Graefe’s Arch Clin Exp Ophthalmol(2005) 243: 798–802
DOI 10.1007/s00417-005-1149-2
CLINICAL INVESTIGATION
Kelechi C. OgbuehiTurki M. Almubrad
Received: 11 November 2004Revised: 7 January 2005Accepted: 14 January 2005Published online: 15 March 2005# Springer-Verlag 2005
Repeatability of central corneal thicknessmeasurements measured with the TopconSP2000P specular microscope
Abstract Background: The non-contact specular microscope has be-come the method of choice for aquick, accurate and non-invasive as-sessment of central corneal thickness(CCT), which is an important variableto monitor before and after refractivesurgery. The consistency of the resultsproduced by such widely used meth-ods/equipment must be assessed todetermine their reliability. The pur-pose of this study was to assesswithin- and between-observer repeat-ability of, and to determine if asystematic bias exists in the measure-ments made by, the Topcon SP2000Pspecular microscope. Methods: TheCCT of the right eyes of 70 adultsubjects, divided equally between menand women, was assessed on twoseparate occasions (4–7 days apart) byeach of two examiners using the low-intensity auto mode of the SP2000Pspecular microscope. Results: Theaverage CCT values for men andwomen, measured by one observer,were 0.52±0.03 mm (mean ± SD) and0.52±0.04 mm, respectively. Theaverage for the entire sample was 0.52±0.04 mm. Within- and between-observer repeatability were assessedby plotting the mean difference (foreach subject) between two readingsmade by the same observer or oneeach by both observers against thecombined average CCT reading ofboth sessions; the mean difference
between two sets of observations wasnot significantly different from zero(P<0.05). For the first observer, the95% limits of repeatability werebetween −0.015 and 0.017 mm. Forthe second observer, the 95% limits ofrepeatability were between −0.018and 0.018 mm. For the between-observer repeatability, the 95% limitsof agreement were between −0.016and 0.016 mm. For both within- andbetween-observer repeatability, wefound no systematic bias of the meandifference with the average CCTreading. Conclusion: The within-and between-observer limits of agree-ment we found were similar to thosepreviously reported for the TopconSP2000P specular microscope; therange of the 95% limits of repeat-ability were within ±1 SD of theaverage CCT reading for both ses-sions. We suggest that a technique beconsidered reliable if: (1) the meandifference between two measurementsdoes not vary significantly from zero,(2) there is no systematic bias of themean difference with the magnitude ofthe measured quantity and (3) the errorinherent in a measurement techniqueis within ±1 SD of the averagemeasurement of the two sessions.
Keywords Corneal thickness . CCT .Non-contact specular microscope .SP2000P. Limits of agreement . LoA .Pachometry
K. C. Ogbuehi (*) . T. M. AlmubradDepartment of Optometry andVision Sciences,College of Applied Medical Sciences,P.O. Box 10219,Riyadh, 11433, Kingdomof Saudi Arabiae-mail: [email protected].: +966-1-4355010Fax: +966-1-4809150
Introduction
Evaluation of central corneal thickness (CCT) -and cornealthickness profiles- is an important indicator of the patencyof the corneal endothelial pump [18], and it can be used as asensitive indicator for a wide range of disorders that affectthe anterior segment of the eye (and, as a consequence, thecornea), including glaucoma [4], corneal ectasias such askeratoconus and keratoglobus [8], and dry eye [5, 7, 11].The assessment of corneal central thickness and profile hasbecome even more important with the advent of modernrefractive surgical techniques, which have made cornealsurgery more widespread. Repeated measures of cornealcentral thickness and profile have thus become even morerelevant. Avariety of modern equipment is available for theassessment of corneal thickness, the most versatile andpopular of which is the non-contact specular microscope,which has rapidly become amethod of choice for pachometryassessments. In addition to its non-invasive method, it com-bines an assessment of the mid-peripheral thickness of thecornea in four locations with a differential focusing of cor-neal epithelial and endothelial surfaces, which consequentlyallows for the simultaneous assessment of corneal endothe-lial cell density. The first description of the corneal endo-thelium in vivo viewed by specular reflection was made byVogt [17]. Since then, advances in computer technologyhave simplified the tedious nature of corneal endothelialmorphometric analysis, making non-contact specular mi-croscopy a convenient and popular method for assessingcorneal endothelial morphology while simultaneously ana-lysing corneal thickness profiles.
To the knowledge of the authors, only one previous studyhas assessed the intra-observer and inter-observer reproduc-ibility of repeated corneal thickness assessments with thenon-contact specular microscope SP2000P [3]. The presentstudy was undertaken to confirm the findings of Cho andCheung [3], and we go further to suggest criteria for val-idating repeat measures of the same quantity made by amethod.
