- Email: [email protected]
Comparison of corneal thickness measurements using ultrasound and Orbscan slit-scanning topography in normal and post-LASIK eyes
Comparison of corneal thickness measurements using ultrasound and Orbscan slit-scanning topography in normal and post-LASIK eyes H. Soma Chakrabarti, FRCOphth, Jennifer P. Craig, PhD, MCOptom, Arun Brahma, MD, FRCOphth, Tahira Y. Malik, FRCOphth, Charles N.J. McGhee, PhD, FRCOphth ABSTRACT Purpose: To compare corneal thickness measurements made by ultrasonic and slitscanning techniques in normal eyes and in eyes after laser in situ keratomileusis (LASIK). Setting: Corneal Diseases and Excimer Laser Research Unit, University of Dundee, Dundee, Scotland. Methods: Central corneal thickness (CCT) was measured in 101 eyes of 59 normal subjects and in 30 eyes of 21 post-LASIK patients. Measurements were made with an Orbscan slit-scanning elevation topographer and immediately afterward with an ultrasound pachymeter. Results: The difference in mean CCT between ultrasound (538.0 m ⫾ 36.7 [SD]) and Orbscan (566.6 ⫾ 40.7 m) pachymetry was statistically significant (P ⬍ .001) in the normal eyes; the Orbscan measurement was approximately 28 m higher than that of the ultrasound pachymeter. The difference in mean CCT between the ultrasound and the slit-scanning techniques was also statistically significant in the post-LASIK eyes (mean values 475.3 ⫾ 50.3 m and 461.9 ⫾ 74.2 m, respectively; P ⬍ .0001). Differences in CCT in individual subjects were much more variable in the post-LASIK eyes than in the normal eyes. The Bland and Altman method for assessing clinical agreement between 2 instruments showed that in 95% of cases, the CCT measurements with both instruments would be within 65 m in normal eyes and 150 m in post-LASIK eyes. Conclusion: Central corneal thickness measurements were, on average, 28 m higher with the Orbscan than with the ultrasound pachymeter in normal eyes and 13 m lower in post-LASIK eyes. The degree of variability within each group indicated that these 2 techniques are not clinically comparable, precluding interchangeable use of their data in planning or assessing corneal surgery. J Cataract Refract Surg 2001; 27:1823–1828 © 2001 ASCRS and ESCRS
© 2001 ASCRS and ESCRS Published by Elsevier Science Inc.
0886-3350/01/$–see front matter PII S0886-3350(01)01089-6
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
T
he widespread use of ultrasound (US) in medicine has revolutionized many aspects of clinical practice. Ultrasound has been used for over 30 years to assess and quantify ocular structures.1,2 To date, clinical measurement of central corneal thickness (CCT) has primarily been by ultrasonography or by optical methods. Ultrasound techniques provide the current gold standard for measuring corneal thickness,3 but they have the disadvantages of the need for topical anesthesia and contact of a probe with the cornea. Optical pachymetry avoids contact but is considered less accurate (10 to 100 m nonlinear error).4 Computerized slit-scanning elevation topographical devices, such as the Orbscan (Orbtek Inc.), have become more widely available in the clinical setting. They give not only advanced topographical information5,6 but also wide-field pachymetry (corneal thickness measurements across the entire central and midperipheral cornea) without direct corneal contact. Accurate corneal thickness measurements are important in managing corneal diseases such as keratoconus, in which there is progressive corneal apical thinning,7 and in corneal refractive surgery in which knowledge of the depth of residual corneal tissue after ablation is critical.8 Access to accurate noninvasive pachymetry techniques would provide invaluable information, particularly in the assessment of refractive surgery patients in whom the risk of corneal epithelial damage or flap displacement from contact with a probe may, on occasion, preclude the measurement of CCT by Accepted for publication May 8, 2001. From the Department of Ophthalmology, University of Dundee, Ninewells Hospital and Medical School, Dundee, Scotland (Chakrabarti, Craig, Brahma, Malik, McGhee); the Discipline of Ophthalmology, University of Auckland, Auckland, New Zealand (Craig, Malik, McGhee); and the Department of Ophthalmology, Manchester Royal Infirmary, Manchester, England (Brahma). Supported in part by an unrestricted grant from the Speed Pollock Memorial Research Fund and the Dundee Teaching Hospitals NHS Anonymous Trust, Dundee, Scotland. None of the authors has a financial or proprietary interest in any product mentioned. Mr. G. Cairns provided mathematical advice for the Orbscan slit-scanning topographical analysis. Reprint requests to Dr. Jennifer P. Craig, Discipline of Ophthalmology, University of Auckland, Private Bag 92019, Auckland 1001, New Zealand. 1824
contact US. This prospective study was undertaken to compare corneal thickness measurements with the Orbscan and with US pachymetry in normal eyes and in eyes that had had excimer laser in situ keratomileusis (LASIK) for myopia.
