Technique selection, orientation marking, separation mechanics, and the way the scroll is managed can each convert a viable donor specimen into a biologically compromised one before it ever reaches the anterior chamber.
Descemet membrane endothelial keratoplasty tissue is only approximately 10–15 microns thick. It contains donor corneal endothelium and Descemet membrane, without donor stroma. That absence of stromal support is precisely why DMEK can deliver a more anatomically selective transplant than thicker endothelial grafts—and why mechanical errors become disproportionately expensive. There is little tissue available to absorb force, compensate for a tear, or conceal endothelial injury.
The practical question is not whether DMEK tissue is delicate. Everyone agrees on that. The useful question is more adversarial: which parts of current preparation practice are genuinely protective, and which merely make the workflow feel controlled?
The mechanics of DMEK graft integrity: why thickness matters
A DMEK graft is not a miniature version of a DSEK or DSAEK graft. It is a different mechanical object. In DSEK-based procedures, donor posterior stroma adds thickness and some structural stiffness. In DMEK, the surgeon or technician is working with a composite of Descemet membrane and endothelial cells, a tissue layer that naturally scrolls and resists manipulation in ways that are not always intuitive.
That distinction matters at three stages:
1. During tissue separation, when traction or fluid pressure can injure the endothelial monolayer.
2. During orientation marking, when the marker itself can create localized mechanical trauma.
3. During unfolding and positioning, when the scroll must be opened without repeatedly contacting instruments, the chamber surface, or the recipient cornea.
The endothelial layer is not a passive coating. It is the functional component responsible for maintaining corneal deturgescence. Endothelial cell density declines with age and disease, and mature human endothelial cells have limited proliferative capacity in vivo. A graft can therefore appear structurally intact while carrying a biologically meaningful reduction in reserve.
This is the first methodological trap. Gross integrity is not the same as endothelial viability. A membrane without an obvious tear may still have suffered cell loss at its periphery, along the manipulation path, or around an orientation stamp. If assessment focuses only on whether the graft can be inserted, it answers the least demanding question.
A DMEK graft can survive the preparation table and still fail the biological test.
The thinness of the tissue also explains why “gentler” techniques cannot be judged by appearance alone. A method that produces a clean, easily handled membrane may still generate more endothelial cell death than a less visually reassuring alternative. The relevant endpoint is not elegance of the scroll. It is the preservation of functioning donor endothelium.
The integrity problem is cumulative
Mechanical injury rarely arrives as one dramatic event. More often, it accumulates through small contacts:
- lifting or peeling the membrane with uneven traction;
- allowing the tissue to fold against itself under pressure;
- repeatedly repositioning a scroll that has already been oriented;
- applying a larger orientation mark than necessary;
- exposing the endothelium to fluid movement that creates shear;
- delaying viability assessment until after multiple handling steps.
This cumulative pattern complicates quality control. If the final cell count is lower than expected, the cause may not be a single identifiable mistake. The preparation workflow itself may have created several minor insults that become significant in combination.
For researchers and tissue technicians, that means the protocol should be designed around minimizing total manipulation rather than optimizing one isolated maneuver. Reducing the number of contacts is often more meaningful than making each contact appear technically impressive.
Manual peeling versus hydrodissection: the decisive comparison
Hydrodissection is attractive because it appears to offer a clean separation mechanism. A liquid bubble can seem less invasive than direct manual peeling: fluid enters the interface, the tissue lifts, and the operator avoids some instrument-to-membrane contact. It is a persuasive story—and the comparative evidence is inconvenient for it.
Studies of graft harvesting have reported significantly higher endothelial cell death with hydrodissection, sometimes described as the liquid bubble technique, than with manual peeling. The point is not that fluid is intrinsically harmful or that every hydrodissection attempt produces catastrophic damage. The point is that the method creates a mechanical and hydrodynamic environment in which endothelial injury can be greater than with controlled manual separation.
That finding should disrupt the common assumption that fewer visible instrument movements necessarily mean less biological trauma. Fluid pressure is still force. A bubble advancing along the posterior corneal surface can generate shear, uneven separation, and localized stress at the interface. The absence of a forceps tip in the field does not mean the tissue is being treated gently.
