Predicting graft thickness in ultrathin DSAEK using a mechanical microkeratome system Authors: Giacomo Visioli, Luca Lucchino, Fabio Scarinci, Valentin Jünger, Giulio Pocobelli, Gemma Lodato, Antonella Falchi, Marco Marenco, Alessandro Lambiase, Augusto Pocobelli, Rossella Anna Maria Colabelli Gisoldi Published: 2026 ABSTRACT AbstractPurposeTo examine pre-cut factors associated with post-cut graft thickness in ultrathin Descemet stripping automated endothelial keratoplasty (UT-DSAEK) prepared with a mechanical microkeratome system and to develop a nomogram to support blade selection.MethodsDonor corneas processed at a single eye bank were included. All tissues were dissected using a standardised mechanical microkeratome protocol with fixed pressure and cutting speed. Pre-cut variables included central pachymetry, microkeratome blade size, endothelial cell density, cause of death and demographic characteristics. Multivariable regression was used to evaluate predictors of post-cut graft thickness. Adjusted predictions across the observed pachymetry range were generated to construct a nomogram estimating expected graft thickness for each blade.ResultsA total of 107 donor corneas were analysed. Mean post-cut graft thickness was 92.7±24.3 µm. In the multivariable model, blade size (p<0.001) and pre-cut pachymetry (p<0.001) showed the strongest associations with graft thickness, accounting together for approximately 45.7% of model variance. The cut-to-blade ratio analysis indicated that all blades produced cuts deeper than their nominal thickness, with ratios increasing across blade sizes. The resulting nomogram enables selection of the blade more likely to achieve grafts within the 70–100 µm range based on the donor’s initial pachymetry.ConclusionsBlade size and initial pachymetry show measurable associations with post-cut graft thickness in UT-DSAEK prepared under standardised mechanical conditions. The proposed nomogram offers a practical tool to assist in selecting the blade expected to achieve ultrathin grafts with greater reproducibility. WHAT IS ALREADY KNOWN ON THIS TOPIC DSAEK (Descemet stripping automated endothelial keratoplasty) graft thickness influences postoperative visual outcomes, with thinner grafts associated with reduced posterior corneal higher-order aberrations and improved best-corrected visual acuity. DSAEK (Descemet stripping automated endothelial keratoplasty) graft thickness influences postoperative visual outcomes, with thinner grafts associated with reduced posterior corneal higher-order aberrations and improved best-corrected visual acuity. Several donor-related and technical factors have been investigated as potential determinants of post-cut graft thickness; however, ultrathin graft preparation remains variable, and most available predictive models were not designed to predict outcomes under standardised mechanical cutting conditions. Several donor-related and technical factors have been investigated as potential determinants of post-cut graft thickness; however, ultrathin graft preparation remains variable, and most available predictive models were not designed to predict outcomes under standardised mechanical cutting conditions. WHAT THIS STUDY ADDS Under constant pressure and cutting speed, pre-cut central pachymetry and microkeratome blade size were the only variables independently associated with post-cut graft thickness. Under constant pressure and cutting speed, pre-cut central pachymetry and microkeratome blade size were the only variables independently associated with post-cut graft thickness. Based on these parameters, a practical nomogram was developed to estimate the expected graft thickness for each blade size, with preliminary testing showing close agreement between predicted and measured values. Based on these parameters, a practical nomogram was developed to estimate the expected graft thickness for each blade size, with preliminary testing showing close agreement between predicted and measured values. