Keratoconus is a noninflammatory ectatic disease of the cornea. In recent years, the popularity of refractive surgery has made it more important to identify suspect and definite cases of keratoconus. Advances in technology have supported these efforts.1 Assessing the corneal curvature, corneal thickness, and corneal surface topography patterns and elevation readings are integral parts of a patient’s workup.2,3 Among various indices, keratometry has more applications in diagnosing and grading the disease, tracking disease progression, and planning treatment and patient follow-up.2,4,5 Keratoconus patients usually have periodic examinations to determine the progression of the disease or the effect of treatment after surgery.4,6,7 Therefore, the accuracy and precision of keratometry is very important in keratoconus patients. To rely on results of repeated examinations, acceptable device repeatability is needed.
The repeatability of keratometry measurements with different devices has been studied in normal eyes.8,9 In eyes with keratoconus, however, the cornea has an irregular surface and measurements might not be as valid and reliable. There are also a few studies on the repeatability of 1 or 2 devices in measuring some anterior segment indices in keratoconus patients. Szalai et al.10 found lower repeatability with 2 new devices in keratoconus patients than in normal cases. A study by McMahon et al.11 showed poor repeatability of 3 topography systems in keratoconus patients.
One limitation of previous repeatability studies of keratoconus patients is that they compared 2 or 3 devices in each study. Interindividual differences, such as those in the tear-film layer, can influence imaging results in certain situations. Thus, the repeatability of corneal power measurements of various devices is better studied in a single population.
To our knowledge, corneal power measurement repeatability with different measuring techniques has not been studied in keratoconus patients. Thus, we designed the present study to examine the repeatability of 5 devices that use 5 different techniques to measure the corneal power in a sample of keratoconus cases. In addition, because the corneal power varies by the grade and severity of keratoconus, we explored the level of repeatability with different grades of keratoconus.
Patients and methods
The case series covered the period from June to October 2013. Patients were selected from the keratoconus clinic of an eye hospital in Tehran, Iran. All patients who presented at the clinic over the previous 3 months and who were diagnosed with keratoconus using clinical and imaging examinations were approached. Patients provided informed consent after receiving an explanation of the research and study methodology. Exclusion criteria included a history of eye surgery and the use of contact lenses over the preceding month.
Imaging was performed in each patient after he or she enrolled in the study. Imaging was performed using 5 devices by the same experienced optometrist under standard conditions between 10 am and 3 pm. Three acquisitions were taken at 10-minute intervals with each device. Erroneous acquisitions were repeated after 5 minutes.
The first measurement of corneal power was taken with the Pentacam HR (software version 1.17r72, data management version 6.03r11, Oculus, Inc.), which is a high-resolution computerized corneal topographer that uses a rotating digital Scheimpflug camera. Data collected from 25 000 points are used to create a 3-dimensional image of the anterior segment. Display information includes elevation, curvature, and pachymetry maps as well as simulated keratometry (K) readings, namely the K flat, which is equivalent to the minimum K reading, and K steep, which is equivalent to the maximum K reading.12
After a 10-minute interval, measurements were taken with the Eyesys Placido topographer (EyeSys Vision), which captures video images of the Placido disk reflections on the corneal surface. To measure the corneal power, the deviation of the reflected Placido rings on the corneal surface is determined to compute the K values in the flat and steep meridians.13 The accuracy of this technique in measuring the central and peripheral radius of curvature is dependent on the shape of the cornea; measurement accuracy is reduced for aspheric surfaces that show sudden flattening, although this appears to be clinically acceptable.14
Next, Orbscan II (Bausch & Lomb Surgical) scanning-slit corneal topography was used. This system provides a complete analysis of the corneal surface. During the imaging process, 20 slit-light beams are projected onto the cornea from each side at an angle of 45 degrees as they scan the surface in 0.75 second, and images are captured from 9000 points. Simulated K readings are among the many indices measured or calculated by this topography machine.
