Fifty-eight papers described PMD, of which 14 were deemed relevant, excluding non-clinical studies and those not relevant to the specific disease. Table 1 provides the studies on CXL for pellucid marginal degeneration. While initial studies were mainly case reports, Mamoosa et al. (2014) performed a study to treat 21 eyes of 15 patients with PMD using the “Dresden” protocol (epithelium removed, riboflavin 0.1%, 3 mW/cm2 of UV-A for 30 min = 5.4 J/cm2 total) (Table 2) [11]. This did not significantly improve uncorrected vision after 6 months, but BCVA improved, and K1 (flat) and K2 (steep) were reduced considerably. This study was the largest of its type thus far, but the follow-up was limited to 6 months. Cagil et al. [12] treated 20 eyes of 15 patients with PMD with accelerated, transepithelial CXL [12]. Over the follow-up of 36 months, the astigmatism was reduced (4.97 ± 2.00 to 3.28 ± 3.12 D, p = 0.01) alongside Kavg (47.12 ± 4.66 to 46.27 ± 4.46 μm, p = 0.01), and BCVA was stable (0.37 ± 0.35 to 0.33 ± 0.33 logMAR, p = 0.48). The use of both transepithelial and accelerated CXL (ACXL) limits the interpretation of the effects of the changes separately.

Pircher et al. [17] performed a retrospective study on 16 eyes that underwent CXL for PMD [17]. Pachymetry of the other 14 initially decreased from 456 ± 74 μm to 447 ± 77 μm and increased to 495 ± 70 μm after 12 months, and BCVA increased from 0.38 to 0.29 (logMAR). This both confirms stabilization demonstrated elsewhere and shows improvement in VA with a method similar to the Dresden technique, albeit decentered inferiorly.

Steppat et al. in 2008 reported the stabilization of the topography of all 13 eyes of 8 patients observed over 18 months with CXL [18]. Stojanovic et al., using methodology analogous to the “Dresden” technique, treated both keratoconus and PMD with the excimer laser and CXL, showing improvement in UNVA (uncorrected visual acuity) and BCVA (best-corrected visual acuity) and 12-month stability in their ectasia, although this study did not differentiate the two conditions [19]. The rest of the studies included were small case series/reports. Bayraktar et al. had a patient with PMD who underwent CXL in addition to other surgical interventions [15]. Similarly, Spadea et al. presented a 43-year-old female with PMD in the left eye, treated with Dresden protocol CXL with inferior treatment decentration of 1-mm clearance from the limbus [14]. In a further case study by Spadea et al. [16], one 59-year-old woman with bilateral PMD had the left eye treated with a 400-µm-deep incision tunnel and pocket through 160° of the inferior cornea and sliding the created flap with suturing to reduce the radius of curvature and correct high astigmatism [16]. Three months later, iontophoretic CXL was performed. One patient with Axenfeld-Reiger syndrome and PMD was treated with a similar decentered Dresden technique with a 28-month follow-up [20]. All these reported favorable results.

In 2014, Hafezi et al. published a case of a 47-year-old man with bilateral Terrien marginal degeneration (TMD) that was treated with eccentric epithelium-off CXL in both eyes while protecting the corneal limbus [5]. Irradiation was performed using the Dresden protocol. Five years of postoperative follow-up showed regression of the Kmax values (59.0 to 55.4 D in the right eye and 81.5 to 67.7 D in the left eye), a local thickening of the corneal stroma, and bilateral improvement of corrected distance visual acuity. A case report was published in 2018 by Lamarca et al. of a 28-year-old man with bilateral progressive TMD [21]. The team performed eccentric epithelium-off CXL analogous to that of Hafezi et al. [5]. After a 9-year follow-up, it has also been observed improvement in keratometry (Kmax in the right eye regressed from 51.8 to 43.9 D and from 46.2 to 45.7 D in the left eye) and visual acuity findings.

