Confocal Microscopy Analysis of Corneal Changes AfterPhotorefractive Keratectomy Plus Cross-Linking for

Keratoconus: 4-Year Follow-up

GIOVANNI ALESSIO, MILENA L’ABBATE, CLAUDIO FURINO, CARLO SBORGIA, ANDMARIA GABRIELLA LA TEGOLA

� PURPOSE: To analyze corneal confocal microscopychanges after combined photorefractive keratectomy(PRK) plus the cross linking (CXL) procedure.� DESIGN: Prospective interventional case series.� METHODS: At the Department of Basic Medical Sci-ences, Neuroscience, and Sense Organs of the Universityof Bari, Bari, Italy, 17 eyes of 17 patients with progres-sive keratoconus underwent confocal microscopy exami-nation before and after 1, 3, 6, 12, 18, and 48 monthsfollowing PRK plus the CXL procedure. The mainoutcome measures were mean superficial epithelial celldensity; mean basal epithelial cell density; mean anterior,mid and posterior keratocyte density; qualitative analysisof stromal backscatter; sub-basal and stromal nerve den-sity parameters; and mean endothelial cell density.� RESULTS: During the 4-year follow-up, themean superfi-cial epithelial cell density, mean basal epithelial cell densityand mean endothelial cell density remained unchanged(P > 0.05). The anterior mid-stromal keratocyte densityshowed a significant decrease (P < 0.05) as comparedwith preoperative values, and the posterior stromal kerato-cyte density showed a significant increase at 1 and 3 monthsof follow-up.Sub-basal and stromalnervedensityparameterswere significantly decreased until postoperative month 6(P<0.05at1,3, and6months) and then tended to increaseup to preoperative values by the 18th postoperative month.� CONCLUSION: Corneal changes after the PRK plusCXL procedures seem to be pronounced and long lastingas far as keratocyte density of the anterior and mid stromais concerned. Sub-basal nerve densities tend to reachpreoperative values 6 months after surgery. (Am JOphthalmol 2014;158:476–484. � 2014 by ElsevierInc. All rights reserved.)

CORNEAL CROSS-LINKING (CXL) WITH RIBOFLAVIN

and ultraviolet-A (UVA) is currently the predom-inant treatment for mild and moderate progressive

keratoconus. It is generally preferred to keratoplastic

Accepted for publication May 21, 2014.From the Clinica Oculistica, Department of Basic Medical Sciences,

Neuroscience, and Sense Organs, University of Bari, Bari, Italy.Inquiries to Maria Gabriella La Tegola, Department of Basic Medical

Sciences, Neuroscience and Sense Organs, University of Bari, PiazzaGiulio Cesare, 11, Bari, Italy; e-mail: mariagabriella.lategola@uniba.it

476 � 2014 BY ELSEVIER INC.

surgeries because it is noninvasive and requires a shorter vi-sual rehabilitation time and, in particular, because there isno risk for graft failure. However, there are instances inwhich keratoplasty is preferred to CXL for the treatmentof patients with severe keratoconus with corneal scarring.Since CXL was introduced, the technique has been

enhanced by combining certain elements of photorefrac-tive keratectomy (PRK) to improve the refractive out-comes. Instead of mechanical removal of the epithelium,according to the technique described by Wollensak andassociates,1 ablation of the epithelium and superficialstroma is used in the PRK plus CXL procedure to remodelthe corneal surface and reduce irregular astigmatism. In ourexperience, as compared with standard CXL, the PRK plusCXL procedure offers better visual outcomes in terms oflower keratometric values, changes in the elevation ofthe anterior corneal surface and greater reductions in rootmean square values.2 Some studies have reported changesin corneal structure investigated by in vivo confocal micro-scopy after CXL alone,3–6 but only a few studies have beenmade of the outcomes of the PRK plus CXL procedure,7 anda follow-up of 4 years has never been reached previously.The aim of this study was to analyze morphologic

changes in all corneal layers, as examined by confocalmicroscopy, over a 4-year follow-up period after PRK plusCXL for keratoconus.

METHODS

� SUBJECTS: The study protocol was approved and moni-tored by the Local Ethics Committee of Azienda Ospeda-liera Policlinico di Bari (protocol no. 839/C.E.) andconformed to the tenets of the Declaration of Helsinki.Informed consent was obtained from all patients takingpart in the study. This is a registered clinical trial: clinicaltrial registration at http://www.controlled-trials.com/ISRCTN57262986.Between July 2008 and June 2009, 17 patients with kera-

toconus (34 eyes) (mean age, 31.17 6 8.12 years; range,21–46 years) were recruited at the Cornea Service of theDepartment of Basic Medical Sciences, Neuroscience,and Sense Organs of the University of Bari to take partin this prospective nonrandomized clinical study. The 34

0002-9394/$36.00http://dx.doi.org/10.1016/j.ajo.2014.05.024

ALL RIGHTS RESERVED.

eyes were graded as stage I-III according to the Alio andShabayek classification8 and were assigned to 1 of 2 groups:the worse eye underwent PRK followed by CXL (17 eyes),and the better eye (fellow eye) underwent CXL alone (17eyes). The diagnosis of keratoconus was based on cornealtopography (OrbscanIIz; Bausch & Lomb, Rochester,New York, USA) as an asymmetric bowtie pattern,9 withor without skewed axes, and a paracentral inferior-superior dioptric difference of more than 1.4 diopters (D).10

