Article Text

Download PDFPDF

Spontaneous long-term course of persistent peripheral graft detachments after Descemet’s membrane endothelial keratoplasty
  1. Franziska Bucher,
  2. Deniz Hos,
  3. Stephan Müller-Schwefe,
  4. Philipp Steven,
  5. Claus Cursiefen,
  6. Ludwig M Heindl
  1. Department of Ophthalmology, University of Cologne, Cologne, Germany
  1. Correspondence to Dr Franziska Bucher, Department of Ophthalmology, University of Cologne, Kerpener Strasse 62, Cologne 50924, Germany; franziskabucher{at}


Background Peripheral corneal graft detachment after Descemet's membrane endothelial keratoplasty (DMEK) is a frequently occurring postoperative complication. The natural course of these persistent peripheral detachments over time is not known.

Methods 166 patients were surveyed by slit-lamp-adapted optical coherence tomography (SL-OCT) directly after surgery, during first postoperative week, 4 weeks, 3, 6 and 12 months, postoperatively. Patients with a persistent peripheral graft detachment 4 weeks after DMEK (n=16) were observed for their spontaneous course up to 1 year postoperatively.

Results Persistent graft detachments could be characterised into two phenotypes: peripheral roll (n=11; 69%) and laminar detachment (n=5; 31%). Maximal length of the detachment did not change in peripheral rolls during observation period (12 months vs 4 weeks, 578±122 µm vs 593±106 µm, p=0.74), whereas laminar detachments spontaneously attached to the host's stroma (12 months vs 4 weeks, 0 µm vs 1088±295 µm, p≤0.001). Central corneal thickness and (peripheral) corneal thickness above the detached area did not significantly change in either group.

Conclusions Persistent peripheral graft detachments after DMEK occurred in 10% of patients and had two distinct OCT-phenotypes. Peripheral rolls did not change during the first 12 months, postoperatively. By contrast, peripheral laminar detachments attached spontaneously even months after surgery. Corneal thickness reduction was only observed above peripheral laminar detachment, but not above peripheral rolls.

  • Cornea
  • Dystrophy
  • Treatment Surgery

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Posterior lamellar techniques have nowadays replaced full-thickness penetrating keratoplasty (PK) for the treatment of corneal endothelial disorders.1–7 Furthermore, lamellar techniques, such as Descemet's membrane endothelial keratoplasty (DMEK), obtain faster and better visual rehabilitation, less postsurgical astigmatism and improved surface topography compared with PK.4 ,6 ,8 ,9 Moreover, the selective implantation of donor endothelium without stroma (eg, in Fuchs’ corneal endothelial dystrophy, bullous keratopathy and pseudoexfoliation syndrome keratopathy) leads to a lower risk of immune reactions and transplant rejection, postoperatively, with the possibility of using one transplant for two recipients (‘split-cornea transplantation’).4 ,5 ,8–12 However, DMEK is a challenging surgical procedure with a diverse risk profile.4 ,13 ,14 Unlike rare complications such as immune reactions,15 graft dislocation and partial graft detachment is a frequent complication seen after posterior lamellar keratoplasty.14 ,16

After DMEK surgery, the position of the graft can be visualised by slit-lamp-adapted optical coherence tomography (SL-OCT).17 Hereby, even small detachments of the graft, which cannot be seen by slit-lamp biomicroscopy, can be exactly located and monitored over time.17 Due to corneal graft detachments in the early postoperative phase, air injections into the anterior chamber are often performed to attach the thin layer of Descemet membrane and endothelium onto the host's stroma. However, in spite of repetitive air injections (‘rebubbling’), a complete graft attachment cannot always be achieved.17

So far, it has not been described how often persistent graft detachments, which are refractory to repetitive air injections, occur and further develop over time. Furthermore, as retinal detachments of the upper hemisphere are more harmful in terms of faster progression of the detachment, it is of interest whether corneal graft detachments of the upper hemisphere show a similar behaviour.

Therefore, we were interested in the characteristics and course of persistent corneal graft detachments despite rebubbling after DMEK.

