Individualized Corneal Cross-linking With

Riboflavin and UV-A in Ultrathin Corneas: The

Sub400 Protocol

FARHAD HAFEZI, SABINE KLING, FRANCESCA GILARDONI, NIKKI HAFEZI, MARK HILLEN,
REYHANEH ABRISHAMCHI, JOSE ALVARO P. GOMES, COSIMO MAZZOTTA, J. BRADLEY RANDLEMAN, AND
EMILIO A. TORRES-NETTO

PURPOSE: To determine whether corneal cross-linking significant changes were found in CDVA (P [ .611),
(CXL) with individualized fluence (‘‘sub400 protocol’’) is sphere (P [ .077), or cylinder (P [ .915).
able to stop keratoconus (KC) progression in ultrathin
CONCLUSIONS: The ‘‘sub400’’ individualized fluence
corneas with 12-month follow-up. CXL protocol standardizes the treatment in ultrathin cor-

DESIGN: Retrospective, interventional case series. neas and halted KC progression with a success rate of

METHODS: Thirty-nine eyes with progressive KC and 90% at 12 months. The sub400 protocol allows for the
corneal stromal thicknesses from 214 to 398 mm at the treatment of corneas as thin as 214 mm of corneal stroma,
time of ultraviolet irradiation were enrolled. After epithe- markedly extending the treatment range. The DL depth
lium removal, ultraviolet irradiation was performed at 3 did not predict treatment outcome. Hence, the depth is
mW/cm2 with irradiation times individually adapted to unlikely related to the extent of CXL-induced corneal
stromal thickness. Pre- and postoperative examinations stiffening but rather to the extent of CXL-induced micro-
included corrected distance visual acuity (CDVA), refrac- structural changes and wound healing. (Am J
tion, Scheimpflug, and anterior segment optical coherence Ophthalmol 2021;224:133–142. Ó 2020 Elsevier Inc.
tomography imaging up to 12 months after CXL. Outcome All rights reserved.)
measures were arrest of KC progression at 12 months post-
operatively and stromal demarcation line (DL) depth.

K

RESULTS: Thirty-five eyes (90%) showed tomographi- ERATOCONUS (KC) IS A CHRONIC PROGRESSIVE
cal stability at 12 months after surgery. No eyes showed ophthalmic disease that leads to bulging and thin-
signs of endothelial decompensation. A significant corre- ning of the cornea, resulting in increasing irregular
lation was found between DL depth and irradiation time astigmatism and, ultimately, poor vision. Because of a study
(r [ D0.448, P [ .004) but not between DL depth published in 1986 showing a prevalence of 0.05% in a sin-
and change in Kmax (r [ L0.215, P [ .189). On gle state of the US population, KC has been classified as a
average, there was a significant change (P < .05) in thin- rare disease.1 However, recent studies including modern
nest stromal thickness (L14.5 ± 21.7 mm), Kmax (L2.06 screening methods have demonstrated a prevalence 5 to
± 3.66 D) and densitometry (D2.00 ± 2.07 GSU). No 10 times higher in Western Europe and up to almost 100
times higher in certain regions of the Middle East compared
with the historic 1986 study.2,3
Supplemental Material available at AJO.com. First introduced in 2003, corneal cross-linking (CXL) is
Accepted for publication Dec 7, 2020. a treatment that can prevent KC progression.4,5 CXL has
From the Laboratory for Ocular Cell Biology, Center for Applied radically changed the visual prognosis of patients with
Biotechnology and Molecular Medicine, University of Zurich, Zurich
(F.H., F.G., R.A., E.A.T.-N.); Department of Ophthalmology, ELZA KC and other ectasias, namely, postoperative ectasia and
Institute, Dietikon (F.H, F.G., N.H., M.H., R.A., E.A.T.-N.), pellucid marginal degeneration.4,5 CXL acts to halt KC
Switzerland; Department of Ophthalmology, USC Roski Eye Institute, progression by increasing the biomechanical stiffness of
University of Southern California, Los Angeles, California (F.H.),
USA; Faculty of Medicine, University of Geneva, Geneva, Switzerland the cornea.6 In the original standard CXL (‘‘Dresden’’) pro-
(F.H., E.A.T.-N.); Department of Ophthalmology, University of tocol, the corneal stiffening was achieved first by debriding
Wenzhou, Wenzhou, China (F.H.); Department of Information the corneal epithelial cells, saturating the corneal stroma
Technology and Electrical Engineering, Swiss Federal Institute of
Technology Zurich, Zürich, Switzerland (S.K.); Department of with riboflavin, and then irradiating the corneal stroma
Ophthalmology, Paulista School of Medicine, Federal University of Sao with ultraviolet (UV)-A light at 365 nm.6 The riboflavin
Paulo, Sao Paulo, Brazil (J.A.P.G., E.A.T.-N.); Department of in the stroma absorbs the UV energy, resulting in a photo-
Medicine, Surgery and Neurosciences, University of Siena (C.M.);
Department of Ophthalmology, Siena Crosslinking Center (C.M.), chemical reaction that generates reactive oxygen species
Siena, Italy; Department of Ophthalmology, The Cole Eye Institute, (ROS). The ROS induce covalent bonds between the
Cleveland Clinic, Cleveland, Ohio (J.B.R.), USA. collagen fibers and the proteoglycans of the extracellular
Inquiries to Farhad Hafezi, ELZA Institute, Webereistrasse 2, 8953
Dietikon, Switzerland; e-mail: farhad@hafezi.ch matrix (EM).7,8 At the same time, the riboflavin acts to