Enrolment depended on documented progression of kera-toconus in the previous 6 months, a corneal thickness of atleast 450mmat the thinnest point in theworse eye, hard con-tact lens and full spectacle correction intolerance, and ageolder than 18 years. Progression of keratoconus was definedbased on serial differential topography, as an increase inthe apex keratometry by more than 1.0–1.5 D and a corre-sponding change (>1.0–1.5 D) in the refractive cylinderin the previous 6 months. Exclusion criteria were a cornealthickness of less than 450 mm at the thinnest point in theworse eye, corneal scarring orVogt striae, any ocular disorderother than keratoconus, any history of eye surgery, any sys-temic disease, pregnancy, and contact lens wear. Patientswho failed to attend the follow-up visits were excluded.

Visual acuity, refractive, topographic, and corneal higherorder aberrations outcomes after PRK plus CXL and afterCXL alone up to 24 months of follow-up have alreadybeen reported.2 The current study investigated the confocalmicroscopy findings in 17 corneas of 17 patients after thePRK plus CXL procedure over a 4-year follow-up period.In this trial the eyes treated only by CXL were not takeninto consideration because there are already many publica-tions concerning confocal microscopy analyses in theseeyes, reporting similar results in all cases.3–6

� TOPOGRAPHIC-GUIDED PRK PLUS CXL PROCEDURE:

The steps of topographic-guided PRK plus CXL are re-ported briefly below. After acquisition of the corneal shapeby means of the Scheimpflug-based Precisio tomographer(Ligi Tecnologie Medicali, Taranto, Italy), elevation datatogether with the patient’s subjective refraction were im-ported into Corneal Interactive Programmed TopographicAblation (CIPTA) software to plan the ablation proce-dure. To ensure the maximum reduction of irregularitieswith minimum tissue consumption, the Restored Morpho-logical Axis strategy of the CIPTA software was used. Thecenter of the planned ablation was the corneal apex in alltreated eyes. The stromal ablation depth was between 18and 49mm (mean, 31.16 9.5mm). A supplementary depthof 50 mm was selected by the surgeon for epithelium abla-tion because of transepithelial PRK. The corneal epithe-lium was removed by laser within a 9 mm diameter.Transepithelial topography-guided PRK was performedwith the high-resolution 1-KHz flying spot laser iRES(Ligi Tecnologie Medicali). Immediately after PRK, ribo-flavin 0.1% solution (10 mg riboflavin-5-phosphate in20% dextran T500 solution; Ricrolin; SOOFT Italia,

VOL. 158, NO. 3 CONFOCAL MICROSCOPY AFTER EXC

Montegiorgio, Italy) was administered topically every 2minutes for 30 minutes, under topical anesthesia. Ribo-flavin absorption throughout the corneal stroma and ante-rior chamber was confirmed on slit-lamp examination.Then the cornea was exposed to UVA 365 nm light atan irradiance of 3.0 mW/cm2 for 30 minutes. DuringUVA exposure, riboflavin drops were continued every2 minutes. At the end of the procedure, a soft bandage con-tact lens was applied and ofloxacin and 0.1% indomethacindrops were administered until re-epithelialization was com-plete. After epithelial healing and removal of the contactlens, 0.1% fluorometholone was instilled 4 times daily forthe first month. The dosage was tapered by 1 drop monthlyover the next 3 months.

� CONFOCAL MICROSCOPY AND OUTCOMES: The Confo-scan 4 (NIDEK Technologies, Padova, Italy) confocalmicroscope equipped with a 403 objective lens was usedin all eyes. After the instillation of local anesthetic(benoxinate hydrochloride 0.4 g eye drops, Alfa IntesIndustria Terapeutica Splendore, Casoria, Italy), theinstrument lens was advanced until the high-viscosity gel(hydroxypropyl methylcellulose 2.5%) contacted thecornea. A fixed device setting was used for all examina-tions: full-thickness mode, 72% light intensity; 7 mmscan step; autoalignment function. The Z-ring was usedin all cases. Each image represented a coronal section of460 3 345 mm with magnification of 3500. A total of 4scans were obtained at the optical center of each corneafor all patients at each follow-up visit. The confocal micro-scopy examination was performed by the same researcher inall study participants. The best-focused 3 frames per layerwere selected for analysis of each cornea. Evaluationswere repeated at 1, 3, 6, 12, 18, and 48 months postopera-tively. Image parameters were:

(1) Superficial epithelial cell density (cells/mm2); a mini-mum of 35 cells were counted manually by Confoscan4NAVIS analysis software in a randomly selected rect-angular area. The software highlights the results asunreliable if fewer than 35 cells are present in theselected rectangle.

(2) Basal epithelial cell density (cells/mm2). At least 75cells were counted manually by Confoscan 4 NAVISanalysis software in a randomly selected polygonalarea.