Patients and Methods Group 1: Normal Subjects One hundred one eyes from 59 normal subjects with a mean age of 37 years ⫾ 12 (SD) were included in the study. Ethical approval for the study was obtained from the local medical ethics committee, and all subjects provided informed consent for participation in the study. All subjects completed a questionnaire to exclude history of ocular disease, family history of ocular disease, or systemic disease with ocular manifestations. One investigator (H.S.C.) performed an ophthalmic examination of all eyes to exclude corneal pathology. Eyes that did not have a best spectacle-corrected visual acuity of 6/6 or better were not included in the study. No subjects had worn contact lenses for the 3 months prior to the study. The mean refractive error (spherical equivalent) was ⫺1.29 ⫾ 2.01 diopters (D) (range ⫹3.00 to ⫺8.50 D); 40.6% of eyes were emmetropic (41 eyes within ⫾0.50 D). Group 2: Post-LASIK Patients Another 30 eyes of 21 patients who had LASIK for myopia were recruited. In this group, LASIK had been performed by a single surgeon (C.N.McG.) at a single center using the Chiron Technolas 217 scanning-spot excimer laser and Automated Corneal Shaper威 microkeratome. The interval between LASIK treatment and pachymetric assessment was at least 1 month (range 1 to 12 months). Corneal topography and wide-field pachymetry were assessed by observers experienced in the use of the Orbscan. The illuminated optical section of the Orbscan scanned the anterior and posterior surfaces of the cornea simultaneously, and computer algorithms generated anterior and posterior elevation maps, depicting, by convention, areas above the best-fit sphere for each map as warm colors (oranges and reds) and areas below as cool colors (blues and purples). The distance between these anterior and posterior surfaces was calculated by Orbscan proprietary software to provide a single corneal
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
thickness measurement. Each eye had a single scanning procedure. Immediately afterward, CCT was measured by US pachymetry (Micropach, Sonomed Inc.). Corneal anesthesia was achieved with 1 drop of preservative-free oxybuprocaine 0.4% applied immediately prior to measurement. Patients were required to observe a fixation target during measurement to ensure optimal alignment and centration of the ultrasonic probe. Consistent with the authors’ standard clinical practice, 5 measurements of each eye were taken and the mean was calculated. To confirm that the results of the 2 instruments were comparable clinically, a statistical analysis method described by Bland and Altman9 was performed. This compared the difference in measured values between the 2 instruments with the mean value.
Results Group 1: Normal Subjects The mean CCT by US pachymetry in the 101 eyes was 538.0 ⫾ 36.7 m and by Orbscan pachymetry, 566.6 ⫾ 40.7 m. Pachymetric data from both instruments were not significantly different from those in a normal distribution (Kolmogarov-Smirnoff test, P ⬎ .05); therefore, the data were deemed suitable for parametric statistical analysis. Central corneal thickness values were statistically significantly different between the 2 tests (P ⬍ .001, Student paired t test). The Orbscan measurement was, on average, 5.3% greater than that of the US pachymetry (Figure 1). Plotting the Orbscan results against the US results and observing them relative to the line of perfect agreement (y ⫽ x), most points fall above the line; statistically, the correlation between the 2 sets of measurements is good (r2 ⫽ 0.827; P ⬍ .001). Figure 2 depicts the difference between readings against the mean of the 2 readings for each subject and the 95% confidence intervals (1.96 ⫻ standard deviation). The 2 measurements obtained by these 2 techniques differed by between ⫺5 and 60 m in 95% of cases. Neither topographic power of the cornea (mean apical K) nor refractive error (spherical equivalent) at the spectacle plane were found to influence CCT as measured with the Orbscan (r2 ⫽ 0.077, P ⬎ .05 and r2 ⫽ 0.012, P ⬎ .05, respectively) or by
Figure 1. (Chakrabarti) The CCT measured by Orbscan pachymetry plotted against that measured by US pachymetry in normal eyes (n ⫽ 101). The best-fit line is designated by the dotted line and the line of equivalence (y ⫽ x), by the solid line.