The fourth device was the IOLMaster (Zeiss AG). This system is based on partial coherence interferometry (PCI) and uses image analysis to measure the corneal curvature and anterior chamber depth. For keratometry, 6 points of light are projected onto the central 3.0 mm of the cornea. Reflections from these points are recorded by the PCI device; the toroidal surface curvatures are calculated by measuring point separation.15
Finally, measurements were taken using the Javal manual keratometer (Haag-Streit), whose operation is based on the concept of a fixed image size and variable object size. To obtain minimum K and maximum K readings and their axes, 4 points are focused on the 3.0 to 4.0 mm central corneal zone; the operator subjectively aligns the mires. Although the device can be highly accurate for regular spherocylindrical surfaces, readings have limited reliability when testing irregular corneas.16
Collected data were analyzed using SPSS software (version 22, IBM SPSS). From each device, the displayed readings for minimum K and maximum K, which are from the flat axis and steep axis of the cornea, respectively, were recorded. Because these variables are the main indices used in clinical practice, their repeatability was examined using separate variables rather than their mean value. Repeatability of keratometry was studied in 3 patient groups. Group 1 comprised keratoconus patients with a maximum K value of less than 50.0 diopters (D). Group 2 comprised patients with a maximum K value between 50.0 D and 55.0 D. Group 3 comprised patients with a maximum K value of more than 55.0 D.
First, the mean and standard deviation (SD) of the 3 minimum K and maximum K readings in each group were determined with each device. Repeated-measures analysis of variance (ANOVA) was used to compare mean values. Considering the number of comparisons, the level of significance was reduced to 0.005 after Bonferroni correction as follows:
The effect of both eyes was controlled as a covariate in the repeated-measures analysis.
With G-power software (version 3.1, Franz Faul, Kiel University, Kiel, Germany), the power was assessed for each test using sample size, type 1 error, effect size, correlation between repeated measures, and number of measurements. The power for all repeated-measures ANOVA tests was calculated.17
The intrasession test–retest variability was also calculated. Then the within-subjects SD was determined and multiplied by 2.77 to calculate the repeatability index.18 The repeatability index is an indicator of repeated measurement error; the lower the index, the better the repeatability. To determine the coefficient of variation (CoV), the SD was divided by the mean value of the measurements and expressed as a percentage; a lower value is more desirable with this index as well. To examine the variance between repeated data, the intraclass correlation coefficient (ICC) was determined. An ICC value of 1.00 means there is no difference in the variance of repeated data, a value less than 0.75 indicates a weak correlation, a value between 0.75 and 0.90 shows a moderate correlation, and a value over 0.90 shows a strong correlation. To determine the agreement of Scheimpflug pachymetry, scanning-slit corneal topography, Placido topography, and PCI with the manual keratometer, Bland-Altman plots with 95% limits of agreement (LoA) were used. The horizontal and vertical axes indicate the mean value of a variable with 2 devices and the interdevice difference (bias) for the variable, respectively. The 95% LoA was calculated as
of the interdevice difference. All agreement analyses were done using maximum K readings.
The study evaluated 78 eyes of 45 keratoconus patients. The mean age of the 21 women and 24 men was 28.6 years ± 6.5 (SD). Twenty-seven eyes had a maximum K of less than 50.0 D. In 26 eyes, the maximum K was between 50.0 D and 55.0 D, and 25 eyes had a maximum K of more than 55.0 D. The difference in age or sex was not statistically significant between groups (P>.05).
In Group 1, repeated-measures ANOVA showed no statistically significant differences in minimum K readings (P=.291 to P=.701; the statistical power was 100% in all comparisons) or maximum K readings between devices (P=.400 to P=.987); the statistical power was 100% in all comparisons. Table 1 shows the mean and SD of the 3 measurements of minimum K and maximum K readings of the 5 devices as well as the repeatability index for these readings in the 3 groups. According to the ICC values (Table 2), the minimum K and maximum K readings were highly repeatable with all devices (0.949 to 0.996). Based on the SD and repeatability indices in this group, the minimum K readings were most repeatable when measured with Scheimpflug pachymetry, while with scanning-slit corneal topography, the minimum K readings were the least repeatable. This was also true for the maximum K readings.