Subjects and methods
The subjects were drawn from patients scheduled forroutine refraction at the King Saud University Optometryclinic. All subjects were pre-presbyopic patients presentingfor a routine eye exam. Those subjects with a positive his-tory for contact lens wear, anterior segment disease or sur-gery, trauma or amblyopia were excluded. Seventy subjects,divided equally between men and women, participated inthis study. Only the right eye was assessed for each subject.The subjects’ age range was 18–35 (24.0±4.3; mean ± SD)years. Informed consent was obtained from each subjectbefore themeasurements were carried out, and the studywasconducted in conformance with the ethical considerationslaid out in the 1964 Declaration of Helsinki. To avoid theeffects of diurnal corneal thickness variations, all measure-ments were carried out in the afternoon [10]. Triplicate CCTassessments using the specularmicroscope (TopconSP2000P;Abdulrehman Al-Gosaibi GTB, Riyadh, Saudi Arabia) wereobtained. Corneal images were captured with the subjectfixating on the central target, the subject’s chin on the chinrest and his/her head on the forehead rest (in a position iden-tical to that for a slitlamp examination). The automatic imagecapture, low-intensity mode of the specular microscope wasemployed in this study.
Data analysis
The average corneal thickness for each subject formed thedata point for that subject. The data points for all the subjectswere collected and analyzed. The level of significance for allcomparisons was set at 5%.
Assessment of intra- and inter-observer reproducibility
To determine the reliability of amethod, repeated readings bythe same observer and by different observers were compared.
Table 2 Mean differenceswithin- and between-observers
Observer 1(1st–2nd readings)
Observer 2(1st–2nd readings)
Observer 1–observer 2
Mean difference (mm) 0.001 0.0001 0.000Standard deviation 0.008 0.009 0.00895% CI (±1.96 SD) −0.015 to 0.017 −0.018 to 0.018 −0.016 to 0.016
Table 1 Central corneal thick‐ness (CCT) readings determinedby both observers
Observer 1 (1st reading +2nd reading)/2
Observer 2 (1st reading +2nd reading)/2
(Observer 1 +observer 2)/2
Mean CCT (mm) 0.522 0.524 0.522Standard deviation 0.036 0.036 0.03695% CI (±1.96 SD) 0.451–0.593 0.453–0.595 0.451–0.593
799
The Bland and Altman [1] statistical analysis method wasemployed to assess the limits of agreement (LoA) betweentwo measurements of CCT (either by the same observer orby two different observers) using the same technique. Thedifference between the means of two readings taken on twoseparate days (separated by a time frame of 1 week or less)was plotted against the average readings of the two ses-sions to determine the limits of agreement. Pearson’s orSpearman rank correlation coefficient analyses were car-ried out on the scatter plot (“differences between means”against the “combined mean CCT”) to exclude a systemat-ic bias of the “differences between means” with the mag-nitude of the measured CCT.
Results
There was no statistically significant difference (P>0.05)between the CCT values for men and women. Therefore,only the values for one set of measurements (by one ob-server) are quoted for male and female CCT values, and theresults of the entire sample are pooled in the statisticalanalyses of each set of measurements by each observer. Theaverage CCT values for men and women, measured by thefirst observer, were 0.52±0.03 mm (mean ± SD) and 0.52±0.04 mm, respectively. The average for the entire samplewas 0.52±0.04 mm. The average readings determined byboth observers are summarised in Table 1.
The 95% LoA for two sets of CCT measurements madeby each observer and for one measurement each made byboth observers is summarised in Table 2. The mean dif-
Fig. 1 Graph plot of the meandifference, first reading minussecond reading, (ordinate)against the mean corneal thick-ness readings (abscissa) of twoseparate sets of measurementstaken by the first observer. APearson correlation coefficientanalysis (P>0.05) excluded asystemic bias of the “meandifference” with the magnitudeof the measured central cornealtickness (CCT).
Fig. 2 Mean difference plot forthe second observer. A Spear-man rank correlation coefficientanalysis (P>0.05) excluded asystemic bias of the “meandifference” with the magnitudeof the measured central cornealthickness (CCT).
800
ference of the two readings made by the first observer was0.001±0.008 mm (mean ± SD). For the second observer, themean difference was 0.0001±0.009 mm. The between-observer mean difference was 0.000±0.008 mm.
