Photoactivated chromophore-corneal cross-linking accelerates corneal healing in fungal keratitis: an updated meta-analysis

This study was conducted in line with the Cochrane Handbook for Systematic Reviews of Interventions and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA 2020; Supplementary materials) [15, 16].

Relevant articles, limited to human and RCTs in Embase, MEDLINE with PubMed, Web of Science, and Cochrane Library, were searched and published up to April 5, 2022, by two authors (S. Y. Liu and S. F. Fang). The search strategy contained three components: clinical condition (“keratitis,” “corneal ulcer”), intervention (“cross-linking reagents,” “riboflavin,” “anti-infective agents”, “ultraviolet therapy,” “photosensitizing agents,” “ultraviolet rays,” “collagen”), and study type (randomized clinical trial). Detailed search strategies are provided in the Supplementary materials. Further, the reference lists in the eligible RCTs were examined manually in case that there were other eligible studies. This procedure was repeated until no more studies were discovered.

Based on predefined criteria, two authors (S. Y. Liu and S. F. Fang) independently selected the title and abstract of the 1256 studies identified by the search. Following that, we downloaded the full texts of these articles and conducted a review. Another author (L. J. Zhang) examined the data. Discussions were also performed if there was a divergence.

Two authors (S. Y. Liu and S. F. Fang) independently extracted data in the included RCTs: first author, publication year, sample size, patient characteristics, the protocol of adjuvant PACK-CXL, drug regimen, and other outcomes data. Another author (L. J. Zhang) examined the data. Discussions were also performed if there was a divergence.

Cochrane Collaboration’ s tool was utilized to identify the risk of bias [17]. Trials were scored as high, low, or unclear based on the following items: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. The study with high risk for ≧ 1 item was regarded as high risk, and the study with low risk for all items was regarded as low risk. Otherwise, the study was regarded as unclear risk [18].

Certainty of evidence was examined using Grading of Recommendations, and Assessment, Development and Evaluations (GRADE) [19, 20], and summary tables were obtained through GRADE profiler online (https://​gradepro.​org/​).

A total of 1256 studies were screened, in which 278 duplicate publications (22%) and 958 irrelevant studies (76%) were excluded after the initial screening. Full text of the twenty (2%) studies were acquired for further evaluation. Thirteen studies were excluded: two conference abstracts [21, 22], one editorial [23], one protocol [24], four studies without RCT design [25‐28], two letters to the editor [29, 30], one study with duplicate data [31], one secondary analysis [32], one compares PACK-CXL only with SAT only [33]. Finally, seven RCTs were included [11‐14, 34‐36], and the procedure is shown in Fig. 1.

The characteristics of these RCTs were shown in Table 1. The publication year ranged from 2015 to 2022, and the sample size ranged from 13 to 111 (together, 283). Eighty patients were diagnosised as bacterial keratitis, and 203 patients were diagnosised as fungal keratitis. One-hundred thirty-nine patients received adjuvant PACK-CXL and SAT, while 144 patients received SAT only. All the trials recorded adverse events [11‐14, 34‐36], three trials reported the duration of corneal healing performed [11, 13, 34], two trials reported the size of corneal epithelial defect at 1 week [34, 36], three trials reported the size of corneal infiltrate at 1 week [11, 34, 36], only one trial reported the depth of corneal infiltrate at final follow-up [11], and six trials reported visual acuity at final follow-up [11‐14, 35, 36].

Risk-of-bias assessment for the seven trials was presented in Fig. 2. All the trials were classified as with a high risk of bias.

Fifty-two patients with fungal keratitis in two trials were included (27 patients with adjuvant PACK-CXL) [11, 13]. Thirty-two patients with bacterial keratitis in one trial were included (16 patients with adjuvant PACK-CXL) [34]. Adjuvant PACK-CXL significantly reduced the time needed to perform corneal healing in fungal keratitis (MD =  − 1.13, 95% CI, − 1.83 to − 0.42, P < 0.05), with low heterogeneity among the two trials (I² = 0%, P = 0.33) (Fig. 3).

Eighteen patients with fungal keratitis in one trial were included (eight patients with adjuvant PACK-CXL) [36]. Forty-four patients with bacterial keratitis in two trials were included (23 patients with adjuvant PACK-CXL) [34, 36]. All these trials found that adjuvant PACK-CXL could not reduced the size of corneal epithelial defect at 1 week in fungal and bacterial keratitis. However, Kasetsuwan et al. presented this outcome with “median, quartile,” and we failed to perform a meta-analysis.

Fifty-one patients with fungal keratitis in two trials were included (29 patients with adjuvant PACK-CXL) [11, 36]. Fifty-four patients with bacterial keratitis in two trials were included (23 patients with adjuvant PACK-CXL) [34, 36]. For fungal keratitis, all these trials found that adjuvant PACK-CXL could not reduced the size of corneal infiltrate at 1 week. For bacterial keratitis, it was controversial. Bamdad et al. found that adjuvant PACK-CXL could reduced the size of corneal infiltrate at 1 week, while Kasetsuwan et al. found that it could not. However, Kasetsuwan et al. presented this outcome with “median, quartile,” and we failed to perform a meta-analysis.

Only one trial was included; the ulcer depth did not reduce with the administration of PACK-CXL (P > 0.05) [11]. Therefore, analysis was not carried out.

Two-hundred three patients with fungal keratitis in five trials were included (99 patients with adjuvant PACK-CXL) [11‐13, 35, 36]. Forty-eight patients with bacterial keratitis in two trials were included (24 patients with adjuvant PACK-CXL) [14, 36]. All these trials found that adjuvant PACK-CXL could not improve visual acuity, even might result in decreased visual acuity. However, some RCTs presented this outcome with “median, quartile,” and we failed to perform a meta-analysis.

