The accurate assessment of central corneal thickness (CCT) is of clinical relevance before refractive surgery, in the management of keratoconic patients, in the assessment of donor corneas before penetrating keratoplasty, and when interpreting intraocular pressure (IOP) measurements.1–5
Ultrasound pachymetry (USP) is a contact form of imaging relying on traditional US principles. Advantages of USP include its transportability, accuracy in recording CCT, and proven repeatability.6 Disadvantages include the necessity for use of topical anesthesia, contact with the cornea, difficulty in probe placement, and user dependence.7,8 There are numerous corneal pachymetry techniques available, with USP being the currently accepted gold standard.9
The development of new measurement techniques has encouraged a shift toward noncontact imaging. Fourier domain optical coherence tomography (FD-OCT) is a new technology offering clinicians the ability to image both anterior and posterior structures of the eye. Whereas earlier forms of OCT used a time domain technology, FD-OCT instead combines a broadband light source with interferometry to determine the echo time delay of transmitted light,10 leading to faster acquisition time and improved intersession reliability when compared with time domain OCT.11,12
Evidence supporting FD-OCT use for corneal and anterior segment imaging has been growing, with numerous CCT accuracy, repeatability, and reproducibility studies performed.8,13,14 The FD-OCT has displayed high repeatability, with intraclass correlation coefficients (ICCs) reported within the 0.98 to 0.99 range.6,8,13,15,16 The OCT CCT measurements often differ from USP results; however, there remains a strong linear correlation between the two devices.13,15,17 Knowledge of this difference in measurement may enable application of an algorithm to produce the USP equivalent.
The purpose of this study was to establish the within-rater repeatability and agreement of CCT measured using the Topcon three-dimensional OCT-2000 (Topcon Medical Systems, Oakland, NJ) against USP.
A further purpose of this study was to establish the minimum number of scans and meridians within a single scan required to achieve a reliable assessment of CCT measurement from FD-OCT. To the best of our knowledge, there are no published reports on the within-rater repeatability and agreement of the Topcon three-dimensional OCT-2000.
MATERIALS AND METHODS
This prospective cross-sectional study was composed of 20 participants with normal eyes. Ethics approval was received from the institutional review board, and the study followed the tenets of the Declaration of Helsinki. Before enrolling in the study, the methods and objectives were explained to the participants, and informed consent was obtained.
Participant exclusion criteria were significant corneal pathology, history of previous refractive or ophthalmic surgery, elevated IOP, and myopia of more than −6.00 diopters. Intraocular pressures were measured using a noncontact tonometer (Tomey, Phoenix, Ariz), with an IOP greater than 24 mm Hg considered to be elevated.18 Visual acuity was assessed at 6 m using a visual display unit Snellen chart and standard subjective refractive techniques. Before enrollment, ocular health was assessed by an optometrist.
After recording baseline demographics, all participants underwent imaging with FD-OCT. The Topcon three-dimensional OCT-2000 is a recently released tomographer that has the ability to take 27,000 A scans per second over a maximum scanning range of 6.0 by 6.0 mm to a depth of 2.3 mm.19 A superluminescence diode delivers an infrared light source with a wavelength of 840 nm. The longitudinal resolution is reported by the manufacturer as 5 to 6 μm.19
Each participant was positioned comfortably on the FD-OCT headrest and requested to direct his or her gaze through the viewing path at the internal fixation point. The subject’s pupil was used in combination with the real-time image available on the FD-OCT display monitor to center the scan. Images were taken from one eye (OD), using the anterior segment option, available on FastMap software (Topcon Medical Systems), set to provide a radial scan with 12 equally spaced meridional lines around the central cornea. Five separate scans were taken to enable calculation of within-rater repeatability and for comparison with those of USP. The combination of 12 meridional lines with five scans resulted in a total of 60 imaged corneal sections that could be used for calculating CCT. All scans were conducted by a single investigator (L.N.).