Patients and methods


One hundred and sixty-six consecutive eyes of 166 patients (87 women and 79 men; 69±12 years (range 31 years–92 years), at the time of surgery) who underwent successful DMEK surgery between 1 July 2011 and 31 December 2012, at the Department of Ophthalmology, University of Cologne, Cologne, Germany, were included in this study. The indication for all DMEK procedures was poor spectacle-corrected visual acuity (≤20/40) due to Fuchs’ endothelial dystrophy or pseudophakic bullous keratopathy. Eyes with prior failed endothelial keratoplasty were excluded. This prospective, non-randomised, clinical, single-centre study was carried out in conformance with the tenets of the Helsinki Declaration. Written informed consent was obtained from all patients before surgery. The protocol was approved by the institutional review board and adheres to all German federal and state laws.

DMEK surgery and management of graft detachment in the early postoperative period

All DMEK procedures were performed after standardised protocols under general anaesthesia by the same experienced surgeon (CC), as described previously.2 ,12 ,18 Preoperatively, all patients received neodymium:yttrium-aluminium-garnet laser (VisuLas YAG II plus; Carl Zeiss Meditec, Jena, Germany) iridotomy to avoid pupillary block or Urrets–Zavalia syndrome. A triple procedure was performed in 60 patients (in case of clinically significant coexisting cataract), with a conventional phacoemulsification and implantation of an acrylic intraocular lens into the capsular bag. Briefly, a 4 mm clear corneal tunnel was prepared in addition to three small paracenteses.5 Thereafter, the central 9 mm Descemet membrane was removed under air filling with a miotic pupil using a Price hook (Moria, Doylestown, Pennsylvania, USA). Then, the previously prepared donor endothelium–Descemet membrane graft (8 mm in diameter) was inserted into the anterior chamber using a conventional lens injector cartridge (Acri.Tec, Hennigsdorf, Germany) with endothelial side outward.5 After removing air form the anterior chamber, the graft was inserted into the fluid-filled anterior chamber and then unfolded by an injected air-bubble on top of the graft. The graft was attached to the host stroma by completely filling the anterior chamber with air for at least 5 min and subsequently deflating the air to approximately 90% of the anterior chamber volume. Criteria for additional air injections into the anterior chamber during the first postoperative week included central or paracentral graft detachment with consecutive corneal stromal oedema in the optical zone and impaired visual acuity. Air reinjection into the anterior chamber was performed once in 110 patients (66%), twice in 15 patients (11%) and thrice in 3 patients (2%) within the first postoperative week. Thereafter, none of the patients received additional reinjections, and the remaining graft detachments were followed-up closely by SL-OCT (Heidelberg Engineering, Lübeck, Germany) and slit-lamp biomicroscopy.

Patients were positioned in a supine position until there was no remaining air in the anterior chamber. All patients received postoperative topical treatment with ofloxacin eye drops (Floxal EDO; Mann, Berlin, Germany) four times daily, lubricants (Hylo-Care; Ursapharm, Saarbrücken, Germany) every 2 h, hyperosmotic eye drops (NaCl 5% Dispersa; Omnivision, Neuhausen, Switzerland) every 2 h, pilocarpine 2% eye drops (Bausch & Lomb, Irvine, California, USA) five times daily for as long as air was in the anterior chamber, and corticosteroid eye drops (Inflanefran forte; Pharm Allergan, Ettlingen, Germany) every 2 h. Topical treatment was reduced gradually based on clinical condition.

Clinical assessment

Preoperative and postoperative standardised eye examinations included measurement of Snellen uncorrected visual acuity and best spectacle-corrected visual acuity, slit-lamp examination, tonometry, funduscopy and SL-OCT of the anterior segment. Standardised eye examinations were repeated on each visit at 4 weeks, 3, 6 and 12 months, postoperatively.

Analysis of graft detachment

SL-OCT was performed before as well as on the first postoperative day after DMEK. Postoperative SL-OCT scans of all patients were analysed, and persistent corneal graft detachments observed in this series were characterised in position (inferior, temporal, superior and nasal) as well as in shape of the detachment (peripheral roll or laminar detachment). All patients with a persistent peripheral graft detachment 4 weeks after DMEK and at least 3 weeks after last air injection were monitored over 12 months follow-up. The maximal length of the detachment was measured in the SL-OCT scan with the greatest detachment. Corneal thickness was measured in the centre of the cornea as well as in the centre of graft detachment.

Main outcome measures and statistical analyses

Main outcome measures were spontaneous course and characteristics of refractory corneal graft detachment (peripheral roll vs laminar detachment), area of graft detachment, maximal diameter of graft detachment, localisation of the detachment (inferiorly, temporally, superiorly and nasally), as well as central corneal thickness (CCT) and corneal thickness above detachment. Statistical significance was tested using the Mann–Whitney U test and Fisher's exact test. A p value of <0.05 was considered statistically significant. All analyses were performed with Microsoft Excel 2011 (Microsoft, Redmond, Washington, USA) and Prism6, V.6.02 (GraphPad Software, San Diego, California, USA).