(3) Keratocyte density (cells/mm2) in the anterior, mid,and posterior thirds of stroma. The stroma was dividedinto 3 equal-thickness anteroposterior regions, and 3frames were selected from each of these 3 regions,expressed as percent stromal depth (1%–33%, 34%–66%, and 67%–100%). Images of the anterior bound-ary of the stroma immediately posterior to the Bowmanlayer (the layer with the highest concentration of kera-tocyte nuclei9) were included in the analysis of eacheye at each follow-up time. At least 35 keratocyte

477IMER LASER AND CROSS-LINKING

TABLE 1. Baseline and 4-Year Postoperative Clinical Datafor Keratoconic Corneas that Underwent Photorefractive

Keratectomy Plus the Cross-Linking Procedure

Baseline Data 4-year Postoperative Data

CCT, mm 516 6 26.05 486.8 6 28.3

CTT, mm 489.2 6 23.6 451.8 6 29.7

SimK1, D 44.83 6 11.67 42.98 6 3.01

SimK2, D 47.41 6 12.44 44.86 6 4.23

SimKavg, D 46.12 6 12.04 43.98 6 3.34

Kapex, D 54.46 6 14.3 49.87 6 4.67

CCT¼ central corneal thickness; CTT¼ corneal thinnest thick-

ness; D ¼ diopters; Kapex ¼ apex keratometry; SimKavg ¼simulated keratometry average; SimK1 ¼ flattest simulated

meridian keratometry; SimK2 ¼ steepest simulated meridian

keratometry.

Notes:Mean6 SD. CCT and CTT values were obtained by the

AS-OCT Visante system.

SimK1, SimK2, SimKavg, and Kapex values were obtained by

CSO Eye Top tomographer (Costruzione Strumenti Oftalmici,

Firenze, Italy).

nuclei were counted manually, marking each cell ornucleus inside a randomly selected rectangular area.Cells that touched the top or left edge of the rectanglewere counted, but cells that touched the bottom orright edge were not.

(4) Sub-basal nerve density, defined as the mean totallength (mm) within the region of interest (by default,81,972.17 mm2), mean total length of the nerves perimage (mm/mm2), and mean total number of nerveswithin the region of interest (including main nervetrunks and branches). The semiautomated NervesTracking Tool of NAVIS software was used in all cases.

(5) Stromal nerve density, defined as the total number ofnerves.

(6) Endothelial cell density (cells/mm2). In a randomlyselected polygonal area, at least 75 cells were countedby Confoscan 4 NAVIS software.

Qualitative analysis of corneal haze was also performed.Evaluation of stromal backscatter served to identify thecorneal layer associated with pathologic scatter and forfollow-up of corneal haze.

All counts were performed independently, by 2 experi-enced examiners blinded to the study identifiers. Compar-isons between preoperative and postoperative values ateach time point were performed for all parameters bypaired-samples t test. A P value of less than 0.05 wasconsidered statistically significant.

RESULTS

THE CLINICAL FEATURES OF THE SUBJECTS ARE SHOWN IN

Table 1.Themean values and P values for comparisons of the pre-

operative and the postoperative results are shown inTable 2.

Compared with baseline values, mean superficial epithe-lial density, mean basal epithelial cell density, and meanendothelial cell density were not significantly different ateach postoperative time point. Anterior stromal keratocytenuclei were present in no eyes (n¼ 0) at 1, 3 and 6 postop-erative months; in 23.5% of eyes (n ¼ 4) at 12 months; in35.2% of eyes (n ¼ 6) at 18 months; and in 41.1% of eyes(n ¼ 7) at 48 months. Mid stromal keratocyte nuclei werepresent in no eyes (n ¼ 0) at 1, 3 and 6 postoperativemonths; in 23.5% of eyes (n ¼ 4) at 12 months; and in35.2% of eyes (n ¼ 6) at 18 and 48 months. The meananterior and mid stromal keratocyte densities were signifi-cantly lower at each follow-up as compared with the preop-erative values and in no case reached the preoperativevalue. Compared with baseline values, mean posteriorkeratocyte density at 1 and 3 months was significantlyincreased, while at 6, 12, 18, and 48 months it was not sta-tistically different. Postoperatively, sub-basal nerves were

478 AMERICAN JOURNAL OF

present in no (n ¼ 0) eyes at 1 and 3 months; in 82.3% ofeyes (n ¼ 14) at 6 months; in 94.1% of eyes (n ¼ 16) at12 months; and in 100% of eyes (n ¼ 17) at 18 and48 months. Postoperatively, stromal nerves were presentin no eyes (n ¼ 0) at 1 and 3 months; in 88.2% of eyes(n ¼ 15) at 6 months; and in 100% of eyes (n ¼ 17) at12, 18 and 48 months. The mean total length of nerves,mean total length of nerves per image, mean total numberof sub-basal nerves, and mean total number of stromal nerveswere drastically lower and undetectable at 1 and 3 monthspostoperatively and then tended to increase up to 18monthspostoperatively. By 4-year follow-up, the preoperative valuesfor mean total length of nerves, mean total length of nervesper image, mean total number of sub-basal nerves, and meantotal number of stromal nerves were restored.Confocal images of corneal stroma at different time

points in a 32-year-old patient are illustrated in Figure 1and Figure 2. Confocal microscopy findings in keratoconiceyes before treatment are not reported because they havealready been described11 and are very similar to ours.