The USP was performed using the Corneo-Gage ultrasound pachymeter (Sonogage, Cleveland, Ohio) on the same eye (OD) within 15 minutes of FD-OCT imaging. The eye was anesthetized using a topical local anesthetic (proxymetacaine hydrochloride 0.5%). The ultrasound probe was placed perpendicular to the central cornea, and five readings were taken of the CCT. All USP measurements were conducted by a single investigator (G.B.).
Reports on the influence of local anesthetic on corneal thickness are varied, with some suggestions that application of topical local anesthetic may cause a transient corneal edema.20,21 To avoid this potentially confounding variable, all participants underwent imaging with FD-OCT before application of topical local anesthetic and USP.
The FD-OCT CCT analysis was performed by a single investigator (L.N.) using the calipers available on the FastMap software. The position of the central cornea was taken to be the point of normal reflection (Fig. 1). The CCT was taken to be the distance between the air-epithelial interface and the endothelial–aqueous humor interface. The calipers were applied at the point of normal reflection between these interfaces. A thickness measure was then calculated by the FastMap software, and results were recorded in Microsoft Excel 2008 (Microsoft, Redmond, Wash). This method was repeated for the 12 meridians per scan. A total of 60 CCT measures were produced per participant.
To minimize bias in caliper measurement, L.N. was masked to the USP results until after analysis of the FD-OCT scans. The USP was performed by G.B. in a separate room from FD-OCT, and results independently recorded by G.B. in Microsoft Excel.
Descriptive statistics were calculated for FD-OCT and USP CCT measurements using SPSS version 20 (SPSS, Chicago, Ill). “Within-rater repeatability” is defined as the amount each measurement varies when performed by the same rater on the same subject during a single session, with all other factors kept constant.22 To assess within-rater repeatability, the ICC, within-subject variance (S w), and within-subject coefficients of repeatability (COR) and variation (COV) were calculated.
The S w provides guidance on the within-subject repeatability for measurement using a single instrument during one session and is a means of representing measurement error. The S w was calculated by taking the square root of the mean of the variances for each participant.23 For FD-OCT measurements, variances were calculated based on the five scans and 12 meridians for each participant. For USP measurements, variances were calculated based on the five readings.
The COR was calculated using the following formula: √2 × 1.96 × S w.23 It is expected that the difference between two measurements for the same subject will be less than the COR for 95% of observations. The COV indicates the measurement error in proportion to the mean and was calculated as the ratio of the S w and the mean of the measurements.24
The ICC was calculated using the single measures one-way random effects model. The ICC ranges from a value of 0 to 1, with 0 indicating no agreement and 1 indicating absolute agreement between repeated measurements. The ICC approaches 1 as the S w approaches 0.
To assess the agreement of FD-OCT compared with USP, analysis of variance and simple linear regression were performed in addition to Bland-Altman analysis. The difference between means was regressed against the average of the means to assess for proportional bias. To assess absolute agreement between instruments, the ICC was calculated using the average measures two-way random effects model. Statistical significance was set at a value of p < 0.05.
Determining the Minimum Number of Scans Required to Achieve Reliable CCT Assessment
Subanalysis was undertaken to determine the minimum number of FD-OCT scans and meridians within a single scan required to achieve a reliable CCT assessment. This analysis was conducted by calculating the mean and 99% confidence intervals (99% CIs) for each participant using all 60 FD-OCT measurements. The 99% CI limits describe the range within which 99% of measurements are likely to fall. The mean CCT measures were then calculated for each participant first using all of the meridians from the first scan and then the horizontal meridian section from each of the five individual scans in the order of first to last measured. Meridians from the single scan were assessed in the order of horizontal first followed by inclusion of further meridians in a counterclockwise direction. The CCT measurements from the individual sections were added in turn and the mean was recalculated until the mean fell within the 99% CI limits previously calculated from all 60 corneal sections. In this manner, the minimum number of sections required from a single scan and the minimum number of scans required using analysis from a single section could be established. Agreement between FD-OCT and USP CCT was revised using the minimum number of scans and calculation of the ICC using the average measures two-way random effects model.
