LITERATURE SEARCH

We searched for articles in PubMed, Cochrane Library, ScienceDirect, Embase, and ClinicalTrial.gov databases until October 1, 2021 with the following keywords: Descemet*

OR endothelial keratoplasty OR DSAEK OR DSEK AND thickness OR thin OR ultrathin OR

UT-DSAEK (details of search strategy used within each database are presented in Appendix 1). The search was not limited to specific years. No language restrictions were applied. We imposed no limitation on sample size or regional origin. To be included, articles had to report clinical outcomes (ie, visual acuity, pachymetry, endothelial cell count, spherical equivalent, or rebubbling or rejection rates) of T/UT DSAEK, defined as DSAEK with preoperative graft thicknesses <130 µm and <100 µm, respectively. Animal studies were excluded. In addition, references of all publications meeting the inclusion criteria were manually searched to identify any further studies. The search strategy is presented in Figure 1 . Two authors (L.B., V.N.) conducted literature searches, collated and separately reviewed the abstracts and, based on the selection criteria, decided the suitability of the articles for inclusion. A third author (F.D.) was asked to review the articles where consensus on suitability was debated. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (Appendix 2).

Literature search strategy. DSAEK = Descemet stripping automated endothelial keratoplasty; DSEK = Descemet stripping endothelial keratoplasty; SD = standard deviation; UT = ultrathin.

DATA COLLECTION

The data collected included first author’s name, publication year, country, study design, aims, outcomes of included articles, inclusion and exclusion criteria, sample size, number of eyes, age, percentage of males, endothelial disease, graft cutting technique, type of surgery (simple or triple EK), characteristics of donors (age, sex), measurement time, central graft thickness, central corneal thickness, spherical equivalent, best spectacle-corrected visual acuity, endothelial cell count, graft rejection rate, and rebubbling rate.

OUTCOMES

The primary outcome was best spectacle-corrected visual acuity (logMAR). Secondary outcomes were pachymetry as central corneal thickness (µm), endothelial cell count (cells per square millimeter), spherical equivalent (diopters [D]), graft rejection rate (%), and rebubbling rate (%).

QUALITY OF ASSESSMENT

We used the Scottish Intercollegiate Guidelines Network (SIGN) criteria to assess the methodological quality of the included articles, both for randomized clinical trials and cohort studies, with the dedicated evaluation grids. SIGN Controlled Trials and SIGN Cohort Studies checklists consisting of respectively 10 and 14 items. We presented an overall quality score for each included study, based on the main causes of bias evaluated in section 1 of the checklist through 4 possibilities of answers (yes, no, can’t say, or not applicable).

STATISTICAL CONSIDERATIONS

Statistical analyses were conducted using Stata software (StataCorp). Characteristics of subjects and parameters evaluated were reported for each study sample as mean ± standard deviation (SD) and number (%) for continuous and categorical variables, respectively. We computed random-effects meta-analysis (DerSimonian and Laird approach) on changes between after the intervention and baseline measures, for each efficacy outcome: best spectacle-corrected visual acuity in logMAR, central corneal thickness in µm, endothelial cell count from donor in cell/mm ² , and SE in D. For graft rejection and rebubbling, meta-analysis was expressed in rates (%). When a study presented several time measurements, the last follow-up time was retained for the main analysis. When visual acuity was evaluated with a Snellen chart or a decimal scale, it was converted to logMAR using the formula: logMAR visual acuity = −log10 (decimal visual acuity). Changes were calculated using the method from Borenstein and associates. This approach has the advantage to express results—ie, effect sizes (ESs) as natural values (logMAR, µm, cell/mm ² , and D, respectively—but suffers from statistical issues because the number of eyes is only taken into account from SD, and by choosing an arbitrary correlation coefficient of 0.5. Therefore, to be statistically accurate, we further computed aforementioned meta-analysis using standardized mean differences (SMD) by comparing measures of efficacy after the intervention to baseline measures (before the intervention). SMD with their 95% CIs were represented graphically on forest plots. A SMD was defined as an unitless measure of the effects of T/UT-DSAEK on outcomes. A SMD centered at 0 reflects the absence of effect, 0.2 a small effect, 0.5 a moderate effect, and 0.8 a large effect. We stratified our meta-analyses on the thickness of corneal grafts in T/UT-DSAEK (<80 μm, 80-100 μm, and 100-130 μm), according to the preoperative thickness (or the earliest postoperative graft thickness). Lastly, we performed several sensitivity analyses to verify the strength of our results. First, we repeated meta-analyses with all measurement times. Then, as recovery from surgery is likely to be achieved over the long term, we computed previous meta-analyses using only studies with a follow-up >1 year. As the most common surgery is the microkeratome technique for Fuchs endothelial dystrophy, we also computed sensitivity meta-analyses on those studies. Heterogeneity in our results was evaluated by examining forest plots, 95% CIs, and the I ² statistic. Heterogeneity between studies is considered low for 0% < I ² <25%, modest for 25% < I ² < 50%, and high for 50% < I ² < 100%. For rigor, funnel plots (metafunnels) of meta-analyses were used to search for publication bias, and meta-analyses were conducted excluding studies not evenly distributed around the base of the funnel. When possible (sufficient sample size), metaregressions were proposed to study the relationship between clinical outcomes and clinically relevant parameters such as characteristics of subjects and donors (age, sex), pathology causing endothelial failure, time, and characteristics of surgery. Results of meta-regressions were expressed as regression coefficients and 95% CIs. P values < .05 were considered statistically significant.