Objectives: In recent years, there have been rapid advances in the field of keratoconus. Penetrating keratoplasty and deep anterior lamellar keratoplasty are the standard surgical procedures. Nevertheless, controversy remains regarding the outcomes of both procedures in the treatment of keratoconus. Therefore, we conducted a meta-analysis comparing postoperative outcome measures of penetrating keratoplasty versus deep anterior lamellar keratoplasty for keratoconus.
Materials and Methods: We searched PubMed, Embase, Web of Science, and the Cochrane Database of Systematic Reviews for eligible studies comparing best-corrected visual acuity, spherical equivalent, refractive cylinder, topography cylinder, and graft rejection episodes and complications of penetrating keratoplasty and deep anterior lamellar keratoplasty. Seven risk domains from software Review Manager 5.3 (The Cochrane Collaboration, Oxford, UK) were applied as quality assessments for the eligible studies. A random-effects model was used for data synthesis.
Results: Thirteen eligible studies were included in our meta-analysis, which encompassed 530 eyes that underwent penetrating keratoplasty and 568 eyes that underwent deep anterior lamellar keratoplasty. With regard to best-corrected visual acuity, refractive cylinder, and topography cylinder, we found no significant differences in results between the 2 procedures (P = .49 and .47, respectively). However, spherical equivalent results were significantly greater in the deep anterior lamellar keratoplasty group than in the penetrating keratoplasty group (P < .001). The risk of graft rejection episodes was more prominent in the penetrating keratoplasty than in the deep anterior lamellar keratoplasty group (odds ratio = 2.69; P = .001). The odds ratio for complications was 1.79 (P = .03). Three studies showed moderate risk of bias, and the other 10 showed high risk of bias.
Conclusions: Deep anterior lamellar keratoplasty is preferred over penetrating keratoplasty for the treatment of keratoconus because of its low risk of rejection and complications.
Key words : Corneal topography, Graft rejection, Visual acuity
Keratoconus is a common degenerative, noninflammatory corneal disease characterized by corneal protrusion, progressive stromal thinning, and highly irregular astigmatism, which eventually results in significant visual impairment. It often occurs bilaterally. The prevalence of keratoconus differs depending on the region; prevalence is 0.1% to 0.5% in Europe and the United States, 0.7% to 0.8% in Japan, and 0.2% to 0.5% in China.1 Ramdas and Vervaet2 reported that wearing rigid gas-permeable contact lenses may halt early-stage keratoconus progression. However, there have been inadequate long-term and large sample size research studies to support this view.3,4 Rigid gas-permeable contact lenses have been shown to reduce corneal curvature, irregular astigmatism, and high-order aberrations, thereby improving vision. The corneal cross-linking technique has been widely performed to strengthen corneal stromal fibers so as to block corneal distortion.5 The intrastromal corneal ring segment technique has also been utilized to flatten the steep area of the cornea, reducing corneal refractive errors.6 When keratoconus develops to moderate or advanced stage, corneal transplant is a better option for patients.7
To reduce the risk of postoperative graft rejection as much as possible, deep anterior lamellar keratoplasty (DALK) has been widely used to treat moderate or even advanced-stage keratoconus.8 Penetrating keratoplasty (PK) has been used to treat acute keratoconus, Descemet membrane rupture, and maximum keratometry on corneal topography > 60 diopters.9 Moreover, it has been found that DALK induced complications that did not occur in PK, including corneal perforation and double anterior chamber.10 Therefore, some believe that PK may become the preferred procedure for the treatment of keratoconus in the future.11 In studies that compared best-corrected visual acuity (BCVA) and refractive errors after PK and DALK for keratoconus, postoperative BCVA and residual refractive errors with DALK were better than shown with PK.8-12 Reinhart and associates13 suggested that postoperative BCVA and refractive errors between these 2 procedures were comparable, and Pantanelli and colleagues14 found that patients who underwent PK had better logarithm of the minimum angle of resolution (logMAR) visual acuity than those who underwent DALK over a 25- to 50-month follow-up. As with graft rejection episodes and graft survival status, postoperative BCVA and residual refractive errors are also recognized as major indicators for evaluating corneal transplant efficacy.
In recent years, there have been new studies comparing the postoperative outcomes of PK and DALK for the treatment of keratoconus. Thus, these new comparisons require a new meta-analysis. In this present study, we hope to fill the gap through a comparison of clinical outcomes of PK versus DALK, including postoperative BCVA, refractive errors, graft rejection episodes, and complications, to evaluate the safety and efficacy of these 2 procedures for the treatment of keratoconus.
