Interface quality of endothelial keratoplasty buttons obtained with optimised femtosecond laser settings
- Antoine Rousseau1,
- Abdelkader Bensalem2,
- Virginie Garnier3,
- Michèle Savoldelli1,4,
- Jean-Jacques Saragoussi1,5,
- Gilles Renard6,
- Jean-Louis Bourges6
- 1Assistance Publique-Hôpitaux de Paris, Department of Ophthalmology, Hôtel-Dieu Hospital, Paris, France
- 2Abbott Medical Optics France SAS, Biot, France
- 3Pierre et Marie Curie University, Department of Electron Microscopy, Institut de Biologie Intégrative IFR 83, Paris, France
- 4INSERM U872 team 17, CRC des Cordeliers, Paris, France
- 5Clinique de la Vision, Paris, France
- 6Université Sorbonne Paris Cité, Paris Descartes, Faculty of Medicine, Assistance Publique-Hôpitaux de Paris, Department of Ophthalmology, Hôtel-Dieu Hospital, Paris, France
- Correspondence to Dr Jean-Louis Bourges, Department of Ophthalmology, Hôtel-Dieu Hospital, 1 place du parvis Notre-Dame, 75004 Paris, France;
- Accepted 1 September 2011
- Published Online First 15 October 2011
Aim To optimise interfaces of endothelial buttons created with femtosecond (FS) lasers.
Setting Department of Ophthalmology, Hôtel-Dieu Hospital, Paris, France.
Methods Forty-two corneas were divided into five groups of various cutting patterns and a control group of 100 μm laser in situ keratomileusis flap creation. A single path full lamellar cut (500 μm) was applied to groups 1 and 2. The same full lamellar cut was applied twice to groups 3 and 4. Two successive lamellar cuts were performed in group 5 (350 and 150 μm). 60 kHz and 150 kHz were used respectively in groups 1, 3, 5, 6 and 2, 4. In each group, different laser settings were tested to obtain the best interface quality while delivering minimal energy to the stroma. The quality of stromal interfaces from created endothelial lenticules was observed using a scanning electron microscope.
Results Stromal adherences persisted after both the single- and double-path procedure, creating central irregularities on the endothelial lenticule. Among all groups and settings tested, the double-layer pattern (group 5) with FS full lamellar cut parameters set for diameter (mm), depth (μm), energy (μJ) and spot size/step (μm) respectively on 9.0 mm, 350 μm, 2.1 μJ, 4:4 μm and 8.3 mm, 150 μm, 0.9 μJ, 4:4 μm created the smoothest interfaces with the best reproducibility.
Conclusions Buttons for endothelial keratoplasty can be created with FS laser with a stromal interface quality comparable with that of refractive surgery.
- endothelial keratoplasty
- femtosecond laser
- Descemet stripping endothelial keratoplasty (DSEK)
- sclera and episclera
- treatment lasers
- experimental and laboratory
- optics and refraction
- ocular surface
- experimental and animal models
- treatment surgery
- medical education
- wound healing
- eye (tissue) banking
- anterior chamber
Endothelial lamellar keratoplasty (ELK) involves the selective replacement of damaged endothelium. The endothelial button (or lenticule) can be prepared with either a microkeratome or a femtosecond laser (FS). Microkeratome lamellar dissection is cheap and well standardised, and provides buttons with smoother interfaces compared with FS precutting.1–3 Smoother stromal interfaces can be used for long-term visual outcomes. On the other hand, lamellar cuts are restricted to depth adjustments by microkeratome headsizes, and the button thickness is poorly reproducible.4 Microkeratomes are subject to possible blockade during the cutting phase, which impairs the regularity of the button's interface.5
The FS has many advantages over microkeratome in precutting endothelial keratoplasty lenticules. It cuts corneal tissue at every chosen profile.6 It can be standardised and reproducible, thus reducing any problems related to dissection.7 8 Both the safety and reliability of corneal lamellar FS cuts have been demonstrated extensively for laser in situ keratomileusis (LASIK) and recently for ELK.9–11 Buttons created with FS are more planar-shaped12 and thinner when obtained with FS, which could be beneficial for visual outcomes.2 However, the smoothness and regularity of the stromal interface still need to be improved, and the laser settings optimised for femtosecond laser endothelial keratoplasty.2 7 To address this, we tested various FS precutting profiles with customised settings to generate a smoother and regular interface on lenticules.