Figure 2. (Chakrabarti) The difference in CCT measurements between Orbscan pachymetry and US pachymetry in normal eyes (n ⫽ 101) plotted against the mean CCT measured with the 2 techniques. The mean difference is represented by a solid line and the 95% confidence limits, by the dashed lines.
ultrasonography (r2 ⫽ 0.007, P ⬎ .05 and r2 ⫽ 0.012, P ⬎ .05, respectively). Group 2: Post-LASIK Patients Pre-LASIK CCT was identified, retrospectively, for the refractive surgery corneas to determine whether the
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
1825
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
sample was representative of the same population as the normal subjects in Group 1. The mean CCT by US pachymetry (n ⫽ 30) was 527.1 ⫾ 36.2 m and by Orbscan pachymetry (n ⫽ 19), 563.1 ⫾ 41.3 m. Preoperatively, neither differed statistically from the respective means of the 101 control eyes (Student unpaired t test, P ⬎ .05 in both cases). The CCT measurements obtained by Orbscan pachymetry were significantly higher than those obtained by US pachymetry (Student paired t test, P ⬍ .001). In the prospective limb of this post-LASIK study, the mean US pachymetry CCT measurement was 475.3 ⫾ 50.3 m and the mean Orbscan pachymetry measurement, 461.9 ⫾ 74.2 m. The Orbscan measurement was therefore, on average, 2.9% lower than that of the US pachymetry, a statistically significant difference (Student paired t test, P ⬍ .001). The correlation between the results of the 2 techniques was poorer than that found in the normal subjects, although still statistically significant (r2 ⫽ 0.774; P ⬍ .001). Figure 3 shows the US results plotted against the Orbscan results, together with the line of equivalence and linear regression line. Relative to US pachymetry, the Orbscan had a tendency to underestimate CCT in thinner corneas and overestimate corneal thickness in thicker corneas in the post-LASIK group. In this in-
stance, plotting the CCT differences against the means in each eye shows that the 95% confidence interval was 150 m. The variability was therefore substantially greater in the post-LASIK eyes than in the Group 1 eyes, although the mean difference of zero was within the confidence limits (⫺88 to 62 m) (Figure 4).
Figure 3. (Chakrabarti) The CCT measured by Orbscan pachym-
Figure 4. (Chakrabarti) The difference in CCT measurements between Orbscan pachymetry and US pachymetry in post-LASIK eyes (n ⫽ 30), plotted against the mean CCT measured with the 2 techniques. The mean difference is represented by a solid line and the 95% confidence limits, by the dashed lines.
etry plotted against that measured by US pachymetry in post-LASIK eyes (n ⫽ 30). The best-fit line is designated by the dotted line and the line of equivalence (y ⫽ x), by the solid line.
1826
Discussion The Orbscan slit-scanning elevation topographer provides information on anterior corneal topography and pachymetry and additional information on posterior corneal curvature and keratometric power. Theoretically, it could replace existing computerized videokeratography and US pachymetry by a single device if it proved to be accurate and reliable. Repeatability (intrapatient variability) has been reported as ⫾8.42 m by Yaylali and coauthors10 and ⫾9.08 m by Lattimore and coauthors11 for pachymetry in normal human eyes. This prospective study, assessing a large number of eyes and conducted by experienced observers, found that CCT measurements were, on average, 28.0 m (5.3%) higher with a slit-scanning technique (Orbscan) than with US pachymetry in normal eyes and 13.4 m (2.9%) lower in post-LASIK eyes.
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
The Orbscan’s clinical accuracy in comparison to US pachymetry has not been well established. Liu and coauthors6 developed a classification system for corneal pachymetry maps but did not compare their individual measurements with US pachymetry. Yaylali and coauthors10 compared the techniques and found that, on average, the Orbscan overestimates the pachymetry value by 5.15%. They state that regression analysis might be used to determine equations that would enable direct conversion of measurements from 1 technique to the other. Indeed, the proprietary software available with the Orbscan allows a linear correction factor (the default of which is 8% of the thickness measurement) to be included at the clinician’s request and modified to suit clinician preference to compensate for this discrepancy. However, although in the present study pachymetry measurements were found to be, on average, 5.3% higher with the Orbscan, which compares well with the published results, the range was between ⫺2.6% and ⫹12.8%. This indicates that in some cases the measurements were much greater and in some they were less than those measured by US pachymetry. Therefore, no simple relationship was found between the measurements of the 2 instruments, and a simple linear transformation does not appear sufficient to allow direct comparison of the data. A definitive reason for the greater corneal thickness typically reported by the Orbscan has not been found, and the error tends to be attributed to the Orbscan measurements since US is considered the gold standard. However, given the difference in the nature of measurements between the 2 techniques, it may be possible that despite being a gold standard, US actually underestimates corneal thickness through its contact with the ocular surface and potential compression of the corneal tissue during measurement. However, the reversal of this phenomenon of overmeasurement when assessing post-LASIK corneas in the current study, in which the same experienced observers performed all measurements, would tend to refute this argument. The high degree of variability in differences between the techniques in individual subjects, which is possibly of more concern than the absolute difference, is highlighted by comparing the 2 techniques in the manner described by Bland and Altman.9 This shows that for 95% of normal corneas, the Orbscan CCT measurement would be no more than 60 m greater than that