Materials and Methods
We screened studies comparing clinical outcomes of PK versus DALK for the treatment of keratoconus. Randomized clinical trials (RCT), retrospective comparisons, prospective comparisons, and cross-sectional studies were included. Initial included studies were required to mention at least one of the outcome parameters mentioned above. We excluded studies that only demonstrated 1 of the 2 procedures, and inclusion criteria were patients with diagnosis of keratoconus treated with PK or DALK. Age, male/female patient, region, and postoperative follow-up time were not restricted. Exclusion criteria were patients with coexisting corneal disorders, uveitis, retinal disorders, and angle-closure glaucoma and those who received 2 or more corneal transplants. Keratoconus was diagnosed using slit lamp and corneal topography.
Definition of terms
We defined PK as full-thickness corneal transplant and DALK as procedures in which the corneal endothelium was retained. Methods to dissociate the anterior corneal stroma from Descemet membrane in DALK procedures included manual dissection and the big-bubble technique. The big-bubble technique is defined as a technique that uses air injection, viscoelastic dissection, and hydrodelamination. Graft rejection episodes were defined as presence of graft edema and thickening, neovascularization, anterior chamber reaction, and Khodadoust lines.15 Complications included all diseases deriving from corneal transplant. The outcome measures included BCVA, spherical equivalent (SE), refractive cylinder, topography cylinder, graft rejection episodes, and complications. We recorded BCVA with logMAR and recorded SE, refractive cylinder, and topography cylinder as positive or negative refractions. Graft rejection episodes and complications were expressed as absolute events.
We analyzed studies and conducted our analyses in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (http://www.prisma-statement.org/).16 In our meta-analysis, only articles published in English were considered, and there were no restrictions on types or grade of journals. To enable our search to focus on recent therapeutic progress of keratoconus and to access sufficient numbers of studies, we set the search intervals as January 2008 to December 2018. We searched 4 electronic databases: PubMed, Embase, Web of Science, and the Cochrane Database of Systematic Reviews. The key words and search strategy are summarized in Table 1. In addition, we manually searched references listed in the preliminarily included studies. No suitable studies in these references met our requirements. This protocol was registered at the Prospective Register for Systematic Reviews (PROPERO, registration number: CRD42018104412).
The study design and outcome parameters were independently extracted by 2 authors (Yaowen Song and Jing Zhang). We identified the study design by screening titles, abstracts, and methods in main texts. If primary data appeared as the formation of ratios rather than integers or decimals in an eligible study, we converted the data in a suitable manner using calculation. All extracted information was examined twice and by 2 authors independently. Disagreements were resolved by discussion and consensus.
The extracted information and data included author names, publication year, study design, sample numbers, surgical age, country, follow-up time, BCVA, SE, refractive cylinder, topography cylinder, graft rejection episodes, and complications. If a series of follow-up records coexisted in 1 study, we adopted the last follow-up results. Graft rejection episodes and complications were expressed as dichotomous variables; therefore, we used odds ratio (OR) and 95% confidence intervals (95% CI) to measure the intervention effects. For continuous variables, we used mean difference (MD) and 95% CI to measure intervention effects. It was impossible to extract all outcome measures in one included study; therefore, we only needed to merge study items containing relevant outcome measures rather than including all eligible studies for compulsory analysis in the subsequent analysis of corresponding outcome measures.
We assessed the quality of eligible studies using 7 risk domains from Review Manager 5.3 (The Cochrane Collaboration, Oxford, UK). Briefly, random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants (performance bias), blinding of personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other biases were evaluated.17 We reviewed all studies carefully and evaluated them in turn. The risk of bias for each domain was graded as low risk, unclear risk, and high risk. If domains were graded as low risk, the study was defined as low risk bias. If 1 to 2 domains were graded as high or unclear risk, the study was defined as moderate risk of bias. If more than 2 domains were graded as high or unclear risk, the study was defined as high risk of bias.
Forest plots were employed for data synthesis. I2 was used to assess heterogeneity, with I2 < 40% considered low heterogeneity and I2 > 75% understood as considerable heterogeneity. We applied a random-effects model for all outcome measures because its conclusions were more conservative and tended to be safer than those of the fixed-effects model. Moreover, sensitivity analysis was performed under the circumstance of I2 ≥ 40%. Briefly, we excluded every study in turn to observe whether heterogeneity changed significantly. If heterogeneity changed significantly, the corresponding studies would be treated as the source of heterogeneity and were examined again. This was followed by a subgroup analysis based on follow-up time so as to determine the origin of heterogeneity to the greatest extent possible. Funnel plots were used to assess publication bias, and analyses based on at least 10 were included in one outcome measure. Begg’s and Egger’s tests derived from software Stata 12.0 (StataCorp, College Station, TX, USA) were introduced if the funnel plot was ambiguous. P < .05 was considered statistically significant.
