Journal of Refractive Surgery

Biomechanics 

Tomographic and Biomechanical Scheimpflug Imaging for Keratoconus Characterization: A Validation of Current Indices

Johannes Steinberg, MD; Marlene Siebert; Toam Katz, MD; Andreas Frings, MD; Juliane Mehlan, MD; Vasyl Druchkiv, MSc; Jens Bühren, MD; Stephan J. Linke, MD

Abstract

PURPOSE:

To analyze the potential benefit of the newly developed Tomography and Biomechanical Index (TBI) for early keratoconus screening.

METHODS:

In this retrospective study, the discriminatory power of the corneal tomography Belin/Ambrósio Enhanced Ectasia Display (BAD-D) index and the newly developed Corvis Biomechanical Index (CBI) and TBI to differentiate between normal eyes, manifest keratoconus eyes (KCE), very asymmetric keratoconus eyes with ectasia (VAE-E), and their fellow eyes with either regular topography (VAE-NT) or regular topography and tomography (VAE-NTT) were analyzed by applying the t test (for normal distribution), Wilcoxon matched-pairs test (if not normally distributed), and receiver operating characteristic curve (ROC). The DeLong test was used to compare the area under the ROC (AUROC). Further, the cut-offs of the analyzed indices presented in a study by Ambrósio et al. from 2017 were applied in the study population to enable a cross-validation in an independent study population.

RESULTS:

All indices demonstrated a high discriminative power when comparing normal and advanced keratoconus, which decreased when comparing normal and VAE-NT eyes and further when analyzing normal versus VAE-NTT eyes. The difference between the AUROCs reached a statistically significant level when comparing TBI versus BAD-D analyzing normal versus all included keratoconic eyes (P = .02). The TBI presented with the highest AUROCs throughout all conducted analyses when comparing different keratoconus stages, although not reaching a statistically significant level. Applying the cut-offs presented by Ambrósio et al. to differentiate between normal and VAE-NT in the study population, the accuracy was reproducible (accuracy in our study population with an optimized TBI cut-off: 0.72, with the cut-off defined by Ambrósio et al. 0.67).

CONCLUSIONS:

The TBI enables karatoconus screening in topographical and tomographical regular keratoconic eyes. To further improve the screening accuray, prospective studies should be conducted.

[J Refract Surg. 2018;34(12):840–847.]

Abstract

PURPOSE:

To analyze the potential benefit of the newly developed Tomography and Biomechanical Index (TBI) for early keratoconus screening.

METHODS:

In this retrospective study, the discriminatory power of the corneal tomography Belin/Ambrósio Enhanced Ectasia Display (BAD-D) index and the newly developed Corvis Biomechanical Index (CBI) and TBI to differentiate between normal eyes, manifest keratoconus eyes (KCE), very asymmetric keratoconus eyes with ectasia (VAE-E), and their fellow eyes with either regular topography (VAE-NT) or regular topography and tomography (VAE-NTT) were analyzed by applying the t test (for normal distribution), Wilcoxon matched-pairs test (if not normally distributed), and receiver operating characteristic curve (ROC). The DeLong test was used to compare the area under the ROC (AUROC). Further, the cut-offs of the analyzed indices presented in a study by Ambrósio et al. from 2017 were applied in the study population to enable a cross-validation in an independent study population.

RESULTS:

All indices demonstrated a high discriminative power when comparing normal and advanced keratoconus, which decreased when comparing normal and VAE-NT eyes and further when analyzing normal versus VAE-NTT eyes. The difference between the AUROCs reached a statistically significant level when comparing TBI versus BAD-D analyzing normal versus all included keratoconic eyes (P = .02). The TBI presented with the highest AUROCs throughout all conducted analyses when comparing different keratoconus stages, although not reaching a statistically significant level. Applying the cut-offs presented by Ambrósio et al. to differentiate between normal and VAE-NT in the study population, the accuracy was reproducible (accuracy in our study population with an optimized TBI cut-off: 0.72, with the cut-off defined by Ambrósio et al. 0.67).

CONCLUSIONS:

The TBI enables karatoconus screening in topographical and tomographical regular keratoconic eyes. To further improve the screening accuray, prospective studies should be conducted.

[J Refract Surg. 2018;34(12):840–847.]

Early keratoconus screening remains an important topic in corneal and refractive surgery. An early keratoconus diagnosis, ideally even before topographic changes occur (so-called subclinical keratoconus), might lead to a timely stabilizing cross-linking therapy. It might also help to avoid ectasia after corneal refractive surgery procedures by excluding patients with abnormal corneal characteristics.1

To date, the gold standard in keratoconus screening is topographic and tomographic (ie, “morphometric”) analyses. In 2015, we analyzed the keratoconus screening potential of in vivo biomechanical corneal analyses using the corneal visualization Scheimpflug technology Corvis ST (Oculus Optikgeräte GmbH, Wetzlar, Germany).2 We used single Corvis ST parameters to differentiate between normal and keratoconic eyes. Unfortunately, we had to conclude that “using (single) static parameters derived from only one predefined frame might not be sufficient to analyze the complex, dynamic biomechanical properties of the cornea (p. 8).” We also determined that it was not possible to demonstrate statistically significant differences between normal and subclinical keratoconic eyes (ie, keratoconic eyes with regular topography).