DISCUSSION

THIS STUDY INVESTIGATED CONFOCAL MICROSCOPY

changes in corneal layers after the PRK plus CXL procedureup to 4 years after surgery. The most noteworthy structuralchanges concerned the anterior andmid stromal keratocytedensity and the sub-basal and stromal nerve density param-eters. The mean anterior and mean mid keratocyte den-sities were markedly decreased in the early postoperativemonths, then showed a slow, gradual increase until 4 years

SEPTEMBER 2014OPHTHALMOLOGY

TABLE 2. In Vivo Confocal Microscope Values and P Values for Comparison of Preoperative and Postoperative Results at Each Time Point in Eyes Treated by PhotorefractiveKeratectomy Plus Cross-Linking for Progressive Keratoconusa

Preoperatively

Postoperatively

1 mo. 3 mo 6 mo 12 mo 18 mo 48 mos.

Superficial epithelial cell

density (cells/mm2)

1481 6 597.5

(636–2561)

1414.4 6 448.6

(831–2127)

1414 6 484.4

(841–2110)

1294.3 6 465.9

(820–2009)

1385.6 6 423

(879–2078)

1234.8 6 371.4

(724–2069)

1308.7 6 414.3

(897 to 2523)

P value 0.74 0.74 0.21 0.48 0.15 0.28

Basal epithelial cell density

(cells/mm2)

5759.5 6 1785.3

(2891–8743)

5459.3 6 1300.3

(3256–7958)

5412.7 6 1324.4

(3674–7700)

5653.4 6 2057.5

(1775–8835)

6150.1 6 1513.2

(3984–8041)

5735.7 6 1467.8

(3827–8020)

6020.7 6 1094.5

(4034 to 7839)

P value 0.60 0.51 0.87 0.58 0.96 0.65

Anterior stromal keratocyte

density (cells/mm2)

880.9 6 199.7

(643–1345)

NA NA NA 138.3 6 247.4

(0 to 518)

210.3 6 262.3

(0–768)

253 6 295.3

(0 to 798)

P value 1.98 10�9 1.98 10�9 1.98 10�9 3.88 10�7 2.81 10�6 8.08 10�5

Mid stromal keratocyte

density (cells/mm2)

618.3 6 90.7

(534–710)

NA NA NA 68.54 6 127.5

(0–305)

99.2 6 139.2

(0–315)

102.2 6 143.5

(80 to 308)

P value 4.78 10�15 4.78 10�15 4.78 10�15 8.53 10�11 3.18 10�9 3.54 10�9

Posterior stromal keratocyte

density (cells/mm2)

736.2 6 97.6

(516–909)

831.3 6 86.3

(635–969)

830.2 6 88.7

(642–956)

782.8 6 139.7

(590–1006)

695.9 6 73.8

(603–832)

740.6 6 157.5

(547–1080)

647.2 6 167.2

(339 to 920)

P value 0.04 0.02 0.35 0.20 0.93 0.06

Endothelial cell density

(cells/mm2)

2945 6 115

(2731–3012)

2932.2 6 269.7

(2171–3365)

2965.8 6 216.9

(2180–3152)

2947.7 6 105

(2734–3105)

2978.2 6 65.3

(2856–3098)

2984 6 67.3

(2828–3084)

2981.3 6 60.1

(2876 to 3087)

P value 0.84 0.63 0.93 0.23 0.12 0.11

Total length of sub-basal

nerves (mm)

588 6 97

(419.6–724.8)

NA NA 300.6 6 161

(280.7–529.8)

397.9 6 145.7

(332.9–600.6)

487.7 6 119.8

(187.9–671.8)

554.6 6 137.5

(237.6 to 712.7)

P value 3.03 10�14 3.03 10�14 1.22 10�5 7.01 10�3 0.01 0.35

Length density of sub-basal

nerves (mm/mm2)

9610.6 6 1325.9

(7865–11765)

NA NA 5245.2 6 931.8

(3829–6463)

5983.4 6 1327.9

(3127–7489)

8085.4 6 2875.9

(2293–11232)

9921.9 6 1173.6

(7865 to 11977)

P value 1.82 10�15 1.82 10�15 4.46 10�8 6.66 10�7 0.05 0.51

Total number of sub-basal

nerves (no.)

6 6 2.5 (2–10) NA NA 4.5 6 1.8 (2–9) 4.8 6 1.3 (3–7) 6 6 1.8 (3–8) 6.2 6 1.9 (3 to 10)

P value 4.75 10�8 4.75 10�8 0.1 0.16 1 0.68

Total number of stromal

nerves (no.)

2.1 6 1 (1–4) NA NA 1.9 6 1.1 (0–4) 2.2 6 1.1 (1–4) 2.4 6 1 (1–4) 2.3 6 1 (1 to 4)

P value 1.47 10�7 1.47 10�7 0.58 0.75 0.48 0.66

NA ¼ not accounted.aMean 6 standard deviation (range).

VOL.