Materials and methods
Precut setting profiles
We divided the experimental corneas into five groups on the basis of the lamellar cut profiles under which they were operated and the frequency of the FS (Intralase; AMO) used. Table 1 details the FS settings applied. Groups 1 and 2 had one single lamellar cut at 500 μm depth (single path) with a 60 or 150 kHz FS, respectively. Groups 3 and 4 had one lamellar cut at 500 μm depth performed twice (double path) within a single applanation procedure, with a 60 or 150 kHz FS laser, respectively. We were able to reduce the spot size and steps for 150 kHz groups (2 and 4). We cut group 5 with two successive lamellar cuts set at two different depths (350 μm and 150 μm) with two successive applanation procedures (double layer). Within the fifth group, the lamellar cut energies were initially set up on the basis of energies delivered for LASIK and were progressively set to a lower amount still compatible with an easy lamellar dissection. The anterior side cut for step 1 was angled at 45° and oversized by a 0.5 mm diameter as compared with the side cut for step 2. Side-cut parameters were initially set close to the maximum energy available minus 0.5 μJ, 20 μm depth over the lamellar cut depth and 3:3 for spot size and step.
Finally, as a control, LASIK free flaps were created on corneas of group 6. Cutting designs are displayed in figure 1. When setting profiles were considered as achieving a good-quality dissection with the lowest energy possible, we performed the procedure in triplicate to ensure its reproducibility.
Creation of endothelial buttons
Human corneas for experimental use
Forty-two human corneas were obtained for experimental use from the Banque Française des Yeux eye bank (Paris, France). After being stored at 31°C in a tissue-culture medium, corneas were placed for 48 h at room temperature in dextran 5% (CorneaJet, Eurobio, France) for deswelling and to reproduce the conditions of standard grafting procedures. The central corneal thickness (CCT) was measured by ultrasound pachymetry (Tomey SP 100, Nagasaki, Japan) before procedures.
Once the FS was set to the desired parameters, we secured the experimental cornea on a Barron disposable artificial anterior chamber (AC) (Katena, Denville, New Jersey) filled with deswelling fluid. The AC pressure was adjusted. We proceeded only if the epithelium was considered normal, and no opacity could be observed in the cutting area, by applanation on the epithelial side of the cornea. The procedure carried out was based on the cutting profile exposed in table 1, with a lamellar cut and a posterior side cut (cut 1) followed, for double path profiles, by an identical second full lamellar cut (cut 2) within the same applanation step (step 1). For double-layer profiles, we performed a 45° angulated anterior side cut instead of a posterior side cut. The AC pressure was readjusted after step 1. We measured intraoperatively the thickness of the residual stromal bed (RSB) at 12 separate locations with an ultrasound pachymeter equipped with sterile disposable probe tips. We proceeded to step 2 with a cornea only if the thinnest RSB exceeded 200 μm. Step 2 was carried out with a second applanation procedure, a 150 μm depth full lamellar cut and a 90° angulated posterior side cut. We completed the lamellar dissection of the anterior stroma with a double-ended LASIK spatula. The remaining collagen bridges were evaluated on a five-grade scale: absent (0), easy to remove and located in the peripheral area (1), easy to remove and located in the central area (2), strong (3), invincible adherences (4). The endothelial button was eventually separated from the mid-stroma. The procedures were videotaped. At the end of each procedure, we inspected every part of the cornea under a surgical microscope before further preparation for scanning electron microscope (SEM) examination.
Corneal sample analysis
In preparation for SEM examination, samples were fixed overnight in 2.5% glutaraldehyde and 1% paraformaldehyde in 0.1 M sodium cacodylate buffer at 4°C. Samples were rinsed in cacodylate buffer and postfixed in 2% osmium tetraoxide for 45 min on a rotator. Samples were dehydrated in a graded ethanol series and dried by rinsing with hexamethyldisilazane followed by air-drying. Samples were mounted on a carbon stub and sputter-coated. Stromal interfaces that obtained a clinical grade other than 4 were systematically observed. We produced images with a Cambridge Instruments Stereoscan 260 SEM (Leica, Cambridge, UK) equipped with a digital camera. We examined both sides of obtained interfaces, except for three lenticules from group 5 on which only the endothelial side was observed.
All procedures were completed uneventfully except for two from group 5 which were cancelled before step 2 after the RSB was measured at less than 200 μm. The optimal procedure yielding the lowest clinical grade with the lowest energy delivered that could be reproduced three times using identical settings to ensure reproducibility is detailed for each group in table 2. The lowest energy levels delivered to yield grade 3 or lower for the lamellar dissection of the posterior lenticule were 2.4 μJ, 1.05 μJ and 0.9 μJ respectively for groups 1–2, 3–4 and 5.