We utilized the search strategy and key words shown in Table 1 to search for studies in 4 electronic databases. A total of 1783 studies were initially identified. Of these, 1498 did not meet our selection criteria. We then browsed titles and abstracts of the remaining 285 studies, and 252 did not fit our inclusion criteria or were duplicates. We read the full texts of the screened 33 studies. Of these, 4 were reviews and 16 were difficult to extract overall DALK data because they divided DALK into subgroups according to surgical methods or patients in some of these 16 studies underwent keratoplasty twice or more. Finally, 13 studies published between 2010 and 2018 were recognized as eligible for our meta-analysis.18-30 A study flow diagram is shown in Figure 1.
Of the 13 eligible studies, 3 were RCTs and the remaining 10 were non-RCTs. The non-RCT studies included 7 retrospective reviews, 2 cross-sectional studies, and 1 prospective study. Regarding total sample size, 530 eyes underwent PK and 568 eyes underwent DALK. In the DALK group, surgeons performed layer-by-layer manual stromal dissection, the Anwar big-bubble technique, viscoelastic dissection, or hydrodelamination to expose Descemet membrane. In both groups, the sample size of the single studies ranged from 3 to 100 eyes. Follow-up after PK and DALK varied across the eligible studies. In the PK group, the shortest postoperative follow-up was 12 months and the longest was 93.1 months. Likewise, in the DALK group, the shortest postoperative follow-up was also 12 months and the longest was 93.5 months. The extracted information and data from these 13 studies are shown in Table 2 and Table 3, respectively.
Best-corrected visual acuity
For BCVA, 10 studies were included: 378 eyes underwent PK and 375 eyes underwent DALK. With regard to BCVA, no statistically significant differences were shown between the PK and DALK groups (P = .07). Forest and funnel plots are shown in Figure 2 and Figure 3, respectively. Begg’s (P = .929) and Egger’s (P = .957) tests were consistent with the funnel plots (not shown). Sensitivity analyses showed that the exclusion of the study from MacIntyre and colleagues28 significantly changed the heterogeneity (I2 changed from 58% to 11%; MD changed from -0.03 to -0.02).
For SE, 8 studies were included: 288 eyes underwent PK and 301 eyes underwent DALK. There was a significant difference in SE between the PK and DALK groups (P < .001). Based on the forest plot, from the perspective of numerical properties, the mean SE value for the PK group was larger than that for the DALK group. However, the extent of SE was measured using diopters, and the mean SE results of the 2 groups were shown as negative refractions. Therefore, we concluded that SE in the DALK group was actually greater than that in the PK group (MD = 0.86; 95% CI, 0.49-1.24) (Figure 4).
For refractive cylinder, 8 studies were included: 306 eyes underwent PK and 342 eyes underwent DALK. We found no statistically significant difference in refractive cylinder between the PK and DALK groups (P = .49; Figure 5). Sensitivity analysis showed that the exclusion of the study from MacIntyre and colleagues28 significantly changed heterogeneity (I2 changed from 50% to 10%, and MD changed from 0.18 to -0.06).
For topography cylinder, 7 studies were included: 259 eyes underwent PK and 289 eyes underwent DALK. We observed no statistically significant difference in topography cylinder between the PK and DALK groups (P = 0.47; Figure 6). Sensitivity analysis showed that the exclusion of the study from Huang and colleagues24 significantly changed heterogeneity (I2 changed from 50% to 0%, and MD changed from 0.16 to -0.05).
Graft rejection episodes
With regard to analysis of graft rejection episodes, 7 studies were included: 291 eyes underwent PK and 324 eyes underwent DALK. We found a statistically significant difference in graft rejection episodes between the PK and DALK groups (P < .001). Based on the forest plot, the risk of subjecting to graft rejection episodes in the PK group was more prominent than the risk shown in the DALK group (OR = 2.69; 95% CI, 1.62-4.48) (Figure 7).
With regard to analysis of complication, 5 studies were included: 153 eyes underwent PK and 186 eyes underwent DALK. We found a statistically significant difference in complications between the PK and DALK groups (P = .03). Based on the forest plot, the risk of having complications was more prominent in the PK group than in the DALK group (OR = 1.79; 95% CI, 1.05-3.05) (Figure 8). Complication details for both groups are summarized in Table 4.
The quality assessment of 13 eligible studies showed that 3 studies had a score of moderate risk of bias and the other 10 had a high risk of bias. A summary of the 7 risk domains for each study is illustrated in Figure 9.