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY The proposed nomogram may assist eye banks and surgeons in selecting the most appropriate microkeratome blade to obtain grafts within the target ultrathin range, improving reproducibility while minimising the risks of underdissection or overdissection, perforation and donor tissue wastage. The proposed nomogram may assist eye banks and surgeons in selecting the most appropriate microkeratome blade to obtain grafts within the target ultrathin range, improving reproducibility while minimising the risks of underdissection or overdissection, perforation and donor tissue wastage. Introduction Descemet’s stripping automated endothelial keratoplasty (DSAEK) is a widely performed option for the treatment of corneal endothelial dysfunction. Although Descemet membrane endothelial keratoplasty (DMEK) has gained popularity, DSAEK is often favoured because of its shorter learning curve, greater technical simplicity and applicability in more complex surgical scenarios.1 2 These advantages have supported its continued use, but postoperative visual outcomes are strongly influenced by graft thickness, with thinner grafts (particularly those <100 µm) being associated with reduced posterior corneal higher-order aberrations and improved best-corrected visual acuity.3 4 To enhance optical performance while retaining the procedural benefits of DSAEK, refinements such as thin and ultrathin DSAEK (UT-DSAEK) have been introduced.5 Donor tissue for DSAEK is typically prepared using a microkeratome, and the resulting lenticule thickness is influenced by several factors, including the blade size, the initial corneal thickness and the pressure within the artificial anterior chamber.6 However, despite attempts to standardise the procedure, manual dissection remains subject to variability and complications such as buttonholing or perforation, especially when targeting ultrathin grafts.7 8 In recent years, mechanical microkeratome systems have been adopted to improve reproducibility by automating the cut and maintaining a constant internal pressure.6 9 Several studies have explored donor-related and technical parameters that might influence post-cut thickness, including donor age, initial pachymetry, cause of death, pressure settings and microkeratome speed.6 10–13 Some groups have developed nomograms to support blade selection, but many are based on retrospective data, limited sample sizes or post-implantation measurements, which do not necessarily reflect the true post-dissection thickness.14–16 However, most available models lack external validation and are not designed to predict outcomes under standardised mechanical conditions. The purpose of this study was to identify predictive factors for achieving the desired graft thickness during UT-DSAEK preparation using a mechanical microkeratome system under fixed pressure and cutting speed. Based on these variables, we developed a simple nomogram to support intraoperative decision-making and improve the reproducibility of outcomes. Methods Donor corneas were procured within 24 hours from death from donors aged under 80 years and processed at the Eye Bank of San Giovanni Addolorata Hospital, Rome, Italy, between January and December 2024. All tissues were preserved using a two-phase storage protocol. After procurement, corneas were transported under hypothermic conditions in Eusol-C (AL.CHI.MI.A. S.R.L., Ponte San Nicolò, Italy) at 2–8°C. On arrival at the eye bank, they were transferred to Tissue-C medium (AL.CHI.MI.A. S.R.L.) and cultured at 31°C, in accordance with the European Eye Bank Association Technical Guidelines and the manufacturer’s instructions for use.17 Forty-four hours before tissue dissection, all tissues were transferred to Carry-C medium (AL.CHI.MI.A. S.R.L.) to obtain deswelling.18 Tissue dissection was performed using a mechanical microkeratome system consisting of the Artificial Chamber Pressuriser (ACP; Moria, Antony, France) and the One-Use Plus Automated (OUP-A) unit. The artificial chamber was filled with balanced salt solution and maintained at a constant intra-chamber pressure, set between 199 and 207 mm Hg throughout the dissection. The anterior surface of each cornea was moistened prior to cutting; no epithelial removal or corneal dehydration was performed. Each cornea underwent a single microkeratome pass at a fixed cutting speed of 3.0 mm/s. The blade size was selected by the technician according to internal protocols, primarily based on the initial pachymetry value of the donor cornea. Microkeratome head sizes included 350, 400, 450, 500 and 550 µm. For each case, the dataset included donor age, sex, cause of death, endothelial cell density (ECD), central corneal thickness measured before dissection, storage temperature, blade size used for cutting, chamber pressure during dissection, cup diameter and graft thickness immediately after cutting. In addition to these primary variables, derived metrics such as the difference between corneal thickness and blade size (‘blade gap’) and the cut-to-blade ratio were calculated. ECD was obtained from the routine pre-cut eye-bank assessment performed to confirm tissue eligibility before microkeratome dissection. After vital staining with soluble lutein/trypan blue, the endothelium was examined by bright-field inverted light microscopy at 100×magnification (Nikon Eclipse TE200, Nikon, Tokyo, Japan). ECD was manually calculated in the central endothelial area using a calibrated 100×100 µm reticule, averaging at least five separate counts and expressing the result as cells/mm². Pachymetric measurements were obtained using high-resolution anterior segment optical coherence tomography (AS-OCT; Casia SS-1000, Tomey Corporation, Japan), both before and after dissection. 