Study design

This retrospective single-center cohort study was conducted in the Metz-Thionville Regional Hospital (Grand Est, France) between October 2015 and April 2021. It was approved by the Ethics Committee of the French Society of Ophthalmology (Approval No. 00008855). All procedures were conducted in accordance with the guidelines of the Declaration of Helsinki. All patients were informed before surgery that their surgery-related data might be used for research. All consented in writing to this possibility. The consent procedure was conducted in accordance with the reference methodology MR-004 of the National Commission for Information Technology and Liberties of France (No. 588909 v1).

Patient selection and data collection

The medical records, which are maintained prospectively, were searched retrospectively for all consecutive patients who met the following inclusion criteria: (1) adult (≥18 years); (2) eye with FECD; (3) underwent DMEK alone (if the affected eye was pseudophakic) or triple-DMEK (cataract surgery followed by DMEK) between October 2015 and April 2021; and (4) followed for at least 12 postoperative months. Patients were excluded from the cohort if: (1) they had another ophthalmological disease that could interfere with visual acuity (e.g. retinal detachment, central retinal vein occlusion, severe age-related macular degeneration, or advanced glaucoma); (2) they had a history of previous corneal graft; (3) preoperative BSCVA data were missing; or (4) the patient was lost to follow-up. Eyes were excluded from analysis if: (1) they were the second operated eye in bilateral DMEK cases; or (2) the treated eye exhibited graft failure in the 12 postoperative months. Note that because we routinely use DSAEK to treat patients with very thick and extremely decompensated corneas due to long-standing bullous keratopathy, the study cohort did not include such patients.

Preoperative and postoperative tests and data collected

All patients underwent the following tests before surgery and 8 and 15 days and 1, 3, 6, and 12 months after surgery: BSCVA (expressed in logMAR), CCT with non-contact ultrasonic pachymetry (Tono pachymeter NT-530P; Nidek Co., Gamagori Aichi, Japan), and anterior-segment optical coherence tomography (OCT) (NIDEK with a special module; Nidek Co., Gamagori Aichi, Japan). Mean anterior keratometry was also conducted before surgery (Visionix Luneau L67 auto-kerato-refractometer, Pont-de-l’Arche, France). Graft endothelial cell density (ECD) was provided by the eye bank.

The following data were collected: patient age and gender; preoperative and postoperative BSCVA and CCT; preoperative mean anterior keratometry, K1, and K2; donor age; preoperative (graft) ECD; graft origin; operative time; use of triple-DMEK; and number of rebubblings, graft rejections, Irvine-Gass syndrome cases, and graft failures requiring regrafting.

BSCVA recovery between day 8 and 3 months was calculated as (day 8 BSCVA—3 month BSCVA). Postoperative change in CCT relative to preoperative CCT was denoted Delta CCT. For example, Delta CCT Day 8 = (day 8 CCT—preoperative CCT).

Surgical technique and postoperative care

To prevent pupillary block during and after surgery, all patients underwent iridotomy at the 6 o’clock position with Nd: YAG Laser (Laser ex-super Q; Ellex Europe, Medical Quantel, Cournon d’Auvergne, France) during a preoperative consultation. All surgeries were performed by the same experienced microsurgeon (JMP). DMEK was performed as described by Melles [2]. In all cases except two, surgery was performed under general anesthesia due to surgeon preference; in the exceptions, general anesthesia was contraindicated and surgery was conducted under locoregional anesthesia (peribulbar block using 23-gauge Atkinson cannula with a 50/50 mixture of 2% lidocaine and 7.5 mg/mL ropivacaine).

Unprepared grafts were obtained from the tissue banks in Besançon or Nancy, France. All grafts were preserved by organ culture in Eurobio organ culture medium at 31°C. They were transported in medium with dextran and were exposed to dextran for 48 to 72 hours. All grafts had a requested ECD greater than 2200 cells/mm². The following steps have been described in detail previously [22]. Thus, the graft was prepared by the microsurgeon in the operating room just before surgery, after which the patient’s cornea was prepared for transplantation and the graft was injected into the anterior chamber. After positioning the graft, a sterile air (75% of cases) or 20% sulfur hexafluoride (25%) bubble was injected into the anterior chamber. In approximately half of the surgeries, the main incision was sutured with one point of Nylon 10.0 that was secondarily buried and removed 1 month later. In the case of triple-DMEK, phacoemulsification was performed before DMEK as described previously [23]. Intracameral steroids were not applied.