In Group 2, repeated-measures ANOVA showed no statistically significant differences in minimum K readings (P=.312 to P=.595); the statistical power was 100% in all comparisons except for the Placido topographer and scanning-slit corneal topographer (76% and 93%, respectively). There were also no statistically significant differences in maximum K readings (P=.460 to P=.852); the statistical power was 100% in all comparisons except scanning-slit corneal topography (96.3%). In this group, ICC values ranged between 0.795 for Placido topography readings and 0.986 for Scheimpflug pachymetry minimum K measurements. The highest and lowest ICC values for maximum K readings were with the PCI device and scanning-slit corneal topographer, respectively. The repeatability index for minimum K was best with Scheimpflug pachymetry and worst with scanning-slit corneal topography (range 0.53 to 2.11). The same was true for the maximum K readings (range 0.60 to 1.92).
In Group 3, repeated-measures ANOVA did not find statistically significant differences in minimum K readings (P=.103 to P=.819); the statistical power was 100% in all comparisons except for the manual keratometer (93%). There were also no statistically significant differences in maximum K readings (P=.135 to P=.500); the statistical power was 100% in all comparisons except the for the Placido topographer (99.4%). In this group of patients, the highest and lowest ICCs in measuring minimum K were with Scheimpflug pachymetry and Placido topography, respectively. For maximum K, the highest and lowest values were from PCI and Placido topography measurements, respectively. The repeatability index showed that all devices performed significantly differently when compared with the results in the other 2 groups of patients. For minimum K, the repeatability index ranged between 1.66 with Scheimpflug pachymetry to 2.98 with Placido topography. For maximum K, the range was from 2.15 with PCI to 2.81 with manual keratometry.
Device Agreement with Manual Keratometer
Figures 1 to 3 show the agreement of the manual keratometer with the other 4 devices in measuring maximum K. These Bland-Altman plots also indicate the 95% LoA. In Group 1, the least bias (mean interdevice difference in measuring maximum K) from the manual keratometer was seen between the manual keratometer and the Scheimpflug pachymetry device. The highest bias was between the manual keratometer and the Placido topography device. Nonetheless, based on the SD of the bias (ie, the 95% LoA), the best and worst agreement with Javal was with Placido topography and Eyesys scanning-slit corneal topography, respectively. Figure 1 shows the Bland-Altman plot in Group 1.
In Group 2, the least bias was seen between the manual keratometer and the Scheimpflug pachymetry device. The highest bias was with the PCI device. In this group, the best and worst 95% LoA with the manual keratometer were with the Scheimpflug device and the PCI device, respectively. Figure 2 shows the Bland-Altman plot in Group 2.
Figure 3 shows the agreement between these devices using the Bland-Altman plots in Group 3. The least and highest bias from the manual keratometer in Group 3 was with Scheimpflug pachymetry and PCI, respectively. Similarly, the best and worst 95% LoA with the manual keratometer were with Scheimpflug pachymetry and PCI, respectively.
To our knowledge, this is the first study to examine the repeatability of 5 imaging devices in measuring the corneal power in keratoconic eyes. Another new aspect of the study was the comparison of the levels of repeatability in relation to the severity of keratoconus.
Javal manual keratometry is recognized as the gold standard method, and although previous studies have shown otherwise when measurements are made in keratoconic patients,16,19 we used it as the reference and determined the agreement of the other 4 devices with the Javal keratometer. Because there is no gold standard keratometry method for keratoconic eyes, results in this section of our study should be interpreted with caution.