We conducted a meta-analysis comparing postoperative outcome measures of PK versus DALK for keratoconus. We found no significant differences with regard to BCVA, refractive cylinder, or topography cylinder between the procedures. Spherical equivalent was greater in the DALK group than in the PK group. The risk of a graft rejection episode was 2.69 times greater in the PK group than in the DALK group. Three studies showed moderate risk of bias, and the other 10 showed high risk of bias.
Corneal transplant is almost irreplaceable in the treatment of advanced keratoconus. Approximately 12% to 20% of patients with keratoconus eventually require keratoplasty.31 For patients with keratoconus, vision rehabilitation after corneal transplant has been shown to be significantly better than in patients with infectious and immunologic corneal diseases, even achieving 0 logMAR visual acuity. Young patients with keratoconus have hope of not only stabilizing the disease but also improving vision and quality of life. Therefore, investigating the differences in postoperative BCVA and refractive errors between the procedures is crucial to further clarify their corrective efficacies. We included 13 studies in this meta-analysis that were published between 2010 and 2018, 6 of which were published between 2015 and 2018. In our analysis, there were no statistically significant differences in BCVA, refractive cylinder, and topography cylinder outcomes between PK and DALK. The I2 results for these 3 outcome measures were 58%, 50%, and 50%, respectively, all of which were scored as moderate heterogeneity. Funnel plot and Begg’s and Egger’s tests indicated no publication bias in the BCVA outcomes. Although some conventional views have held that DALK causes higher corneal astigmatism,32 we found that postoperative refractive cylinder and topography cylinder between PK and DALK groups were comparable based on inspection of forest plots. The disappearance of an astigmatic difference might be attributed to recent improvements in surgical techniques; in other words, surgeons may dissect stroma uniformly during DALK, and adjust sutures appropriately to reduce refractive errors induced by the graft-host interface.33
Moreover, I2 for SE and graft rejection episodes were both 0, suggesting that there was no heterogeneity; therefore, the results of these 2 outcome measures could be considered highly credible.
There was a statistically significant difference in SE between the groups, and SE in the DALK group was greater than that in the PK group. In addition, refractive and topography cylinder results in both groups were comparable; therefore, it could be stated that different spherical diopters led to the diversity of SE between the groups. This visible myopic shift was mainly caused by oversizing of the graft, which is usually performed during DALK to integrate it firmly with the host bed.34 The tendency for corneal steepening was more substantial after DALK with the big-bubble technique than after PK.35
Similar to SE, we observed a significant difference in graft rejection episodes between the PK and DALK groups. Briefly, patients who underwent PK were more susceptible to graft rejection episodes than those undergoing DALK, primarily because of endothelial rejection and the high rate of loss of endothelial cells with PK.28 Endothelial rejection often resulted in visible stromal edema and opacity; therefore, it would lead to more severe visual impairment than epithelial and stromal rejection.36 Corneal endothelial cells derived from grafts were more vulnerable to host inflammatory mediators and immune system attack,37 whereas DALK prevented these adverse events by preserving original endothelium. Most instances of DALK rejection were due to subepithelial rejection,38 and it was demonstrated that epithelial defects mildly hindered visual acuity. Fortunately, most rejection episodes were treated with regular topical and systemic steroid and immunosuppressive regimens, and some common complications in the past including secondary glaucoma could be reversed or inhibited.39
The risk of complications was higher in patients who underwent PK than in those who underwent DALK. Table 4 shows that rejection and high refractive errors were the most common complications after the 2 procedures. Table 4 also shows that endothelial rejection comprised a larger proportion of rejection in the PK group. Obviously, it was necessary to administer standard and long-term therapeutics for graft rejection in cases of graft failure.
There have been other systematic reviews and meta-analyses of clinical outcomes comparing PK and DALK for the treatment of keratoconus. However, we found that these reviews only included studies published before 2014. Liu and associates40 reported a meta-analysis consisting of 16 studies (including 1 RCT). They showed that BCVA was significantly better in the DALK group than in the PK group, and there were no significant differences in SE and astigmatism between the procedures. Chen and associates41 synthesized 5 RCTs in which they compared PK versus DALK for the treatment of corneal stromal disorders. They found that BCVA was better in the DALK group than in the PK group and that there were no significant differences in SE and astigmatism between the PK and DALK groups. The conclusions of these 2 meta-analyses were comparable. However, in the review from Liu and colleagues,40 there was only 1 RCT and only 4 studies that reported postoperative BCVA outcomes. Regarding SE and astigmatism parameters, forest plots showed considerable heterogeneity (I2 = 91% and 77%, respectively). The conclusion of the meta-analysis of RCT studies from Chen and colleagues41 may have provided strong evidence; however, they were concerned with postoperative safety and efficacy in the treatment of corneal stromal diseases, including but not limited to keratoconus, and only 3 studies mentioned keratoconus. Keane and associates42 conducted a meta-analysis that consisted of 2 RCTs. They found that BCVA and keratometric stabilization were comparable between the 2 procedures. However, they believed that this result lacked sufficient evidence because of the small sample number. Our results were consisted with those of Keane and colleagues, and we included additional studies to support their conclusion.
