Malapposition of graft-host interface after penetrating keratoplasty (PK) and deep anterior lamellar keratoplasty (DALK): an optical coherence tomography study

Patients with advanced keratoconus usually have very low vision because to progressive high myopia and irregular astigmatism. Therefore, the primary objective of keratoplasties for these patients should be rebuilding normal corneal curvature to provide acceptable vision [1, 2]. Penetrating keratoplasty (PK) and deep anterior lamellar keratoplasty (DALK) are two main surgical treatments for keratoconus. Despite progress in surgical and examination techniques, postoperative ametropia, and the attendant suboptimal vision, continues to occur; the reasons for this are unclear [3]. Approximately 40% of patients experienced astigmatism after keratoplasty. Among those who had astigmatism, 19 to 38% suffered from high astigmatism (> 5 diopters), which could not be satisfactorily corrected with spectacles or contact lenses [4‐7]. Because keratoconus usually affects young and middle-aged people, achieving satisfactory vision for a lifetime should be the ultimate goal of surgical treatment [2, 3, 5, 6, 8].

Anterior-segment optical coherence tomography (AS-OCT) can penetrate through deep tissues with a stretched wavelength (1.3 μm). AS-OCT can safely be applied at the perioperative stage to obtain cross-sectional images of the anterior segments, which is critical for preoperative evaluation and postoperative follow-ups [9]. Therefore, AS-OCT has been used to observe alignment patterns of the posterior graft-host junctions after PK. However, similar observations after DALK have rarely been reported [10, 11].

To-date, knowledge of the factors influencing postsurgical refractive errors has remained inconclusive because of insufficient relevant reports. Kaiserman and Bahar [12] reported that internal graft-host malappositions were associated with increased postoperative ametropia. However, the precise way in which each malapposition affected postsurgical visual outcomes has rarely been reported [13]. Thus, the aim of this study was to analyze the associations between characteristics of the posterior graft-host interfaces (GHI) and the postoperative visual outcomes in keratoconus patients.

All scans were performed in dark environment. Subjects were required to look straight at a fixation target inside of the device with their eyes opened widely by a blepharostat and to sit motionless throughout the examination. Only when subject’s eyeball position was stable, did we perform scanning.

A total of 360 graft-host sections from 45 AS-OCT images of 45 eyes (40 male and 5 female) were acquired and analyzed, including 22 post-penetrating keratoplasty (post-PK) eyes and 23 post-deep anterior lamellar keratoplasy (post-DALK) eyes. Demographics, postoperative visual outcomes and part of AS-OCT parameters were compared between post-PK and post-DALK (Table 2). The mean follow-up time for the DALK group was significantly shorter than that for the PK group (20.82 ± 10.54 vs. 38.18 ± 22.91, respectively, P = 0.003). DALK patients had relatively higher (worse) logMAR BCVA and lower SE and DS values than did the PK patients. Tg was greater in the post-DALK group than in the post-PK group. Correspondingly, the |Tg-Th| value in the DALK group was larger. Among other GHI metrics, GHT, Pm, and Sm showed no significant differences between groups. However, the frequency of different types of apposition differed between the groups. Post-DALK eyes had more step and graft-step alignments, whereas post-PK eyes had more host, hill and tag junctions. Hill alignment was not observed in post-DALK eyes. The frequencies of regular-apposition and gape did not differ significantly between groups (Table 2 and Table 3; Additional file 5: Figure S5).

Combined data from both post-PK and post-DALK cohorts were used for the following analysis. Although GHIs had quite smooth epithelial surfaces in all scans, malappositions were found on the internal surfaces in 58.3% of scans. In the complete set of internal GHIs, there were 43.9% steps (122 graft steps and 36 host steps), 11.7% protrusions (33 hills and 9 tags), and 2.8% gapes (10 scans). GHT values varied with alignment types as follows: compared with the regular-apposition, internal step and gape significantly reduced GHT, whereas the opposite effect was observed with the hill and tag patterns (Table 4).