During the past 2 years, substantial improvement has been gained in the field of in vivo biomechanical analyses of the cornea. Joda et al.3 developed a biomechanical corrected intraocular pressure. Based on these biomechanically less biased and therefore more accurate intraocular pressure data, Vinciguerra et al.4 defined normative Corvis ST data enabling a better compensation of the influence of the biomechanical corrected intraocular pressure and the corneal thickness and finally introduced the Corvis Biomechanical Index (CBI), a complex index calculated from different single Corvis ST parameters, to enable a more robust and accurate biomechanical in vivo keratoconus screening.5,6 The latest improvement has been the combination of tomographical and biomechanical data derived from Scheimpflug analyses (Pentacam and Corvis ST; both Oculus Optikgeräte GmbH) by calculating a Tomography and Biomechanical Index (TBI) that, according to the authors, provided sensitivity/specificity for screening topographical normal subclinical keratoconic eyes of up to 96%.7 Because these developments raised the hope of improving current keratoconus screening beyond the level of topographic and tomographic analyses, we decided to recalculate our Scheimpflug data to analyze the discriminative ability of the CBI and TBI and compare them to the Belin/Ambrósio Enhanced Ectasia Display (BAD-D) index, a tomography-based index that has been established for objective and reproducible early keratoconus screening.8–11 Further, this study was designed as an independent cross-validation of the results of Ambrósio et al.7 comparing TBI, CBI, and BAD-D cut-offs to differentiate between topographically normal keratoconic eyes.

The motivation to initiate our study was to evaluate whether the addition of in vivo biomechanical analyses can help to improve keratoconus screening beyond the current standard of topographic and tomographic analyses and thereby help with the important but seemingly still difficult task of identifying eyes with subclinical keratoconus.

Patients and Methods

This study was performed as a cooperative effort between the Department of Ophthalmology, University Medical Center Hamburg-Eppendorf, the zentrumsehstärke Clinic for Ophthalmology at the University Campus Hamburg-Eppendorf, and the CareVision Eye Clinic Hamburg. Our study adhered to the tenets of the Declaration of Helsinki. Informed consent for retrospective data analysis and approval of the local ethics committee for the study were obtained.

The Corvis ST analyses of normal eyes, keratoconic patients with asymmetric ectasia (ie, one eye with regular topography, the other eye with distinct irregular topography due to keratoconus), and eyes with clinically and topographically manifest keratoconus in both eyes were reanalyzed to calculate the CBI and additionally combine the Corvis ST analyses with Pentacam HR Scheimpflug analyses obtained within 1 hour before the Corvis ST data to calculate the new TBI. For recalculating the Pentacam data, software version V1.21r17 was used. The CBI and TBI analyses were conducted using the Oculus software version V1.3r1476. Due to our methodology of recalculating our data, the definition for normal and keratoconic eyes remained the same as previously published.5 We based our definition of normal and keratoconic eyes on topography data according to the well-established KISA% index.12 To improve the definition of normal and subclinical keratoconic eyes, we further added cut-offs for the paracentral inferior-superior (I-S) dioptric difference and the maximum keratometry (Kmax).5 Because this study was partially designed as a cross-validation, we applied the group names of the study by Ambrósio et al.7:

  • Normal eyes: Our objective topographic criteria were: both eyes with a KISA% index of less than 60%, Kmax of 47.00 D or less, and I-S difference of less than 1.45 D. Because no truly established tomographic parameter(s)/cut-off(s) for differentiating normal from keratoconus suspect eyes exist, we adapted our classification for normal eyes to the recent publication by Ambrósio et al. by adding the criterion of “overall subjective normal topography and tomography examinations” based on the evaluation of experienced refractive surgeons (JS, SJL). Only one eye was randomly selected for further statistical analysis.
  • Very asymmetric eyes with normal topography (VAE-NT): The less affected eye (fellow eye) of a keratoconic patient was included if the following criteria were met: KISA% index of less than 60%, I-S difference of less than 1.45 D, and Kmax of 47.00 D or less (ie, same topographic criteria as in normal eyes, except than in normal eyes, both eyes of the patient met the criteria).
  • Very asymmetric eyes with ectasia (VAE-E): The fellow eyes of the VAE-NT displaying a KISA% index of greater than 100% and at least one of the following biomicroscopic signs: Vogt striae, Fleischer ring, or focal stromal thinning.
  • Clinically manifest keratoconic eyes (KCE): The inclusion criteria were the same as for VAE-E, except that both eyes of the patient met the ectasia criteria. Only one eye was randomly selected for further statistical analysis.

Different from our previous publication2 and the study of Ambrósio et al.,7 we defined another group for additional subanalyses concentrating on the potential screening of “sub-morphometric” keratoconic eyes (ie, keratoconic eyes with regular topography AND tomography):

  • Very asymmetric eyes with normal topography and tomography (VAE-NTT): The less affected eye (fellow eye) of a keratoconic patient was included if the following criteria were met: KISA% index of less than 60%, I-S difference of less than 1.45 D, Kmax of 47.00 or greater D, and a BAD-D overall index of less than 1.6.9. The VAE-NTT group displays a subgroup or a sample of the VAE-NT eyes.

In this regard, it is important to recognize that the TBI analyses combine biomechanical and tomographical data. Analyzing corneal data after excluding tomographically suspicious keratoconic fellow eyes (ie, BAD-D < 1.6) includes the risk of falsely decreasing the keratoconus screening ability of the TBI.