Because a measurement’s repeatability is a requisite for its accuracy and reliability, the present study can help determine a gold standard keratometry method for keratoconic eyes. However, before discussing the results in this study, the difference in the measurement axes of this study’s instruments should be kept in mind. Because the Eyesys is a Placido-based system and relies on a reflected image, the Pentacam is an optical cross-section, and the Orbscan II is a combined Placido scanning-slit system, it is possible that the reference axis, where all axial curvature measurements are made, will be different. Therefore, this difference could be a reason for the difference in K readings between the different devices. Nevertheless, we compared each device with itself for repeatability.
As shown in Group 1 (maximum K <50.0 D), all 5 devices had ICCs greater than 0.9; these high numbers support the reliability of the devices. According to the repeatability index for minimum K and maximum K readings, the Pentacam device had the highest repeatability and Orbscan II device the lowest, albeit results with Javal keratometer were close to those of the Pentacam device in this group. Because the cornea is less irregular in cases of keratoconus with a maximum K reading of less than 50.0 D, the corneal curvature is still similar to that of the normal cornea. Thus, the Javal keratometer might produce reliable results.16 The repeatability index calculated for minimum K and maximum K readings was 0.360 and 0.395, respectively, with Scheimpflug pachymetry, while these values were 1.179 and 1.243, respectively, with scanning-slit corneal topography. Szalai et al.10 studied the repeatability of Pentacam Scheimpflug pachymetry and an anterior segment optical coherence tomography (AS-OCT) device in measuring keratometry in the flat and steep meridians and reported repeatability indices of 0.472 and 0.468, respectively, for Scheimpflug pachymetry measurements in normal eyes; these values were 2.084 and 1.560, respectively, in keratoconic eyes. Because of the range of the K readings, the sample of keratoconus patients in that study is not comparable to our first group of patients; nonetheless, the Pentacam’s repeatability index in their normal cases was close to, or even slightly better than, that in our first group. The repeatability index for measurements of normal eyes with the AS-OCT device used in their study was similar to that of our cases of mild keratoconus.10 The Pentacam Scheimpflug pachymetry repeatability CoVs reported in other studies of normal cases were close to the values in our cases of mild keratoconus.20,21 Other studies have shown excellent repeatability for the Scheimpflug system in normal cases. Crawford et al.8 reported acceptable K reading repeatability for Orbscan scanning-slit corneal topography and Pentacam Scheimpflug pachymetry in normal subjects. As for the other devices in our study, the next highest repeatability indices were seen with IOLMaster PCI device and the Eyesys Placido topography system, while Orbscan scanning-slit corneal topography had the worst repeatability index.
The repeatability of the K readings in Group 2 (maximum K 50.0 to 55.0 D) was similar to that in the first group. The best and worse repeatability indices were with Scheimpflug pachymetry and scanning-slit corneal topography, respectively, and in measuring maximum K in this group, only the manual keratometer had results close to the Scheimpflug pachymetry device. In this group of patients, the repeatability indices with all 5 devices were worse than in the first group. In addition to the repeatability index, which is an indicator of measurement error and intraobserver error in this group of patients, the ICCs for scanning-slit corneal topography and Placido topography were less than 0.9; this shows the average variance between repeated measurements and overall, the results point to decreased device repeatability as the maximum K value increases.22
Our observations in Group 3 were especially noteworthy. All repeatability indices were worse in this group than in the other 2 groups. The best repeatability was with Scheimpflug pachymetry maximum K readings and PCI minimum K readings. Nonetheless, as the repeatability index values indicate, the measurement error was very high in this group while the ICCs appeared to be acceptable. The main and new finding in the present study is that all 5 devices had low repeatability in cases of severe keratoconus. Szalai et al.10 also observed low repeatability with AS-OCT and Scheimpflug pachymetry measurements in keratoconic patients.