LogMAR BCVA showed a positive correlation with GHT (r = 0.030) and Tg (r = 0.001), but a negative correlation with the frequency of protrusion [F (protrusion)] (r = 0.01). SE was found to have a positive correlation with F (step) (r = 0.001), F (graft step) (r = 0.028), Pm (r = 0.018), Sm (r = 0.037), Tg (r = 0.022), and |Tg-Th| (r = 0.007). SE was negatively correlated with GHT (r = 0.021). Very similar results were found for DS. We did not find any relationships between DS and the OCT parameters. Astig value, representing keratometric astigmatism, was found to have slight positive correlation with Sm (r = 0.047). Details are listed in Table 5. Scatter plots with LOWESS curves are listed in Fig 1, 2, 3 and 4.

Based on the graft-host characteristics with significant correlations, linear regression equations for the visual outcomes were successfully established, except for the one between logMAR BCVA and F (protrution) (F = 3.937, P = 0.054). SE increased on average by 6.851, 5.428 and 5.164 diopter per 1% increase in F (step), F (graft step) and Pm, respectively. SE also increased on average by 0.31 for every 10-μm increment in |Tg-Th|. LogMAR BCVA increased on average by 0.01 for each 10-μm increment in both GHT and Tg. DS increased on average by 6.319, 5.067 and 4.518 diopter for every 1% increase in F (step), F (graft step) and Pm, respectively. DS decreased on average by 0.2 diopter per 10-μm increment in GHT. With each 10-μm increase in |Tg-Th|, DS increased by 0.27 diopter averagely. Astig value increased averagely by 0.17 diopter for each 10-μm increment in Sm. Complete linear regression equations are listed in Table 6.

We found that misalignments of internal GHIs existed in both postsurgical groups, similar to those in previous reports [12, 13]. Smoothness of the anterior GHI can result from several factors. First, the surgeon can align the epithelial surface of a corneal wound under direct vision during an operation. Other factors include suture traction and the powerful regeneration capacity of corneal epithelium. It has been reported that posterior GHI discontinuity exists universally, but related studies are very limited [13]. Lang et al. [15] found misalignments of the DM layer in 22 eyes in post-mortem examinations of 25 patients (30 eyes) who had undergone PK. Kaiserman et al. [12] reported more post-PK internal graft-host malappositions (76.4%) in keratoconic patients compared with those who received corneal transplantations for other corneal diseases (50%). For keratoconic patients, the higher probability of irregular wound healing might be due to uneven thinning of the cornea before surgery. In our study, 58.3% graft-host junctions were misaligned. The major reasons for the relatively low incidence of misalignment may be the expertise of our experienced surgeons and the use of equal graft-host size. Moreover, unlike previous studies, we enrolled both post-PK and post-DALK patients. The different healing processes of these two procedures might account for the discrepancy in misalignment rates among studies [16].

Until now, there have been conflicting opinions about the causes of postoperative astigmatism; these include small intraoperative trephinations, uneven suture tension and graft-recipient misalignments [8, 11]. Limberg et al. [17] proposed that imprecise graft-host matching might result in astigmatism of about 4 to 6 diopters [1]. Kaiserman et al. [11] reported slight misalignment-associated astigmatism after PK. Jhanji et al. [18] categorized the alignment patterns as step and ledge, but failed to describe the relevance of malappositions to postoperative visual outcomes. However, they proposed a theory that an oversized graft would affect the GHI alignment due to the curled shape of the internal surface of the larger graft. In the current study, graft and host beds were prepared isometrically in all cases so size difference was not an issue. Suture tension within each case proved to be almost even, based on the symmetry of postoperative corneal morphology under AS-OCT observation.

Referring to lamellar keratoplasty, the GHI manifest as a moderate-to- highly reflective interface in AS-OCT images [19]. However, the incision depth of corneal stroma in our DALK procedure virtually reached the DM layer. Hence, the very thin residual stroma made the AS-OCT cross-sectional images from the PK and DALK groups very similar [19]. In our study, the PK group had better logMAR BCVA than did DALK group [(0.25 ± 0.24) vs. (0.47 ± 0.20), P = 0.002] despite similar astigmatism. This is probably due to the shorter postoperative recovery time in the DALK cohort, resulting in thicker junctional graft, with the consequent irregular or optically less clear GHI resulting in lower visual acuity. There have been different opinions about the relative merits of these two surgeries for years. However, there was strong evidence, from a register of controlled trials, suggesting better logMAR BCVA at ≥6 months with PK in a recent systematic review [4]. Our results concur with that conclusion [10]. The more severe SE and DS results in our study might be attirubuted to the relatively worse keratoconus and higher degree of myopia before surgery in our patients [2].