For all eyes included in the study, no contact lenses had been worn for at least 4 (rigid contact lenses) or 2 (soft contact lenses) weeks prior to examination, and there was no history of further eye disease or previous ocular surgery.

The techniques for Pentacam and Corvis ST analyses have been described previously.13–15

Statistical Analysis

Data were analyzed with StataCorp software (Stata Statistical Software: Release 14; StataCorp, College Station, TX, and R Core Team; 2017). For descriptive analysis, we reported medians and quartiles of the variables and visualized them using box plots. To analyze the discriminatory power of any continuous explanatory variable, we performed receiver operating characteristics curve (ROC) analyses. Cut-offs were defined by minimizing the difference between sensitivity and specificity, thus considering each of them equally important. Pairwise comparisons of the area under the ROC (AUROC) were accomplished with the non-parametric DeLong test.16

We also adapted the cut-offs published by Ambrósio et al.7 to analyze their discriminatory power between two of our groups (NE vs VAE-NT) in a subanalysis.

Results

The statistical analysis included data of 105 eyes in the normal group, 32 eyes in the VAE-NT group, 18 eyes in the VAE-NTT group, 28 eyes in the VAE-E group, and 96 eyes in the KCE group. Descriptive values for age, topography and tomography parameters, and the tomographical BAD-D overall index, the CBI, and the combined TBI are displayed in Table A (available in the online version of this article).

Study Population Details

Table A:

Study Population Details

Due to our inclusion and exclusion criteria, distinct differences in topographic measurements between topographically appearing normal eyes (VAE-NT and VAE-NTT) and clinically manifest keratoconic eyes (VAE-E and KCE) could be demonstrated. Simultaneously, next to the topographical homogeneity of the normal, VAE-NT, and VAE-NTT eyes, the data also displayed an apparently homogenous deviation of the tomographical parameters (ie, thinnest point in corneal thickness and back surface elevation at the thinnest point between these groups. To demonstrate and analyze potential differences between normal and the other groups based on tomographic and biomechanical indices, we displayed the deviation of the indices in the corresponding groups in boxplots (Figure 1) and performed ROC analyses (Tables BC, available in the online version of this article, and Figure 2). The analyses regarding potential statistically significant differences between the ROC results (ie, the AUROCs) are displayed in Table C.

Boxplots to demonstrate the deviation of the (A) Belin/Ambrósio Enhanced Ectasia Display (BAD-D), (B) Corvis Biomechanical Index (CBI), and (C) Tomographic and Biomechanical Index (TBI) between the analyzed groups. NE = normal eyes; KCE = clinically manifest keratoconus eyes; VAE-E = very asymmetric eyes with ectasia; VAE-NT = very asymmetric eyes with normal topography; VAE-NTT = very asymmetric eyes with normal topography and tomography

Figure 1.

Boxplots to demonstrate the deviation of the (A) Belin/Ambrósio Enhanced Ectasia Display (BAD-D), (B) Corvis Biomechanical Index (CBI), and (C) Tomographic and Biomechanical Index (TBI) between the analyzed groups. NE = normal eyes; KCE = clinically manifest keratoconus eyes; VAE-E = very asymmetric eyes with ectasia; VAE-NT = very asymmetric eyes with normal topography; VAE-NTT = very asymmetric eyes with normal topography and tomography

Results of the ROC Analysesa

Table B:

Results of the ROC Analyses

DeLong Test Results for Pairwise Comparison of the AUROCsa

Table C:

DeLong Test Results for Pairwise Comparison of the AUROCs

Receiver operating characteristics curves (ROC) of the Belin/Ambrósio Enhanced Ectasia Display (BAD-D), Corvis Biomechanical Index (CBI), and Tomographic and Biomechanical Index (TBI) to differentiate between normal eyes (NE) and keratoconic eyes (KCE) (all keratoconic eyes combined and separated into groups of different stages). VAE-NT = very asymmetric eyes with normal topography; VAE-NTT = very asymmetric eyes with normal topography and tomography

Figure 2.

Receiver operating characteristics curves (ROC) of the Belin/Ambrósio Enhanced Ectasia Display (BAD-D), Corvis Biomechanical Index (CBI), and Tomographic and Biomechanical Index (TBI) to differentiate between normal eyes (NE) and keratoconic eyes (KCE) (all keratoconic eyes combined and separated into groups of different stages). VAE-NT = very asymmetric eyes with normal topography; VAE-NTT = very asymmetric eyes with normal topography and tomography

The boxplots display a distinct difference in the distribution of BAD-D, CBI, and TBI values between the topographically normal groups (normal, VAE-NT, VAE-NTT) and the topographically defined keratoconus groups (VAE-E, KCE). Comparing normal versus VAE-NT and VAE-NTT, a difference in the distribution in all analyzed indices could be demonstrated between the three groups with an apparently lowest discriminatory power of the BAD-D, followed by the CBI and the TBI.