In the postoperative period, all patients were placed in the supine position for the first 12 hours. All patients underwent treatment with a topical antibio-corticosteroid (Maxidrol, Dexamethasone + Neomycin Polymyxine B; ALCON, Rueil Malmaison, France) four times a day. An ophthalmic ointment (Vitamin A dulcis; ALLERGAN, Courbevoie, France) was also applied two times a day to promote healing. Two months after surgery, Maxidrol was replaced with long-term low-dose corticosteroid eye drops (FLUCON; NOVARTIS Pharma, Rueil Malmaison, France).

Patients who developed Irvine-Gass syndrome, as confirmed by anterior-segment optical coherence tomography (AS-OCT) (NIDEK; Nidek Co., Gamagori Aichi, Japan), and loss of visual acuity were treated with oral acetazolamide (Diamox®; Sanofi, Gentilly, France; one 250 mg tablet 3 times/day for one month) and NSAID (indomethacin 0.1%; Chauvin, Montpellier, France; 4 times/day for one month). Cases of allograft rejection were never observed in our series.

If AS-OCT revealed detachment of 20% or more of the graft or the detachment was threatening the visual axis [24], rebubbling was performed by injecting a sterile air bubble in the anterior chamber under topical anesthesia.

A graft was considered to have failed if edema persisted after 3 months, or nonadhesion of the graft was observed despite up to four rebubblings. In these cases, regraft with DMEK or DSAEK was conducted. Graft failure due to technical problems such as graft inversion was included in this definition.

Statistical analyses

Since graft failure rates are an important measure of the success of DMEK, patients with graft failure were included in the cohort but then excluded from all statistical analyses. The second-operated eyes in bilateral DMEK cases were also excluded from all analyses. Continuous data were expressed as mean±standard deviation and range while categorical variables were expressed as n (%). Pearson’s correlation tests followed by Bonferroni correction were conducted to determine the relationships between (i) preoperative CCT and pre/postoperative BSCVA, day 8-month 3 BSCVA recovery, or postoperative CCT and (ii) 12-month BSCVA and postoperative CCT or Delta CCT. Eyes were also divided into two subgroups according to the ROC-determined 625-μm preoperative CCT threshold reported by Brockmann et al. [19]. The subgroups were compared in terms of pre/perioperative variables, postoperative BSCVA, or day 8–3 month BSCVA recovery by Student’s t-test or Fisher’s exact test, as appropriate, followed by Bonferroni correction. All statistical analyses were performed with SAS software (version 9.4, SAS Inst., Cary, NC, USA). P-values <0.05 were considered to indicate statistical significance.

Patient disposition in the study

During the study period, 185 eyes underwent DMEK. Of these, 24 were excluded because they had another ophthalmological disease that could interfere with visual acuity (n = 8; two with epiretinal membrane, two with post-uveitic endothelitis, one with severe AMD, and three with retinal vein occlusion), they had undergone corneal grafting previously (n = 11), preoperative BSCVA data were missing (n = 4), or they were lost to follow-up (n = 1). Moreover, 18 eyes that required a second graft during the first postoperative year due to primary graft failure and 21 second-operated eyes were excluded from statistical analysis. Thus, 124 eyes (124 patients) were included in the statistical analysis (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart depicting the disposition of the subjects during the study.

AMD, age-related macular degeneration; DMEK, Descemet membrane endothelial keratoplasty; DSAEK, Descemet stripping automated endothelial keratoplasty.

https://doi.org/10.1371/journal.pone.0282594.g001

Baseline clinical and operative characteristics of the cohort

The patients were on average 72 years old and 63% were women. The average preoperative CCT, BSCVA, and mean anterior keratometry were 626 μm, 0.66 logMAR, and 44 D, respectively. The donors were on average 74 years old and the grafts had a mean preoperative ECD of 2542 cells/mm². The grafts were slightly more likely to come from Besançon rather than Nancy (57% vs. 43%). Triple-DMEK was conducted in 45% of the cases. DMEK surgery lasted on average 35 min (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Baseline and operative characteristics of the cohort (n = 124 eyes and 124 patients).

https://doi.org/10.1371/journal.pone.0282594.t002

Postoperative outcomes

The DMEK graft failure rate was 10% (18 of the 185 total DMEK surgeries during the study period). This reflects the learning curve of the surgeon and the fact that we switched from air to SF6 for bubbling only recently. In the last study year, our primary failure rate was 3%. Rebubbling was required in 37 of the 124 analyzed study eyes (30%), of which 11 (9% of the whole cohort) required multiple rebubblings. Six eyes (5%) presented with Irvine-Gass syndrome during follow up. All cases resolved with treatment. There were no cases of graft rejection (Table 2).