What manual peeling does better
Manual peeling has its own risks. It demands tactile control, stable visualization, and restraint. Excessive traction can produce tears, peripheral defects, or endothelial stripping. The technique is not protective by definition; it is protective only when the operator controls the direction and magnitude of force.
Its advantage is more specific: manual peeling allows separation to proceed incrementally, with the tissue response visible at each stage. That makes it possible to stop, adjust the angle, or reduce traction before a small resistance becomes a larger defect. Hydrodissection can create a more distributed and less predictable force pattern, particularly when the liquid interface advances unevenly.
A useful comparison looks like this:
| Preparation factor | Manual peeling | Hydrodissection |
|---|---|---|
| Primary separation force | Direct, controlled traction | Fluid pressure and interface propagation |
| Main mechanical risk | Localized tearing or excessive pull | Shear, uneven separation, and pressure-related injury |
| Visual feedback | Continuous response of the tissue to traction | Less direct control over the advancing fluid plane |
| Reported endothelial effect | Lower cell death than hydrodissection in comparative harvesting studies | Significantly higher endothelial cell death in comparative studies |
| Technical demand | High tactile and visual control | Appears simpler, but does not eliminate mechanical risk |
| Practical implication | Minimize traction and unnecessary repetitions | Do not assume a liquid bubble is biologically gentler |
The conclusion is not a sentimental defense of traditional technique. Manual peeling should be preferred because the available comparative evidence favors it for endothelial preservation, not because older methods deserve deference. In corneal tissue handling, tradition is not a mechanism of action.
A controlled peeling workflow
A preparation protocol intended to preserve endothelial viability should treat manual peeling as a low-force, low-repetition process:
1. Establish a stable field before initiating separation.
Poor visualization encourages compensatory traction. If the operator cannot see the interface clearly, the tissue is already being asked to tolerate more force than the protocol can reliably monitor.
2. Separate progressively rather than pursuing speed.
The membrane should respond to controlled advancement. Resistance is information, not an invitation to pull harder.
3. Avoid repeated lifting and regripping.
Every additional grasp introduces another opportunity for focal endothelial injury. The goal is not merely to complete the peel, but to reduce the number of mechanical events required to complete it.
4. Treat peripheral resistance as a warning condition.
Abrupt changes in resistance may indicate attachment, an irregular interface, or a developing tear. Continuing with the same force because the protocol is “almost complete” is precisely how small defects become unusable tissue.
5. Keep orientation decisions separate from unnecessary manipulation.
Marking, unfolding, and repositioning should not become a sequence of repeated corrective movements. Each action should solve a defined problem.
This is where many DMEK graft preparation challenges become operational rather than theoretical. A technically successful peel can still be an inefficient peel, and inefficiency is often another name for avoidable exposure to injury.
Orientation marking: the smallest mark with the largest argument
Orientation marking is usually treated as a minor procedural detail. That is a mistake. The mark may be small, but it is applied directly to a graft whose endothelial surface has almost no tolerance for localized trauma.
Comparative evidence shows that orientation stamping induces endothelial cell damage. The amount of loss is lower with a smaller F-stamp than with a larger S-stamp. This is not a cosmetic preference between two symbols. It is a direct demonstration that the marking strategy changes the biological cost of orientation control.
The question should therefore be framed plainly: how much mark is actually needed to prevent inversion?
If a smaller F-stamp provides adequate orientation information, a larger mark has no obvious biological justification. It increases the contact area and potentially expands the zone of damaged or stressed cells. The larger symbol may be easier to recognize, but recognizability is not the only endpoint. A transplant tissue protocol should not reward visual convenience at the expense of endothelial reserve.
Why orientation errors are not solved by larger marks
The usual defense of a larger stamp is operational: it is easier to identify under difficult conditions. But that argument assumes the problem is a binary choice between visibility and safety. In practice, orientation control can be improved through the entire workflow:
- consistent marking location;
- standardized inspection before the graft is folded or scrolled;
- clear documentation of the mark used;
- minimizing the time between marking and implantation;
- avoiding repeated contact with the marked region;
- confirming orientation through more than one visual cue when appropriate.
A mark should provide orientation information, not become a second preparation event.
This is another point at which the consensus can be misleading. The aim is not to eliminate marking at all costs. The aim is to reduce marking-related injury while preserving reliable orientation. The smaller F-stamp is therefore not simply a preferred symbol; it is a practical compromise between navigation and cell preservation.