The ROC analyses (Tables BC and Figure 2) were in line with the results of the boxplots. Whereas all indices demonstrated a high discriminatory power between normal and KCE and normal versus VAE-E, the AUROCs decrease significantly analyzing normal versus VAE-NT and VE versus VAE-NTT. Comparing the three analyzed indices, the TBI demonstrated the highest AUROC values in all conducted analyses. Despite the supremacy of the TBI when it comes to comparing the absolute AUROC values, a statistically significant difference could only be demonstrated comparing TBI and BAD-D when analyzing normal versus all keratoconic eyes combined. Besides, as displayed in Figure 2 and Table B, when analyzing normal versus VAE-NTT, the TBI still demonstrated a discriminative power on a “subtomographical” level yielding a sensitivity/specificity ratio of 0.67/0.65. Despite the difference in our methodology, we applied the optimized cut-offs for the comparison of normal and VAE-NT of Ambrósio et al.7 to our study population to enable a direct cross-validation of their cut-offs in a new (our) study population.

As displayed in Table 1, applying the cut-offs of Ambrósio et al.,7 the accuracies were almost identical compared to the results using our optimized cut-offs to differentiate between normal and VAE-NT.

Comparison of the Sensitivity, Specificity, and Accuracy Between NE and VAE-NT

Table 1:

Comparison of the Sensitivity, Specificity, and Accuracy Between NE and VAE-NT

Discussion

Comparing the normal and KCE groups, all indices demonstrated a high discriminative ability (AUROCs/sensitivity/specificity for normal vs KCE: BAD-D = 0.984/0.97/0.98; CBI = 0.970/0.93/0.95; TBI = 0.998/0.98/1.0). Theoretically, the integration of biomechanical in vivo analyses could improve keratoconus screening ideally to a level of separating keratoconic eyes with apparently regular topography and tomography from normal eyes. To explore the ability of the CBI and TBI, we analyzed the discriminative ability of these indices to differentiate between the normal and VAE-NT groups and additionally created the VAE-NTT group. Compared to the results for normal versus KCE, the discriminating power of all analyzed indices decreased when comparing normal and VAE-NT (AUROC/sensitivity/specificity) for BAD: 0.748/0.69/0.69; CBI: 0.787/0.69/0.69; TBI: 0.825/0.72/0.71) and noticeably further when trying to separate normal from VAE-NTT (AUROC/sensitivity/specificity) for BAD-D = not analyzed, because the index was used for the definition of the VAE-NTT eyes; CBI: 0.704/0.67/0.67; TBI: 0.732/0.67/0.65). Throughout all conducted analyses, the tomography and biomechanical analyses combining the TBI presented the highest discriminatory power, but reached a statistically significant level only for the comparison of the AUROCs analyzing TBI versus BAD-D in normal versus all keratoconic eyes (P = .02). When analyzing normal versus VAE-NTT, both biomechanical indices demonstrated a noticeably decreased discriminatory power with a sensitivity/specificity of 0.67/0.67 for the CBI (AUROC: 0.704) and 0.67/0.65 for the TBI (AUROC: 0.732).

The BAD-D was established several years ago,9 but the CBI and especially the TBI were introduced only recently by an international study group.6,7 The CBI has been generated by using logistic regression analyses combining different biomechanical parameters and one tomographical index (Ambrósio's Relational Thickness to the horizontal profile; all measured with the Corvis ST) comparing “normal” and “clear keratoconus” eyes.6 Because of a high variety of formerly published results regarding keratoconus screening using the Corvis ST, we recently conducted a “proof of principle” study that confirmed the ability of the Corvis ST to differentiate between normal and advanced keratoconic eyes.5 Analyzing the ability of the CBI to differentiate between normal eyes and much more subtle keratoconus stages, the discriminatory power decreased noticeably (CBI AUROC for normal vs KCE: 0.970; normal vs VAE-NT: 0.787; normal vs VAE-NTT: 0.704). This result is not surprising considering the definition of the VAE-NT and VAE-NTT groups (ie, only keratoconic eyes of early stages were included that still demonstrated at least a normal topography [VAE-NT] or even a normal topography + tomography [VAE-NTT]).

Ambrósio et al.7 introduced the TBI to discriminate normal eyes from keratoconic eyes of all stages. They used the “leave-one-out cross-validation” (LOOCV) method to generate this new index. Citing the study of Ambrósio et al.7: “In this method, a new model is built as many times as the number of cases included in the study. Each different model is built for all cases, excluding one patient in whom the model is tested” (p. 8). Using the LOOCV strategy and performing ROC analyses to evaluate the discriminatory power of the TBI to distinguish between normal and VAE-NT, they could demonstrate an AUROC of 0.838 for the BAD-D. Simultaneously, the CBI displayed an AUROC of 0.822 and the TBI of 0.985. Ambrósio et al.7 further discussed the risk of creating an algorithm/model that might be prone to some degree of “overfitting” and suggested using the model in a novel population “to provide a more realistic estimation of the performance.”

Applying the BAD-D, CBI, and TBI indices in our study population, the discriminatory power of the three indices decreased noticeably when comparing normal and VAE-NT (AUROCs current study/Ambrósio et al. for BAD-D = 0.748/0.838; CBI = 0.787/0.822; TBI = 0.825/0.985). The differences between the AUROCs in our study never reached a statistically significant level comparing normal to the CBI VAE-NT (BAD-D vs TBI: P = .067; CBI vs TBI: P = .454). Comparing normal eyes and keratoconic eyes with a regular BAD-D (ie, normal tomography) and regular topography, we could demonstrate a noticeably discriminative power of the TBI. However, the following facts should be considered when evaluating the potential benefit of the TBI (AUROC normal vs VAE-NTT for the TBI: 0.704).