Starting 15 days after surgery, BSCVA improved markedly from 0.66 logMAR at baseline to 0.08 logMAR at 12 months. This associated with a gradual decrease in CCT from 626 μm at baseline to 538 μm at 12 months (Fig 2 and S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Change in BSCVA and CCT After DMEK.

BSCVA, best spectacle-corrected visual acuity; CCT, central corneal thickness.

https://doi.org/10.1371/journal.pone.0282594.g002

Correlations between preoperative CCT and postoperative BSCVA and CCT

Preoperative CCT did not correlate with preoperative BSCVA or BSCVA at 12 months or earlier time points. It also did not correlate with early recovery in BSCVA (defined as change in BSCVA at 3 months relative to day 8) (Table 3) or postoperative CCT (S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Correlations of preoperative CCT with postoperative BSCVA or BSCVA recovery.

https://doi.org/10.1371/journal.pone.0282594.t003

Comparison of eyes with low and high preoperative CCT in terms of postoperative BSCVA

The ROC analysis of Brockmann et al. [19] showed that a preoperative CCT threshold of 625 μm had the greatest discriminatory power. When we applied this threshold to our 124 eyes, we found that 70 and 54 had preoperative CCTs of <625 and ≥625 μm, respectively. The two subgroups did not differ significantly in terms of patient or donor age, patient sex, or graft ECD. Importantly, they did not differ in either preoperative or postoperative BSCVA (Table 4 and Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Change in BSCVA in eyes with low (<625 μm) and high (≥625 μm) preoperative CCT.

BSCVA, best spectacle-corrected visual acuity; CCT, central corneal thickness.

https://doi.org/10.1371/journal.pone.0282594.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Comparison of eyes with low preoperative CCT (<625 μm) and high preoperative CCT (≥625 μm) in terms of pre/perioperative variables, postoperative BSCVA, and BSCVA recovery.

https://doi.org/10.1371/journal.pone.0282594.t004

As could be expected given the lack of correlations between preoperative and postoperative CCT (S2 Table), the low and high preoperative CCT subgroups did not differ in postoperative CCT (S3 Table).

Correlations between postoperative CCT and 12-month BSCVA

We then explored the relationships between CCT at various postoperative timepoints and 12-month BSCVA. Notably, postoperative CCT at later time points (1, 3, 6, and 12 months) correlated significantly with 12-month BSCVA (r = 0.29, 0.45, 0.49, and 0.45; p = 0.020, 0.001, 0.001, and 0.001, respectively). However, this was not observed for CCT at days 8 and 15 (Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Correlations of 12-month BSCVA with CCT at various postoperative timepoints.

https://doi.org/10.1371/journal.pone.0282594.t005

We also examined how change in CCT (relative to baseline; designated Delta CCT) at various time points correlated with 12-month BSCVA. Change in CCT at 3, 6, and 12 months also correlated significantly with 12-month BSCVA (r = 0.33, 0.34, and 0.28; p = 0.002, 0.001, and 0.018, respectively) (Table 5).

Correlation between postoperative CCT and 12-month BSCVA

Our failure to find a correlation between preoperative CCT and post-DMEK BSCVA has also been observed by three other studies, including two with large cohorts (n = 160 and 1084) [14, 15, 21]. However, a study by Peraza-Nieves et al. on 393 DMEK eyes found with multivariate analysis that 1-μm higher preoperative CCT predicted 0.0008 higher logMAR BSCVA at 24 months [6]. Since their mean±standard deviation preoperative CCT was 667±92 μm, that meant an eye with 92-μm higher CCT (i.e. the average difference between the CCT of a given eye and the average CCT of all eyes) was predicted to have 0.074 logMAR higher BSCVA than average at 24 months (Table 1).

The reasons for the discrepancy between the studies of Brockmann et al. and Peraza-Nieves et al. versus our and the other three studies are not clear. One possibility is that the eyes in the former studies had longer standing disease and had developed stromal haze that decreased postoperative BSCVA recovery. However, the FECD stage ranges of these studies seem similar to those in our study.