158,N

O.3

479

CONFO

CALM

ICROSC

OPYAFT

EREXCIM

ERLASER

ANDCROSS-L

INKIN

G

FIGURE 1. Corneal confocal images 1, 3 and 6 months after the photorefractive keratectomy plus cross-linking procedure in a 32-year-old patient with keratoconus. (Top row) anterior stromal (left), mid stromal (middle left) boundary between treated and un-treated stromal (middle right) and posterior stromal (right) layers 1 month after surgery. (Intermediate row) anterior stromal(left), mid stromal (middle left) boundary between treated and untreated stromal (middle right) and posterior stromal (right) layers3 months after surgery. (Bottom row) anterior stromal (left), mid stromal (middle left) boundary between treated and untreated stro-mal (middle right) and posterior stromal (right) layers 6 months after surgery. One month postoperatively (top row), anterior stromashowed dishomogeneous hyper-reflectivity without recognizable cells and nerve fibers; mid stroma was edematous and hyporeflectivewith rare activated keratocytes; the boundary of the cross-linked area was outlined by a reticular hyper-reflective pattern with elon-gated keratocytes and hyper-reflective needle-like structures; deeper, posterior stroma showed keratocyte nuclei. Three months post-operatively (middle row), anterior stroma showed a honeycomb appearance, and mid stroma hyporeflectivity without detectablecellular nuclei; the boundary of the treated area was underlined by hyper-reflective bands, while the deeper stroma was unchanged.Six months postoperatively (bottom row), structural changes are similar to those in the third month, but tiny nerve fibers exhibiting afragmented appearance are recognizable in the anterior-mid stroma.

postoperatively but never regained preoperative values.Sub-basal and stromal nerve densities were not evaluableat the first and third postoperative month examinationsbut gradually returned to the preoperative values bythe fourth year postoperatively. By the eighteenth post-operative month, preoperative values for nerve densitywere restored, and no later significant changes wereobserved.

Computer-assisted manual measurement of cell andnerve density on confocal microscope images is a time-consuming and subjective method, with obvious concernsabout intra- and interexaminer repeatability. Not only isthe evaluation of each image parameter affected by the skillof the operators, but the analysis of each corneal layer raisesadditional issues.With regard to stromal keratocyte density,cell nuclei with higher contrast and sharper edges are moreeasily identifiable and countable than those with lowercontrast and blurred edges.11,12 In 2010, McLaren and

480 AMERICAN JOURNAL OF

associates13 developed a program to detect automaticallystromal cell nuclei in Confoscan 4 images, but no one hasyet used this consistently in clinical studies. Moreover,the confocalmicroscope is unable to distinguish keratocytesfrom bonemarrow-derived cells and other cells that are pre-sent in the corneal stroma, and these cell types will affectestimates of keratocyte density. With regard to endothelialcell density, in this study, semiautomated image analysiswas performed. In this way, the operator verifies the cellborder outline and can make manual corrections. A greateraccuracy of endothelial cell density analysis has been re-ported using the semiautomated method.14 Moreover,methods of objective measurement of the light backscatterby corneal layers have also been developed, but none ofthese has been used extensively in clinical studies.15,16

Since 2010, solutions to the problem of standardizationof the measurements of brightness of light backscatteredby the cornea have been proposed so as to ensure

SEPTEMBER 2014OPHTHALMOLOGY

FIGURE 2. Corneal confocal microscope images 12, 18 and 48 months after the photorefractive keratectomy plus the cross-linkingprocedure in a 32-year-old patient with keratoconus. (Top row) anterior stromal (left), mid stromal (middle left) boundary betweentreated and untreated stromal (middle right) and posterior stromal (right) layers 12 months after surgery. (Intermediate row) anteriorstromal (left), mid stromal (middle left) boundary between treated and untreated stromal (middle right) and posterior stromal (right)layers 18 months after surgery. (Bottom row) anterior stromal (left), mid stromal (middle left) boundary between treated and un-treated stromal (middle right) and posterior stromal (right) layers 48 months after surgery. Twelve months postoperatively (toprow), the anterior and mid stroma showed a dishomogeneous hyper-reflectivity with recognizable nerve structures (trunks andbranches); the boundary of the treated area can be identified by the presence of hyper-reflective bands that retained cell nuclei;the deep stroma was normoreflective and populated by cells. Eighteen months postoperatively, too (middle row), anterior and midstroma were unformed and widely acellular, while the sub-basal nerve plexus was restored to the preoperative state; at the boundaryof the treated area hyper-reflective band-like structures remained; the deep stroma was unchanged. Four years postoperatively, thesub-basal nerve plexus formed an interconnected network; anterior and mid stroma remained largely acellular and dishomogeneouslyhyper-reflective; hyper-reflective bandlike structures with keratocytes through mesh were evident between the cross-linked and un-treated areas; deep stroma was unchanged.

comparable results at different times and in differentlaboratories.16,17 In this study, however, only a qualitativeassessment of the backscattered light was made to detectthe possible corneal layers involved. For the nerve-densityparameters, we used the semiautomated Nerves TrackingTool of NAVIS software, which requires interaction withthe operator in discriminating visible nerves from back-ground data. To the best of our knowledge, there are nopublished data regarding the repeatability of the NervesTracking Tool of NAVIS software. Despite all limitationsrelated to the measurement methods, the results of thisstudy unequivocally show a depletion of the anterior andmid stromal keratocytes that persists over time as well as areduction of the nerve density parameters but with restora-tion of the baseline values by 18 months after surgery.

Various authors have described corneal changes byconfocal microscopy after the standard CXL procedure.