Elling et al. (2017) performed CXL on 26 eyes of 14 patients to reduce myopia [22]. Over 6 months, uncorrected vision improved (0.51 ± 0.29 logMAR to 0.14 ± 0.16) as did spherical equivalent (− 1.75 ± 0.60 to − 0.78 ± 0.51) with no reduction in the endothelial count and no serious complications. This study shows stability for 6 months in patients with low myopia.

Stodulka et al. (2019) performed a pilot study using patterned CXL to correct low hyperopia rather than myopia [23]. Twenty-two hyperopic eyes were enrolled for epithelium-on CXL on the mid-periphery to increase curvature. Uncorrected vision improved (0.30 logMAR to 0.10, p = 0.0005), refractive error was reduced (+ 0.75 to + 0.25 D, p > 0.0001), and BCVA was unchanged (0.00 logMAR, p = 0.5). This demonstrated the extension of therapeutic effects, although the correlation of attempted versus achieved refractive outcomes was weak in this pilot study (R2 = 0.1163) and may improve with further research. Kanellopoulos (2014) proposed the use of high-fluence transepithelial CXL to treat myopia, which demonstrated a consistent reduction in myopia (+ 1.44 D) in four cases, stabilizing after 1 month [24]. The long-term hyperopic effects of CXL due to axial flattening were shown in keratoconic eyes by O’Brart et al. [25]. Wollensak and Spoerl (2004) proposed the use of cross-linking in scleral collagen to constrain axial growth [26]. This could limit the progression of pathologic myopia, a growing health burden, particularly in parts of East Asia, but research is thus far limited to animal models. The topographic changes induced by the CXL have also been extended to use as a refractive procedure in itself, with clear benefits of tissue preservation and no penetration [24].

The first reported use of CXL for the management of BK was by Krueger et al. [10] and published a case report with one patient and prior results from cadaverous experimental models, using a femtosecond laser to form 2 pockets in the stroma at 150 and 350 µm depth. In each bag, 200 µl of 0.1% riboflavin was injected before UV-A exposure [10]. In the experimental model, there was both an increase in corneal clarity and a reduction in corneal thickness. In the clinical case, the patient’s VA improved from counting fingers (CF) to 20/80 in 1 month and reduced the pain, and the vision remained stable for more than 6 months, allowing the postponement of keratoplasty.

Saim Khan et al. (2016) used the unmodified Dresden protocol and an accelerated protocol (A-CXL) for 24 eyes of 24 bullous keratopathy patients [27]. While VA did not improve (2.09 logMAR ± 0.23 to 2.13 ± 0.22), the mean central corneal thickness (CCT) reduced (753.96 ± 55.16 μm to 641 ± 29.25), and non-visual symptoms reduced, if temporarily. The thickness of the corneas (~ 700 μm) and severity of edema may have limited the penetration of the riboflavin and UV light, as Bottós et al. (2010) reported [28], and thus, deeper edema may have prevented VA improvements.

The histological analysis of BK post-CXL was reported by Arora et al. [29] as they performed penetrating keratoplasty either 1 or 3 months after CXL in both groups, allowing histological analysis [30,31,32]. They also found insufficient improvements to VA post-CXL (1.925 ± 0.173 logMAR to 1.75 ± 0.296 at 1 month (p = 0.01) to 1.81 ± 0.23 at 3 months). Similarly, CCT fell initially and then rose (846.46 ± 88.741 to 781.0 ± 98.788 μm at 1 month (p < 0.01) to 805.08 ± 136.06 μm at 3 months) (Table 3). There was notable stromal compaction in 50% of cases measured by ultrasound pachymetry, but the effect was temporary, and keratocyte repopulation was complete within 1 month, measured through histological staining nuclei.

Other papers with CCTs of 700 μm using the Dresden protocol found a similar marginal improvement in VA [33, 35, 36]. There were more significant but similarly temporary improvements in pain, including 2 cases with associated corneal ulcers treated by Kozobolis et al. [37]. Kasai et al. [34] accelerated the protocol with only temporary improvements in pain and edema compared to hypertonic saline [34]. Choy et al. (2020) questioned the reliability of pain improvements and noted an increase in recurrent epithelial defects, requiring amniotic membrane transplants [38].