Within this subanalysis, we also randomized the order of analyzing whole scans or meridians from a single scan to investigate whether bias was introduced from the systematic order that we initially adopted. Randomization was provided from an online random sequence generator (Randomness and Integrity Services Limited, Ireland, UK),25 and the same methodology as previously described, for detecting the minimum number of scans and meridians within a scan, was conducted.
Twenty participants (11 women and 9 men) underwent imaging, with mean age ± SD of 33 ± 14 years, range of 18 to 61 years. Table 1 displays the within-rater repeatability measures. Kendall τ rank correlation coefficient was calculated for FD-OCT and USP measurements to ensure that the subject’s SD was unrelated to the CCT. Kendall τ was 0.105 (p = 0.516) and 0.185 (p = 0.256) for FD-OCT and USP, respectively. This demonstrated no relationship and justified the use of the S w.
The CCT measured using FD-OCT (528 ± 27 μm) was significantly less than that when measured using USP (544 ± 29 μm, p < 0.001). However, the intraclass correlation between FD-OCT and USP was high (0.99; 95% CI, 0.97–0.99).
A significant linear correlation was seen between USP and FD-OCT CCT measurements (Pearson correlation, r = 0.96; p < 0.05) (Fig. 2). Bland-Altman analysis demonstrated FD-OCT measurement to be thinner than USP, with a mean difference of −16.08 μm (95% limits of agreement, −32.06 to −0.11 μm) (Fig. 3). There was no significant proportional bias found for CCT measurements of USP and FD-OCT (Pearson correlation, r = 0.27; p = 0.257).
Subanalysis of the minimum number of FD-OCT scans and meridians required to maintain CCT within the 99% CI was performed. A mean of 3.55 scans or 5.75 meridians from a single scan were required to maintain CCT within 99% CI. Therefore, a single meridian from four FD-OCT scans was deemed to be the minimum required to achieve a similar level of accuracy to CCT calculated from all 60 corneal sections that were captured for each subject. Measures of agreement between FD-OCT and USP were revised using a single meridian from the four scans to reveal that the ICC between FD-OCT and USP remained high (0.99; 95% CI, 0.97–0.99). Table 2 indicates the ICC and 95% CI range for measures of agreement between FD-OCT and USP from using only the horizontal meridian captured from between one and five FD-OCT scans.
When the subanalysis was repeated using randomized sequencing, the minimum number of FD-OCT scans and meridians required to maintain the CCT within 99% CI was calculated as 2.80 scans or 5.40 meridians from a single scan.
This study has found the Topcon 3D-OCT 2000 to achieve high within-rater repeatability and high correlation with USP for measurement of CCT. The FD-OCT was found to be superior to USP across several measures of repeatability (ICC, S w, and COR). However, both instruments displayed good repeatability, with an ICC of 0.99 and 0.96 for FD-OCT and USP, respectively. The S w was small, 5.07 μm and 5.49 μm for FD-OCT and USP, respectively. The COV was equal for both USP and OCT, demonstrating no difference when the measurement error is reported in proportion to the mean.
Although attaining the true value of CCT is important, it is also important that the measured value is able to be reproduced on repetition. An instrument with a high repeatability is of immense clinical value if the deviation from the true value is known. Repeated measures using the same instrument on the same patient will produce a range of readings reflective of variation in the patient and the measurement process. It is expected that this variation is around the true value being measured. To date, USP is the most accurate instrument at approximating the true CCT,26 hence, its use in this study as the benchmark against which to judge FD-OCT measurements.