VOL. 158, NO. 3 CONFOCAL MICROSCOPY AFTER EXC

Structural changes involve the superficial and mid cornealstroma, but the reports are conflicting. In fact, after CXLsome authors have reported a full-thickness stromal kerato-cyte repopulation as soon as 1 year postoperatively,3

whereas other authors have noted a lower density of kera-tocytes after variable follow-up times. Knappe and associ-ates4 pointed out that until 12 months postoperatively,the number of hyper-reflective keratocyte nuclei in theanterior stroma was reduced as compared to before treat-ment. Croxatto and associates5 reported an incipient repo-pulation of keratocytes in the superficial and mid stroma,together with a normal pattern of the deep stroma assoon as 1 month after cross-linking, but examination at36 months disclosed hypocellularity of the anterior andmid corneal stroma with a moderate hyper-reflectivity ofthe extracellular matrix as compared with the preoperativefindings.

481IMER LASER AND CROSS-LINKING

After the PRK plus CXL procedure, full-thicknesscorneal stromal keratocyte repopulation has not beendemonstrated7 until 12 months.

Our study confirms previous confocal reports7 in terms ofincomplete keratocyte repopulation and deeper acellularstromal thickness after the PRK plus CXL procedure, andit adds a longer follow-up time (4 years). The PRK plusCXL procedure includes ablation of Bowman membranetogether with the epithelium, so perhaps this could leadto a deeper penetration of riboflavin into the cornealstroma. It has been shown that in animals without Bowmanmembranes (such as rabbits, in which the mean cornealthickness is 407 6 20 mm), 24 hours after the standardepithelium-off CXL there is extensive keratocyte and endo-thelial cell death.18 All this invites speculation about thebiomechanical role of the Bowman membrane. The biome-chanical role of the Bowman membrane in humans was theobject of reports in the 1990s. In fact, Seiler and associates19

claimed that the Bowman membrane made no significantcontribution to the mechanical stability of the cornea,andWilson and associates20 asserted that the Bowman layerhad no critical function in corneal physiology, being only anindicator of stromal-epithelial interactions. Otherwise, in-direct confirmation of the nonsubstantial function of theBowman membrane in corneal biomechanics was gainedfrom data about the risk for ectasia after refractive surgery.In photorefractive keratectomy, the Bowman membrane,together with a certain amount of anterior stroma, isremoved by excimer laser, but the risk for postsurgical kera-tectasia is significantly less than after LASIK21 and is attrib-utable to many factors.

After CXL, the increase in rigidity of the human corneawas estimated to be about 300%.22 We hypothesize that anincreased biomechanical strength of the cornea, induced byCXL, together with the effect of a deeper penetration ofriboflavin into the corneal stroma, could explain the pro-nounced keratocyte depletion of the stromal cornea post-operatively. Perhaps after an initial apoptosis period,keratocytes have great difficulty in repopulating and pene-trating the corneal stroma because of a greater compactnessand stiffness induced by cross-linking. In fact, in our expe-rience, in a minority of the corneas examined (23.5% at12 months [n ¼ 4]; 35.2% at 18 months [n ¼ 6]; 41.1%at 48 months [n ¼ 7]) at various times after surgery, therewas keratocyte repopulation only in the anterior portionof the cornea, immediately beneath the epithelium. Inthis minority of corneas, the cross-linked area appearsalmost completely acellular and is preceded and followedby stroma-containing cells that are probably unable tomake their way into the cross-linked stroma. There are 2indirect confirmations of this hypothesis. First, the kerato-cyte repopulation in epikeratoplasty lenticules was notedto be of variable extent and more common in the anteriorthan in the posterior region of central lenticules.23 Second,by contrast, biosynthetic corneas made of recombinanthuman collagen, synthesized in yeast and chemically

482 AMERICAN JOURNAL OF

cross-linked,24 and the sponge skirt surrounding a centralcore in AlphaCor (Argus Biomedical, Pty. Ltd., Perth,Australia), are cell-free when implanted, and they allowbiointegration by in-growth of endogenous cells, nervesand collagen deposition due to emulation of the scaffoldingfunction of the natural extracellular matrix of the cornea.Keratocytes produce the collagen and proteoglycans

necessary to maintain the turnover of corneal tissue so,theoretically, a keratocyte depletion could play a role inthe health of the cornea. However, in our experience sofar, the corneal keratocyte depletion appears to have noconsequences on corneal clarity or curvature.An acute turnover of corneal collagen occurs rapidly

over the first 2 weeks of wound healing,25 but the rate ofcollagen turnover in healthy and keratoconic corneas isnot yet known. Cornea is presumed to be a very slow-turnover tissue. A potential limitation of this study isthat follow-up is still not long enough to provide a clearanswer to the question of the effects over time of stromalkeratocyte depletion.After CXL, transitory corneal opacities similar to haze