18 Figure 1 shows representative endothelial images obtained before cutting and after repeat trypan blue staining following microkeratome dissection, with no appreciable post-cut staining suggestive of gross endothelial damage. Representative endothelial images before and after microkeratome dissection. (A) Pre-cut endothelial appearance after trypan blue staining in an organ-cultured cornea from a 64-year-old female donor, showing a regular endothelial mosaic, pre-cut endothelial cell density of 2600 cells/mm², and no evident endothelial mortality. (B) Post-cut endothelial appearance after repeat trypan blue staining in the same tissue. Dissection was performed with a 450 µm blade, pre-cut pachymetry of 560 µm, and artificial anterior chamber pressure of 200 mm Hg, obtaining a 60-µm-thick, 9-mm-diameter graft. No appreciable post-cut staining suggestive of gross endothelial damage is evident. The post-cut thickness, measured immediately after cutting, was used as the primary outcome. Grafts were categorised into three groups according to the measured thickness: less than 70 µm, between 70 and 100 µm, and greater than 100 µm. The range between 70 and 100 µm was defined as optimal based on current literature and surgical experience, corresponding to the ultrathin DSAEK category.3 5 Based on the identified predictors, a nomogram was subsequently developed to support microkeratome blade selection under standardised cutting conditions. To provide a preliminary internal validation of the proposed nomogram, an additional subset of 30 donor corneas prepared under the same standardised protocol was analysed separately from the model-development cohort. In this validation subset, the microkeratome blade was selected according to the nomogram-derived prediction. The measured post-cut thickness, obtained by AS-OCT immediately after dissection, was then compared with the thickness predicted by the nomogram. This study was based on the retrospective analysis of donor corneas processed within an eye bank setting, with technical and procedural outcomes related to graft preparation. Given the nature of the research question and outcomes, patient or public involvement was not considered applicable. Statistical analysis All statistical analyses were performed using Stata V.18.0 (StataCorp LLC, College Station, Texas, USA). Continuous variables were summarised as mean±SD or as median and IQR, depending on distribution. Categorical variables were reported as absolute frequencies and percentages. Comparisons across graft-thickness groups were performed using one-way analysis of variance or Kruskal-Wallis tests for continuous variables, and χ2 or Fisher’s exact tests for categorical variables. A panel regression model was first used to estimate, for each microkeratome blade size, the adjusted ratio between the actual cut depth and the nominal blade thickness. Marginal effects, together with their 95% CIs, were calculated and used to generate a plot illustrating the expected cut-to-blade ratio for each blade size. A multivariable regression model was then constructed to identify predictors of post-cut graft thickness. Candidate independent variables were selected based on univariate findings and clinical plausibility. Model assumptions and multicollinearity were assessed using standard diagnostic procedures. Model performance was quantified using the pseudo-R² derived from the ratio between explained and total model variance. Finally, adjusted predictions of graft thickness were calculated across the observed range of donor pachymetry for each blade size, and the regression equation from the final reduced model was used to construct a nomogram to support blade selection under standardised cutting conditions. Results A total of 107 corneas from 107 donors were included. The median donor age was 63 years (IQR 16), 57.0% were male, and the mean endothelial cell density was 2605±144 cells/mm². Cardiovascular disease represented the most common cause of death (55.1%). The median pre-cut central pachymetry was 562 µm (IQR 78). Table 1 summarises donor characteristics and technical parameters across the three thickness groups; the mean post-cut graft thickness was 92.7±24.3 µm. Based on