We found FD-OCT to underestimate CCT by a mean of −16.08 μm (95% limits of agreement, −32.06 to −0.11 μm) when compared with USP. There was good agreement between USP and FD-OCT CCT measures, with an ICC of 0.99 (95% CI, 0.97–0.99) and significant linear correlation (Pearson correlation, r = 0.96; p < 0.05). The finding of OCT underestimating CCT is comparable to those of numerous previous studies (Table 3).1,8,13,14,16,17,27,28 The underestimation of CCT by OCT has been suggested to possibly be an instrument-dependent issue.28 The relationship between instruments has remained difficult to predict. However, the likelihood of FD-OCT CCT underestimation being a model-dependent issue is low, considering the variability of previous results.6,13
Despite FD-OCT significantly underestimating CCT, the high ICC demonstrates that this underestimation was consistent across participants. In addition, there was no proportional bias, with no relationship between the difference between the means and CCT. That is, the underestimation of CCT by FD-OCT demonstrated a linear relationship with USP, in keeping with the results of the regression analysis. This finding was consistent with those of previous studies that commented on agreement between the OCT and USP16,27,30 and gives confidence that the Topcon FD-OCT provides a reliable measure of CCT once underestimation is taken into consideration.
Underestimation of CCT by FD-OCT carries clinical implications. Refractive surgery requires adequate residual corneal stromal tissue, at least 250 μm, under the flap.7 Reliance on CCT measurement using the Topcon tomographer in the preoperative workup would result in higher exclusion rates for refractive surgery. Therefore, use of FD-OCT for this purpose may even reduce the risk of iatrogenic keratectasia. In cases where FD-OCT CCT results in exclusion from refractive surgery, use of alternative measurement devices may be considered.
Corneal thickness measures are also of importance in the diagnosis and management of keratoconic patients.3 Sole dependence on FD-OCT for the diagnosis and follow-up of keratoconic patients may result in elevated false-positive rates. It is, therefore, necessary to consider the significant underestimation of CCT by FD-OCT in clinical practice to ensure safe and effective patient management.
There is a need for discussion of the number of scans required for FD-OCT CCT measurement. In our study, despite a small sample size of 20 participants, we produced a large number of CCT measurements. This was a result of the Topcon OCT image analysis constructing 12 full meridian optical sections at 15 degrees of spacing for each scan. We found that analysis of a single meridian from four scans was sufficient to maintain CCT measures within 99% CI. The number of scans required to maintain CCT accuracy was less than the respective number of meridians from a single scan, which is likely to be a result of reliance on the quality of each scan. Analyzing CCT measurements from multiple meridians of a single scan relies on the scan being of high quality. A scan with poor pupil centration14 will result in 12 inadequately positioned meridians. Therefore, use of a single meridian from scan repeats potentially minimizes centering errors.
It could be argued that the systematic order in which we analyzed data may introduce bias. To test this, we also randomized the order of scan and meridian sequencing to find that this led to a reduction in the minimum number of FD-OCT scans and meridians required to maintain CCT within the 99% CI limits. This gives confidence that systematic ordering did not lead to an underestimation of the number of scans or analyzed meridians required to achieve CCT measurements within the 99% CI limits. The purpose of our subanalysis was to determine the mode of practice for practitioners to minimize the number of measurements required in clinical practice, and in this regard, this additional randomized analysis gives confidence that averaging the CCT measurements from the horizontal meridian of the first four captured scans provides an accurate measurement of CCT with the Topcon FD-OCT.
The revised ICC, comparing a single meridian from four FD-OCT scans with USP results, remained high at 0.99 (95% CI, 0.97–0.99). This was despite a reduction from 60 to four FD-OCT CCT measurements per participant. In our study, analysis of a minimum of four FD-OCT scan repeats produced accurate CCT readings when compared with those of USP. Interestingly, analysis of fewer than four FD-OCT scans also produces good agreement with USP, as demonstrated in Table 2. Individual preference may direct the balance between efficiency and accuracy of FD-OCT CCT measurement in clinical practice.