Post-keratoplasty alignment patterns can be classified into four basic types: regular apposition, step, protrusion and gape. Moreover, graft step and host step are subtypes of step, while hill and tag are subtypes of protrusion [12]. Studies have shown that various preoperative corneal pathologies influence the wound alignment patterns [12, 20]. Sung et al. [13] used AS-OCT to observe the posterior surface of corneal wounds from 13 post-PK keratoconic eyes, and found 78.8% had malapposed junctions, including 22.1% with gape, 22.1% protrusion and 34.6% step. Among the 360 graft-host sections, confined to only keratoconic eyes in the current study, the most common malapposition was graft step (122 cases, 33.9%). Since it has been reported that preoperative corneal pathology can influence the graft-host apposition patterns [14, 21], we hold the opinion that asymmetrical pre-operative thinning of the cornea in different disease stages caused increased graft step numbers, because the normal corneal grafts from donors were generally thicker than the recipient beds, which had already thinned.

GHT represented the contact area between the graft and recipient bed. Generally, step and gape significantly reduced GHT, whereas hill increased GHT. In the current study, tag pattern also slightly increased GHT. “Tag” refers to a small piece of DM layer protruding from the corneal wound, while “hill” is a protrusion of both the DM layer and deep corneal stroma [12, 13]. Hence, we think that tag should increase GHT, but only to a very limited extent.

In present study, various associations between visual outcomes and characteristics of the corneal alignment patterns were specifically evaluated and quantified. Generally, we found the malappositions that decreased GHT significantly increased the postoperative SE and DS. However, the decreased GHT lowered logMAR BCVA. We speculate that expansion of GHT might enhance stability of the corneal wound, and thus reduce the postoperative SE and DS. However, logMAR BCVA increased slightly with GHT in some unknown way, which needs further exploration. SE and DS significantly increased under the influence of F (step), F (graft step), Pm, Tg, and |Tg-Th|. Specifically, Astig value and SE both showed a slight tendency to increase along with increments of Sm. We speculate that F (step), F (graft step) and Pm may be the major factors responsible for post-keratoplasty ametropia. Moreover, the relationships between |Tg-Th| and visual outcomes indicated that larger graft-host disparity might lead to more-serious ametropia. Hence, we assume that the thinner the recipient bed is, the worse the visual outcomes will be. The increase in graft thickness mainly occurs shortly after surgery, because of tissue edema; hence, this could explain the increase of logMAR BCVA, SE, and DS with Tg, because most cases were observed shortly after operation. Sung et al. [13] evaluated the characteristics of GHI after PK using AS-OCT and reported that the graft-host thickness disparity, which closely related to the wound alignment state, showed a positive correlation with keratometric astigmatism (r = 0.56, P < 0.01). Kaiserman et al. [12] analyzed 204 post-PK graft-host sections from 27 eyes, and found that Sm correlated negatively with postoperative SE (r = − 0.2, P = 0.02) and positively with postoperative DC (r = 0.26, P = 0.006). Although the correlations were slight, Sm could explain, in part, the astigmatism and ametropia. Herein, Sm increase the degree of astigmatism, if only to a minor extent.

Our study has several limitations. Because it only represented a single-center experience, with relatively few subjects, we may have missed some statistical conclusions that might otherwise have been significant. Next, the study subjects were restricted only to keratoconic patients, so it may not be possible to extrapolate from our results to cases with other corneal diseases. Moreover, we unavoidably missed some information about internal GHIs because data for only eight positions per eye were acquired within a raster scan.

We verified the widespread existence of internal graft-host malappositions using AS-OCT and quantified the relationships between GHI characteristics and visual outcomes in postoperative keratoconic patients for the first time. GHT increased in protrusion alignment and decreased in step and gape patterns. F (step), F (graft step), Pm and |Tg-Th| influenced SE and DS positively. Central corneal keratometric astigmatism increased along with Sm increment. LogMAR BCVA rose with increasing GHT and Tg.

In conclusion, investigating of the characteristics of GHI is valuable for explaining varied ametropia in keratoconic eyes. Our study has potential reference value for future technological advancement. Further studies are warranted to determine better ways of achieving optimal graft-host apposition.