0002-9394/$36.00 © 2020 ELSEVIER INC. ALL RIGHTS RESERVED. 133

https://doi.org/10.1016/j.ajo.2020.12.011
shield deeper corneal layers (particularly the corneal endo-
thelium) from UV-induced damage and cell death.7,8 TABLE 1. Technical Specifications Used in the Sub400
Before CXL could be used clinically, the concern for Protocol
endothelial cell safety had to be addressed. The Dresden
CXL Technical Settings
protocol specified that a minimal corneal thickness of
400 mm after epithelial debridement needs to be present Parameter Individualized CXL
for CXL to be performed. This measurement was made Treatment target Keratoconus
based on riboflavin diffusion calculations and the total Fluence (total) (mJ/cm2) Variable
Soak time (min) and interval 20 (q2)
amount of UV energy that would be delivered to the
Intensity (mW/cm2) 3
cornea, especially at the endothelial level. As a result,
Treatment time (min) Variable
this 400 mm corneal thickness limitation has been Epithelium status Off
excluding many corneas with ectasias like KC that may Chromophore 0.1% riboflavin
have benefit from CXL-induced strengthening from Light source CXL-365
receiving that treatment. Irradiation mode Continuous
Since 2009, various modifications of the epithelium-off
Dresden protocol to overcome the 400 mm limit have CXL ¼ corneal cross-linking.
been developed.9–11 These techniques aimed at
modifying stromal thickness to allow for a safe and
effective CXL treatment. Two examples include hypo- principles of Good Clinical Practice, the Human Research
osmolaric riboflavin used to swell a thin cornea to a thick- Act, the Human Research Ordinance, and local regulations.
ness of more than 400 mm, and in contact lens-assisted Male and female patients with progressive corneal ectasia
CXL (CACXL), the stroma is artificially ‘‘thickened’’ by and a corneal stromal thickness <400 mm were enrolled in
placing a contact lens over the cornea. A third approach the study. Progressive ectasias were considered to be KC eyes
has been to leave islands of epithelium over the thinnest with an increase in anterior keratometry > _1 D within the
areas of the corneal stroma. Although all 3 were promising, last 12 months and/or primary ectasia in patients aged 9-
each of these techniques has major limitations and was not 19 years,13,14 who were, based on age at presentation, also
standardized. For these reasons, our group developed and considered to be progressive and, therefore, were treated at
published an algorithm that rather adapted the overall the initial presentation. The maximum increase in the ante-
fluence in the CXL procedure based on the patient’s indi- rior sagittal keratometry was evaluated through the differen-
vidual stromal thickness (sub400 protocol) to cross-link tial maps via comparison of the Scheimpflug tomography or
the stroma, still protecting the corneal endothelium from Placido examinations. Comprehensively, differential maps
damaging amounts of UV-A irradiation.12 of the anterior sagittal curvature were electronically gener-
In this study, the algorithm12 was used to individualize ated and evaluated. In cases in which the differential maps
irradiation settings based on each patient’s minimal were not available due to software incompatibility or
corneal thickness at the end of riboflavin soaking but because the patient had a previous examination stored in
immediately before administering UV-A irradiation. We a nonelectronic format, maximum keratometry was used
then investigated whether CXL with individualized fluence to compare the 2 examinations. In all cases, progression of
was able to stop KC progression in ultrathin corneas at 1 ectasia was characterized by an increase of at least 1 D in
year after treatment. the anterior sagittal curvature preoperatively.
Exclusion criteria included a history of >10 pack-years of
tobacco smoking,15–17 pregnancy or lactating women, pre-
existing ocular trauma, previous ocular surgery, inability to
PATIENT AND METHODS understand the nature of the study and/or give consent, and
patients under guardianship.
THE STUDY WAS PERFORMED IN PATIENTS WHO PRESENTED Clinical data, including corrected distance visual acuity
with progressive KC and corneal stromal thicknesses of (CDVA), refraction, and biomicroscopy, were recorded
<400 mm. Surgeries were performed between May 2016 before surgery and postoperatively at 1 month and
and December 2018 at the ELZA Institute in Dietikon/ 12 months after CXL. To assess the depth of the demarca-
Zurich, Switzerland, and the data were collected retrospec- tion line (DL) after CXL, anterior segment optical coher-
tively. Approval from Cantonal ethics committee of the ence tomography (OCT) was performed during the 1-
Canton of Zurich was granted for retrospective data collec- month postoperative consultation using spectral-domain
tion (BASEC number 2018-02369), data were collected OCT technology (Spectralis HRA version 1.10.0.0; Heidel-
through a search in the patient database system, and written berg Engineering, Heidelberg, Germany). The distances
consent was received from all patients. This study was con- from the DL to the anterior stroma and from the DL to
ducted in accordance with the Declaration of Helsinki, the the endothelium were recorded. In all patients, such

134 AMERICAN JOURNAL OF OPHTHALMOLOGY APRIL 2021

FIGURE 1. Schematic drawing of the principle of the sub400 protocol. DL [ demarcation line.

measurements were performed by the same examiner and determine the patient’s individualized fluence require-
distances were measured at the thinnest point of the cornea. ment—per our published nomogram—aiming to obtain a
All subjects had corneal evaluations performed using a DL 70 mm above the corneal endothelium.12
rotational Scheimpflug system (Pentacam HR; Oculus, To facilitate the clinical application of individualized
Wetzlar, Germany) by the same trained individual. The fluence, irradiation intensity/irradiance was kept fixed at
standard resolution setting was used to capture images 3 mW/cm2, whereas treatment time was modified. A table
(25 images per scan), and the following parameters were was created depicting the individual fluence applicable to
recorded: thinnest corneal stromal thickness, anterior thin corneas in increments of 10 mm (Table 2). Then,
radius of curvature in the 3.0 mm zone centered on the CXL was performed at 365 nm using a commercially avail-
thinnest location of the cornea (ARC3 mm), maximum able CXL device (CXL-365; Schwind Eye-Tech-Solutions,
anterior keratometry (Kmax), total anterior densitometry Kleinostheim, Germany) at an intensity of 3 mW/cm2.
(AntDens), total central densitometry (CenDens), total Postoperatively, a bandage contact lens was placed, and
posterior densitometry (PostDens), and total average antibiotic and corticosteroid drops were administered. Pa-
densitometry (TotalDens). According to the Scheimpflug tients were re-examined at the slit lamp on postoperative
system’s standard parameters, AntDens corresponds to day 1, and daily until the epithelium was closed, as well
the 120 mm most superficial corneal layers and PostDens as after 1 month and 12 months after the procedure. The
corresponds to the 60 mm closest to the endothelium. contact lens was removed on day 4. Postoperative drops
included fourth-generation fluoroquinolone antibiotics