Our study was limited to the assessment of CCT in normal participants. Application of pachymetry techniques to clinical practice frequently involves patients with corneal pathologies. In addition, imaging for USP and FD-OCT was conducted by separate investigators during a single session. Although this enabled accurate measurement of single-session within-rater repeatability, it does not provide information on between-rater repeatability and intersession repeatability. Consequently, there is a need for further studies evaluating CCT measurements using the Topcon 3D-OCT 2000 in a variety of patient populations.
The Topcon 3D-OCT 2000 has been shown to be an instrument with excellent repeatability for CCT measurements. The 16.08 μm underestimation of CCT by FD-OCT compared with USP means that the instruments are not interchangeable. Although this underestimation was consistent throughout the CCT range measured, the absolute difference between instruments must be considered in clinical practice.
Luke C. Northey
School of Medicine Sydney, The University of Notre Dame, Australia
PO Box 944 Broadway
New South Wales 2007, Australia
The authors thank the Research in Orthokeratology Group, School of Optometry and Vision Science, The University of New South Wales, for loan of the ultrasound pachymeter.
Received June 6, 2012; accepted August 13, 2012.
1. Grewal DS, Brar GS, Grewal SP. Assessment of central corneal thickness in normal, keratoconus, and postlaser in situ keratomileusis eyes using Scheimpflug imaging, spectral domain optical coherence tomography, and ultrasound pachymetry. J Cataract Refract Surg 2010; 36: 954–64.
2. Cheng AC, Rao SK, Lau S, Leung CK, Lam DS. Central corneal thickness measurements by ultrasound, Orbscan II, and Visante OCT after LASIK for myopia. J Refract Surg 2008; 24: 361–5.
3. Aurich H, Pham DT, Wirbelauer C. Biometric evaluation of keratoconic eyes with slit lamp-adapted optical coherence tomography. Cornea 2011; 30: 56–9.
4. Choi CY, Youm DJ, Kim MJ, Tchah H. Changes in central corneal thickness of preserved corneas over time measured using anterior segment optical coherence tomography. Cornea 2009; 28: 536–40.
5. Day AC, Machin D, Aung T, Gazzard G, Husain R, Chew PT, Khaw PT, Seah SK, Foster PJ. Central corneal thickness and glaucoma in East Asian people. Invest Ophthalmol Vis Sci 2011; 52: 8407–12.
6. Nam SM, Im CY, Lee HK, Kim EK, Kim TI, Seo KY. Accuracy of RTVue optical coherence tomography, Pentacam, and ultrasonic pachymetry for the measurement of central corneal thickness. Ophthalmology 2010; 117: 2096–103.
7. Reinstein DZ, Archer TJ, Gobbe M. Repeatability of intraoperative central corneal and residual stromal thickness measurement using a handheld ultrasound pachymeter. J Cataract Refract Surg 2012; 38: 278–82.
8. Chen S, Huang J, Wen D, Chen W, Huang D, Wang Q. Measurement of central corneal thickness by high-resolution Scheimpflug imaging, Fourier-domain optical coherence tomography and ultrasound pachymetry. Acta Ophthalmol 2012; 90: 449–55.
9. Swartz T, Marten L, Wang M. Measuring the cornea: the latest developments in corneal topography. Curr Opin Ophthalmol 2007; 18: 325–33.
10. Maeda N. Optical coherence tomography for corneal diseases. Eye Contact Lens 2010; 36: 254–9.
11. Huang JY, Pekmezci M, Yaplee S, Lin S. Intra-examiner repeatability and agreement of corneal pachymetry map measurement by time-domain and Fourier-domain optical coherence tomography. Graefes Arch Clin Exp Ophthalmol 2010; 248: 1647–56.
12. Prakash G, Agarwal A, Jacob S, Kumar DA, Banerjee R. Comparison of Fourier-domain and time-domain optical coherence tomography for assessment of corneal thickness and intersession repeatability. Am J Ophthalmol 2009; 148: 282–90.
