plane forceps. Eight mm-diameter donor lenticules were prepared from the second cornea in each pair, using an artificial anterior chamber pressurized to 65 mm Hg and a microkeratome with a 350-μ head (Moria). Donor lenticules were placed beneath the host cornea within the artificial chamber prior to sealing, then floated into apposition with the host cornea using an air bubble injected beneath the donor as in standard DSAEK surgery (Fig. 1a). 

Figure 1. 

 

OCT imaging of DSAEK model. (a) Position of graft, air bubble and host matches that of the surgical setting. (b) The area of interface fluid was outlined manually and calculated using RTVue 4.0 software (Optovue Inc.).

Using a specially designed rig, the artificial anterior chamber was secured with the cornea opposite the wide-angle objective lens of an RTVue AS-OCT. The RTVue 4.0 AS-OCT can acquire 26,000 axial scans per second and has a depth resolution of 5 μ (full width, half maximum) with the wide-angle anterior segment lens attached. ¹⁰ Use of OCT to measure interface fluid in EK has already been established. ¹¹  

Central images were taken serially every 2 minutes for 14 minutes. The area of fluid trapped in the interface between donor and host corneas during air compression was measured in each image by outlining the pockets of visible fluid, seen as areas of low intensity on AS-OCT, using the integrated software (Fig. 1b). Two pressure conditions were tested: 18 mm Hg (normal pressure) and 60 mm Hg (high pressure), each in 6 pairs of corneas. Anterior chamber pressure was confirmed using an analog pressure gauge connected to 1 of the anterior chamber ports. The effect of pre-placed venting incisions was then examined in 6 additional pairs of corneas. Four paracentral, full-thickness, stab incisions were made in the host cornea using a 15° blade in the manner described by Price et al. ⁵ Scans were taken prior to anterior chamber air fill, after air fill to a pressure of 18 mm Hg, and then after opening of the venting incisions with a blunt 30-gauge cannula.  

Measurement of Graft–Host Adhesion

Direct measurement of graft adhesion force was performed using an Instron 5566 universal testing machine (Instron, Norwood, MA), equipped with a 2.5 N static load cell and Blue Hill software (Instron). The reproducibility and linearity of the load cell used is 0.25% (in the range 0.001–2.5 N), according to the product supplier specifications. ¹²  

Measurement of graft adhesion in the standard artificial anterior chamber was not possible. A modified artificial anterior chamber was therefore designed (Fig. 2), in which 1 cornea of each test pair (designated as the host cornea) was supported, epithelial side down, and stabilized by vacuum through a perforated supporting base with a curvature matched to the corneal curvature, such that no visible tissue deformation was produced by the application of vacuum stabilization (negative pressure applied = 30 mm Hg). A descemetorhexis was performed on the host and a donor lenticule was prepared as described in the section on OCT analysis. A 6-0 nylon suture fixed to a 2 mm-diameter disc of aluminum (50 μ thick) was passed through the center of the donor lenticule, such that the disc was sandwiched in the donor host interface (Fig. 2a). 

Figure 2. 

 

Illustrations showing orientation of graft material in adhesion testing apparatus. (a) Graft and host separated by aluminum disc. (b) The host cornea is held in place by vacuum (lower chamber) and pressurized (upper chamber) to simulate anterior chamber gas fill. (c) After a defined duration and pressure air tamponade, the chamber and suture are attached to the grips of the universal testing machine to allow the measurement of adhesion force by distraction.

This defined area of non-adhesion allowed for a consistent initiation and propagation of graft–host detachment during the adhesion test. The host cornea and graft button were dried using cellulose sponges, and a standard amount of interface fluid (0.05 mL of BSS) was dropped onto the host prior to the donor being positioned above and the pressure chamber sealed and inverted (Fig. 2b). After applying the test air compression conditions, the pressure chamber and the 6-0 nylon suture were fixed to the lower and upper grips of the tensile test machine (Fig. 2c), and an increasing force of distraction was applied at a standard rate. The force required to initiate peeling of the donor lenticule away from the host (adhesion force) was defined for each test condition by the peak of the force/displacement curve recorded, as a constant tensile deformation of 1 mm/minutes was applied (Fig. 3). 

Figure 3. 

 

Graph showing force (in Newtons) increasing as the corneas are pulled apart in the universal testing machine. A measurement of peak adhesion (arrow) is taken from the force displacement curve.

Several variations of this experiment were utilized and are described below. The corneal tissue was returned to the Optisol/organ culture storage medium for 3 minutes between tests. 

Increasing Pressure and Time of Anterior Chamber Gas Fill

Pressure and duration of gas fill were tested in the following combinations: 60 mm Hg for 8 minutes, 60 mm Hg for 1 minute, 18 mm Hg for 8 minutes, and 18 mm Hg for 1 minute. Each variable was tested in 8 pairs of corneas with the order rotated, using a Latin square design to exclude any systematic errors related to order in the testing sequence. 