Orientation marking is not neutral. In a 10–15 micron graft, the label is part of the injury profile.
For tissue banks and research laboratories, the implication extends beyond the individual technician. If different operators use different stamp sizes, pressure, locations, or timing, the resulting grafts are not being prepared under one protocol. They are being exposed to several undocumented mechanical environments. Standardization should specify the marking method as carefully as it specifies the tissue source and preservation pathway.
Descemet membrane scroll management without unnecessary trauma
The scroll is often described as a defining inconvenience of DMEK. That description is accurate but incomplete. The scroll is not merely a nuisance to be flattened; it is the natural mechanical behavior of a thin, elastic tissue layer. Attempts to force it into a cooperative shape can produce more damage than the scroll itself.
The central principle is simple: do not turn every orientation problem into a manipulation problem.
A tightly rolled graft may tempt the operator to open it through repeated contact, sweeping motions, or aggressive fluid movement. Yet every additional attempt increases the opportunity for endothelial cell loss. The tissue should be managed with the smallest number of corrective actions compatible with reliable positioning.
This is particularly important when interpreting viability data. Reported endothelial cell loss after DMEK averaged approximately 37% at six months, 40% at one year, and 55% at five years in the cited cohorts. Those figures reflect the entire clinical pathway, not preparation alone. They should not be used to assign all postoperative loss to peeling or scrolling. But they do establish a sobering baseline: the graft enters surgery with no excess biological capacity to waste.
The early postoperative decline is substantial, followed by continued attrition over time. A reported linear decline in endothelial cell density of 118 cells/mm² per year between six months and eight years after DMEK further underlines the need to preserve as much viable endothelium as possible before implantation. Preparation is not the only determinant of long-term function, but it is the first avoidable source of loss.
What to record during graft preparation
A serious DMEK graft viability assessment should not rely on a single final observation. The preparation record should, where the laboratory workflow permits, distinguish between:
- donor tissue condition before peeling;
- presence of visible tears or peripheral defects;
- separation method used;
- orientation marker type and size;
- number of major handling or repositioning events;
- visible endothelial abnormalities after preparation;
- final tissue suitability for the intended application.
The purpose is not bureaucratic accumulation. It is causal reconstruction. Without a record of how the graft was prepared, a low postoperative or post-thaw cell count becomes difficult to interpret. Was the donor already marginal? Was there a mechanical insult during separation? Did the tissue undergo repeated attempts at unfolding? Did the marking technique add avoidable damage?
A database that records only donor identity and final status is not a quality system. It is an inventory list with better formatting.
Long-term graft survival: what the 10-year comparison actually says
The long-term data are often presented as a victory lap for DMEK, and the broad direction is justified. But the numbers deserve a more exact reading.
At ten years, median longitudinal endothelial cell loss was reported as:
- 63% for DMEK
- 68% for DSEK
- 76% for penetrating keratoplasty
The cumulative risk of graft failure between six months and ten years was reported as:
- 5% with DMEK
- 11% with DSEK
- 19% with penetrating keratoplasty
DMEK therefore showed lower long-term endothelial cell loss and lower cumulative graft failure than the compared techniques in that analysis. The result is clinically important, but it does not mean the procedure is biologically effortless. A 63% median endothelial cell loss at ten years is still profound attrition. The advantage is relative, not magical.
That distinction matters because triumphalist interpretations can produce a second dogma: if DMEK performs better than DSEK and penetrating keratoplasty, preparation details must be secondary. The evidence points in the opposite direction. A technique with superior long-term outcomes still depends on the viability of the initial graft. Better downstream performance does not make upstream injury irrelevant.
The comparison also needs to be kept anatomically honest. DMEK has no donor stromal tissue, whereas DSEK-type grafts contain additional posterior stromal tissue. The procedures differ in tissue composition, surgical manipulation, interface biology, and optical consequences. A simple ranking of failure percentages cannot explain every mechanism. It can, however, challenge the idea that a thinner graft is automatically a weaker graft in the long term.
Paradoxically, the tissue that is hardest to handle may be the tissue that produces the more favorable long-term biological profile—provided the preparation does not squander its endothelial reserve.