In another study of keratometry repeatability in cases of keratoconus, Montalbán et al.23 found high repeatability for the Sirius system (Costruzione Strumenti Oftalmici S.r.l.), which combines Scheimpflug photography and Placido disk technology. They also found acceptable repeatability for posterior corneal measurements. Some believe that repeatability of K measurements in keratoconus patients is not different from that in normal subjects,22,24 while others report weak repeatability of such measurements.11 Savini et al.22 found medium repeatability (CoV 4.9% and ICC 0.868) of posterior corneal measurements with the Sirius system.
In summary, in cases in our study with a maximum K of 55.0 D or less, the Pentacam Scheimpflug pachymeter had better repeatability than the other 4 devices, with the Javal manual keratometer being the next most repeatable. Scheimpflug imaging is a direct method of measuring the corneal power; this is a newer technique than manual keratometry and Placido disk topography.12 One of its advantages is that it measures more points from more areas of the cornea because of its non-central rotating camera. With this imaging technique, the posterior surface of the cornea can be measured as well.25 Manual keratometers and Placido disk topographers have an unmeasured central scotoma because of the central location of the eyepiece or camera. This scotoma can be as large as 3.2 mm in a 44.0 D cornea, and although smaller, the scotoma with Placido disk topographers can be 1.6 mm. Thus, these traditional devices are not capable of examining the central cornea and by measuring the peripheral mires, the corneal power might be overestimated. The IOLMaster PCI device and Orbscan II scanning-slit corneal topographer analyze fewer points of the corneal surface; thus, less information is acquired from the corneal surface than with other devices. In the third group of patients (maximum K >55.0 D), however, all devices had worse CoV repeatability indices and measurement error was significantly higher than in the other 2 groups. In cases of severe keratoconus, fewer points are analyzed because the cornea is more irregular. In each measurement, only a limited number are measured, especially in maximum K measurements. This could explain the weak repeatability in this group.
The present study has limitations. A potential limitation is using devices in a nonrandomized order. We avoided a random order because it is not clear whether measurements with 1 device can affect readings of the other. However, we believe this should not affect the overall results in the study. Also, because technicians were not blinded to the results, the repeatability of manual keratometry might have been overestimated.
What Was Known
- Repeatability of keratometry measurements of normal cases have been reported as acceptable.
What This Paper Adds
- In keratoconus, when the maximum K is greater than 55.0 D, all imaging systems had weak repeatability as a result of measurement error and unreliable K results.
1. Duffey RJ, Leaming D. Trends in refractive surgery in the United States. J Cataract Refract Surg
2. Rabinowitz YS. Keratoconus. Surv Ophthalmol. 42, 1998, p. 297-319, Available at: http://www.keratoconus.com/resources/Major+Review-Keratoconus.pdf
. Accessed January 22, 2105.
3. Muftuoglu O, Ayar O, Ozulken K, Ozyol E, Akıncı A. Posterior corneal elevation and back difference corneal elevation in diagnosing forme fruste keratoconus in the fellow eyes of unilateral keratoconus patients. J Cataract Refract Surg
4. Vega-Estrada A, Alió JL, Brenner LF, Burguera N. Outcomes of intrastromal corneal ring segments for treatment of keratoconus: five-year follow-up analysis. J Cataract Refract Surg
5. Sloot F, Soeters N, van der Valk R, Tahzib NG. Effective corneal collagen crosslinking in advanced cases of progressive keratoconus. J Cataract Refract Surg
6. Daxer A, Mahmoud H, Venkateswaran RS. Intracorneal continuous ring implantation for keratoconus: one-year follow-up. J Cataract Refract Surg
7. Torquetti L, Berbel RF, Ferrara P. Long-term follow-up of intrastromal corneal ring segments in keratoconus. J Cataract Refract Surg
8. Crawford AZ, Patel DV, McGhee CNJ. Comparison and repeatability of keratometric and corneal power measurements obtained by Orbscan II, Pentacam, and Galilei corneal tomography systems. Am J Ophthalmol
9. Chen D, Lam AKC. Reliability and repeatability of the Pentacam on corneal curvatures. Clin Exp Optom. 92, 2009, p. 110-118, Available at: http://onlinelibrary.wiley.com/doi/10.1111/j.1444-0938.2008.00336.x/pdf
. Accessed January 22, 2015.