Completely Removing Interface Fluid

The effect of complete removal of interface fluid was evaluated in 12 pairs of corneas. Each pair was tested at 18 mm Hg for 1 minute after application of interface fluid as described above. Subsequent adhesion measurements were obtained after compression at 18 mm Hg for 1 minute and 60 mm Hg for 1 minute in corneal pairs where no fluid had been reapplied after drying with a cellulose sponge. 

Roughening of the Host Stroma

The effect of roughening the entire stromal surface was evaluated in 12 pairs of corneas. Adhesion strength was measured after compression at 18 mm Hg for 1 minute with interface fluid in corneal pairs that had been dried and no further interface fluid applied prior to placing the donor onto the host. The entire stromal surfaces of the host corneas were roughened using a 30-gauge needle in a scoring motion until small fibrils of corneal tissue could be seen on the host surface. Adhesion was measured in the roughened group without application of interface fluid after compression at 18 mm Hg for 1 minute. 

Statistical Analysis

Statistical analysis using repeated measures ANOVA was conducted on SPSS version 16.0 (SPSS Inc., Chicago, IL). Significance was set at P = 0.05. 

Results

AS-OCT Evaluation of Interface Fluid in DSAEK

AS-OCT of the non-vented corneas shows that interface fluid diminished during air compression in both high and low pressure groups (Fig. 4). The reduction in interface fluid over time reached statistical significance in both groups (P < 0.0001). Interface fluid area was more variable for the low-pressure condition (18 mm Hg), but the fluid area was not significantly reduced where a high-pressure gas fill (60 mm Hg) was used (P = 0.43). Analysis of the rate of interface fluid dispersion between the 2 groups by regression failed to show a difference between high and low pressure groups (P = 0.56). 

Figure 4. 

 

Graph showing mean interface fluid area against compression time in high pressure and low-pressure air tamponade groups. Bars represent 1 standard deviation.

Complete removal of interface fluid was not seen in any sample, regardless of AC pressure at 14 minutes of air compression (Figs. 5a, 5b). The presence of venting incisions was sufficient to allow spontaneous removal of interface fluid in 1 of the 6 samples. The authors opened the venting incisions in all 6 samples by passing a blunt 30-gauge cannula centrally. This allowed complete removal of residual fluid visible in the 5 samples in which fluid remained (Figs. 5c, 5d). 

Figure 5. 

 

AS-OCT images showing interface fluid at different stages of compression. (a) Initial anterior chamber gas fill. (b) Image taken after 12 minutes of high-pressure air tamponade showing reduction without complete elimination of interface fluid. (c) Interface fluid visible after initial air fill in spite of venting incisions. (d) Interface fluid completely removed after venting incisions are opened with blunt cannula.

Assessment of Graft Adhesion

Increasing the pressure of anterior chamber gas fill had no significant effect on graft adhesion (18 mm Hg vs. 60 mm Hg, P = 0.38). Adhesion after 8 minutes of air fill was higher than after 1 minute. This trend toward stronger adhesion with increased time of air compression did not reach significance (P = 0.11) (Fig. 6). 

Figure 6. 

 

Graph showing mean peak adhesion measured directly by tensile testing in each of the four testing combination. Error bars represent 1 standard deviation.

Effect of Removing Interface Fluid on Adhesion Strength

Removing interface fluid resulted in a significant increase in adhesion strength, with average adhesion more than doubling (P = 0.0001) (Fig. 7a). Increasing compression pressure showed no further increase in adhesion when compared to removal of interface fluid alone (Tukey post test, P = 0.23). 

Figure 7. 

 

Graph showing mean peak adhesion with and without interface fluid, and at standard and high pressure. Error bars represent 1 standard deviation.

Effect of Surface Roughening on Adhesion Strength

Stromal roughening showed an additional beneficial effect over removal of interface fluid alone, with adhesion increasing by an average of 46% in paired corneas after roughening (P = 0.0034) (Fig. 8). 

Figure 8. 

 

Graph showing mean peak adhesion before and after stromal roughening. Error bars represent standard deviation.

Discussion

Air compression alone is insufficient to allow complete removal of interface fluid in DSAEK. Our OCT study shows the most common configuration adopted at the time of air tamponade is that of a peripheral annulus of tissue apposition between graft and host corneas with a lake of interface fluid trapped centrally. During the air compression phase of DSAEK, the eye is a closed chamber with intraocular pressure equalized throughout. Since there is no pressure gradient, increased air compression pressure will, as the authors observed, have no effect on the rate of interface fluid dispersion once a steady pressure is achieved. Initial inflation to a higher pressure may drive more fluid from the interface as a function of shape change in relation to corneal elasticity, and this could explain the trend to more retained interface fluid and greater variability in retained fluid area observed in OCT sections for lower pressure tamponade; but sustained high pressure tamponade will make little or no difference to the volume of retained interface fluid. 