have been described in a minority of eyes,6 mainly in caseswith preoperative confocal evidence of hyperactivated,hyperdense keratocyte nuclei in the anterior stroma, ordeep stromal dark striae. After the PRK plus CXL proce-dure, Kymionis and associates7 found a posterior linear stro-mal haze in 46.42% of the treated eyes up to 1 year aftersurgery. This posterior haze gradually moved anteriorlyand became less dense during the follow-up period. In ourstudy, at slit-lamp examination, 14 eyes (82%) demon-strated a specific CXL demarcation line and 7 eyes (50%)had central subepithelial haze formation (grade 0.5 to126) by 12 months after treatment. Confocal microscopyexamination helped to better localize and characterizethe sources of backscatter (haze) after the PRK plus CXLprocedure. In our experience, confocal microscope exami-nation in all patients showed 2 regions of high cornealbackscatter: at the boundary between the cross-linkedand not treated area and in the subepithelial location.Even at 4 years after surgery, the baseline morphology ofthe corneal stroma had not been restored. The visualimpact of subepithelial haze and abnormalities in thecorneal stroma has yet to be determined. Corneal back-scatter per se does not affect vision because backscatteredlight does not affect retinal image quality, but we speculatethat it might be a factor that could limit the quality ofvision.It is well known that sub-basal nerve density is lower in

keratoconic corneas, and a positive correlation betweensub-basal nerve density and the severity of keratoconushas also been demonstrated.27 Also, in the early weeks ormonths after CXL a sub-basal nerve depletion has been re-ported in confocal microscope studies, based on qualitativeassessment only, with a return to the preoperative statusseveral months after the procedure.3,6 In our experience,after the PRK plus CXL procedure, sub-basal and stromal

SEPTEMBER 2014OPHTHALMOLOGY

nerves were not identified up to 6 months after surgerybecause they were absent or because of edema, which limitsthe analysis to the confocal microscope. This is furthercomplicated by the fact that changes in backscattered lightafter the PRK plus CXL procedure could limit visibility ofthe nerve fibers because of a masking effect, making earlyisolated and short nerve fibers invisible on confocal micro-scope images during the first months after surgery. From theeighteenth postoperative month, nerve density remainedstable over time. The ability of the sub-basal nerve plexusto restore the preoperative density had already been notedafter CXL,28 but in this study we were able to visualizenerve structures only from the sixth postoperative monthonward. The first nerve fibers appeared in the anterior-mid stroma and then in the subepithelial region.

This study suffers from some limitations, in addition tothose related to computer-assisted manual assessment of

VOL. 158, NO. 3 CONFOCAL MICROSCOPY AFTER EXC

cell and nerve densities. First, because of the thick opticalsection produced by the Confoscan as compared to theHeidelberg retinal tomographer, accurate identification andlocalization of specific cell layers could be difficult. Second,examinations were confined to the central 3–4 mm of thecornea, although not in identical regions each time becauseconfocalmicroscopy employsmagnification up to3500.As aresult, positional repeatability is low, and the scans repre-sented random, full-thickness samples of the central cornea.This is complicated by the fact that the eye may move withrespect to the tip of the lens during scanning, so that evenspacing between images is not necessarily ensured.In conclusion, the results of our study indicate that the

PRK plus CXL procedure affects corneal innervation tran-siently and keratocyte density significantly, up to 4 years offollow-up. Further studies with larger patient samples andlonger follow-up periods are needed to confirm our findings.

ALL AUTHORSHAVE COMPLETED AND SUBMITTED THE ICMJE FORM FOR DISCLOSUREOF POTENTIAL CONFLICTS OF INTEREST,and none were reported. Involved in Conception and design of study (G.A., M.L., C.F., C.S., M.G.L.); Collection of data (M.L.); Analysis and interpre-tation of data (M.G.L., M.L.); Preparation and writing of manuscript (M.L.); Critical revision of manuscript (G.A., M.G.L., C.F.); Final approval of manu-script (C.S.); Data collection and literature search (M.L.); and Final revision (M.G.L., M.L.).

REFERENCES

1. Wollensak G, Spoerl E, Seiler T. Riboflavin/ultraviolet-a-induced collagen crosslinking for the treatment of keratoco-nus. Am J Ophthalmol 2003;135(5):620–627.

2. Alessio G, L’Abbate M, Sborgia C, La Tegola MG. Photore-fractive keratectomy followed by cross-linking versus cross-linking alone for management of progressive keratoconus:two-year follow-up. Am J Ophthalmol 2013;155(1):54–65.

3. Kymionis GD, Diakonis VF, Kalyvianaki M, et al. One-yearfollow-up of corneal confocal microscopy after cornealcross-linking in patients with post laser in situ keratosmileusisectasia and keratoconus. Am J Ophthalmol 2009;147(5):774–778.

4. Knappe S, Stachs O, Zhivov A, Hovakimyan M, Guthoff R.Results of confocal microscopy examinations after collagencross-linking with riboflavin and UVA light in patientswith progressive keratoconus. Ophthalmologica 2011;225(2):95–104.

5. Croxatto JO, Tytiun AE, Argento CJ. Sequential in vivoconfocal microscopy study of corneal wound healing aftercross-linking in patients with keratoconus. J Refract Surg2010;26(9):638–645.

6. Mazzotta C, Traversi C, Baiocchi S, et al. Corneal healing af-ter riboflavin ultraviolet-A collagen cross-linking determinedby confocal laser scanning microscopy in vivo: early and latemodifications. Am J Ophthalmol 2008;146(4):527–533.