Another possibility is suggested by our earlier study on the relationship between central graft thickness (CGT) and final BSCVA after DSAEK [30]. Whether thinning DSAEK grafts actually improves visual outcomes has been a vexed question in the literature, with some studies [31–36] but not others [37–44] finding an effect. Our study showed that postoperative, but not preoperative, CGT predicted final BSCVA. Therefore, we suggested that the variable CGT-BSCVA relationships reported in the literature may reflect inaccuracies or variation in the preoperative CGT measurements due to various factors (e.g. graft culture duration) that blur the relationship between CGT and visual acuity; these factors disappear after surgery, allowing the intrinsic relationship between CGT and BSCVA to emerge [30]. Our explorations in the present study suggested that a similar phenomenon may be operating in the relationship between CCT and BSCVA in DMEK: we found that unlike preoperative CCT, postoperative CCT did correlate significantly with final BSCVA. Specifically, thinner corneas in the postoperative period associated with better postoperative BSCVA. The relationship started appearing at 1 month (r = 0.29, p = 0.020) and then grew stronger and persisted through to the end of follow-up (r = 0.45–0.49, all p = 0.001). To our knowledge, this is the first time this relationship has been examined or reported.

Thus, we speculate that preoperative CCT does not always show a relationship with final BSCVA in the literature (Table 1) due to interstudy variation in the preoperative CCT measurements, which may be induced by disease- and/or preoperative setting-related factors. These factors disappear after DMEK, allowing the inherent relationship between CCT and BSCVA to become more apparent. This speculation is supported by the smaller variation in our CCT measurements at 12 months compared to our preoperative CCT measurements (standard deviation: 41 vs. 59 μm). It is also supported by the variable relationships between preoperative and postoperative CCT in the literature: Brockmann et al. [45] found that these two variables correlated strongly whereas we and another study that also did not find a CCT-BSCVA relationship did not observe any such correlation [46] (Table 1).

These findings together suggest that while CCT does have a relationship with post-DMEK visual acuity, preoperative measurement variation means this variable may not be a reliable predictor of visual outcomes after DMEK.

The relationship between CCT and BSCVA likely reflects the fundamental healthiness of the graft corneal endothelial cells, specifically their ability to resorb intrastromal fluid, thereby reducing the thickness of the cornea. This notion is supported by multiple in vitro and in vivo studies showing that corneal ECD dictates the efficiency of the endothelial pump [47–50]. Moreover, the Cornea Donor Study showed that postoperative CCT predicts graft failure (p = 0.002) and associates modestly with endothelial cell loss during follow-up (r = -0.29) [51]. It should be noted, however, that we did not observe a direct relationship between postoperative CCT and postoperative central ECD. Thus, when the eyes were divided according to median 6-month CCT (526 μm), the resulting two subgroups did not differ significantly in terms of 6-month ECD (1328 vs. 1300 cells/mm², p = 0.230 on Wilcoxon test). The same was true for 12-month CCT (median, 529 μm; 1123 vs. 1099 cells/mm², p = 0.440). Similarly, there were no correlations between postoperative CCT and postoperative ECD at either 6 months (r = 0.002, p = 0.981 on Spearman correlation test) or 12 months (r = 0.01, p = 0.881). Thus, while it is possible that postoperative CCT predicts postoperative BSCVA because endothelial cell function shapes CCT and CCT dictates BSCVA, it seems that the ECD variable does not adequately reflect endothelial function in our setting. It is possible that other measures of endothelial function would be able to detect this relationship.

Study limitations

This study had some limitations. First, it was a retrospective study, which lends itself to selection and information bias. However, the data were all recorded prospectively. Second, it was a monocentric study. However, all surgeries were conducted by one surgeon, which may have limited confounding due to surgeon-related variation. Third, the sample size was relatively small (n = 124). However, many other studies in the field have employed such or smaller sample sizes, including that by Brockmann et al. (n = 108) [19]. Fourth, our study focused on moderate to severe FECD cases only: in our institution, very advanced cases that have intrastromal scars (which are irreversible structural changes and lead to poorer postoperative visual outcomes [52, 53]) are generally treated with DSAEK. Thus, our finding that postoperative BSCVA does not depend on preoperative CCT cannot be extrapolated to cases of very severe endothelial disease. Fifth, we used noncontact ultrasound pachymetry to measure CCT rather than contact ultrasound pachymetry. However, noncontact pachymetry is easily conducted during consultation and several studies have shown that these two methods differ negligibly in terms of CCT measurements [54, 55]. Sixth, since Brockmann et al. and Peraza-Nieves et al. respectively used AS-OCT and Pentacam to measure CCT [6, 45], it is possible that the CCT-measuring technique explains the different findings of our study. However, Schrittenlocher et al., who used both Pentacam and ultrasound pachymetry to measure CCT, also did not find a relationship between preoperative CCT and final visual acuity after DMEK regardless of which method they used [15] (Table 1). Moreover, while ultrasound pachymetry, AS-OCT, and Pentacam measurements of CCT are not interchangeable, they correlate closely [56–59].