First, the TBI algorithm incorporates tomographical data for the analyses of the cornea. Preselecting tomographically (BAD-D defined) “normal” eyes for the comparison with (completely) normal eyes tends to falsely impair the screening ability of the TBI. We still used this methodology because we aimed to analyze whether the TBI could screen any kind of differences in eyes which, by the current keratoconus screening standard, (falsely) demonstrated to be equal.

Second, the BAD-D is an already well tested and established index for keratoconus screening that has been optimized throughout by increasing the number of analyzed eyes. The TBI has just been developed by an ambitious study group (Ambrósio et al.) analyzing less then 100 keratoconic eyes with regular topography. On the one hand, this is a high number considering the rarity of this entity; on the other hand, analyzing more data of early keratoconic eyes might help to further adjust the index. Therefore, we transferred our data to Oculus Optikgeräte GmbH for further analysis.

Analyzing differences in the results between our study and Ambrósio et al.'s,7 the definition of the cutoffs has to be considered. Other than Ambrósio et al.,7 we did not define our cut-off for the ROC analyses yielding the highest accuracy in our study population, but for minimizing the difference between sensitivity and specificity. This is important when comparing groups of different sizes. When aiming for the highest accuracy in unequally balanced group sizes, the result is biased by the group with more included samples (ie, we included 105 normal and 18 VAE-NTT and choosing a cut-off favoring the specificity could lead to an accuracy of up to 85% (> 105/123) just by correctly classifying all normal [and concurrently misclassifying all VAE-NTT]).

Analyzing the results in Table 1, despite the discussed differences in our methodology, their and our cut-offs demonstrated almost equal accuracies for all analyzed indices. Interestingly, whereas both study groups defined almost equal cut-offs for the tomographical index BAD-D, our cut-offs for the CBI and TBI were noticeably lower. The cut-offs of Ambrósio et al.7 also led to a noticeably higher specificity in our population than our cut-offs aiming for a minimized difference between sensitivity and specificity.

Despite the discussed methodological differences, considering identical topography inclusion criteria and the practically identical tomographical characteristics, the difference of the discriminatory power of the biomechanical indices between the two study groups is of interest. To enable this cross-validation of the biomechanical indices, we defined our inclusion and exclusion criteria based on the methodology of Ambrósio et al. Therefore, comparing the deviation of “key” topographic and tomographic indices such as the KISA% index and the BAD-D to characterize the corneal morphology/keratoconus stages in both studies, a high conformity within the defined groups exist (ie, median of the KISA% index of the VAE-NT group = 7.51 for Ambrósio et al.7 for the CBI 8.0 for our study).

However, we noticed a distinct difference regarding the deviation of age in both study populations. Whereas we included patients with a median age of 33 to 36 years throughout all of our analyzed groups, the study population of Ambrósio et al. consisted of patients with a median age of 33 to 43 years, including patients up to 88 years of age (highest age in our study population: 62 [normal]). Although they did not report a statistically significant difference between their age groups (analysis of variance; P = .273), the inclusion of a large amount of eyes of “older” people might bias the biomechanical findings, which is also indicated by their findings of a clinically minor but statistically significant negative correlation between the TBI and age (P < .0001; Spearman's coefficient of rank correlation [rho] = −0.18). This finding implies the benefit of an age-adapted cut-off definition considering that in our day-to-day clinical practice we use keratoconus screening especially in patients from their early teens up to the age of 40 to 50 years (laser vision correction candidates).

Further explanations for the differences of the AUROCs in both studies and the difficulty of defining one common cut-off optimized beyond the applied study population are the risk of “over-fitting” the index to the original study population (despite using statistical strategies such as the LOOCV) and the widely spread distribution of CBI and TBI values within the keratoconic eyes of early stages (see also Figures 12). On the one hand, this distribution is most probably attributed to the technique of the Corvis ST analyzing biomechanical characteristics of the cornea by using several parameters, which are acquired not by measuring static morphological characteristics of the cornea (eg, the BAD-D), but by measuring the dynamic corneal response to an air-puff impulse. On the other hand, the broad distribution of the biomechanical indices could be caused by potentially conducting (highly) biasing mistakes with our current methodology of defining early stages of keratoconus. Specifically, we could be wrong by defining unsuspicious fellow eyes of keratoconic patients as keratoconic eyes (our VAE-NT and VAE-NTT groups). With increasing knowledge about the disease and especially its biomechanical characteristics, some opinion leaders discuss the possibility of “true-unilateral keratoconus.”17 In addition, we might have analyzed keratoconic eyes in a stage in which even subtle biomechanical changes might not even have occurred.

We curently do not know exactly what causes the disease. A widespread theory about the pathogenesis of keratoconus includes the necessity of some kind of “trigger-factor” in patients with a genetic predisposition to “start” the pathogenesis and enable us to detect any (biomechanical) alterations/keratoconus-specific characteristics of the cornea (“two-hit hypothesis”).7,18,19 Considering the potential biasing factors and the results of Ambrósio et al. and our study, it seems questionable whether the demonstrated accuracy of early keratoconus staging can be further improved by conducting retrospective studies of morphometrically unsuspicious keratoconic fellow eyes. Instead prospective large-scale studies are warranted to clarify this issue. However, our current and future studies hopefully contribute to the knowledge about the biomechanical properties of the cornea in early keratoconus screening, ideally leading to earlier treatment of patients with subclinical keratoconus and improving the safety profile of laser vision correction by excluding suspicious cases (high sensitivity) and simultaneously giving refractive surgeons more certainty about whom to exclude (ie, increase of the specificity). Regarding laser vision correction candidate screening, next to the importance of the preoperative condition of the cornea, factors such as the amount of tissue ablation (percentage tissue altered) should also be considered in respect of the postoperative stability of the cornea.20 Therefore, further studies combining both analyses of the biomechanical state before and after laser vision correction in respect to the percent tissue altered should be forwarded to continuously improve the safety of laser vision correction.