The 95% LoA for the two readings made by the firstobserver ranged from −0.015 to 0.017 mm. For the secondobserver, the 95% LoA ranged from −0.018 to 0.018 mm.The between-observer 95% LoA ranged from −0.016 to0.016 mm. As a percentage of the mean CCT of two as-sessments, the 95% LoA of the mean difference rangedfrom −2.9 to 3.3% for the first observer, −3.4 to 3.4% for thesecond observer and −3.1 to 3.1% for between observers.Figures 1, 2 and 3 are the mean difference plots for two setsof CCT values measured by the first , second and both ob-servers, respectively.
The 95% LoA for the mean differences for each observerand between observers were all within the range of ±1 SD ofthe combined mean of the two observations. For the firstobserver, for example, the 95% confidence interval for themean difference between two assessments ranged from−0.015 to 0.017 mm (range=0.032 mm). The average CCTof two sets of readings made by the first observer was 0.522mm (±0.036 SD). Therefore, 95% of the differences be-tween both readings would be expected to fall within therange of ±1 SD of the mean CCT measured by observer 1.
Discussion
The “mean difference” method described by Bland andAltman [1] is perhaps the most appropriate method forassessment of repeatability and agreement. Thismethodwasemployed in this study to assess the repeatability of CCTmeasures made with the SP 2000 P non-contact specularmicroscope. The measurement error inherent in any tech-
nique must necessarily be significantly smaller than thevariability (between and within individuals) of the ocularcomponent which it assesses. We reason that if 95% of themeasurement error inherent in a technique is accounted forwithin ±1 SD of the mean of the ocular component mea-sured, this error would not significantly bias any assess-ment of the ocular component in question. Therefore, wesuggest that the assessments made by a technique be con-sidered reproducible if the mean difference between tworeadings is not significantly different from zero, there is nosystematic bias of the mean difference with the magnitudeof the measured quantity, and the range between the LoAis within the region of ±1 SD of the combined mean ofboth measurements of the variable (using the same exper-imental sample).
Based on the criteria above, the measurements of CCTmade by the Topcon SP2000P non-contact specular micro-scope are repeatable between sessions (by the same observer)and between observers, indicating that the measurementerror inherent in this technique is small compared with thevariability (between and within individuals) of CCT mea-sures and that there is little or no examiner bias on mea-surements made with this specular microscope. In the onlyother study to have assessed the repeatability of CCT mea-surements between sessions and between observers and inwhich the age group of the subjects (19–38 years) was iden-tical to the age group recruited for this study (18–35 years),Cho and Cheung [3] reported similar (P>0.05) CCT values(538±34 μm; n=19) to those found in this study. More per-tinently, they reported within-observer 95% LoA for CCTof −0.013 to 0.015 mm. For between-observer LoA, theyreported values from −0.014 to 0.017 mm. This compareswell with our results where the 95% LoAwere from −0.015to 0.017 mm for the first observer, −0.018 to 0.018 mm forthe second observer and −0.016 to 0.016 mm between ob-
Fig. 3 Mean difference plot forboth observers. A Spearmanrank correlation coefficientanalysis (P>0.05) excluded asystemic bias of the “meandifference” with the magnitudeof the measured central cornealthickness (CCT).
801
servers. The LoA values reported by Cho and Cheung [3]showed that when within- or between-observer LoA valueswere within the range of ±1 SD of mean CCT, the interclasscorrelation coefficient (ICC) was always above 0.9, the cut-off point to ensure a reasonable validity of repeatedmeasuresof clinical variables suggested by Portney and Watkins [14].However, for the one region of the cornea (inferior cornealthickness) for which Cho and Cheung [3] reported LoAvalues (between observers) and for which the range of theLoA values exceeded ±1 SD, the ICC was less than 0.9.
In another study that assessed the within-observer re-peatability, the Topcon SP2000P non-contact specularmicro-scope showed the smallest variability of repeated measuresof CCTeven though the average CCT values measured withthis technique were significantly less than those measuredwith scanning slit corneal topography and with contact ultra-sonic pachymetry [15]. However, the within-observer rangefor LoA for the Suzuki et al. [15] study was about 60 μmcompared with about 35 μm for the present study eventhough the mean CCT values for both studies were identical(P>0.05). One other study [16] assessed only the within-observer repeatability and found that the variability ofrepeated CCT measures was greatest with the SP2000Pspecular microscope compared with ultrasound pachymetryand ultrasound biomicroscopy. Tam et al. [16] also reportedhigher values for average CCTmeasured with the SP2000P
specular microscope compared with the other two tech-niques, in contrast with the results from other studies [2, 9,13]. Repeatability in the Tam [16] study was computed bycalculating the standard deviation of triplicate CCT mea-surements made for each subject. The standard deviationwas averaged for the entire sample and used to calculate therepeatability. As a result of the different method of analysisused by Tam [16], no “mean difference” plot for within-observer repeatability was generated, which made it impos-sible to compare the results of that study with those of thepresent study.