Our sensitivity analysis showed that the study from MacIntyre and associates28 changed heterogeneity significantly for BCVA and refractive cylinder outcomes. Likewise, the study from Huang and colleagues24 changed the heterogeneity of topography cylinder outcome. Therefore, we reviewed these 2 studies carefully. Both employed retrospective reviews as research means. Table 2 shows the relevant data involved in our sensitivity analysis, including BCVA, refractive cylinder, and topography cylinder. MacIntyre and associates mentioned that the Anwar big-bubble technique and the Melles manual dissection method were used for DALK. Huang and colleagues performed the modified big-bubble technique, which is analogous to the DALK procedure reported in other studies. Postoperative follow-up time was a source of heterogeneity for the refractive cylinder outcome in accordance with subgroup analysis. However, it was only mildly associated with the heterogeneity of BCVA and topography cylinder outcomes. The relationship between postoperative follow-up time and refractive cylinder remains unclear. We speculate that another reason that these 2 studies caused an increase in heterogeneity was due to their retrospective designs, thus allowing an adequate guarantee of data integrity and homogeneity. Fortunately, with regard to BCVA, refractive cylinder, and topography cylinder outcomes, the comparative results did not change irrespective of the presence or absence of both of these studies. Therefore, we kept both in our pilot study.
There are some limitations in our review. Only 3 RCT studies were available; therefore, our outcome measures might not supply strong evidence for treatment of keratoconus. Furthermore, there was moderate heterogeneity in BCVA, refractive cylinder, and topography cylinder outcomes, possibly disrupting the reliability of our conclusion. We adopted a random-effects model with consideration of these 2 aspects. Some meta-analyses that included both RCT and non-RCT studies have been published.40 We reviewed some forest plots published previously that compared clinical outcomes of these 2 procedures and found that moderate or considerable heterogeneity was observed in some outcome measures.43 In addition, there were some high-risk biased studies included in the analysis (n = 10). This is primarily because low- or moderate-risk biased studies based on the 7 risk domains of Cochrane Collaboration are not usually found, especially for non-RCTs. Therefore, practitioners should carry out more RCTs that explore differences in clinical outcomes between PK and DALK to enhance the quality of the evidence.
Our meta-analysis showed that BCVA, refractive cylinder, and topography cylinder were comparable between patients who underwent PK versus DALK. Given that undergoing PK increased the risk of graft rejection episodes and complications, we recommend that DALK should be considered for treatment of keratoconus even if it leads to greater SE than PK.
DOI : 10.6002/ect.2019.0123
From the Beijing Tongren Eye Center, Beijing Tongren Hospital, Capital Medical,
University, Beijing Ophthalmology and Visual Science Key Laboratory, Beijing,
Acknowledgements: This work was supported by grants from National Natural Science Foundation of China, Grant/Award number: 81470608. The authors have no conflicts of interest to disclose.
Corresponding author: Zhiqiang Pan, Beijing Tongren Eye Center, Beijing Tongren Hospital, Capital Medical University, Beijing Ophthalmology and Visual Science Key Laboratory, Dongjiaominxiang 1#, Dongcheng District, Beijing, 100730, China
Figure 1. Study Flow Diagram
Figure 2. Forest Plot Comparing Best-Corrected Visual Acuity After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty
Figure 3. Funnel Plot Investigating Best-Corrected Visual Acuity
Figure 4. Forest Plot Comparing Spherical Equivalent After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty
Figure 5. Forest Plot Comparing Refractive Cylinder After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty
Figure 6. Forest Plot Comparing Topography Cylinder After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty
Figure 7. Forest Plot Comparing Graft Rejection Episodes After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty
Figure 8. Forest Plot Comparing Complications After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty
Figure 9. Summary of Risk Bias for All Included Studies
Table 1. Details of Key Words and Search Strategy
Table 2. Basic Characteristics of Included Studies
Table 3. Data Extracted From Included Studies
Table 4. Details of Complications After Penetrating Keratoplasty And Deep Anterior Lamellar Keratoplasty