THE SUB400 PROTOCOL: Table 1 and Figure 1 summarize twice a day for 7 days, followed by 0.1% fluorometholone
the technical specifications and CXL surgical principles. drops twice a day for 12 weeks, and preservative-free artifi-
CXL was performed by mechanically removing the epithe- cial tears as needed.
lium over 9 mm of the central cornea. After de-
epithelialization, the cornea was soaked with sodium
STATISTICAL ANALYSIS: Statistical analysis was
edetate and trometamol-enriched riboflavin phosphate performed using IBM SPSS Statistics (version 25; SPSS,
0.1% hypotonic solution (Ricrolinþ; Sooft, Montegiorgio, Inc; FML Liquifilm;. Allergan, Inc., Irvine, CA). The
Italy) for 20 minutes. Ultrasound pachymetry was Shapiro-Wilk test was used to test all variables for
performed every 5 minutes during soaking to monitor even- normality. In situations in which both variables were
tual changes in corneal stromal thickness (SP-1000; normally distributed, the t test (2-tailed) was used. In cases
Tomey, Nagoya, Japan). Guided by the preoperative where at least one of variables was not normally distributed,
Scheimpflug images, the intraoperative pachymetry mea- the related-samples Wilcoxon-signed rank nonparametric
surements were performed in the thinnest area of the test was used for further analysis. In cases of normally
cornea. As a routine, 10 measurements were taken in the distributed variables, the results were reported as mean 6
thinnest area and the lowest value was considered. At standard deviation, whereas in variables abnormally
the end of the soaking period, corneas were rinsed with distributed, they were reported as medians and interquartile
balanced salt solution to rinse off any surplus of riboflavin, ranges. Statistical analysis was performed with a confidence
and ultrasound pachymetry was performed to determine interval of 95%. Pearson or Spearman correlation tests
minimal stromal thickness. This intraoperative pachyme- were performed for normally and abnormally distributed
try measurement was performed at the end of the riboflavin variables, respectively. Correlations were considered signif-
instillation, as this corneal thickness value was required to icant at the 0.05 level (2-tailed).

VOL. 224 INDIVIDUALIZED CXL: THE SUB400 PROTOCOL 135


DEMARCATION LINE: Figure 2 display different treated
TABLE 2. Table Describing the Individual Fluence in eyes, with DL markings. The distance from the DL to the
Increments of 10 mm anterior stroma was 275 6 61 mm (range: 126-385 mm)
and from the DL to the endothelium 93 6 47 mm (range:
Individualized CXL
0-250 mm). A significant correlation was found between
Sub400 Protocol DL depth and irradiation time (r ¼ þ0.448, P ¼ .004).
Minimum Stromal Required UV Irradiation Demarcation Line Notably, there was a considerable standard deviation in
Thickness (mm) Duration (min) Depth (mm) DL across the entire set of patients (Figure 3). In most cases,
200 1 130 the distance from the DL to the epithelium was larger than
210 01:20 140 anticipated (Figure 4).
220 01:40 150
230 2 160
CORNEAL THICKNESS AND KERATOMETRY: Intraopera-
240 02:30 170 tive minimum ultrasound pachymetry after riboflavin soak-
250 3 180 ing was 343 6 46 mm (range: 214-398 mm), of which 1 eye
260 03:30 190
(2.6%) had between 214 and 250 mm, 3 eyes (7.7%) be-
270 4 200
tween 251 and 300 mm, 16 eyes (41%) between 301 and
280 5 210
350 mm, and 19 eyes (48.7%) between 351 and 400 mm.
290 6 220
300 7 230
Plots showing pre- and intraoperative thinnest corneal
310 9 250 thickness vs Kmax for each case are available as Supple-
320 10 255 mental Material.
330 12 265 Scheimpflug data showed that, on average, there was a
340 14 275 significant change from baseline at 12 months in thinnest
350 16 283 thickness (14.5 6 21.7 mm, P < .05) and in Kmax
360 18 290 (2.06 6 3.66 D, P ¼ .001), but no difference in
370 20 300 ARC3 mm (0.05 6 0.32 mm, P ¼ .089). ARC3 mm
380 23 310
values were 6.50 6 1.03 mm at baseline and 6.52 6
390 26 320
0.77 mm at 1 year after (P ¼ .089). Mean Kmax values
400 29 330
were 58.5 6 7.6 D at baseline and 56.4 6 7.8 D at 1
CXL ¼ corneal cross-linking; UV ¼ ultraviolet. year after (P ¼ .001).
In the individual eyes, Kmax remained stable (less than 1
D change) or improved at 1 year after CXL in 35 of 39 eyes,
demonstrating that CXL successfully halted progression in
90% of the eyes from this series. Eight eyes had a Kmax flat-
RESULTS tening of up to 1.0 D (20%), 14 eyes between 1.1 and 2.0 D
(36%), 4 eyes between 2.1 and 3.0 D (10%), 3 eyes between
THIRTY-NINE EYES FROM 32 PATIENTS WITH CORNEAL STRO- 4.1 and 8.0 D (8%), and 3 eyes had a Kmax flattening above
mal thickness <400 mm were enrolled in the study. The 8.1 D (8%). Three eyes (8%) showed an increase in Kmax
mean age was 29.1 6 10.1 years (range: 13-55 years): 8 below 1.0 D (range: 0-0.4 D).
eyes (21%) were from patients between 13 and 20 years Four eyes (10%) from 3 patients showed an increase of
old, 14 eyes (36%) from patients between 21 and 30 years >
_1 D in Kmax (range: 1.3-2.8 D), consistent with treatment
old, 11 eyes (28%) from patients between 31 and 40 years failure and continued progression. All 4 failed treatment
old, and 6 eyes (15%) were from patients over 40 years of eyes were highly progressive before CXL, 2 of which had
age. Thirty-four eyes (87%) had KC with an increase in progressed up to 7.4 and 9.9 D within 6 months preopera-
anterior keratometry > _1 D within the last 12 months— tively. Curiously, those 2 mentioned very high preopera-
mean 2.5 6 2.0 D (range: 1.0-9.9 D)—and 5 (13%) had pri- tively progressive eyes were from the same 42-year female
mary ectasia aged 13-19 years. patient, who had hypothyroidism, a factor that can influ-
The CDVA at baseline and 12 months postoperatively ence corneal biomechanics.18,19 In these 4 treatment fail-
were, respectively, 0.39 6 0.22 and 0.41 6 0.25 logMAR ure eyes, the minimum intraoperative pachymetry
(logarithm of the minimum angle of resolution). Spherical averaged 330 mm (range: 320-343 mm) and preoperative
power varied between 4.90 6 5.44 and 4.07 6 6.02 D, Kmax averaged 59.2 D (range: 50.3-64.7 D).
and cylindric power varied between 4.36 6 2.82 There was no correlation between the preoperative Kmax
and 4.00 6 4.18, respectively, at presentation and 1 and postoperative change in Kmax treatment (r ¼ 0.240,
year after. No significant changes were found in CDVA P ¼ .141). Also, preoperative Kmax values were not signif-
(P ¼ .611), sphere (P ¼ .077), and cylinder (P ¼ .915) icantly higher in patients with treatment failure vs patients
from baseline to 12 months postoperatively. No eyes without treatment failure (59.2 6 6.2 vs 63.3 6 9.6 D, P ¼
showed signs of endothelial decompensation. .914). There was no correlation between preoperative