We demonstrated a good discriminatory power of all analyzed indices to differentiate between normal and advanced keratoconic eyes. Analyzing keratoconic eyes with unsuspicious topography (VAE-NT) and the subgroup with unsuspicious topographical and tomographical eyes (VAE-NTT), and thereby screening for patients even before visual symptoms due to the keratoconus occur, the accuracy decreased as expected, most probably at least partially due to a limited methodology by retrospectively analyzing keratoconic fellow eyes. Our study emphasizes that further optimization is necessary to detect sub-morphometric stages on a reliable level. However, the biomechanical analyses (CBI) were equal to the tomographic analyses (BAD-D) in all conducted screening scenarios. Combining biomechanical and tomographic analyses (TBI), the screening ability was slightly better than the current gold standard of topographic and tomographic analyses.

The topography findings are presented in Figures S1S32 (available in the online version of this article).

Corneal topography of patient 1 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S1:

Corneal topography of patient 1 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 2 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S2:

Corneal topography of patient 2 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 3 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S3:

Corneal topography of patient 3 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 4 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S4:

Corneal topography of patient 4 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 5 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S5:

Corneal topography of patient 5 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 6 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S6:

Corneal topography of patient 6 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 7 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S7:

Corneal topography of patient 7 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 8 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S8:

Corneal topography of patient 8 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 9 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S9:

Corneal topography of patient 9 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 10 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S10:

Corneal topography of patient 10 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 11 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S11:

Corneal topography of patient 11 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 12 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S12:

Corneal topography of patient 12 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 13 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S13:

Corneal topography of patient 13 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 14 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S14:

Corneal topography of patient 14 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 15 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S15:

Corneal topography of patient 15 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 16 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S16:

Corneal topography of patient 16 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 17 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S17:

Corneal topography of patient 17 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 18 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S18:

Corneal topography of patient 18 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 19 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S19:

Corneal topography of patient 19 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 20 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S20:

Corneal topography of patient 20 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 21 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S21:

Corneal topography of patient 21 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 22 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S22:

Corneal topography of patient 22 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 23 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S23:

Corneal topography of patient 23 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 24 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S24:

Corneal topography of patient 24 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 25 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S25:

Corneal topography of patient 25 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 26 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S26:

Corneal topography of patient 26 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 27 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S27:

Corneal topography of patient 27 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 28 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S28:

Corneal topography of patient 28 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 29 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S29:

Corneal topography of patient 29 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 30 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S30:

Corneal topography of patient 30 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 31 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S31:

Corneal topography of patient 31 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Corneal topography of patient 32 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

Figure S32:

Corneal topography of patient 32 with very asymmetric keratoconus showing (A) ectasia and (B) regular topography in the fellow eye.