Persistent peripheral graft detachments

SL-OCT scans 4 weeks post-DMEK were analysed for remaining graft detachments. Sixteen patients (10%) showed a persistent peripheral graft detachment outside the visual axis visible on SL-OCT examination, 3 weeks after last air injection (8 after DMEK; 8 after Triple-DMEK); 150 patients showed complete graft attachment (90%). Twelve of the 16 patients with a persistent graft detachment had previously received one additional air injection, and six patients had been treated with a second rebubbling within the first postoperative week. Graft detachments after 4 weeks could be characterised as peripheral rolls (n=11) with wound-up Descemet's membrane and endothelium and as laminar detachments (n=5) with correct position of endothelial cells, Descemet's membrane and host's stroma (figure 1). The mean length of the graft detachments was 215±128 µm in greatest length for laminar detachments and 308±203 µm for peripheral rolls (analysed time point: 4 weeks after DMEK). The occurrence of peripheral rolls or laminar detachments was independent from the performed procedure (DMEK vs triple-DMEK: peripheral roll: p=0.1; laminar detachment: p=0.7). All graft detachments showed identical characteristics in the first postoperative week. Laminar detachments decreased in size, peripheral rolls persisted in measurements, but none of the detachments showed progression.

Figure 1

Characteristics and spontaneous course of peripheral graft detachments. Comparative images of two patients after Descemet's membrane endothelial keratoplasty (DMEK); left column showing graft detachment characterised as peripheral roll; right column demonstrating laminar detachment (red boxes indicate magnified areas). Slit-lamp photographs 4 weeks (A and B) and 3 months (E and F) post-DMEK showing a clear cornea. Slit-lamp adapted optical coherence tomography (SL-OCT) scans of the same patient with a persisting peripheral roll of the Descemet and endothelium (left column: C and G). SL-OCT images of the second patient (laminar detachment) 4 weeks postoperatively (D). Spontaneous graft attachment 3 months after surgery after laminar detachment can be observed (H).

Spontaneous course and characteristics of corneal graft detachments

All 16 graft detachments were analysed regarding their spontaneous course using SL-OCT for up to 1 year. By contrast with peripheral rolls, all laminar detachments resolved clinically and by SL-OCT over a period of 3 months (one patient), 6 months (one patient) and 12 months (three patients) without any further intervention. The maximal length of the detachment (distance of uncovered hosts’ stroma) did not decrease in peripheral rolls over 12 months (4 weeks: 593±106 µm, 3 months: 628±102 µm, 6 months: 518±81 µm, 12 months: 578±122 µm, 12 months vs 4 weeks p=0.74; figure 2A), but did decrease significantly in laminar detachments (4 weeks: 1088±295 µm, 3 months: 997±297 µm, 6 months: 357±151 µm, 12 months: 0 µm, 12 months vs 4 weeks p≤0.001; figure 2B). The maximal length of the detachment in peripheral rolls was significantly smaller than in laminar detachments (p=0.04).

Figure 2

Analysis of spontaneous course of persistent peripheral detachments refractory to air injections after Descemet's membrane endothelial keratoplasty. The maximal length of detached area did not decrease in peripheral rolls (A), but only in laminar detachments (B). (Data shown as mean and 5%–95% percentile; ***p≤0.001).

Localisation of peripheral detachments

Peripheral rolls could be detected inferiorly in two patients, temporally in four patients, superiorly in three patients and nasally in one patient. Laminar detachments were localised inferiorly in three patients, nasally and superiorly in one patient. All peripheral rolls, independent from localisation, did not attach, whereas all laminar detachments, also independent from localisation, attached completely. Even laminar detachment of the upper hemisphere did not increase, but spontaneously attached without need for repeated air injections.

Corneal thickness

CCT measured by SL-OCT did not decrease significantly over the follow-up period in either group when compared with CCT 4 weeks after surgery (12 months vs 4 weeks postoperatively; peripheral roll: 505±48 µm vs 559±85 µm, p=0.2; figure 3A, C; laminar detachment: 517±27 µm vs 528±66 µm, p=1.0). Peripheral corneal thickness above laminar detachments also did not decrease significantly over 12 months in comparison with corneal thickness above peripheral rolls (12 months vs 4 weeks postoperatively; peripheral roll: 730±95 µm vs 754±104 µm, p=0.45; figure 3B, D; laminar detachment: 692±74 µm vs 835±107 µm, p=0.056).