136 AMERICAN JOURNAL OF OPHTHALMOLOGY APRIL 2021

FIGURE 2. Anterior segment optical coherence tomography of a demarcation line in an ultrathin cornea after 1 month of individ-
ualized corneal cross-linking using the sub400 protocol. Thickness measurements were made at the thinnest point of the cornea ac-
cording to pachymetry mapping and, as desired, show different demarcation line depths. Respectively in images A, B, and C, the
corneas present stromal thicknesses of 421, 353, and 216 mm, and distances between the demarcation line and the endothelium
of 84, 105, and 90 mm.

corneal thickness and the postoperative Kmax flattening tocol is based on a model taking into account the diffusion
(r ¼ 0.061, P ¼ .713). of oxygen and also correlations between CXL density and
experimentally determined amount of corneal stiffening,12

DENSITOMETRY: There was a significant increase from so that each cornea can receive an individual amount of to-
baseline at 12 months in AntDens (þ3.12 6 3.36 tal energy.
GSU, P < .05), CenDens (þ1.97 6 2.09 GSU, P < Currently, many corneas with advanced KC that would
.05), PostDens (þ0.90 6 1.47 GSU, P < .05), and Total- likely benefit from undergoing CXL cannot be treated
Dens (þ2.00 6 2.07 GSU, P < .05). Although a signifi- with the Dresden protocol because of a stromal thickness
cant increase in densitometry was observed as expected, of less than 400 mm. This initiated the development of
all patients remained within the typical clinical pattern modified techniques aiming to increase corneal stromal
of mild opacity after CXL. No patient had ‘‘deep stromal thickness artificially. However, such alternatives have lim-
haze’’20 or scarring in the evaluated period, as seen in itations that create variable outcomes and often reduce
Figure 5, showing representative OCT images of corneas efficacy.
with the highest degrees of flattening observed in our The first approach, first published in 2009 by Hafezi and
study. associates, was preoperative swelling of the cornea with
hypo-osmolaric riboflavin.9,21,22 The authors reported on
20 eyes treated with this technique. There were no cases
of endothelial cell loss, and keratectasia was stable at the
DISCUSSION 6-month postoperative follow-up.9 Another approach was
the CA-CXL proposed by Jacob and associates,10 where a
THE RESULTS OF THIS STUDY DEMONSTRATE THAT INDIVID- contact lens soaked in iso-osmolaric riboflavin is used to
ualized CXL with the sub400 protocol was able to success- ‘‘increase’’ the effective thickness of the cornea. A third
fully prevent KC progression of ultrathin keratoconic approach was a customized epithelial debridement
corneas in 90% of cases after 1 year of follow-up. This study approach called ‘‘epithelial island cross-linking’’ proposed
introduces for the first time the concept of individualizing by Mazzotta and colleagues.11 This approach spared epithe-
total energy during CXL, according to intraoperative lial cells around the thinnest point of the cornea; the
pachymetry. The individualized CXL with the sub400 pro- riboflavin-soaked island attenuates the UV-A energy. As

VOL. 224 INDIVIDUALIZED CXL: THE SUB400 PROTOCOL 137

FIGURE 3. Nonlinear relation between the ultraviolet (UV) irradiation time and predicted demarcation line. The measured data
points did follow the prediction, but presented a rather high standard deviation.