Opening paracentral venting incisions induces a pressure differential across the donor lenticule—anterior chamber pressure on the endothelial side and atmospheric pressure on the stromal side. Venting incisions may be self-sealing, and did not clear fluid effectively in the authors' OCT model without probing with a narrow gauge cannula to allow fluid egress. Knecht et al. found that venting incisions resulted in removal of interface fluid in only 2 of 4 cases assessed using intra-operative AS-OCT, but did not state whether these incisions were probed. ¹¹ Other surgical techniques not tested here, including surface sweeping to milk fluid from the interface, may be effective in promoting the elimination of interface fluid during air compression. Venting incisions may not be necessary in cases at low risk for dislocation. 

Complete fluid removal and stromal roughening both increased adhesion in the authors' strain gauge model of DSAEK. These findings fit with a mechanism of adhesion in EK based on weak intermolecular forces. As water is expelled from the interface and the tissue surfaces approximate, weak intermolecular forces, including van der Waals forces, electrostatic attraction, and hydrogen bonds start to act. Because the tissue surfaces are not completely even, microscopic gaps between the 2 surfaces persist between points of contact (asperities) after macroscopic adhesion is established (Fig. 9 1552). Compression of the graft against the host flattens these asperities increasing the surface area of contact, further increasing adhesion. Over time, migration of long polymer chains is likely to occur between the graft and host in a process known as reptation. ¹³ It is possible that stromal roughening enhances the effective area of contact and this process of molecular interdigitation. In their experiments, the authors roughened the entire stromal surface. In clinical practice, only a small annular area in the periphery is scored to allow the optical center to remain smooth for the optimal visual outcome. The beneficial effect of roughening in the clinical setting may therefore be less than that observed in this study. 

Figure 9. 

 

Schematic illustration of probable mechanisms of graft adhesion. (a) Illustration showing apposition of the donor endothelial disc (D) to the host cornea (H). A magnified view of the area indicated by the red arrow (b) shows how after initial donor/host tissue apposition, compression drives fluid from between asperities (protrusions in an uneven microscopic soft tissue landscape) and flattens them, serving to increase the area and proximity of intermolecular contact (arrow) promoting the formation of (c) secondary bonds (van der Waals forces, hydrogen bonds [red dashed line]), and weak electrostatic attraction (between anions and cations) and initial tissue adhesion. Further compression acts to promote reptation (molecular interdigitation probably largely of collagen polymers), which greatly increases the area of close intermolecular contact between the contacting tissue surfaces, promoting further secondary bonding and a stronger adhesion.

Figure 10. 

 

Variation of interface fluid film thickness as a function of time (minutes) for 1 mL (solid blue line) and 0.5 mL (dashed red line) gas bubbles.

Activation of the endothelial pump driving fluid out of the cornea may play an additional role in early graft adhesion, although good endothelial function does not appear to be a requirement. ¹⁴ Spontaneous late graft detachments are rare even in failed grafts. 

In DSAEK, the trapped central lake of interfacial fluid the authors observed in OCT sections may result from a curvature mismatch between host and donor, and may be exacerbated by increased donor rigidity. Central fluid loculation is rarely seen in Descemets Membrane Endothelial Keratoplasty (DMEK), where the transplanted tissue has little rigidity and rapidly conforms to the internal curvature of the cornea. Venting incisions are not used in DMEK surgery, and detachments are likely to occur by other mechanisms that could include the tendency for DMEK donor sheets to scroll when immersed and relatively weak attachment due to smooth interfacial surfaces. ^15,16 The effect of donor hydration and thickness on trapped interface fluid would be a useful channel for further study. By analogy with DMEK, thinner, more compliant DSAEK donor lenticules may be associated with less fluid loculation and relative protection from graft dislocation. 

There are several limitations to the experimentation described here. In the different adhesion experiments, baseline adhesion measured after low-pressure compression for 1 minute varied significantly between corneal pairs. This may relate to differences in sample hydration, storage media, ¹⁷ or microscopic surface smoothness. The authors did not control for variations in graft thickness. Standard tests of adhesion force normally use flat, rigid specimens, whereas the specimens in this study were curved, non-rigid, and would have peeled away rather than detached across the entire interface. It is well established that the inertia of a body to bending is related to the second moment of the cross-sectional area (rectangular in this case), which is a cubic function of the thickness. Consequently, as the graft thickness increases, the measured force increases, even if the adhesion force is invariant. ¹⁸ The authors used repeated measures analysis to compensate for interspecimen variation attributable to these and other uncontrolled influences on the study's adhesion strength measurements. 