7. Kymionis GD, Portaliou DM, Diakonis VF, et al. Posteriorlinear stromal haze formation after simultaneous photorefrac-tive keratectomy followed by corneal collagen cross-linking.Invest Ophthalmol Vis Sci 2010;51(10):5030–5033.

8. Alio JL, Shabayek MH. Corneal higher order aberrations: amethod to grade keratoconus. J Refract Surg 2006;22(6):539–545.

9. Rabinowitz YS. Keratoconus. Surv Ophthalmol 1998;42(4):297–319.

10. Buhren J, Kuhne C, Kohnen T. Defining subclinical kerato-conus using corneal first-surface higher-order aberrations.Am J Ophthalmol 2007;143(3):381–389.

11. Mocan MC, Yilmaz PT, Irkec M, Orhan M. In vivo confocalmicroscopy for the evaluation of corneal microstructure inkeratoconus. Curr Eye Res 2008;33(11):933–939.

12. Patel S, McLaren J, Hodge D, Bourne W. Normal humankeratocyte density and corneal thickness measurement by us-ing confocal microscopy in vivo. Invest Ophthalmol Vis Sci2001;42(2):333–339.

13. McLaren JW, Bourne WM, Patel SV. Automated assessmentof keratocyte density in stromal images from the ConfoScan 4confocal microscope. Invest Ophthalmol Vis Sci 2010;51(4):1918–1926.

14. Kitzmann AS, Winter EJ, Nau CB, McLaren JW, Hodge DO,Bourne WM. Comparison of corneal endothelial cell imagesfrom a noncontact specular microscope and a scanningconfocal microscope. Cornea 2005;24(8):980–984.

15. Morishige N, Takahashi N, Chikamoto N, Nishida T.Quantitative evaluation of corneal epithelial oedema byconfocal microscopy. Clin Experiment Ophthalmol 2009;37(3):249–253.

16. Hillenaar T, Sicam VA, Vermeer KA, et al. Wide-range cali-bration of corneal backscatter analysis by in vivo confocalmicroscopy. Invest Ophthalmol Vis Sci 2011;52(5):2136–2146.

17. McLaren JW, Bourne WM, Patel SV. Standardization ofcorneal haze measurement in confocal microscopy. InvestOphthalmol Vis Sci 2010;51(11):5610–5616.

18. Armstrong BK, Lin MP, Ford MR, et al. Biological andbiomechanical responses to traditional epithelium-off andtransepithelial riboflavin-UVA CXL techniques in rabbits.J Refractive Surg 2013;29(5):332–341.

483IMER LASER AND CROSS-LINKING

19. Seiler T, Matallana M, Sendler S, Bende T. Does Bowman’slayer determine the biomechanical properties of the cornea?Refract Corneal Surg 1992;8(2):139–142.

20. Wilson SE, Hong JW. Bowman’s layer structure and function:critical or dispensable to corneal function? A hypothesis.Cornea 2000;19(4):417–420.

21. Randleman JB, Woodward M, Lynn MJ, Stulting RD. Riskassessment for ectasia after corneal refractive surgery.Ophthalmology 2008;115(1):37–50.

22. Wollensak G, Spoerl E, Seller T. Stress-strain measurementsof human and porcine corneas after riboflavin-ultraviolet-A-induced cross-linking. J Cataract Refract Surg 2003;29(9):1780–1785.

23. Baumgartner SD, Binder PS. Refractive keratoplasty: histopa-thology of clinical specimens. Ophthalmology 1985;92(11):1606–1615.

24. Fagerholm P, Lagali NS, Merret K, et al. A biosyntheticalternative to human donor tissue for inducing corneal

484 AMERICAN JOURNAL OF

regeneration: 24-month follow-up of a phase 1 clinical study.Sci Transl Med 2010;2(46):46ra61.

25. Lass JH, Baker DS, Klein LR. Comparison of collagen degra-dation and synthesis in the peripheral and central areas ofexperimental corneal grafts. Exp Eye Res 1987;45(1):39–43.

26. Fantes FE, Hanna KD, Waring GO 3rd, Pouliquen Y,Thompson KP, Savoldelli M. Wound healing after excimerlaser keratomileusis (photorefractive keratectomy) in mon-keys. Arch Ophthalmol 1990;108(5):665–675.

27. Niederer RL, Perumal D, Sherwin T, McGhee CN. Laserscanning in vivo confocal microscopy reveals reduced inner-vation and reduction in cell density in all layers of the kera-toconic cornea. Invest Ophthalmol Vis Sci 2008;49(7):2964–2970.

28. Kontadakis GA, Kymionis GD, Kankariya VP, Pallikaris AI.Effect of corneal collagen cross-linking on corneal innerva-tion, corneal sensitivity, and tear function of patients withkeratoconus. Ophthalmology 2013;120(5):917–922.

SEPTEMBER 2014OPHTHALMOLOGY

Biosketch

Giovanni Alessio, MD is Associate Professor in Ophthalmology at University of Bari (Italy) Department of Basic Medical

Sciences, Neuroscience, and Sense Organs. His research focuses on refractive surgery, external eye diseases. His clinical and

surgical practice include cataract surgery and corneal transplantations.

VOL. 158, NO. 3 484.e1CONFOCAL MICROSCOPY AFTER EXCIMER LASER AND CROSS-LINKING