a potential consequence, the edge of the epithelial island linking reaction: besides modification of stromal thickness,
would refract the UV-A energy into the intermediate modification of riboflavin concentration would allow us to
stroma, potentially increasing the cross-linking effect in control the amount of chromophore present in the anterior
an undesired manner.11 layers of the stroma. The more riboflavin reacts with the
In essence, each of these approaches has its limitations, photons provided by the UV-A light in the anterior cornea,
because the modifications introduced to ‘‘increase’’ corneal the less energy is available in the deeper layers of the
thickness interfere with some of the fundamental key fac- stroma, making the CXL reaction shallower. In practice,
tors involved in the cross-linking reaction. The swelling such an approach would require a multitude of riboflavin
approach with hypo-osmolaric riboflavin leads to a stiff- solutions with different concentrations, which is not
ening effect similar to CXL using iso-osmolaric riboflavin feasible in daily clinical practice.
in a 400 mm cornea. However, the swelling effect of Another consideration is rather than modifying stromal
hypo-osmolaric riboflavin is variable: some corneas swell thickness or riboflavin concentration, the total irradiation
massively, whereas other corneas have little reaction.9,23 (fluence) could be adapted to the corneal thickness of the
This variability makes this swelling approach highly unpre- individual patient. This approach seems to be the most
dictable. The second approach is CACXL. The greatest logical because it would require just a single type of ribo-
stiffening effect of cross-linking is observed in the anterior flavin solution. Practically speaking, the treating surgeon
cornea.24 Biomechanical stress-strain measurements, ther- would reduce irradiation time to meet the fluence required
mal shrinkage tests,25 and Brillouin microscopy24 have in the individual cornea.
shown that CACXL results in a 30% reduction of the stiff- As logical as this ‘‘sub400’’ approach might seem, it was
ening effect compared with the epi-off Dresden protocol, impossible to implement it back in 2009 because too little
most probably due to a reduction in available oxygen.25,26 was known about the CXL reaction. Specifically, riboflavin
Custom-shaped small-incision lenticule extraction lenti- diffusion and kinetics was unknown at that time, and oxy-
cules have also been used in a similar capacity to the con- gen as a central element of the CXL process had not even
tact lens.27 Finally, the ‘‘epithelial island’’ approach been identified.26,31 The ‘‘sub400’’ protocol is based on a
exhibits an unequal DL between epithelialized and de- published algorithm that accounts for stromal riboflavin,
epithelialized areas,11 where areas of the intact epithelium oxygen, and UV availability during the cross-linking pro-
cause not only UV attenuation but also oxygen restriction cedure.26,31 It is based on estimating diffusion of riboflavin
and further biomechanical loss.28,29 In addition, it appears and oxygen by Fick’s law of diffusion and UV energy by the
that the cross-linking effect is shallower in areas under the Lambert-Beer law of light absorption. The presence of
epi-on region (150 mm) than in the epi-off regions these 3 factors determines the speed and amount of the
(250 mm).30 induced photochemical reactions (types I and II). It is
All current CXL techniques for thin corneas aim to in- assumed that the amount of singlet oxygen (Soxy) produced
crease stromal thickness. In theory, other measures should during the treatment interacts with the available EM and
also be considered like controlling the depth of the cross- thereby forms the relevant cross-links. The concentration

138 AMERICAN JOURNAL OF OPHTHALMOLOGY APRIL 2021

FIGURE 4. Demarcation line depth vs thinnest corneal thickness. The black continuous line is the trend line of the measured data
points, the blue continuous line indicates the location of the epithelium, and the blue dashed line the 70 mm distance margin.

of those additionally induced cross-links [CXL] can then be It is important to emphasize that the pachymetry consid-
estimated from. ered to calculate the irradiation dose of the sub400 protocol
8 9 is based on intraoperative pachymetry, after the riboflavin
> >

 < 
kRFH2ox
$½EMð1ekEMox Dt
=
Þ ; soaking. Despite this, we have also assessed the preopera-
½CXL ¼ ½CXL0  þ Soxy $ 1  e kEMox tive Scheimpflug data of all eyes. Whereas the cornea
>
: >
;
with the thinnest stromal thickness was 214 mm (intraoper-
atively measured with an ultrasound pachymeter, without
where [.] denotes concentration of riboflavin, Dt indicates
the epithelium), the thinnest cornea included showed a to-
the calculation time step, and kRFH2ox and kEMox are the re-
tal thickness of 325 mm (preoperatively measured with
action rate constants for RFH2 oxidation and EM oxida-
Scheimpflug tomography, with the epithelium). Therefore,
tion, respectively. [CXL0] represents the CXL
clinically, a cornea that presents with a Scheimpflug assess-
concentration at the previous time step. This model per-
ment of 325 mm total thickness (with the epithelium) or
mits the prediction of not only the amount of biomechan-
more can be treated with the sub400 protocol.
ical stiffening achieved after CXL but also the duration of
This study has some limitations. Although the vast ma-
UV irradiation required to achieve a CXL concentration
jority of eyes in this study had preoperative progression
that corresponds to the threshold of keratocyte apoptosis
documented by differential corneal imaging maps, a frac-
(when applied to clinical protocols) and as such to predict
tion (15%) of the eyes were from patients over 40 years
the penetration depth of a modified CXL treatment. The
and thus would be less likely to progress naturally. There-
accuracy of the theoretical model has been verified previ-
fore, one would think that they would not progress despite
ously in preclinical experiments, where the predicted
CXL: interestingly, all of these eyes had confirmed preoper-
CXL concentration did strongly correlate (R2 ¼ 0.95)
ative progression, and despite CXL 2 of these eyes showed
with the biomechanical stiffening in porcine, murine,
postoperative progression and were considered to be fail-
and lapin corneas.12 These experimental data suggest that
ures. In other cases, the extreme corneal shape in far-
sub400 can be used to individualize UV-A irradiation dura-
progressed ectasias made Placido-based topography and
tion with standard CXL lamp settings.
Scheimpflug imaging less reproducible. Also, primary ecta-
Based on this algorithm,12 the present study introduces a
sias in children or adolescents were primarily treated.13 A
standardized epithelium-off CXL method for the treatment
further limitation, not inherent specifically to this study,
of corneas with a stromal thickness of less than 400 mm:
are the metrics used to assess the CXL effects. Currently,
rather than artificially modifying corneal thickness, the
there is no clinical consensus on ideal metrics.32 Therefore,
sub400 protocol adapts the total UV-A energy to the
besides using Kmax, like the majority of studies, we have also
patient’s individual stromal thickness. The present study
evaluated the anterior radius of curvature in a 3.0 mm zone.
verified and confirmed that CXL with individualized
Using these 2 indices, we were able to demonstrate that
fluence was able to stop KC progression of corneas as thin
CXL was able to at least halt KC progression. Interestingly,
as 214 mm with 90% of success at 1 year after CXL.
the reduction found in the point of maximum keratometry