determined by US pachymetry. In a clinical setting in which pachymetric assessment is essential before refractive surgery, this relatively poor level of agreement is of concern. In treating patients by LASIK, it has been suggested that a minimum of 250 m of tissue remain beneath the lamellar flap after stromal ablation to ensure maintenance of corneal structural integrity. If treatment were based on measurements by Orbscan, the surgeon would have to leave considerably more tissue untouched to be certain that at least 250 m remained. This is an important issue when treating eyes with higher myopic refractive errors, since the ablation depth is proportional to the magnitude of the attempted refractive correction. Comparing measurements obtained by US pachymetry and slit-scanning techniques in post-LASIK patients was particularly interesting. It demonstrated more extensive variability and a reversal of the trend to overmeasurement identified in normal eyes. In the postLASIK eyes, the CCT appeared thinner when measured by the Orbscan than by ultrasonography. This may be due to the change in corneal shape following LASIK. The natural cornea is generally prolate, with a more steeply curved central portion and a relatively flatter periphery. After excimer laser photorefractive surgery (photorefractive keratectomy or LASIK) for myopia, the curvature of the cornea is altered and tends to become oblate—relatively flatter in the center than in the midperiphery. The reconstruction algorithms that construct the corneal elevation data in the Orbscan, using loworder polynomials, effectively smooth small irregularities in the corneal surface.12 While this smoothing may be ideal in the normal, healthy corneal surface, it may be insufficiently sensitive to reflect the changes in the corneal surface that occur with refractive surgery. Unfortunately, the proprietary algorithms used by the manufacturers of the Orbscan to calculate pachymetry are not available for inspection, but from the results of this study, it is reasonable to suggest that they may require further refinement to take account of the corneal shape after refractive surgery. An alternative hypothesis is that this variability may be the result of changes in the stromal refractive index associated with varying degrees of corneal hydration in response to the surgery. Indeed, it is believed that such refractive index changes may contribute to the discrepancy observed between the change in refraction and the
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
1827
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
changes in central ocular surface power after LASIK.13 However, such a reduction in the refractive index after LASIK, as postulated by Patel and coauthors,13 would result in a slight overestimation of corneal thickness by the Orbscan, which uses 1.376 as its default corneal refractive index in all calculations. Clearly, a reduction in the corneal refractive index cannot account for the disparity between the techniques, although the effect that a change in corneal density and structure may have on US measurements must also be considered. The velocity of the propagation of sound, unlike light, tends to be reduced in less dense media.14 Thus, a pachymeter that has a programmed velocity of propogation will overestimate the thickness of a structure that is of a lower density than expected, since the US radiation will travel more slowly and the echo will take longer to reach the detector. This overestimation in corneal thickness by US, in the presence of a less dense medium, may exceed the overestimation by the Orbscan and contribute to the difference in findings between normal and post-LASIK eyes. In conclusion, the results of this study suggest that rather than using the Orbscan autonomously to assess corneal thickness, it should be used in conjunction with US pachymetry, particularly when measurement of CCT is critical as in corneal refractive surgery. The variability noted in the Orbscan CCT measurements, relative to the current gold standard of US pachymetry, suggests the need to refine the reconstruction algorithms, particularly for assessing corneas with altered surface and/or compositional characteristics.
References
4.
5.
6.
7. 8.
9.
10.
11.
12.
13.
1. Murphy GE, Murphy CG. Comparison of efficacy of longest, average, and shortest axial length measurements with a solid-tip ultrasound probe in predicting intraocular lens power. J Cataract Refract Surg 1993; 19:644 – 645 2. Laroche D, Ishikawa H, Greenfield D, et al. Ultrasound
1828
3.
14.