13. Ishibazawa A, Igarashi S, Hanada K, Nagaoka T, Ishiko S, Ito H, Yoshida A. Central corneal thickness measurements with Fourier-domain optical coherence tomography versus ultrasonic pachymetry and rotating Scheimpflug camera. Cornea 2011; 30: 615–9.
14. Li Y, Tang M, Zhang X, Salaroli CH, Ramos JL, Huang D. Pachymetric mapping with Fourier-domain optical coherence tomography. J Cataract Refract Surg 2010; 36: 826–31.
15. Fukuda S, Kawana K, Yasuno Y, Oshika T. Anterior ocular biometry using 3-dimensional optical coherence tomography. Ophthalmology 2009; 116: 882–9.
16. Rao HL, Kumar AU, Kumar A, Chary S, Senthil S, Vaddavalli PK, Garudadri CS. Evaluation of central corneal thickness measurement with RTVue spectral domain optical coherence tomography in normal subjects. Cornea 2011; 30: 121–6.
17. Wirbelauer C, Thannhauser CL, Pham DT. Influence of corneal curvature on central and paracentral pachymetry with optical coherence tomography. Cornea 2009; 28: 254–60.
18. Kass MA, Heuer DK, Higginbotham EJ, Johnson CA, Keltner JL, Miller JP, Parrish RK 2nd, Wilson MR, Gordon MO. The Ocular Hypertension Treatment Study: a randomized trial determines that topical ocular hypotensive medication delays or prevents the onset of primary open-angle glaucoma. Arch Ophthalmol 2002; 120: 701–13.
20. Mukhopadhyay DR, North RV, Hamilton-Maxwell KE. Effect of a proparacaine 0.50%–sodium fluorescein 0.25% mix and contact ultrasound pachymetry on central and midperipheral corneal thickness measured by noncontact optical pachymetry. J Cataract Refract Surg 2011; 37: 907–13.
21. Nam SM, Lee HK, Kim EK, Seo KY. Comparison of corneal thickness after the instillation of topical anesthetics: proparacaine versus oxybuprocaine. Cornea 2006; 25: 51–4.
22. Bland JM, Altman DG. Measurement error and correlation coefficients. BMJ 1996; 313: 41–2.
23. Bland JM, Altman DG. Measurement error. BMJ 1996; 312: 1654.
24. Bland JM, Altman DG. Measurement error proportional to the mean. BMJ 1996; 313: 106.
26. Marsich MW, Bullimore MA. The repeatability of corneal thickness measures. Cornea 2000; 19: 792–5.
27. Prospero Ponce CM, Rocha KM, Smith SD, Krueger RR. Central and peripheral corneal thickness measured with optical coherence tomography, Scheimpflug imaging, and ultrasound pachymetry in normal, keratoconus-suspect, and post-laser in situ keratomileusis eyes. J Cataract Refract Surg 2009; 35: 1055–62.
28. Li EY, Mohamed S, Leung CK, Rao SK, Cheng AC, Cheung CY, Lam DS. Agreement among 3 methods to measure corneal thickness: ultrasound pachymetry, Orbscan II, and Visante anterior segment optical coherence tomography. Ophthalmology 2007; 114: 1842–7.
29. Williams R, Fink BA, King-Smith PE, Mitchell GL. Central corneal thickness measurements: using an ultrasonic instrument and 4 optical instruments. Cornea 2011; 30: 1238–43.
30. Huang J, Pesudovs K, Yu A, Wright T, Wen D, Li M, Yu Y, Wang Q. A comprehensive comparison of central corneal thickness measurement. Optom Vis Sci 2011; 88: 940–9.
Keywords:© 2012 American Academy of Optometry
central corneal thickness; optical coherence tomography; repeatability; reliability; ultrasound pachymetry