VOL. 224 INDIVIDUALIZED CXL: THE SUB400 PROTOCOL 139

FIGURE 5. Representative optical coherence tomography (OCT) images of corneas of 2 patients (A-C and B-F) with the highest
degrees of flattening observed in our study, before (A, D), at 1 month (B, E), and at 6 months (C, F) after corneal cross-linking
(CXL). None of the eyes treated in this study showed ‘‘deep stromal haze’’ or stromal scarring.

but not in a 3 mm zone could suggest improvement of The fluence of 5.4 J/cm2 at a stromal thickness of 400 mm
corneal regularity after CXL. A further limitation is that that was originally established in the Dresden protocol rep-
we were unable to evaluate endothelial cell density. How- resents the baseline fluence used in our ‘‘sub400’’ protocol
ever, no cornea showed clinical signs of endothelial decom- and is then reduced in thinner corneas following our
pensation: the total fluence in our study never exceeded the published algorithm. Interestingly, recent assessments us-
5.4 J/cm2 used in the original Dresden protocol, and other ing 2-photon imaging tomography indicate that there is a
published studies used fluences up to 14 J/cm2, without discrepancy of a factor of 1.7 between the concentration
observing endothelial damage.33,34 of corneal riboflavin using the new 2-photon imaging tech-
Furthermore, experimental evidence suggests that the nology and the old theoretical estimates.38 This discrep-
actual threshold of endothelial damage might be tradition- ancy might allow for substantially higher baseline
ally overestimated.34 It is important to note that, despite fluences in the near future.39
(1) the aforementioned current indirect evidence of endo- Another technique for the treatment of thin corneas,
thelial security, and (2) that we have not observed any called the ‘‘M’’ protocol, was recently proposed by Mazzotta
signs of clinical decompensation throughout the present and colleagues.40 This ‘‘M’’ protocol gathers all published
study, decompensation would be just an end stage sign of and validated clinical data on the penetration depth of
endothelial compromise; thence, in light of the lack of various epi-off and epi-on CXL techniques that had been
endothelial count, subtle endothelial changes could not published over the years. The ‘‘M’’ protocol matched the
be verified. Finally, another limitation was that DL depths in vivo scanning laser confocal microscopy and OCT
demonstrated considerable variability preventing the iden- morphological data5 with the mathematical assessment of
tification of a systematic deviation from the predictions the cross-link concentration threshold according to the
with the limited number of patients included in this study. measured DL, assuming the Dresden protocol as bench-
However, it seems likely that the algorithm somewhat mark. It demonstrates that the maximum interaction be-
underestimated the demarcation depth, which could be tween UV-A, riboflavin, oxygen, and collagen-
overcome in the future by applying higher irradiances, or proteoglycans complex would be in the first 200 mm—
prolonging the irradiation time. where the 70% of riboflavin-UV-A interactions occur,
The ocular structures are particularly sensitive to light- whereas the remaining 30% of CXL photo-oxidative reac-
induced damage.35 The main reason why the Dresden tion would be dissipated in the deep stroma between 200
protocol imposed a stromal thickness of more than and 300 mm.41
400 mm was to protect the corneal endothelium. So, The CXL techniques include using continuous or pulsed
aiming to protect sensitive structures such as the corneal light, with and without iontophoresis, and a range of
endothelium, rough estimates of riboflavin concentration different intensities, ranging from 3 to 30 mW/cm2. In
were calculated before the introduction of CXL in contrast, the ‘‘sub400’’ protocol introduced here uses 1 sin-
2003.34,36,37 From such estimates, the ‘‘400mm rule’’ was gle intensity in an epi-off setting, based on our published
created and globally disseminated as the minimal required algorithm. Although both the ‘‘M’’ and the ‘‘sub400’’ proto-
stromal thickness for epithelium-off CXL. Excess exposure col may achieve similar results, the ‘‘sub400’’ protocol re-
of the corneal endothelium to UV irradiation (above a quires less sophisticated technology.
threshold of 0.35 mW/cm2) would lead to cell death by The ‘‘sub400’’ individualized fluence CXL protocol stan-
apoptosis,38 putting cornea homeostasis and transparency dardizes the treatment in ultrathin corneas and is able to
at stake. halt KC progression with a success rate of 90% at

140 AMERICAN JOURNAL OF OPHTHALMOLOGY APRIL 2021

12 months, allowing treatment of corneas as thin as light CXL that fluences of 10 J/cm2 and even 15 J/cm2
214 mm of corneal stroma. This finding extends the clinical can achieve higher penetration of the cornea with a
range of cases that can be safely cross-linked to far reduced exposure time compared with the Dresden pro-
progressed KC stages. Furthermore, a significant correla- tocol.42,43 Even though the algorithm relies on many
tion was found between DL depth and irradiation but not constants that could be modified, we believe that future
between DL depth and change in Kmax. In other words, development will not be based on changing the curve of
the DL depth did not predict treatment outcome. Hence, irradiation calculations, but rather adjusting overall
the DL depth is not likely to relate to the extent of CXL- fluence, using higher values consistent with the latest
induced corneal stiffening but rather to induced wound published studies. Therefore, it is possible that extending
healing. In particular, it is also not a measure of how our algorithm to other CXL protocols could be used to
much ROS are created, given that there was a considerable perform more effective CXL in the future, rendering cor-
variation across patients of similar irradiation times. Still, neas stronger and more resistant to corneal ectasia
DL depth might be a clinically relevant parameter for retro- progression.
spective patient-specific assessment of the susceptibility to The introduction of CXL has changed the natural course
CXL-induced damage and its penetration. of corneal ectatic disease, and as a result, reduced the need
Finally, the principles behind the sub400 protocol for corneal transplants.44 Our new ‘‘sub400 protocol’’
apply to all corneas thickness and not just for corneas approach in ultrathin corneas will broaden the range of in-
with a thickness under 400 mm. The irradiance of 3 dications of CXL and further decrease the need for corneal
mW/cm2 was chosen to allow for every UV illumination transplantations. However, stabilizing extremely
device on the market—even the oldest—to be used with progressed corneal ectasias will only be beneficial if visual
this nomogram. The next step is to investigate how this rehabilitation can be achieved. The recent advances in
algorithm could be adapted to use baseline fluences sub- contact lens designs, particularly the rise of scleral contact
stantially higher than 5.4 J/cm2; it has been previously lenses, allow for a satisfactory CDVA even in stromal
shown in topography-guided high-irradiance pulsed- thicknesses of less than 250 mm.