biomicroscopic localization and evaluation of intraocular foreign bodies. Acta Ophthalmol Scand 1998; 76:491– 495 Salz JJ, Azen SP, Bernstein J, et al. Evaluation and comparison of sources of variability in the measurement of corneal thickness with ultrasonic and optical pachymeters. Ophthalmic Surg 1983; 14:750 –754 Olsen T, Nielsen CB, Ehlers N. On the optical measurement of a corneal thickness. I. Optical principle and sources of error. Acta Ophthalmol 1980; 58:760 –766 Oshika T, Tomidokoro A, Tsuji H. Regular and irregular refractive powers of the front and back surfaces of the cornea. Exp Eye Res 1998; 67:443– 447 Liu Z, Huang AJ, Pflugfelder SC. Evaluation of corneal thickness and topography in normal eyes using the Orbscan corneal topography system. Br J Ophthalmol 1999; 83:774 –778 Rabinowitz YS. Keratoconus. Surv Ophthalmol 1998; 42:297–319 Pe´rez-Santonia JJ, Ayala MJ, Sakla HF, et al. Retreatment after laser in situ keratomileusis. Ophthalmology 1999; 106:21–28; discussion by ME Whitten, 28 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307–310 Yaylali V, Kaufman SC, Thompson HW. Corneal thickness measurements with the Orbscan Topography System and ultrasonic pachymetry. J Cataract Refract Surg 1997; 23:1345–1350 Lattimore MR Jr, Kaupp S, Schallhorn S, Lewis R. Orbscan pachymetry; implications of a repeated measures and diurnal variation analysis. Ophthalmology 1999; 106:977–981 Snook RK. Pachymetry and true topography using the Orbscan system. In: Gills JP, Sanders DR, eds, Corneal Topography; the State of the Art. Thorofare, NJ, Slack Inc, 1995; 89 –103 Patel S, Alio´ JL, Pe´rez-Santonja JJ. A model to explain the difference between changes in refraction and central ocular surface power after laser in situ keratomileusis. J Refract Surg 2000; 16:330 –335 Kremkau FW. Diagnostic Ultrasound; Principles, Instrumentation, and Exercises, 2nd ed. London, Grune and Stratton, 1984; 5–31
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
© 2001 ASCRS and ESCRS Published by Elsevier Science Inc.
0886-3350/01/$–see front matter PII S0886-3350(01)01089-6
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
T
he widespread use of ultrasound (US) in medicine has revolutionized many aspects of clinical practice. Ultrasound has been used for over 30 years to assess and quantify ocular structures.1,2 To date, clinical measurement of central corneal thickness (CCT) has primarily been by ultrasonography or by optical methods. Ultrasound techniques provide the current gold standard for measuring corneal thickness,3 but they have the disadvantages of the need for topical anesthesia and contact of a probe with the cornea. Optical pachymetry avoids contact but is considered less accurate (10 to 100 m nonlinear error).4 Computerized slit-scanning elevation topographical devices, such as the Orbscan (Orbtek Inc.), have become more widely available in the clinical setting. They give not only advanced topographical information5,6 but also wide-field pachymetry (corneal thickness measurements across the entire central and midperipheral cornea) without direct corneal contact. Accurate corneal thickness measurements are important in managing corneal diseases such as keratoconus, in which there is progressive corneal apical thinning,7 and in corneal refractive surgery in which knowledge of the depth of residual corneal tissue after ablation is critical.8 Access to accurate noninvasive pachymetry techniques would provide invaluable information, particularly in the assessment of refractive surgery patients in whom the risk of corneal epithelial damage or flap displacement from contact with a probe may, on occasion, preclude the measurement of CCT by Accepted for publication May 8, 2001. From the Department of Ophthalmology, University of Dundee, Ninewells Hospital and Medical School, Dundee, Scotland (Chakrabarti, Craig, Brahma, Malik, McGhee); the Discipline of Ophthalmology, University of Auckland, Auckland, New Zealand (Craig, Malik, McGhee); and the Department of Ophthalmology, Manchester Royal Infirmary, Manchester, England (Brahma). Supported in part by an unrestricted grant from the Speed Pollock Memorial Research Fund and the Dundee Teaching Hospitals NHS Anonymous Trust, Dundee, Scotland. None of the authors has a financial or proprietary interest in any product mentioned. Mr. G. Cairns provided mathematical advice for the Orbscan slit-scanning topographical analysis. Reprint requests to Dr. Jennifer P. Craig, Discipline of Ophthalmology, University of Auckland, Private Bag 92019, Auckland 1001, New Zealand. 1824
contact US. This prospective study was undertaken to compare corneal thickness measurements with the Orbscan and with US pachymetry in normal eyes and in eyes that had had excimer laser in situ keratomileusis (LASIK) for myopia.