Quality of the stromal interface
The mean CCT before step 1 was 558 μm (SD±115 μm, range 551–609 μm). The mean central RSB measured after step 1 was 228 μm (SD±18 μm, range 202–261 μm). Three sets of parameters out of four procedures led to an incomplete dissection (grade 4): two in group 2 where both had identical FS settings (559 and 600 μm CCT; 500 μm depth; 0.8 μJ, 2:2 μm spot/step sizes), one in group 3 (555 μm CCT; 500 μm depth; 1.0 μJ, 5:5 μm spot/step sizes) and one in group 5 (215 μm RSB; 150 μm depth; 1.0 μJ, 4:4 μm spot/step sizes). Noticeably in group 5, the cutting failure was attributed to some fluid remaining on the stromal bed before step 2. Small residual bridges of uncut corneal stromal bed tissue (grade 1 and 2) were observed in the central area in groups 1 (2/6), 2 (1/3) and 5 (3/16) for full lamellar cut energies below 2.4 μJ, 1.2 μJ and 0.9 μJ, respectively, and in the peripheral area in group 1 (1/6) at 2.1 μJ and below. Overall, all procedures using double-layer profiles created interfaces that were easy to dissect and smooth with FS 60 kHz lamellar cut set on: 350 μm depth, 9.0 mm diameter, 2.1 μJ, spot size/step=4:4 for step 1 and 150 μm depth, 8.3 mm diameter, 0.9 μJ, spot size/step=4:4 for step 2 (figure 2).
Figure 3 displays the stromal sides of endothelial buttons created with the optimal setting based on clinical analysis for each group. Interfaces created after a single path profile show marked lamellar irregularities in both the central and peripheral area of the buttons' stromal side (figure 3A/a, arrowheads), although this was noticeably reduced with 150 kHz Intralase femtosecond Laser compared with 60 kHZ Intralase femtosecond Laser (figure 3aa versus 3bb). In contrast, in group 2, straight irregularities were observed crossing each other across the interface area (figure 3b, arrows). Moreover, spheroid irregularities of various diameters dotted the stromal interface in group 1 and, to a lesser extent, in group 3 (figure 3A/aa arrows). Interfaces appeared smoother in groups 3 and 4 after a double path pattern (figure 3cc/dd). However, some adherences still remained and generated irregularities in the central area of the button (figure 3c/d arrowheads). With appropriate FS settings (table 2), the double-layer profile created smooth and even interfaces across both the mid-stroma and the buttons' side (figure 3E/e/F/f). Such a quality of interface was not obtained with other profiles. It was possible to generate interfaces as smooth as the LASIK flap from procedures in group 6 (figure 4). No ultramicroscopic damage was observed on endothelial cells by SEM.
Endothelial lenticule interfaces created by mechanical dissection are reputed to be smoother than when precut with lasers. The advantages of the FS for cutting into corneal tissue are now broadly acknowledged for LASIK surgery. Should such advantages over microkeratome-assisted procedures be applicable to ELK button preparation, three issues would still remain.
The first issue is that donors' corneas are slightly oedematous and optically diffractive as a consequence of the eye-banking process in organ culture media, which induces a near quiescent state for endothelium. The optical aberrations accumulate proportionally as the beam is focused deeper through the stromal layers,13 by nonlinear self-focusing effects.14 This jeopardises the laser efficacy at cutting and probably initiated the adherences observed in groups 1 to 4. In addition, shock waves in oedema enlarged the cavitation bubbles and could explain the entities observed at the interface of group 1 lenticules (figure 3aa), causing cell damage.15 16 While performing a double path profile at 500 μm, we observed fewer interface adherences at 150 kHz, although both frequencies generated residual bridges across the central area of the lenticule (figure 3cc/dd). A high frequency allows a decrease in spot size and steps, typically to 2:2 μm instead of 4:4 μm. This lowered the energy required to achieve the lamellar cut efficiently and preserve endothelial cell viability.
The second issue is to determine the optimal amount of energy for a lamellar cut. The setting should be intense enough to penetrate deep towards the posterior stroma and overcome oedematous diffraction. At the same time, it should be limited to prevent any keratocyte activation or inflammatory reactions.17 The induced endothelial toxicity is debatable,10 15 since FS does not generate detectable thermal diffusion in corneal tissues.18
Third, the deeper into the corneal stroma the full lamellar cut, the rougher the interface created.13 More than light diffraction, the lamellae architecture across the cornea may explain this in part. The anterior stroma is made of large bundles of collagen fibrils19 20 with dense interlamellae branching, creating bridges between collagen lamellae for shear resistance.21 This prevents collagen lamellae bending out of shape after the anterior lamellar cut. Posterior collagen lamellae are less interweaved and less randomly distributed,20 22 which impairs the regularity of lamellar cuts performed in the posterior stroma.