Postoperative adherence: gas tamponade is support, not repair
DMEK graft adherence depends on positioning the donor tissue against the recipient posterior cornea and supporting it with a gas bubble. A bubble of 20% sulfur hexafluoride, or air, is used in the anterior chamber to maintain contact while the graft adheres.
This step is sometimes discussed as though the gas bubble can compensate for imperfect preparation. It cannot. Tamponade supports an appropriately oriented graft; it does not restore cells lost during peeling, reverse a damaged endothelial surface, or repair a preparation-induced defect.
The distinction is operationally important:
- Preparation determines the starting biological quality.
- Positioning determines whether the tissue can function in the intended location.
- Gas tamponade supports contact during adherence.
Confusing these stages leads to poor troubleshooting. If a graft does not adhere, the cause may involve orientation, positioning, interface fluid, or insufficient support. If endothelial function is poor despite apparent attachment, the problem may instead be the quality of the donor tissue or the extent of cell injury sustained before and during implantation. The bubble is not a general-purpose correction mechanism.
The use of 20% SF6 also illustrates why protocol language must remain precise. “Use gas” is not a sufficient instruction for reproducibility. The selected gas, concentration, timing, and chamber management belong in the documented workflow. At the same time, no gas protocol should be presented as a substitute for atraumatic graft handling. The most carefully controlled postoperative support cannot recover endothelial cells that were lost on the preparation bench.
Building a more defensible DMEK preparation protocol
The strongest protocol is not the one with the most steps. It is the one that makes unnecessary injury difficult and makes deviations visible.
A practical framework for donor corneal tissue handling techniques should include five commitments:
1. Use manual peeling when the objective is endothelial preservation.
Comparative harvesting evidence reports higher endothelial cell death with hydrodissection than with manual peeling. The liquid bubble should not be treated as the default gentle option simply because it reduces direct instrument contact.
2. Minimize total manipulation.
Avoid repeated lifting, regripping, unfolding, and reorientation. Every maneuver needs a reason that is more compelling than operator convenience.
3. Use the smallest effective orientation mark.
A smaller F-stamp has been associated with lower endothelial cell loss than a larger S-stamp. Orientation should be reliable, but the marking strategy should not consume unnecessary endothelial reserve.
4. Assess viability beyond gross appearance.
An intact membrane is not automatically a healthy graft. Where the laboratory has the capability, endothelial assessment should be integrated into the workflow rather than postponed until a problem appears.
5. Separate preparation quality from postoperative mechanics.
Graft adherence, gas support, and final positioning matter, but they should not obscure the condition of the tissue before implantation.
The resulting protocol may look less dramatic than a technique built around speed and visual confidence. That is a virtue. DMEK is not improved by theatrical manipulation. It is improved by reducing the number of opportunities for invisible injury.
The research gap is not another heroic technique
The field does not need more enthusiasm for thin grafts without equally rigorous accounting of how those grafts are prepared. The evidence already supports several uncomfortable conclusions: hydrodissection can produce more endothelial cell death than manual peeling; orientation stamping is not biologically neutral; larger marks can be worse than smaller ones; and long-term DMEK outcomes, while favorable relative to DSEK and penetrating keratoplasty, still involve major endothelial attrition.
The unresolved questions are therefore narrower and more useful. How should donor age, preservation interval, baseline endothelial condition, and preparation stress be modeled together? Which viability assays best predict long-term graft performance rather than merely immediate post-preparation appearance? How much of the early cell loss reflects preparation, implantation, postoperative positioning, or recipient factors?
Those questions demand linked records, not isolated snapshots. A donor cornea should be traceable from procurement and preservation through peeling, marking, viability assessment, implantation, and follow-up. Otherwise, the field will continue to report endpoint survival while remaining uncertain about where the cells were actually lost.
DMEK graft integrity is not secured by treating the tissue as fragile. Fragility is a description, not a protocol. Integrity is secured by choosing the lower-injury separation method, reducing manipulation, limiting orientation trauma, documenting the preparation pathway, and refusing to confuse successful insertion with preserved biology.
The prevailing paradigm says DMEK is difficult because the graft is thin. The more consequential interpretation is that DMEK exposes how imprecise the preparation paradigm has been all along. The next advance will not come from making the membrane appear easier to handle. It will come from proving, step by step, which handling choices allow more endothelial cells to remain alive.