With respect to the variation of Goldmann intraocularpressuremeasurements with CCT,Miglior et al. [12] consid-ered a variation between repeated measures of CCT ≥15 μmto be clinically significant based on the conversion factorproposed by Ehlers et al. [6] of 0.7 mmHg/10 μm change inCCT. In that study, the mean variation between repeatedCCT measures was an order of magnitude less than 15 μm.This suggests that the variability inherent in non-contactspecular microscopy is not likely to introduce any additionalerrors to clinical applanation tonometry assessments.
Our results are relevant to the interpretation of futurerepeatability and agreement studies necessary for the assess-ment of the reliability of techniques for measuring ocularcomponents.
References
1. Bland JM, Altman DG (1986) Statisti-cal methods for assessing agreementbetween two methods of clinical mea-surement. Lancet 1:307–310
2. Bovelle R, Kaufmann SC, ThompsonHW, Hamano H (1999) Corneal thick-ness measurements with the TopconSP-2000P specular microscope and anultrasound pachymeter. Arch Ophthal-mol 117:868–870
3. Cho P, Cheung SW (2000) Central andperipheral corneal thickness measuredwith the Topcon specular microscopeSP-2000P. Curr Eye Res 21(4):799–807
4. Copt RP, Thomas R, Mermoud A(1999) Corneal thickness in ocularhypertension, primary open-angle glau-coma, and normal tension glaucoma.Arch Ophthalmol 106:2154–2160
5. De Paiva CS, Harris LD, Pflugfelder SC(2003) Keratoconus-like topographicchanges in keratoconjunctivitis sicca.Cornea 22:22–24
6. Ehlers N, Bramsen T, Sperling S (1975)Applanation tonometry and centralcorneal thickness. Acta Ophthalmol(Copenh) 53:34–43
7. Guzey M, Satici A, Karadede S (2002)Corneal thickness in trachomatous dryeye. Eur J Ophthalmol 12:18–23
8. Insler MS, Baumann JD (1986) Cornealthinning syndromes. Ann Ophthalmol18:74–75
9. Kawana K, Tokunaga T, Miyata K,Okamoto F, Kiuchi T, Oshika T (2004)Comparison of corneal thickness mea-surements using Orbscan II, non-con-tact specular microscopy, and ultrasonicpachymetry in eyes after laser in situkeratomileusis. Br J Ophthalmol88:466–468
10. Lattimore MR Jr, Kaupp S, SchallhornS, Lewis R IV (1999) Orbscan pachy-metry: implications of a repeated mea-sures and diurnal variation analysis.Ophthalmology 106:977–981
11. Liu Z, Pflugfelder SC (1999) Cornealthickness is reduced in dry eye. Cornea18:403–407
12. Miglior S, Albe E, Guareschi M,Mandelli G, Gomarasca S, Orzalesi N(2004) Intraobserver and interobserverreproducibility in the evaluation ofultrasonic pachymetry measurements ofcentral corneal thickness. Br JOphthalmol 88:174–177
13. Modis L Jr, Langenbucher A, Seitz B(2001) Scanning-slit and specular mi-croscopic pachymetry in comparisonwith ultrasonic determination of cornealthickness. Cornea 20:711–714
14. Portney LG, Watkins MP (2000) Statis-tical measures of reliability. Foundationsof clinical research: applications topractice, 2nd edn. Appleton and Lange:New Jersey
15. Suzuki S, Oshika T, Oki K, Sakabe I,Iwase A, Amano S, Araie M (2004)Corneal thickness measurements: scan-ning-slit corneal topography and non-contact specular microscopy versusultrasonic pachymetry. J Cataract Re-fract Surg 29:1313–1318
16. Tam ES, Rootman DS (2003) Com-parison of central corneal thicknessmeasurements by specular microsco-py, ultrasound pachymetry, and ultra-sound biomicroscopy. J CataractRefract Surg 29:1179–1184
17. Vogt A (1920) Die Sichtbarkeit deslebenden Hornhautendothels; Ein Bei-trag zur Methodik der Spaltleampen-mikroskopie. Graefes Arch Clin ExpOphthalmol 101:123–144
18. Waring GO III, Bourne WM, Edelhau-ser HF, Kenyon KR (1982) The cornealendothelium: normal and pathologicstructure and function. Ophthalmology89:531–590
802