Patients and Methods Group 1: Normal Subjects One hundred one eyes from 59 normal subjects with a mean age of 37 years ⫾ 12 (SD) were included in the study. Ethical approval for the study was obtained from the local medical ethics committee, and all subjects provided informed consent for participation in the study. All subjects completed a questionnaire to exclude history of ocular disease, family history of ocular disease, or systemic disease with ocular manifestations. One investigator (H.S.C.) performed an ophthalmic examination of all eyes to exclude corneal pathology. Eyes that did not have a best spectacle-corrected visual acuity of 6/6 or better were not included in the study. No subjects had worn contact lenses for the 3 months prior to the study. The mean refractive error (spherical equivalent) was ⫺1.29 ⫾ 2.01 diopters (D) (range ⫹3.00 to ⫺8.50 D); 40.6% of eyes were emmetropic (41 eyes within ⫾0.50 D). Group 2: Post-LASIK Patients Another 30 eyes of 21 patients who had LASIK for myopia were recruited. In this group, LASIK had been performed by a single surgeon (C.N.McG.) at a single center using the Chiron Technolas 217 scanning-spot excimer laser and Automated Corneal Shaper威 microkeratome. The interval between LASIK treatment and pachymetric assessment was at least 1 month (range 1 to 12 months). Corneal topography and wide-field pachymetry were assessed by observers experienced in the use of the Orbscan. The illuminated optical section of the Orbscan scanned the anterior and posterior surfaces of the cornea simultaneously, and computer algorithms generated anterior and posterior elevation maps, depicting, by convention, areas above the best-fit sphere for each map as warm colors (oranges and reds) and areas below as cool colors (blues and purples). The distance between these anterior and posterior surfaces was calculated by Orbscan proprietary software to provide a single corneal
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
thickness measurement. Each eye had a single scanning procedure. Immediately afterward, CCT was measured by US pachymetry (Micropach, Sonomed Inc.). Corneal anesthesia was achieved with 1 drop of preservative-free oxybuprocaine 0.4% applied immediately prior to measurement. Patients were required to observe a fixation target during measurement to ensure optimal alignment and centration of the ultrasonic probe. Consistent with the authors’ standard clinical practice, 5 measurements of each eye were taken and the mean was calculated. To confirm that the results of the 2 instruments were comparable clinically, a statistical analysis method described by Bland and Altman9 was performed. This compared the difference in measured values between the 2 instruments with the mean value.
Results Group 1: Normal Subjects The mean CCT by US pachymetry in the 101 eyes was 538.0 ⫾ 36.7 m and by Orbscan pachymetry, 566.6 ⫾ 40.7 m. Pachymetric data from both instruments were not significantly different from those in a normal distribution (Kolmogarov-Smirnoff test, P ⬎ .05); therefore, the data were deemed suitable for parametric statistical analysis. Central corneal thickness values were statistically significantly different between the 2 tests (P ⬍ .001, Student paired t test). The Orbscan measurement was, on average, 5.3% greater than that of the US pachymetry (Figure 1). Plotting the Orbscan results against the US results and observing them relative to the line of perfect agreement (y ⫽ x), most points fall above the line; statistically, the correlation between the 2 sets of measurements is good (r2 ⫽ 0.827; P ⬍ .001). Figure 2 depicts the difference between readings against the mean of the 2 readings for each subject and the 95% confidence intervals (1.96 ⫻ standard deviation). The 2 measurements obtained by these 2 techniques differed by between ⫺5 and 60 m in 95% of cases. Neither topographic power of the cornea (mean apical K) nor refractive error (spherical equivalent) at the spectacle plane were found to influence CCT as measured with the Orbscan (r2 ⫽ 0.077, P ⬎ .05 and r2 ⫽ 0.012, P ⬎ .05, respectively) or by
Figure 1. (Chakrabarti) The CCT measured by Orbscan pachymetry plotted against that measured by US pachymetry in normal eyes (n ⫽ 101). The best-fit line is designated by the dotted line and the line of equivalence (y ⫽ x), by the solid line.
Figure 2. (Chakrabarti) The difference in CCT measurements between Orbscan pachymetry and US pachymetry in normal eyes (n ⫽ 101) plotted against the mean CCT measured with the 2 techniques. The mean difference is represented by a solid line and the 95% confidence limits, by the dashed lines.
ultrasonography (r2 ⫽ 0.007, P ⬎ .05 and r2 ⫽ 0.012, P ⬎ .05, respectively). Group 2: Post-LASIK Patients Pre-LASIK CCT was identified, retrospectively, for the refractive surgery corneas to determine whether the
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
1825
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
sample was representative of the same population as the normal subjects in Group 1. The mean CCT by US pachymetry (n ⫽ 30) was 527.1 ⫾ 36.2 m and by Orbscan pachymetry (n ⫽ 19), 563.1 ⫾ 41.3 m. Preoperatively, neither differed statistically from the respective means of the 101 control eyes (Student unpaired t test, P ⬎ .05 in both cases). The CCT measurements obtained by Orbscan pachymetry were significantly higher than those obtained by US pachymetry (Student paired t test, P ⬍ .001). In the prospective limb of this post-LASIK study, the mean US pachymetry CCT measurement was 475.3 ⫾ 50.3 m and the mean Orbscan pachymetry measurement, 461.9 ⫾ 74.2 m. The Orbscan measurement was therefore, on average, 2.9% lower than that of the US pachymetry, a statistically significant difference (Student paired t test, P ⬍ .001). The correlation between the results of the 2 techniques was poorer than that found in the normal subjects, although still statistically significant (r2 ⫽ 0.774; P ⬍ .001). Figure 3 shows the US results plotted against the Orbscan results, together with the line of equivalence and linear regression line. Relative to US pachymetry, the Orbscan had a tendency to underestimate CCT in thinner corneas and overestimate corneal thickness in thicker corneas in the post-LASIK group. In this in-
stance, plotting the CCT differences against the means in each eye shows that the 95% confidence interval was 150 m. The variability was therefore substantially greater in the post-LASIK eyes than in the Group 1 eyes, although the mean difference of zero was within the confidence limits (⫺88 to 62 m) (Figure 4).