these values, 9 grafts (8.4%) were <70 µm, 59 (55.1%) were between 70 and 100 µm, and 39 (36.4%) were >100 µm. In the 70–100 µm group, the most frequently selected microkeratome heads were 450 and 500 µm, with a mean blade–cornea gap of 126.4±21.9 µm. The <70 µm and >100 µm groups showed mean gaps of 112.6±25.0 and 144.1±22.7 µm, respectively (p<0.001). Donor age, sex, cause of death, endothelial density and cutting pressure did not differ statistically across groups. Figure 2 shows the adjusted ratio between the actual cut depth and the nominal blade size. All blades produced cuts deeper than their stated thickness, with the cut-to-blade ratio increasing progressively as blade size increased. Table 1Donor characteristics and technical parameters of graft preparation stratified by final graft thicknessTotal eyes (107)Graft thickness <70 µm (n. 9)Graft thickness ≥70 and <=100 µm (n. 59)Graft thickness >100 µm (n. 39)P valueAge, median (IQR)63 (16)67 (22)63 (15)62 (19)0.939Sex (female), n (%)46 (43.0)1 (11.1)28 (47.5)17 (43.6)0.131Cardiovascular death, n (%)59 (55.1)6 (66.7)29 (49.2)24 (61.5)0.428Endothelial count, mean±SD2605±1442589±1902616±1382590±1430.623Cold storage temperature (∼4°C)31 (29.0)4 (44.4)20 (33.9)7 (17.9)0.131Pre-cut thickness (µm), median (IQR)562 (78)590 (78)586 (91)552 (70)0.162Microkeratome (µm), mean±SD443.5±51.4488.9±41.7452.5±52.9419.2±39.1 <0.001 Actual cut (µm), median (IQR)479 (97)560 (91)496 (94)441 (55) <0.001 Cutting pressure (mm Hg), mean±SD201.6±1.8200.3±1.1201.7±1.9201.9±1.60.057Cup diameter (mm), mean±SD10.0±0.510.0±0.210.0±0.69.9±0.40.332Blade-cornea thickness gap (µm), mean±SD131.7±21.9112.6±25.0126.4±21.9144.1±22.7 <0.001 Post-cut thickness (µm), mean±SD92.7±24.337.2±28.286.4±8.3114.9±8.8<0.001Cutting thickness difference (µm), mean±SD39.0±25.375.3±28.640.0±22.129.2±21.3<0.001Actual cut-blade ratio, mean±SD1.08±0.051.15±0.051.09±0.051.07±0.05 <0.001 The table reports demographic data, pre-cut corneal measurements, microkeratome settings and post-cut outcomes for each thickness group. Bold values indicate statistically significant results (p<0.05).Bold values indicate significant results (p<0.05). Donor characteristics and technical parameters of graft preparation stratified by final graft thickness The table reports demographic data, pre-cut corneal measurements, microkeratome settings and post-cut outcomes for each thickness group. Bold values indicate statistically significant results (p<0.05). Bold values indicate significant results (p<0.05). Adjusted cut-to-blade ratio for each microkeratome blade size. The plot shows the predicted ratio between the actual cut depth and the nominal blade thickness, with 95% CIs. All blades produced cuts deeper than their stated thickness, and the cut-to-blade ratio varied across blade sizes, with the highest predicted ratio observed for the 550 µm blade. Wider CIs for the 350 and 550 µm blades may reflect the smaller number of observations available for these blade sizes. Multivariate regression identified microkeratome blade size (p<0.001) and pre-cut pachymetry (p<0.001) as the only variables independently associated with post-cut graft thickness. Sex and cardiovascular death were retained in the model but did not reach statistical significance. The final population-averaged model showed a pseudo-R² of 0.457. The corresponding regression equation was: Regression coefficients and CIs are reported in table 2. Table 2Population-averaged multivariable regression model for predictors of post-cut graft thicknessVariableCoefficient (β)SEZP value95% CIMicrokeratome blade (µm)–0.73140.0789–9.27 <0.001 –0.8860 to –0.5768Pre-cut pachymetry (µm)+0.48910.07106.98 <0.001 +0.3517 to +0.6266Female sex+5.46493.78171.450.148–1.9471 to +12.8770Cardiovascular death+5.09563.85051.320.186–2.4511 to +12.6422Intercept130.7118.357.12 <0.001 94.75 to 166.67Pseudo-R²: 45.73%. Bold values indicate statistically significant results (p<0.05). Population-averaged multivariable regression model for predictors of post-cut graft thickness Pseudo-R²: 45.73%. Bold values indicate statistically significant results (p<0.05). A linear nomogram was generated from the regression equation to facilitate intraoperative blade selection, and its structure is shown in figure 3. Adjusted predictions of post-cut graft thickness across the observed range of donor central pachymetry (pre-cut central corneal thickness) for each microkeratome blade size. For any given pre-cut thickness, the model estimates the expected post-cut graft thickness associated with each blade, generating a prediction curve for every blade size. These curves form the basis of the nomogram: by identifying