PACK-CXL (Table 4), using an epithelium-off method with 30-min UV light exposure at 3 mW/cm2 intensity, has generally been performed in the late stage of infectious keratitis where antibiotics have not resolved the infection, and tissue loss and perforation are concerns. One hundred eleven articles on PubMed were searched with bacterial keratitis and CXL variations, of which 11 were considered relevant here.

Iseli et al. [39] treated corneal melts of infectious origin in five patients with PACK-CXL, where intensive topical and systemic antibiotics had failed to stop the progression [39]. These cases had bacterial (2), fungal (2), and bacterial/fungal (1) etiologies, and although the melts ceased, all required keratoplasties to improve vision. This offered early evidence for wide-ranging effectiveness in late-stage, last-resort cases.

Said et al. [41] performed a randomized controlled trial with bacterial and/or fungal keratitis. They observed significantly reduced complications in the PACK-CXL with antibiotics group versus antibiotics alone. However, the final VA achieved was similar in both groups (1.64 ± 0.62 with PACK-CXL versus 1.67 ± 0.48 in control) (Table 5). This may be partly due to the severity of cases (initial VA worse than LogMAR 2 in both groups) [41].

Makdoumi et al. (2010) treated a further seven patients, demonstrating epithelialization and resolution of both hypopyons, which symptomatic improvement in six within 24 h [46]. Corneal melting ceased in all patients. Notably, Makdoumi et al. (2017) also performed an in vitro study showing effectiveness against methicillin-resistant S. aureus, and visible blue light rather than UV-A may allow deeper penetration into the cornea. However, the risk to endothelium must be weighed against those of emergency keratoplasty [47].

Idrus et al. [43] used the Dresden protocol on 27 eyes unresponsive to multiple topical broad-spectrum antibiotics [43]. This resolved all hypopyons, allowed re-epithelialization (although 20 required amniotic membrane transplant), and resulted in VA improvement (majority HM or worse to majority 0.7 logMAR or better). Notably, one patient with cowpox virus isolates resolved as well, with a final VA of 0.5.

An interventional study by Knyazer et al. [42] of 20 eyes with resistant bacterial keratitis demonstrated the effectiveness of accelerated PACK-CXL, including one case with Candida [42] (Table 6). One patient required a tectonic graft, but otherwise, there were no adverse effects, suggesting that accelerated protocols are similarly effective [6].

The other studies included smaller patient numbers and had variable success rates. Notably, Mattila et al. [40] combined PACK-CXL with an amniotic membrane transplant to reduce scarring [40]. Of the seven cases with follow-up data, all were re-epithelialized, and the infections were resolved. Two patients required later surgery (PKP and DALK) to improve visual function. Bamdad et al. (2015) and Knyazer et al. [6] have performed studies on moderate keratitis (ulcer diameter 2 to 7 mm and stromal depth < 300 µm)) [6, 51]. These showed a significant reduction in treatment time (24.7 to 17.2 days), faster re-epithelialization (12.7 to 9.3 days), and a reduction in emergency tectonic keratoplasty (19.4 to 0%) (Table 7).

Such cases undergo PACK-CXL where pre-existing treatment, typically including fluoroquinolones, is not paused or stopped. There are in vitro reports of phototoxicity from fluoroquinolones, suggesting switching such antibiotics in anticipation of CXL [56].

There were 124 articles on PubMed when searched with fungal, keratitis, and variations of CXL, of which 23 were considered relevant here. These methods are generally similar to Dresden, with fewer keratoplasties in treated groups.

Iseli et al. [39] were the first to use PACK-CXL in fungal keratitis with moderate success; one patient was successfully treated but later required topography-guided LASIK, and the other suffered progression of infiltrates after 3 weeks, requiring penetrating keratoplasty. Using methods analogous to the Dresden method, they resolved the Acremonium infection in one patient, who later needed DALK, but failed to resolve another with Fusarium infection 33 (Table 8).