Failure to find a significant effect for increased time of compression in the strain gauge experiments may simply be a product of the low number of corneal specimens. The authors chose to measure adhesion after 8 minutes of complete AC fill, as this corresponds to their current clinical regimen adapted from Price et al. ⁵ As suggested by the study's OCT data, showing continued interfacial fluid dispersion with time (at least until 14 minutes), it is possible that measuring adhesion at a later time point may have yielded significant results. Further experimentation is required to determine the point at which additional air compression time no longer exerts any significant protective effect against donor dislocation. In the authors' clinic practice, they routinely leave a low pressure, partial air fill and encourage the patient to lie supine for 1 hour to utilize the buoyancy of the residual bubble to support the donor lenticule. The authors' mathematical model, which suggests that interface fluid should continue to disperse with time, and clinical data showing fluid present at the end of surgery disperses by the following day, both suggest the value of leaving the anterior chamber partially filled with air. ¹¹  

In conclusion, central loculation of fluid by a peripheral annulus of donor contact may be common in DSAEK. Mathematical and experimental models described here support the use of a complete air fill during surgery for an as yet undefined duration. Sustained supraphysiological intraocular pressures during tamponade do not appear to promote interfacial fluid dispersal or greater graft–host adhesion strength, whereas venting incisions, when splinted open to allow fluid egress, are effective in clearing interface fluid quickly, and interface drying and stromal roughening both appear to promote adhesion. Stromal roughening, venting incisions, and extended intraoperative air compression durations can all be considered as adjunctive measures to promote adhesion in DSAEK, particularly in cases with risk factors for detachment. 

Acknowledgments

The authors thank Sue Perkins from University College London for her advice on the mechanisms of adhesion and the process of reptation. 

Footnotes

 Supported in part by the UK National Institute for Health Research Biomedical Research Centre in Ophthalmology at Moorfields Eye Hospital and the UCL Institute of Ophthalmology.

Footnotes

 Disclosure: M.S. Bhogal, None; R.I. Angunawela, None; E. Bilotti, None; I. Eames, None; B.D. Allan, None

Appendix

Fluid Dynamics Modeling of DSAEK

To examine the influence of pressure, bubble size and time on the squeezing of fluid from between the donor and host cornea, a mathematical model, based on standard lubrication analysis, was used. This applies scaling analysis to estimate the characteristic gap separation between the transplanted cornea and the stroma. This type of analysis enables different aspects of the surgical procedure to be assessed. 

The bubble size determines how much of the bubble is in contact with the cornea. For a small bubble, (<0.2 mL) the bubble is spherical, and only a small fraction is in contact with the cornea. For larger bubbles, the surface tension is small in comparison to buoyancy forces. The bubble has a spherical cap and the base is approximately flat. For a closed system, injecting gas into the eye increases the IOP because fluid cannot escape. The increase in IOP is felt everywhere in the eye but this is not the force driving fluid from the cornea, which instead arises from gradients of pressure that are determined by hydrostatic forces. The effect of IOP is mainly to force the front of the eye to have a rounded appearance, which itself facilitates the drainage of fluid. 

Denoting h as a nominal gap separation between the donor cornea and stroma, μ is the dynamic viscosity of water (0.78 × 10⁻³ Pa s), g is the density of water (1000 kg/m³) and ∂ is the acceleration due to gravity (9.8 m/s²). The flow of fluid along the cornea/stroma gap scales as  

 

where p is the pressure and s is the distance from the top of the cornea and the constant λ = 1/12. This thin gap separation between the donated and host stroma (h) compared to the size of the eye means that the flow in the layer is approximately in a thin sheet. In this approximate model, the authors assume that the gap separation is spatially uniform and analyze how this varies with time as the buoyancy force squeezes fluid from the gap. Mass conservation tells us that the reduction in the gap thickness is due to the fluid between squeezed out of the film, expressed here as  

 

 

 

The pressure in the film scales as $p\sim \frac{\mu R^2}{\lambda h^3}\frac{dh}{dt}$, where R is the radius of the transplanted cornea. The force on the cornea $F\sim \pi R^{2}p\sim \frac{\mu R^4}{\lambda h^3}\frac{dh}{dt}$ arises from the buoyancy force created by the constant buoyancy force of the bubble. This can be integrated to show  

 

 

This expression relates the rate at which the fluid layer (of initial thickness h₀ ) between the host and donor cornea is squeezed. This depends on the bubble volume V, time of the compression t and cornea radius (Fig. 10). The analysis breaks down when the gap separation becomes comparable to the asperities in the donor and host cornea.