References

  1. Martinez-Abad A, Piñero DP. New perspectives on the detection and progression of keratoconus. J Cataract Refract Surg. 2017;43:1213–1227. doi:10.1016/j.jcrs.2017.07.021 [CrossRef]
  2. Steinberg J, Katz T, Lucke K, Frings A, Druchkiv V, Linke SJ. Screening for keratoconus with new dynamic biomechanical in vivo Scheimpflug analyses. Cornea. 2015;34:1404–1412. doi:10.1097/ICO.0000000000000598 [CrossRef]
  3. Joda AA, Shervin MM, Kook D, Elsheikh A. Development and validation of a correction equation for Corvis tonometry. Comput Methods Biomech Biomed Engin. 2016;19:943–953. doi:10.1080/10255842.2015.1077515 [CrossRef]
  4. Vinciguerra R, Elsheikh A, Roberts CJ, Mehlan J, Katz T, Linke SJ. Influence of pachymetry and intraocular pressure on dynamic corneal response parameters in healthy patients. J Refract Surg. 2016;32:550–561. doi:10.3928/1081597X-20160524-01 [CrossRef]
  5. Steinberg J, Amirabadi NE, Frings A, et al. Keratoconus screening with dynamic biomechanical in vivo Scheimpflug analyses: a proof-of-concept study. J Refract Surg. 2017;33:773–778. doi:10.3928/1081597X-20170807-02 [CrossRef]
  6. Vinciguerra R, Ambrósio R Jr, Elsheikh A, et al. Detection of keratoconus with a new biomechanical index. J Refract Surg. 2016;32:803–810. doi:10.3928/1081597X-20160629-01 [CrossRef]
  7. Ambrósio R Jr, Lopes BT, Faria-Correia F, et al. Integration of Scheimpflug-based corneal tomography and biomechanical assessments for enhancing ectasia detection. J Refract Surg. 2017;33:434–443. doi:10.3928/1081597X-20170426-02 [CrossRef]
  8. Koc M, Tekin K, Tekin MI, et al. An early finding of keratoconus: increase in corneal densitometry. Cornea. 2018;37:580–586. doi:10.1097/ICO.0000000000001537 [CrossRef]
  9. Villavicencio OF, Gilani F, Henriquez MA, Izquierdo L Jr, Ambrósio R Jr, . Independent population validation of the Belin-Ambrósio Enhanced Ectasia Display implications for keratoconus studies and screening. Int J Keratoconus Ectatic Corneal Dis. 2014;3:1–8.
  10. Reinstein DZ, Archer TJ, Urs R, Gobbe M, RoyChoudhury A, Silverman RH. Detection of keratoconus in clinically and algorithmically topographically normal fellow eyes using epithelial thickness analysis. J Refract Surg. 2015;31:736–744. doi:10.3928/1081597X-20151021-02 [CrossRef]
  11. Shetty R, Rao H, Khamar P, et al. Keratoconus screening indices and their diagnostic ability to distinguish normal from ectatic corneas. Am J Ophthalmol. 2017;181:140–148. doi:10.1016/j.ajo.2017.06.031 [CrossRef]
  12. Steinberg J, Aubke-Schultz S, Frings A, et al. Correlation of the KISA% index and Scheimpflug tomography in ‘normal’, ‘subclinical’, ‘keratoconus-suspect’ and ‘clinically manifest’ keratoconic eyes. Acta Ophthalmol. 2015;93:e199–e207. doi:10.1111/aos.12590 [CrossRef]
  13. Hon Y, Lam AK. Corneal deformation measurement using Scheimpflug noncontact tonometry. Optom Vis Sci. 2013;90:e1–e8. doi:10.1097/OPX.0b013e318279eb87 [CrossRef]
  14. Steinberg J, Ahmadiyar M, Rost A, et al. Anterior and posterior corneal changes after crosslinking for keratoconus. Optom Vis Sci. 2014;91:178–186.
  15. Steinberg J, Katz T, Mousli A, et al. Corneal biomechanical changes after crosslinking for progressive keratoconus with the corneal visualization Scheimpflug technology. J Ophthalmol. 2014;2014:579190. doi:10.1155/2014/579190 [CrossRef]
  16. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837–845. doi:10.2307/2531595 [CrossRef]
  17. Vinciguerra R, Ambrósio R Jr, Roberts CJ, Azzolini C, Vinciguerra P. Biomechanical characterization of subclinical keratoconus without topographic or tomographic abnormalities. J Refract Surg. 2017;33:399–407. doi:10.3928/1081597X-20170213-01 [CrossRef]
  18. Moussa S, Grabner G, Ruckhofer J, Dietrich M, Reitsamer H. Genetics in keratoconus—what is new?Open Ophthalmol J. 2017;11:201–210. doi:10.2174/1874364101711010201 [CrossRef]
  19. Soiberman U, Foster JW, Jun AS, Chakravarti S. Pathophysiology of keratoconus: what do we know today. Open Ophthalmol J. 2017;11:252–261. doi:10.2174/1874364101711010252 [CrossRef]
  20. Santhiago MR. Percent tissue altered and corneal ectasia. Curr Opin Ophthalmol. 2016;27:311–315. doi:10.1097/ICU.0000000000000276 [CrossRef]

Comparison of the Sensitivity, Specificity, and Accuracy Between NE and VAE-NT

ParameterBAD-DCBITBI
AUROC0.7480.7870.825
Current studya
  Cut-off1.090.030.11
  Sensitivity0.690.690.72
  Specificity0.790.680.71
  Accuracy0.710.690.72
Current study vs Ambrósio et al.7
  Cut-off1.080.070.29
  Sensitivity0.690.660.63
  Specificity0.780.830.83
  Accuracy0.710.700.67

Study Population Details

ParameterNE (n = 105)VAE-NT (n = 32)VAE-NTT (n = 18)aVAE-E (n = 28)KCE (n = 96)