Figure 3. (Chakrabarti) The CCT measured by Orbscan pachym-
Figure 4. (Chakrabarti) The difference in CCT measurements between Orbscan pachymetry and US pachymetry in post-LASIK eyes (n ⫽ 30), plotted against the mean CCT measured with the 2 techniques. The mean difference is represented by a solid line and the 95% confidence limits, by the dashed lines.
etry plotted against that measured by US pachymetry in post-LASIK eyes (n ⫽ 30). The best-fit line is designated by the dotted line and the line of equivalence (y ⫽ x), by the solid line.
1826
Discussion The Orbscan slit-scanning elevation topographer provides information on anterior corneal topography and pachymetry and additional information on posterior corneal curvature and keratometric power. Theoretically, it could replace existing computerized videokeratography and US pachymetry by a single device if it proved to be accurate and reliable. Repeatability (intrapatient variability) has been reported as ⫾8.42 m by Yaylali and coauthors10 and ⫾9.08 m by Lattimore and coauthors11 for pachymetry in normal human eyes. This prospective study, assessing a large number of eyes and conducted by experienced observers, found that CCT measurements were, on average, 28.0 m (5.3%) higher with a slit-scanning technique (Orbscan) than with US pachymetry in normal eyes and 13.4 m (2.9%) lower in post-LASIK eyes.
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
The Orbscan’s clinical accuracy in comparison to US pachymetry has not been well established. Liu and coauthors6 developed a classification system for corneal pachymetry maps but did not compare their individual measurements with US pachymetry. Yaylali and coauthors10 compared the techniques and found that, on average, the Orbscan overestimates the pachymetry value by 5.15%. They state that regression analysis might be used to determine equations that would enable direct conversion of measurements from 1 technique to the other. Indeed, the proprietary software available with the Orbscan allows a linear correction factor (the default of which is 8% of the thickness measurement) to be included at the clinician’s request and modified to suit clinician preference to compensate for this discrepancy. However, although in the present study pachymetry measurements were found to be, on average, 5.3% higher with the Orbscan, which compares well with the published results, the range was between ⫺2.6% and ⫹12.8%. This indicates that in some cases the measurements were much greater and in some they were less than those measured by US pachymetry. Therefore, no simple relationship was found between the measurements of the 2 instruments, and a simple linear transformation does not appear sufficient to allow direct comparison of the data. A definitive reason for the greater corneal thickness typically reported by the Orbscan has not been found, and the error tends to be attributed to the Orbscan measurements since US is considered the gold standard. However, given the difference in the nature of measurements between the 2 techniques, it may be possible that despite being a gold standard, US actually underestimates corneal thickness through its contact with the ocular surface and potential compression of the corneal tissue during measurement. However, the reversal of this phenomenon of overmeasurement when assessing post-LASIK corneas in the current study, in which the same experienced observers performed all measurements, would tend to refute this argument. The high degree of variability in differences between the techniques in individual subjects, which is possibly of more concern than the absolute difference, is highlighted by comparing the 2 techniques in the manner described by Bland and Altman.9 This shows that for 95% of normal corneas, the Orbscan CCT measurement would be no more than 60 m greater than that
determined by US pachymetry. In a clinical setting in which pachymetric assessment is essential before refractive surgery, this relatively poor level of agreement is of concern. In treating patients by LASIK, it has been suggested that a minimum of 250 m of tissue remain beneath the lamellar flap after stromal ablation to ensure maintenance of corneal structural integrity. If treatment were based on measurements by Orbscan, the surgeon would have to leave considerably more tissue untouched to be certain that at least 250 m remained. This is an important issue when treating eyes with higher myopic refractive errors, since the ablation depth is proportional to the magnitude of the attempted refractive correction. Comparing measurements obtained by US pachymetry and slit-scanning techniques in post-LASIK patients was particularly interesting. It demonstrated more extensive variability and a reversal of the trend to overmeasurement identified in normal eyes. In the postLASIK eyes, the CCT appeared thinner when measured by the Orbscan than by ultrasonography. This may be due to the change in corneal shape following LASIK. The natural cornea is generally prolate, with a more steeply curved central portion and a relatively flatter periphery. After excimer laser photorefractive surgery (photorefractive keratectomy or LASIK) for