Against this, Said et al. [41] found limited evidence for the benefit (shorter time to re-epithelialization and fewer perforations) [41], and Uddaraju et al. [44] found evidence for poorer outcomes following PACK-CXL for patients with fungal keratitis [44]; the latter was questioned by Richoz for its small sample size, inhomogeneity, late stage of infections, and the use of an unmodified “Dresden” protocol [61].

Wei et al. (2019) performed a blinded trial with 41 patients. They found faster re-epithelialization and infection resolution (2.21 months versus 1.30), more rapid reduction in ulcer depth, and time for loss of hyphae. However, the systematic and standardized analysis of patients on presentation in a large city may imply a more rapid initiation of antifungal therapy [62]. A laboratory study on the successful, early, exclusive use of accelerated PACK-CXL on bacterial keratitis by Richoz et al. (2014) supports this [63].

Lamarca et al. (2016) reported PACK-CXL failure in one case of fungal keratitis, first treated with medroxyprogesterone, moxifloxacin, and tetryzoline [64]. PACK-CXL was performed due to the worsening of hypopyon. The infection progressed despite sensitivity to natamycin and voriconazole. This eventually led to vitreous and penetrating keratoplasty (PK) with further postoperative complications, although the initial use of corticosteroids may have been a factor. Thakur et al. (2019) used the Dresden protocol against a fungal “phaco-tunnel” infection which proved successful, but the partial success of intracameral and intrastromal antifungal injections was likely a factor, and early presentation and treatment may also have roles [65]. There are also more cases [45, 66,67,68] with success attributed to PACK-CXL, but CXL protocols were not provided.

Ten papers were considered relevant in assessing the use of CXL to treat Acanthamoeba keratitis. These papers generally used methods similar to Dresden, although one repeated with higher-intensity UV-A, and one used an accelerated method.

Initial laboratory use of PACK-CXL to treat Acanthamoeba-seeded plates [69,70,71] and Acanthamoeba-seeded rabbit cornea [72] suggested that CXL may not improve and may even accelerate the infection process. However, this latter study was criticized by Richoz et al. (2013) [73], noting that it omitted to state the depth of the injection site and that the study had cyst and trophozoite numbers resembling a chronic infection, as well as omitting antiprotozoal medication. To note, the use of rose Bengal in PACK-CXL in vitro to substitute for riboflavin by Atalay et al. [55] has shown more success in vitro (death rates in control: 12%, riboflavin: 11–14.5%; rose Bengal: 63–66%) [55].

Demirci and Ozdamar (2013) treated a young man with PACK-CXL who had developed an Acanthamoeba corneal infection; he had developed corneal melt and stromal necrosis despite topical propamidine/chlorhexidine [74]. He showed symptomatic improvement within 24 h and return to his pre-infection VA within a month.

Arance-Gil et al. [52] presented a case study of Acanthamoeba keratitis after 1 year of failure to resolve with topical medication [52]. The case improved significantly, initially with no cysts on confocal microscopy, although the cornea required AMT to complete epithelialization. Unfortunately, corneal melting and glaucoma occurred 8 months later, requiring both PK and surgery for glaucoma and subsequent cataract.

Kymionis et al. [53] reported a case of combined Acanthamoeba, Aspergillus, and Candida keratitis, treated with PACK-CXL analogous to the Dresden technique [75]. Two-day post-treatment, a corneal abscess formed extending into half of the cornea before resolution with extensive scarring, with a PK, planned to restore vision. Although extensive scarring occurred, the multi-organism infection resolved over 4 weeks.

Nateghi Petterson et al. [54] performed accelerated PACK-CXL with higher total fluence (7.2 J/cm2) in a case refractive to pharmacology and a previous PACK-CXL procedure (5.6 J/cm2) [