Although procedures from the endothelial side are currently tested,23 ELK lenticules are conventionally prepared from the epithelial side. Our conventional choice was made by our reluctance to generate additional endothelial cell loss,23 either by contact or by pressure of applanation. With a single path profile, the laser penetrates deeper than 400 μm in the stroma, while the posterior stroma swells until the endothelial function is fully recovered. Thus, the single profile procedure both exposes to invincible adherences across the cutting bed and to non-linear lamellar cuts (figure 3A and 3B), thus prompting some surgeons to perform a full lamellar cut twice (groups 3 and 4). Such a profile dramatically decreases the occurrence and the strength of the bridges at the interface. However, this introduces the risk of performing two distinct lamellar cuts and does not address the swelling issue. Furthermore, we observed that it still induces a few irregularities (figure 3d). Although irregularities have been argued to improve the button's adherence to the receiving stromal bed,24 we think that functioning buttons should adhere spontaneously to the posterior stroma by the suction effect of the endothelium. In our opinion, an increased cicatrisation is more likely to impair long-term visual quality and should be prevented as much as possible.
The proposed double-layer profile addresses these issues by removing at first two-thirds of the anterior stroma and by creating the lenticule separately, only 150 μm away from the zero referential point of the laser. At 350 μm depth, the diffraction and optical aberrations are not obstacles to lamellar cuts of good quality. Focused far from the endothelium, the level of energy can be set to a relatively high intensity. We angled the anterior side cut of step 1 to ease the applanation during step 2. This minimised the circular meniscus of fluid accumulating around the edges of the primary cut. During the second step, only a thin part of the stroma remained to be removed. Optimally, for step 2, the lamellar cut, the depth should be customised on the basis of pachymetry measured between the two steps.
Being thin, ELK button can be inserted through short and self-sealing corneal incisions. The double-layer profile can create reproducibly lenticules less than 150 μm thick with little energy and less gas diffusing, thus preventing possible endothelial toxicity. Finally, it is critical to readjust the artificial AC pressure before step 2. The residual mid-stroma, thinner at this point, is better able to bend. It minimises the swelling of the posterior stroma and contributes to stress collagen lamellae. This could explain the unusual regularity observed across the interface area in group 5 (figure 3E/F) which now competes with LASIK flaps (figure 4).
Inevitably, this proof of concept for a new profile to create a lenticule for ELK has several limitations. We did not use software to quantify irregularities. Although none are validated for standard use, this could have provided quantitative data but could also have been misleading by quantifying the gas footprint, lenticule swelling or artefacts from laboratory manipulation as being irregularities. It would therefore have been difficult to form any conclusions. Assessment of thickness regularity was not compatible for lenticules with SEM preparation and interface observations. Ideally, thickness regularity should be evaluated on lenticules once the endothelial function has fully recovered, rather than being estimated with disputable relevance after organ-culture storage and an irregular deswelling process. In this study, we suggest a new method which increases the interface smoothness for endothelial lenticules created with FS lasers. A logical further step is now to compare FS lasers and mechanical microkeratomes set at identical depths for quality and reproducibility of lenticule interfaces.
We did not test all possible settings in this study; this would have multiplied the number of corneas needed. Rather, we chose to focus on lower energy settings when cuts were effective. The double-layer profile should now be tested for optimal settings with other FSs. Additional studies should investigate the endothelium viability of the buttons thus obtained and ultimately analyse the transplanted lenticules. Based on our observations, we claim that it is possible to create thin lenticules with smooth interfaces with a 60 kHz FS laser.
The authors would like to acknowledge I Sourati and P Sabatier from the Banque Francaise des Yeux for her valuable help in providing experimental corneas. The contribution of INSERM U872- team17 is specifically acknowledged and the authors would like to express their deep gratitude to Pr F. Behar-Cohen and her research team for their valuable support.
Funding Abbott Medical Optics (AMO France SAS, Biot, France) and the Centre d'Innovation Thérapeutique en Ophtalmologie (CITO) non-profit organisation (Paris, France) both supported in part the present work. AMO graciously provided femtosecond laser procedures. AMO and the CITO co-funded the scanning electron microscopy sessions.
Competing interests AB is an employee of Abbott Medical Optics France.
Provenance and peer review Not commissioned; externally peer reviewed.