MedianQ25 to Q75 (Range)MedianQ25 to Q75 (Range)MedianQ25 to Q75 (Range)MedianQ25 to Q75 (Range)MedianQ25 to Q75 (Range)
Age3429 to 41 (19 to 62)3626 to 43 (19 to 50)3527 to 43 (19 to 46)3627 to 43 (19 to 50)3327 to 42 (13 to 62)
K1 (D)43.342.2 to 44.0 (39.6 to 45.7)42.941.8 to 43.9 (39.9 to 46.1)42.041.5 to 43.7 (39 to 9 to 45.7)43.842.5 to 45.4 (39.1 to 48.4)45.343.3 to 46.9 (37.7 to 59.1)
K2 (D)44.343.4 to 45.1 (41.1 to 46.4)43.642.6 to 45.0 (40.5 to 47.8)43.341.7 to 44.6 (40.5 to 46.0)47.044.8 to 49.1 (40.5 to 52.3)48.546.4 to 51.0 (39.1 to 61.7)
Astig (D)0.80.5 to 1.4 (0.1 to 3.8)0.70.5 to 1.2 (0 to 4.7)0.60.3 to 1.1 (0.0 to 2.2)2.91.8 to 4.5 (0.7 to 6.6)3.32.2 to 4.5 (0.5 to 14.4)
Kmax (D)44.843.8 to 45.6 (41.2 to 46.8)4543.7 to 45.7 (40.9 to 47)44.342.9 to 45.3 (40.9 to 46.6)51.349.8 to 54.6 (47.3 to 58.1)55.251 to 58.2 (43.6 to 80.3)
TPCT (mm)546529 to 570 (478 to 618)523502 to 539 (459 to 575)532512 to 557 (494 to 575)499473 to 514 (412 to 535)461430 to 487 (349 to 556)
Ele_b_TP (mm)b52 to 7 (−2 to 20)74 to 11 (−1 to 28)52 to 7 (−1 to 8)4939 to 52 (24 to 70)5439 to 69 (20 to 170)
KISA%42 to 9 (0.3 to 53.3)84 to 20 (0.3 to 58)51 to 9 (0.3 to 47.6)369247 to 624 (106 to 1,825)668281 to 1,429 (102 to 999)
BAD-Dc0.750.4 to 1.0 (−0.5 to 2.5)1.50.8 to 1.8 (0.1 to 7.3)0.90.5 to 1.2 (0.1 to 1.57)6.55.8 to 7.6 (3.7 to 10)8.36.3 to 10.9 (0.9 to 25.7)
CBId0.010.0 to 0.0 (0 to 1)0.40.0 to 1.0(0 to 1)0.10.0 to 0.7(0 to 1)11 to 1 (0.1 to 1)11 to 1 (0 to 1)
TBIe0.020.0 to 0.2 (0 to 0.6)0.40.1 to 0.9 (0 to 1)0.19(0.0 to 0.4) (0 to 0.7)11 to 1 (1 to 1)11 to 1 (0.4 to 1)
RMS (CF)1.51.2 to 1.9 (0.8 to 3.8)1.71.4 to 2.4 (1.0 to 4.3)1.41.3 to 1.7 (1.0 to 2.3)9.17.2 to 11.4 (5.5 to 18.6)11.17.9 to 14.4 (2.9 to 45.5)
Z3−1 CF−0.01−0.1 to 0.1 (−0.4 to 0.4)−0.2−0.4 to 0.1 (−1.3 to 0.4)0.04−0.2 to 0.1 (−0.3 to 0.4)−1.8−2.3 to −1.6 (−4.5 to −0.8)−2.4−3.5 to −1.5 (−11.3 to 0.7)

Results of the ROC Analysesa

GroupAUROCSECut-offSensitivitySpecificityAccuracy
NE vs all KCE
  BAD-D0.9380.0151.370.860.890.86
  CBI0.9370.0150.150.880.880.88
  TBI0.9630.010.370.900.910.90
NE vs KCE
  BAD-D0.9840.0092.10.970.980.98
  CBI0.9700.0150.740.930.950.94
  TBI0.9980.0020.570.981.00.99
NE vs VAE-E
  BAD-D1.0000.0003.121.001.001.00
  CBI0.9940.0050.670.960.950.96
  TBI1.0000.0000.751.001.001.00
NE vs VAE-NT
  BAD-D0.7480.0561.090.690.790.71
  CBI0.7870.0500.030.690.680.69
  TBI0.8250.0420.110.720.710.72
NE vs VAE-NTT
  BAD-Db
  CBI0.7040.0740.020.670.670.67
  TBI0.7320.0560.060.670.650.67

DeLong Test Results for Pairwise Comparison of the AUROCsa

GroupsAUROC_1AUROC_2AUROC DifferenceP
NE vs all KCE
  BAD-D vs CBI0.9380.9370.0020.915
  BAD-D vs TBI0.9380.963−0.0250.020
  CBI vs TBI0.9370.963−0.0260.062
NE vs KCE
  BAD-D vs CBI0.9840.9700.0140.312
  BAD-D vs TBI0.9840.998−0.0140.109
  CBI vs TBI0.9700.998−0.0280.056
NE vs VAE-NT
  BAD-D vs CBI0.7480.787−0.0390.484
  BAD-D vs TBI0.7480.825−0.0770.067
  CBI vs TBI0.7870.825−0.0380.454
NE vs VAE-NTT
  CBI vs TBI0.7040.732−0.0280.707
Authors

From the Department of Ophthalmology (JS, MS, TK, JM, VD, SJL) and Care-Vision Germany (JS, TK, SJL), University Medical Center Hamburg-Eppendorf, Hamburg, Germany; zentrumsehstärke, Hamburg, Germany (JS, SJL); University Hospital Duesseldorf, Duesseldorf, Germany (AF); Moorfields Eye Hospital, London, United Kingdom (AF); and Augenpraxisklinik Triangulum, Hanau, Germany (JB).

Dr. Frings was supported by the German Academic Exchange Service (DAAD). Dr. Bühren received lecture fees from Oculus Optikgeräte GmbH, Wetzlar, Germany. The remaining authors have no financial or proprietary interest in the materials presented herein.

AUTHOR CONTRIBUTIONS

Study concept and design (JS, AF, SJL); data collection (MS, JM); analysis and interpretation of data (JS, MS, TK, JM, VD, JB, SJL); writing the manuscript (JS); critical revision of the manuscript (MS, TK, AF, JM, VD, JB, SJL); statistical expertise (JS, MS, AF, VD, JB, SJL); administrative, technical, or material support (MS, TK, JM); supervision (SJL)

Correspondence: Johannes Steinberg, MD, zentrumsehstärke, Martinistrasse 64, 20251 Hamburg, Germany. E-mail: steinberg@zentrumsehstaerke.de

Received: April 12, 2018
Accepted: October 10, 2018

10.3928/1081597X-20181012-01

Sign up to receive

Journal E-contents