the pre-cut pachymetry of a donor cornea and the desired post-cut thickness, the operator can select the blade most likely to achieve the target outcome. Error bars denote 95% CIs. The nomogram was tested on 30 consecutive donor tissues. The measured post-cut graft thickness, the nomogram-predicted thickness and the difference between the two values are reported in online supplemental table 1. The mean difference between measured and predicted thickness was −1.23±13.40 µm. Discussion In this retrospective study, we analysed a consecutive series of 107 donor corneas prepared for DSAEK using a mechanical microkeratome system under constant pressure and cutting speed. The aim was to identify the factors most strongly associated with post-cut thickness and to develop a predictive model to support intraoperative decision-making. The analysis identified blade size and pre-cut central pachymetry as the only variables significantly associated with post-cut graft thickness. Based on these parameters, we generated a linear formula and a practical nomogram to estimate the expected graft thickness. Among the variables considered, the difference between initial corneal thickness and blade size (blade–cornea gap) emerged as a particularly informative parameter. Grafts within the 70–100 µm range, which we considered optimal based on both visual outcomes and surgical manageability, had a mean gap of approximately 126 µm.5 19 This value may serve as a practical threshold for blade selection, especially when using mechanical systems under fixed-pressure conditions. While this concept has been implicitly suggested in previous work, our study provides a direct quantification of the optimal range and supports its utility with regression-derived modelling.12 Our findings are consistent with those of Bae et al regarding the predictive value of pre-cut pachymetry, although their nomogram was generated from a large eye bank database with mixed cutting protocols and did not specifically target ultrathin outcomes.11 However, they also reported that older donor corneas tend to yield thinner grafts, whereas Holland et al observed the opposite trend, with younger corneas producing thinner cuts due to increased stromal pliability.20 In our series, donor age did not emerge as a significant factor, a discrepancy that may be explained by differences in tissue processing conditions as well as variations in donor selection criteria. We did not find a statistically significant effect of cutting pressure on post-cut thickness, despite previous studies suggesting a possible relationship, with higher intra-chamber pressure during microkeratome dissection being associated with deeper cuts.9 20 21 It is plausible that the narrow range of pressures used in our cohort (199–207 mm Hg) limited the ability to detect such an effect.22 In experimental comparisons, higher and less stable pressure conditions were associated with greater numerical endothelial cell-density reduction and cell loss after cutting, although overall endothelial mortality remained below 10%.21 While a large eye-bank series reported the use of 240 mm Hg without increased immediate post-cut trypan blue positivity compared with conventional pressurisation around 200 mm Hg, immediate staining may not capture subtler functional injury or delayed endothelial cell loss, as experimental high-IOP models have shown endothelial pump–barrier disruption, morphological damage, apoptosis and ECD loss after the acute pressure insult.23–25 Higher pressure settings were deliberately avoided to minimise unnecessary endothelial stress while maintaining standardised cutting conditions.9 Similarly, donor sex and cardiovascular cause of death showed non-significant trends and were included in the model for exploratory purposes. Nishisako et al recently reported a deeper average cut in donor corneas from individuals who died of cardiac causes, a finding which we could not replicate with statistical significance, possibly due to sample size limitations or differences in tissue hydration.6 Several studies have proposed alternative strategies to optimise graft thickness, including pre-cut drying techniques, epithelial removal, the double-pass technique and the slow-pass technique.7 9 14 15 26 However, many of these approaches are time-consuming, operator-dependent or not easily transferable across different eye banks. Our method, based on parameters available before dissection, could offer a more accessible and reproducible solution that can be easily incorporated into routine eye bank practice. The preliminary validation subset supports the potential utility of the nomogram as a practical aid for blade