One-hundred eighty-five patients with fungal keratitis in four trials were included (91 patients with adjuvant PACK-CXL) [11‐13, 35]. Sixty-eight patients with bacterial keratitis in one trial were included (33 patients with adjuvant PACK-CXL) [14, 34]. Adjuvant PACK-CXL could not reduce adverse events in both fungal and bacterial keratitis (RR = 0.78, 95% CI, 0.38 to 1.60, P = 0.49; RR = 0.36, 95% CI, 0.08 to 1.71, P = 0.20, respectively) (Figs. 4 and 5).

GRADE evidence is presented in Table 2. In terms of the duration of corneal healing performed, and adverse events risk, the certainty of evidence was low.

We reviewed the relevant studies that compared the effectiveness and safety of adjuvant PACK-CXL versus SAT alone in infectious keratitis comprehensively and systematically. Adjuvant PACK-CXL reduced the time needed to perform corneal healing as compared to SAT alone in fungal keratitis. The risks of adverse events were not significantly different in both fungal and bacterial keratitis.

For the primary outcome, the duration of corneal healing performed, two studies on this subject have been published but reached the opposite conclusion. Davis et al. revealed that it was very uncertain whether adjuvant PACK-CXL was more effective than SAT alone for reepithelialization and complete healing [9]. While Ting et al. found that when compared to SAT alone, adjuvant PACK-CXL resulted in shorter mean time to complete corneal healing [7]. Different types of infectious keratitis may be the underlying reason for this divergence. Therefore, we conducted a subgroup analysis based on infection type and revealed that adjuvant PACK-CXL could reduce the time needed to perform corneal healing as compared to SAT alone in fungal keratitis. For the risks of adverse effects, we also conducted subgroup analyses based on infection type. In accordance with the previous study, adjuvant PACK-CXL could not reduce the risks of adverse effects as compared to SAT alone in both bacterial and fungal keratitis [7].

Certain differences should be highlighted. Firstly, previous studies included case reports, quasi-RCTs, and case series. To achieve reliable estimates, we set strict inclusion criteria. Only the RCTs clarified clearly the enrollment of patients with adjuvant PACK-CXL were included. We further included another four newly published RCTs with 208 patients published in 2019–2021 in this meta-analysis, which promoted statistical power [11‐14]. Secondly, we conducted subgroup analyses based on infection type. Thirdly, due to clinical heterogeneity, a random-effects model was chosen to guarantee a more conservative estimation. Lastly, the GRADE method was employed to evaluate the certainty of evidence in order to assist clinical practice. Therefore, the present study was the most updated and thorough, reinforcing prior results.

In recent years, the antimicrobial effect of PACK-CXL has been investigated in infectious keratitis [37, 38]. However, the administration of PACK-CXL in infectious keratitis was still controversial [39, 40].

We analyzed the suspected reasons, and possible explanations are as follows. Firstly, the clinical outcomes and the risk of adverse events may be related to the severity of infectious keratitis. Uddaraju et al. enrolled patients with culture-positive deep stromal fungal keratitis who had not responded to appropriate treatment for 2 weeks [35]. The poor response could be explained that fungal infections penetrated deeper, and the intensity of UV-A light would not be sufficient to treat. Early and superficial fungal keratitis responded well to PACK-CXL [27, 41]. While in advanced and deep stromal fungal keratitis, it was hard to determine whether the infection reacted to PACK-CXL alone [42]. Secondly, drug regimens of SAT were not uniform in the seven eligible trials. Thirdly, although most of the included studies used the standard Dresden protocol that UV-A radiation exposure of 3 mW/cm² for 30 min, the lengths of wave were different, 365 nm or 370 nm. The efficacy of PACK-CXL followed the Bunsen-Roscoe law of reciprocity, and higher fluence or irradiance substantially increased the killing rates [33, 43], while some study suggested that accelerated PACK-CXL provided an antimicrobial effect similar to the low-intensity, slow setting [44]. Therefore, safety limits for clinical application are required in further studies. Lastly, different causative microorganisms were included in this study. Alio et al. revealed that PACK-CXL decreased corneal melting with the following order from most to least: gram-negative bacteria, gram-positive bacteria, acanthamoeba, and fungus [45]. It was worth noting that PACK-CXL presented a weaker killing effect in acanthamoeba, and even be a risk of activating the latent virus, so PACK-CXL should be applied carefully, in patients with acanthamoeba or viral keratitis [46, 47]. In short, ophthalmologists should pay more attention to the type and severity of infectious keratitis, drug regimens of SAT, and PACK-CXL protocol in clinical practice.

The strength of the present study was that only RCTs were included to reduce heterogeneity. And GRADE method was applied to examine the certainty of evidence. There were some limitations in the present study which may affect the interpretation. Firstly, all the included trials presented detailed data; however, different trials focused on different outcomes or documented the same outcome at different follow-up time. So, it was hard to synthesize the data and perform meta-analysis. Secondly, since only seven trials were included, it was difficult to deny the heterogeneity between the included studies. The heterogeneity may come from the following: population, the type and severity of infectious keratitis, drug regimens of SAT, PACK-CXL protocol, and the judgment of subjective outcomes. All these factors may lead to the lack of precise proof to recommend PACK-CXL as a potential approach in infectious keratitis. For further RCTs trials, we suggested that the type and severity of infectious keratitis, drug regimens of SAT, and PACK-CXL protocol should be in the consistency, making the outcomes be more comparable. Meanwhile, the judgment of subjective outcomes should be performed by at least two ophthalmologists, making the data more precise. Therefore, more RCTs comparing cases with similar baseline are needed.