10. Szalai E, Berta A, Hassan Z, Módis L Jr. Reliability and repeatability of swept-source Fourier-domain optical coherence tomography and Scheimpflug imaging in keratoconus. J Cataract Refract Surg
11. McMahon TT, Anderson RJ, Joslin CE, Rosas GA., the Collaborative Longitudinal Evaluation of Keratoconus (CLEK) Study Topography Analysis Group. Precision of three topography instruments in keratoconus subjects. Optom Vis Sci. 78, 2001, p. 599-604, Available at: http://journals.lww.com/optvissci/Fulltext/2001/08000/Precision_of_Three_Topography_Instruments_in.12.aspx
. Accessed January 22, 2015.
12. Schroeder Swartz T. Pentacam. Agarwal A, Agarwal A, Jacob S Dr, editors. Agarwal’s Textbook on Corneal Topography: Including Pentacam and Anterior Segment OCT. 2nd ed. New Delhi, India: Jaypee Brothers; 2010: pp. 117-136.
13. Koch DD, Foulks GN, Moran CT, Wakil JS. The Corneal EyeSys System: accuracy analysis and reproducibility of first-generation prototype. Refract Corneal Surg
14. Dave T, Ruston D, Fowler C. Evaluation of the EyeSys model II computerized videokeratoscope. Part II: the repeatability and accuracy in measuring convex aspheric surfaces. Optom Vis Sci. 75, 1998, p. 656-662, Available at: http://journals.lww.com/optvissci/Abstract/1998/09000/Evaluation_of_the_EyeSys_Model_II_Computerized.20.aspx
. Accessed January 22, 2015.
15. Fowler CW. Assessment of toroidal surfaces by the measurement of curvature in three fixed meridians. Ophthalmic Physiol Opt
16. Khurana AK. Theory and Practice of Optics and Refraction. 2nd ed. 2008, Elsevier India, Noida, UP, India, p. 157-189
17. Wilhelmus KR. Beyond the P. III: Possible insignificance of the nonsignificant P value. J Cataract Refract Surg
18. Bland JM, Altman DG. Measurement error [Statistics notes]. BMJ. 313, 1996, 744, Available at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2352111/pdf/bmj00560-0056b.pdf
. Accessed January 22, 2015.
19. Tennen DG, Keates RH, Montoya C. Comparison of three keratometry instruments. J Cataract Refract Surg
20. Kawamorita T, Nakayama N, Uozato H. Repeatability and reproducibility of corneal curvature measurements using the Pentacam and Keratron topography systems. J Refract Surg
21. Mao X, Savini G, Zhuo Z, Feng Y, Zhang J, Wang Q, Chen H, Huang J. Repeatability, reproducibility, and agreement of corneal power measurements obtained with a new corneal topographer. J Cataract Refract Surg
22. Savini G, Barboni P, Carbonelli M, Hoffer KJ. Repeatability of automatic measurements by a new Scheimpflug camera combined with Placido topography. J Cataract Refract Surg
23. Montalbán R, Alió JL, Javaloy J, Piñero DP. Intrasubject repeatability in keratoconus-eye measurements obtained with a new Scheimpflug photography-based system. J Cataract Refract Surg
24. Tang M, Chen A, Li Y, Huang D. Corneal power measurement with Fourier-domain optical coherence tomography. J Cataract Refract Surg
25. Holladay JT, Hill WE, Steinmueller A. Corneal power measurements using Scheimpflug imaging in eyes with prior corneal refractive surgery. J Refract Surg
. 2009;25:862-868. erratum, 2010; 26:387.