selection. Rather than replacing eye-bank or surgical judgement, the model provides an objective estimate of expected post-cut thickness when choosing between adjacent blade sizes. The close agreement between predicted and measured values suggests good cohort-level calibration, although residual errors reflect the expected biological and technical variability of donor tissue preparation. However, this represents an initial implementation experience under standardised conditions, and larger independent series are required to confirm its generalisability. Importantly, DSAEK lenticule evaluation should not rely on central graft thickness alone, as this parameter does not capture thickness symmetry, posterior surface regularity or stromal interface quality. This distinction is clinically relevant because thicker DSAEK grafts have been associated with greater posterior corneal asymmetry, which correlates with posterior corneal higher-order aberrations (HOAs) after DSAEK.3 Similarly, increased posterior HOAs have been attributed to uneven donor-lenticule thickness and curvature mismatch between the host cornea and donor lenticule,27 28 while interface opacity may partly result from irregular microkeratome cuts causing stromal fibre disruption.29 Therefore, microkeratome-assisted UT-DSAEK should aim not only for a target central thickness but also for a reproducible graft profile and a regular stromal interface. Experimental studies have shown that artificial chamber pressurisation and cutting systems can affect cut-depth reproducibility, peripheral graft-thickness uniformity and stromal surface smoothness.21 25 30 31 In particular, Nishisako et al showed that controlled high-pressure chamber settings and automated cutting improved reproducibility and peripheral uniformity, although residual graft thickness varied along the microkeratome pass with thinner tissue at the entry side than at the exit side. Moreover, stromal roughness, assessed by scanning electron microscopy, increased with deeper cuts.21 In the present study, all blades produced cuts deeper than their nominal thickness, with an increasing cut-to-blade ratio across blade sizes, suggesting blade-related differences in cutting performance. Although no macroscopic differences in graft profile or stromal surface appearance were observed, subtle effects on final lenticule morphology cannot be excluded.30 Future studies using AS-OCT profile mapping and scanning electron microscopy (SEM)-based roughness analysis in non-transplantable tissues should assess whether blade-related differences in cut depth translate into measurable differences in thickness distribution, graft symmetry, posterior surface regularity and stromal bed smoothness. This study has several limitations. First, the retrospective nature of the analysis may introduce selection or information biases, although all tissues were prepared under homogeneous technical conditions by experienced personnel. Second, the model explains approximately 45% of the variance in graft thickness, indicating that unmeasured factors, such as stromal rigidity, sclerocorneal button, hydration gradients, may also play a role. Third, our sample size was limited compared with large-scale registry studies, although the uniformity of technique strengthens the internal validity of the findings. Although a preliminary nomogram-guided implementation series was performed, validation in larger cohorts remains necessary. Finally, systematic post-cut endothelial viability assessment was not performed, as additional post-cut staining would have required further manipulation of transplantable donor tissues. However, dedicated endothelial safety analyses should assess pre-cut and post-cut endothelial viability to determine whether blade thickness influences endothelial cell loss, particularly given the higher theoretical risk of endothelial trauma during ultrathin graft preparation. In conclusion, this study identifies initial corneal thickness and blade size as the main determinants of graft thickness in DSAEK when using a mechanical microkeratome system under standardised conditions. The nomogram we propose may assist eye banks and surgeons in selecting the most appropriate cutting parameters to obtain grafts within the target ultrathin range, improving reproducibility while minimising the risk of underdissection or overdissection. This may also help minimise complications such as perforation and avoid donor tissue wastage. Prospective validation and integration of additional biomechanical parameters, such as corneal hysteresis and stiffness, may further enhance the predictive capacity of the model. Supplementary material