ALL AUTHORS HAVE COMPLETED AND SUBMITTED THE ICMJE FORM FOR DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST.
Funding/Support: This study was supported in part by separate unrestricted departmental grants from Research to Prevent Blindness, Inc to the USC Roski
Eye Institute, Los Angeles, CA (F.H.) and the Cleveland Clinic Cole Eye Institute (J.B.R.), grants from the Light for Sight Foundation, Zurich, Switzerland
(F.H.), the Velux Foundation, Zurich, Switzerland (F.H.) and grant from International Council of Ophthalmology, San Francisco, CA (E.T.N.). Financial
Disclosures: F.H. is the chief scientific and medical officer of EMAGine AG (Zug, Switzerland) and co-inventor of the PCT applications CH2012/0000090
and PCT2014/CH000075 regarding CXL technology. N.H. is CEO of EMAGine AG, a company producing a CXL device. The remaining authors have no
financial or proprietary interest in the materials presented herein. All authors attest that they meet the current ICMJE criteria for authorship.

REFERENCES 8. Zhang Y, Conrad AH, Conrad GW. Effects of ultraviolet-A

and riboflavin on the interaction of collagen and proteogly-
1. Kennedy RH, Bourne WM, Dyer JA. A 48-year clinical and cans during corneal cross-linking. J Biol Chem 2011;286:
epidemiologic study of keratoconus. Am J Ophthalmol 1986; 13011–13022.
101:267–273. 9. Hafezi F, Mrochen M, Iseli HP, Seiler T. Collagen crosslink-
2. Godefrooij DA, de Wit GA, Uiterwaal CS, Imhof SM, ing with ultraviolet-A and hypoosmolar riboflavin solution in
Wisse RPL. Age-specific incidence and prevalence of kerato- thin corneas. J Cataract Refract Surg 2009;35:621–624.
conus: a nationwide registration study. Am J Ophthalmol 2017; 10. Jacob S, Kumar DA, Agarwal A, Basu S, Sinha P, Agarwal A.
175:169–172. Contact lens-assisted collagen cross-linking (CACXL): a new
3. Torres Netto EA, Al-Otaibi WM, Hafezi NL, et al. Preva- technique for cross-linking thin corneas. J Refract Surg 2014;
lence of keratoconus in paediatric patients in Riyadh, Saudi 30:366–372.
Arabia. Br J Ophthalmol 2018;102:1436–1441. 11. Mazzotta C, Ramovecchi V. Customized epithelial debride-
4. Raiskup F, Theuring A, Pillunat LE, Spoerl E. Corneal ment for thin ectatic corneas undergoing corneal cross-
collagen crosslinking with riboflavin and ultraviolet-A light linking: epithelial island cross-linking technique. Clin
in progressive keratoconus: ten-year results. J Cataract Refract Ophthalmol 2014;8:1337–1343.
Surg 2015;41:41–46. 12. Kling S, Hafezi F. An algorithm to predict the biomechanical
5. Wollensak G, Spoerl E, Seiler T. Riboflavin/ultraviolet-a- stiffening effect in corneal cross-linking. J Refract Surg 2017;
induced collagen crosslinking for the treatment of keratoco- 33:128–136.
nus. Am J Ophthalmol 2003;135:620–627. 13. Chatzis N, Hafezi F. Progression of keratoconus and efficacy
6. Randleman JB, Khandelwal SS, Hafezi F. Corneal cross-link- of pediatric [corrected] corneal collagen cross-linking in chil-
ing. Surv Ophthalmol 2015;60:509–523. dren and adolescents. J Refract Surg 2012;28:753–758.
7. Spoerl E, Huhle M, Seiler T. Induction of cross-links in 14. Ferdi AC, Nguyen V, Gore DM, Allan BD, Rozema JJ,
corneal tissue. Exp Eye Res 1998;66:97–103. Watson SL. Keratoconus natural progression: a systematic