Figure 3

Corneal thickness as measured by slit-lamp-adapted optical coherence tomography over 12 months follow-up. Central corneal thickness in patients with peripheral roll (A) und laminar detachment (B) showed no significant further thinning. (C) Thickness of cornea above detachment showed no significant thinning in patients with peripheral roll and laminar detachment (D). (Data shown as mean and 5%–95% percentile).


In this study, we demonstrate that persistent peripheral detachments after DMEK, despite rebubbling, occurred in 10% of eyes and that there were two different kinds of peripheral corneal graft detachments, which varied in shape (peripheral roll vs laminar detachment) and behaved differently. Laminar detachments always completely reattached over time. By contrast, peripheral rolls, with wound-up endothelium, Descemet's membrane and stroma neither attached nor demonstrated any progression.14

CCT (in laminar detachment and peripheral roll) remained stable when compared with 4 weeks–12 months of postoperative measurements. Apparently, CCT and peripheral corneal thickness reached definite magnitude 4 weeks after DMEK, as we could not detect further thinning after this time-point.

Our study also showed that in contrast with retinal detachments, peripheral corneal laminar detachments of the upper hemisphere that occurred after DMEK did not progress and led to complete graft dislocation. However, as the detachment of the upper hemisphere only occurred in one of our patients, this observation needs further verification in additional follow-up studies. Even months after the last air injection into the anterior chamber peripheral laminar detachments attached without further interventions. This could be due to the pumping function of corneal endothelial cells, which even in late postoperative stages, reduces the remaining fluid between Descemet's membrane and the host's stroma until complete attachment is achieved.19 Another possibility might be that the attached area decreases centrifugally from out the border of the attached graft towards the distant periphery, for example, due to matrix interactions. Further investigations have to analyse the dynamics of these persistent graft detachments more in detail. Investigations by confocal microscopy and Scheimpflug camera might bring additional information on the appearance of different detachment types and their spontaneous clinical course.

So far, we have been unable to define a cut off size for detachments that will attach completely without any additional air injections from those that will need further rebubbling. Additionally, up to now, we could not define whether the localisation, or how central/peripheral, a corneal graft detachment has an influence on visual acuity, postoperative outcome or rebubbling rate.

Dirisamer et al,20 in a previous study, presented an approach as to how to prevent and manage graft detachments after DMEK, with similar findings, that smaller peripheral detachments do not necessarily require further repetitive air injections and can attach spontaneously.

By contrast with that study, we could demonstrate that only laminar detachments attached spontaneously, whereas peripheral rolls did not attach. A study performed by Yeh et al21 describes the postoperative course in relation to the graft detachment area and the OCT attachment status at various time-points after DMEK in a cohort of patients without any postoperative air injections. Here, the 1 h postoperative OCT scans had the highest predictive value on graft attachment outcome.21 In line with our findings, Yeh et al describes that some partial detachments may attach spontaneously, even months after surgery. However, graft detachment area in early postoperative OCT and detachment phenotype may assist in deciding surgical reintervention after DMEK.

Along with current studies, we did not see significant differences in rebubbling and complication rate between DMEK and Triple-DMEK in this cohort of patients.22 DMEK, as well as Triple-DMEK, had no influence on the occurrence of postoperative peripheral rolls or laminar detachments.

In summary, our data showed that peripheral laminar detachments after DMEK can attach spontaneously even months after surgery and that peripheral rolls did not change during the first postoperative year. Furthermore, corneal graft detachments of the upper hemisphere did not seem to be more harmful compared with other localisations.



  • Contributors FBu undertook examination of patients, analysed OCT scans and wrote the article. DH helped with the study design and reviewed the article. SM-S helped with patients’ history. PS helped with OCT analysis. CC and LMH mainly reviewed the article and supervised the research.

  • Funding This work was supported by: German Research Foundation (HE 6743/2-1 to LMH; Priority Research Project SFB 643: B10, CU 47/6-1, CU 47/4-1 to CC; STE 1921/2-1 to PS); GEROK-Programme University Hospital of Cologne (to FB, LMH and DH); Ruth and Helmut Lingen Foundation, Cologne (PS and CC).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Ethikkommission University of Cologne.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • At a glance
    Keith Barton James Chodosh Jost Jonas