myopia, the curvature of the cornea is altered and tends to become oblate—relatively flatter in the center than in the midperiphery. The reconstruction algorithms that construct the corneal elevation data in the Orbscan, using loworder polynomials, effectively smooth small irregularities in the corneal surface.12 While this smoothing may be ideal in the normal, healthy corneal surface, it may be insufficiently sensitive to reflect the changes in the corneal surface that occur with refractive surgery. Unfortunately, the proprietary algorithms used by the manufacturers of the Orbscan to calculate pachymetry are not available for inspection, but from the results of this study, it is reasonable to suggest that they may require further refinement to take account of the corneal shape after refractive surgery. An alternative hypothesis is that this variability may be the result of changes in the stromal refractive index associated with varying degrees of corneal hydration in response to the surgery. Indeed, it is believed that such refractive index changes may contribute to the discrepancy observed between the change in refraction and the
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001
1827
CORNEAL THICKNESS MEASUREMENTS WITH US AND ORBSCAN
changes in central ocular surface power after LASIK.13 However, such a reduction in the refractive index after LASIK, as postulated by Patel and coauthors,13 would result in a slight overestimation of corneal thickness by the Orbscan, which uses 1.376 as its default corneal refractive index in all calculations. Clearly, a reduction in the corneal refractive index cannot account for the disparity between the techniques, although the effect that a change in corneal density and structure may have on US measurements must also be considered. The velocity of the propagation of sound, unlike light, tends to be reduced in less dense media.14 Thus, a pachymeter that has a programmed velocity of propogation will overestimate the thickness of a structure that is of a lower density than expected, since the US radiation will travel more slowly and the echo will take longer to reach the detector. This overestimation in corneal thickness by US, in the presence of a less dense medium, may exceed the overestimation by the Orbscan and contribute to the difference in findings between normal and post-LASIK eyes. In conclusion, the results of this study suggest that rather than using the Orbscan autonomously to assess corneal thickness, it should be used in conjunction with US pachymetry, particularly when measurement of CCT is critical as in corneal refractive surgery. The variability noted in the Orbscan CCT measurements, relative to the current gold standard of US pachymetry, suggests the need to refine the reconstruction algorithms, particularly for assessing corneas with altered surface and/or compositional characteristics.
References
4.
5.
6.
7. 8.
9.
10.
11.
12.
13.
1. Murphy GE, Murphy CG. Comparison of efficacy of longest, average, and shortest axial length measurements with a solid-tip ultrasound probe in predicting intraocular lens power. J Cataract Refract Surg 1993; 19:644 – 645 2. Laroche D, Ishikawa H, Greenfield D, et al. Ultrasound
1828
3.
14.
biomicroscopic localization and evaluation of intraocular foreign bodies. Acta Ophthalmol Scand 1998; 76:491– 495 Salz JJ, Azen SP, Bernstein J, et al. Evaluation and comparison of sources of variability in the measurement of corneal thickness with ultrasonic and optical pachymeters. Ophthalmic Surg 1983; 14:750 –754 Olsen T, Nielsen CB, Ehlers N. On the optical measurement of a corneal thickness. I. Optical principle and sources of error. Acta Ophthalmol 1980; 58:760 –766 Oshika T, Tomidokoro A, Tsuji H. Regular and irregular refractive powers of the front and back surfaces of the cornea. Exp Eye Res 1998; 67:443– 447 Liu Z, Huang AJ, Pflugfelder SC. Evaluation of corneal thickness and topography in normal eyes using the Orbscan corneal topography system. Br J Ophthalmol 1999; 83:774 –778 Rabinowitz YS. Keratoconus. Surv Ophthalmol 1998; 42:297–319 Pe´rez-Santonia JJ, Ayala MJ, Sakla HF, et al. Retreatment after laser in situ keratomileusis. Ophthalmology 1999; 106:21–28; discussion by ME Whitten, 28 Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307–310 Yaylali V, Kaufman SC, Thompson HW. Corneal thickness measurements with the Orbscan Topography System and ultrasonic pachymetry. J Cataract Refract Surg 1997; 23:1345–1350 Lattimore MR Jr, Kaupp S, Schallhorn S, Lewis R. Orbscan pachymetry; implications of a repeated measures and diurnal variation analysis. Ophthalmology 1999; 106:977–981 Snook RK. Pachymetry and true topography using the Orbscan system. In: Gills JP, Sanders DR, eds, Corneal Topography; the State of the Art. Thorofare, NJ, Slack Inc, 1995; 89 –103 Patel S, Alio´ JL, Pe´rez-Santonja JJ. A model to explain the difference between changes in refraction and central ocular surface power after laser in situ keratomileusis. J Refract Surg 2000; 16:330 –335 Kremkau FW. Diagnostic Ultrasound; Principles, Instrumentation, and Exercises, 2nd ed. London, Grune and Stratton, 1984; 5–31
J CATARACT REFRACT SURG—VOL 27, NOVEMBER 2001