VOL. 224 INDIVIDUALIZED CXL: THE SUB400 PROTOCOL 141

 review and meta-analysis of 11 529 eyes. Ophthalmology 2019; nique in thin corneas: a confocal microscopy study. J Refract
126:935–945. Surg 2011;27:444–450.
15. Hafezi F. Smoking and corneal biomechanics. Ophthalmology 31. Richoz O, Hammer A, Tabibian D, Gatzioufas Z, Hafezi F.
2009;116:2259.e1. The biomechanical effect of corneal collagen cross-linking
16. Madhukumar E, Vijayammal PL. Influence of cigarette smoke (CXL) with riboflavin and UV-A is oxygen dependent. Transl
on cross-linking of dermal collagen. Indian J Exp Biol 1997;35: Vis Sci Technol 2013;2:6.
483–486. 32. Lang PZ, Hafezi NL, Khandelwal SS, Torres-Netto EA,
17. Spoerl E, Raiskup-Wolf F, Kuhlisch E, Pillunat LE. Cigarette Hafezi F, Randleman JB. Comparative functional outcomes
smoking is negatively associated with keratoconus. J Refract after corneal crosslinking using standard, accelerated, and
Surg 2008;24:S737–S740. accelerated with higher total fluence protocols. Cornea
18. Lee R, Hafezi F, Randleman JB. Bilateral keratoconus 2019;38:433–441.
induced by secondary hypothyroidism after radioactive iodine 33. Kanellopoulos AJ, Dupps WJ, Seven I, Asimellis G. Toric
therapy. J Refract Surg 2018;34:351–353. topographically customized transepithelial, pulsed, very
19. Tabibian D, de Tejada BM, Gatzioufas Z, et al. Pregnancy- high-fluence, higher energy and higher riboflavin concentra-
induced changes in corneal biomechanics and topography tion collagen cross-linking in keratoconus. Case Report
are thyroid hormone related. Am J Ophthalmol 2017;184: Ophthalmol 2014;5:172–180.
129–136. 34. Seiler TG, Batista A, Frueh BE, Koenig K. Riboflavin Con-
20. Hafezi F, Koller T, Vinciguerra P, Seiler T. Marked remodel- centrations at the endothelium during corneal cross-linking
ling of the anterior corneal surface following collagen cross- in humans. Invest Ophthalmol Vis Sci 2019;60:2140–2145.
linking with riboflavin and UVA. Br J Ophthalmol 2011;95: 35. Marti A, Hafezi F, Lansel N, et al. Light-induced cell death of
1171–1172. retinal photoreceptors in the absence of p53. Invest Ophthal-
21. Hayes S, Boote C, Kamma-Lorger CS, et al. Riboflavin/UVA mol Vis Sci 1998;39:846–849.
collagen cross-linking-induced changes in normal and kera- 36. Wollensak G, Spoerl E, Wilsch M, Seiler T. Endothelial cell
toconus corneal stroma. PLoS One 2011;6:e22405. damage after riboflavin-ultraviolet-A treatment in the rabbit.
22. Richoz O, Mavrakanas N, Pajic B, Hafezi F. Corneal collagen J Cataract Refract Surg 2003;29:1786–1790.
cross-linking for ectasia after LASIK and photorefractive ker- 37. Wollensak G, Sporl E, Reber F, Pillunat L, Funk R. Corneal
atectomy: long-term results. Ophthalmology 2013;120: endothelial cytotoxicity of riboflavin/UVA treatment
1354–1359. in vitro. Ophthalmic Res 2003;35:324–328.
23. Wollensak G, Sporl E. Biomechanical efficacy of corneal 38. Spoerl E, Hoyer A, Pillunat LE, Raiskup F. Corneal cross-
cross-linking using hypoosmolar riboflavin solution. Eur J linking and safety issues. Open Ophthalmol J 2011;5:14–16.
Ophthalmol 2019;29:474–481. 39. Seiler TG, Fischinger I, Koller T, Zapp D, Frueh BE, Seiler T.
24. Zhang H, Roozbahani M, Piccinini AL, et al. Depth-depen- Customized corneal cross-linking: one-year results. Am J
dent reduction of biomechanical efficacy of contact lens– Ophthalmol 2016;166:14–21.
assisted corneal cross-linking analyzed by Brillouin micro- 40. Mazzotta C, Riomani A, Burroni A. Pachymetry-based accel-
scopy. J Refract Surg 2019;35:721–728. erated cross-linking: the ‘‘M nomogram’’ for standardized
25. Wollensak G, Sporl E, Herbst H. Biomechanical efficacy of treatment of all-thickness progressive ectatic corneas. Int K
contact lens-assisted collagen cross-linking in porcine eyes. Kerat Ect Corn Dis 2019;7:137–144.
Acta Ophthalmol 2019;97:e84–e90. 41. Kohlhaas M, Spoerl E, Schilde T, Unger G, Wittig C,
26. Kling S, Richoz O, Hammer A, et al. Increased biomechan- Pillunat LE. Biomechanical evidence of the distribution of
ical efficacy of corneal cross-linking in thin corneas due to cross-links in corneastreated with riboflavin and ultraviolet
higher oxygen availability. J Refract Surg 2015;31:840–846. A light. J Cataract Refract Surg 2006;32:279–283.
27. Sachdev MS, Gupta D, Sachdev G, Sachdev R. Tailored stro- 42. Mazzotta C, Hafezi F, Kymionis G, et al. In vivo confocal mi-
mal expansion with a refractive lenticule for crosslinking the croscopy after corneal collagen crosslinking. Ocul Surf 2015;
ultrathin cornea. J Cataract Refract Surg 2015;41:918–923. 13:298–314.
28. Deshmukh R, Hafezi F, Kymionis GD, et al. Current concepts 43. Mazzotta C, Paradiso AL, Baiocchi S, Caragiuli S,
in crosslinking thin corneas. Indian J Ophthalmol 2019;67: Caporossi A. Qualitative investigation of corneal changes af-
8–15. ter accelerated corneal collagen cross-linking (A-CXL) by
29. Torres-Netto EA, Kling S, Hafezi N, Vinciguerra P, in vivo confocal microscopy and corneal OCT. Clin Experi-
Randleman JB, Hafezi F. Oxygen diffusion may limit the ment Ophthalmol 2013;4:1–6.
biomechanical effectiveness of iontophoresis-assisted transepi- 44. Godefrooij DA, Gans R, Imhof SM, Wisse RPL. Nationwide
thelial corneal cross-linking. J Refract Surg 2018;34:768–774. reduction in the number of corneal transplantations for kera-
30. Kaya V, Utine CA, Yilmaz OF. Efficacy of corneal collagen toconus following the implementation of cross-linking. Acta
cross-linking using a custom epithelial debridement tech- Ophthalmol (Copenh) 2016;94:675–678.

142 AMERICAN JOURNAL OF OPHTHALMOLOGY APRIL 2021