- Email: [email protected]
Can retinoscopy keep up in keratoconus diagnosis?
Contact Lens & Anterior Eye 38 (2015) 234–239
Contents lists available at ScienceDirect
Contact Lens & Anterior Eye journal homepage: www.elsevier.com/locate/clae
Can retinoscopy keep up in keratoconus diagnosis? Susanne Goebels a,∗ , Barbara Käsmann-Kellner a , Timo Eppig b , Berthold Seitz a , Achim Langenbucher b a b
Department of Ophthalmology, Saarland University Medical Center, Kirrberger Strasse 100, Bldg. 22, 66421 Homburg, Germany Experimental Ophthalmology, Saarland University, Kirrberger Strasse 100, Bldg. 22, 66421 Homburg, Germany
a r t i c l e
i n f o
Article history: Received 23 October 2014 Received in revised form 27 January 2015 Accepted 29 January 2015 Keywords: Retinoscopy Tomography ORA Keratoconus Scissors reflex
a b s t r a c t Purpose: To evaluate the diagnostic potential of retinoscopy in comparison with Amsler-grading, Pentacam and Ocular Response Analyzer (ORA) in classifying keratoconus stages. Methods: Clinical examination, retinoscopy, Pentacam and ORA were performed in 126 patients. Data of Amsler, retinoscopy, topographic keratoconus classification (TKC) of Pentacam and keratoconus match probability (KMP) of ORA were analyzed. Each of these four classification techniques quotes keratoconus into stage 0 (normal) to 4 (severe). Descriptive analysis and cross tables were used to compare the different devices. Results: For retinoscopy the distribution in the five keratoconus grades normal/suspect/ mild/moderate/severe (in numbers) was 34/33/34/17/8. For Amsler it was 37/36/35/12/4, for TKC 38/17/34/31/4, for KMP 32/34/32/15/9. The cross tables show large classification differences of all devices. Overall, classification of retinoscopy and Amsler/TKC/KMP is congruent in 51.6%/36.3%/39.8% of the cases. Of all eyes, Amsler was congruent with TKC/KMP in 54.0%/48.4%, and TKC and KMP were congruent in 53.3%. In a binary decision (normal vs. any stage of mild/moderate/severe) matching between retinoscopy and Amsler/TKC/KMP was 98.6%/88.8%/82.4%. Sensitivity/specificity for retinoscopy and Amsler, TKC, KMP was 98.8%/94.0%, 84.4%/100% and 80.0%/79.1%. Conclusions: The congruence of keratoconus classification was very poor of all the techniques tested in our study. This applies to objective measures such as TKC, KMP as well as clinical classification techniques such as Amsler and retinoscopy. Compared to TKC and KMP, retinoscopy underestimates keratoconus stages. In contrast, the performance of binary decisions (normal vs. keratoconus) shows a high sensitivity and specificity. Retinoscopy, however, showed a clear clinical use in confirming the diagnosis of keratoconus. © 2015 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved.
1. Introduction Keratoconus is an ectatic non-inflammatory corneal disorder, in which the cornea forms a conic shape due to thinning of the corneal stroma. This corneal thinning normally induces irregular astigmatism, myopia and protrusion, leading to a mild to marked deterioration of visual performance [1]. Keratoconus is usually bilateral, mostly diagnosed in the 2nd or 3rd decade of life with an incidence of 55:100.000 [1–3], and more prominent in males than females. Diagnosis of keratoconus is easy in the presence of clinical signs such as corneal ectasia or thinning, scars, nodes, Vogt’s striae
∗ Corresponding author at: Department of Ophthalmology, Saarland University Medical Center, Kirrberger Strasse 100, Bldg. 22, 66421 Homburg/Saar, Germany. Tel.: +49 6841 1621240; fax: +49 6841 1622440. E-mail address: [email protected] (S. Goebels).
or Fleischer ring. In early stages (forme fruste) or unclear cases additional diagnostic tools are necessary. High-end technology devices are available to assist the examiner in the diagnosis and staging of keratoconus. Topography or tomography systems, aberrometry or assessment of corneal biomechanics as well as dedicated classification systems are available and actually in focus of research [4]. Beside these highly developed devices classical retinoscopy has been successfully established for decades to classify keratoconus. Although retinoscopy requires training, when having command on it, it is a valuable tool in children and/or handicapped or otherwise non-compliant patients or if sophisticated instruments are not available, it may be the case in developing regions. As keratoconus may be associated with some genetic syndromes as Down syndrome, Leber’s amaurosis or CDL-syndrome, keratoconus needs to be excluded or confirmed in patients who cannot fixate properly. The first idea of retinoscopy evolved in 1859 [5]. Sir William Bowman noted a linear shadow seen in an astigmatic eye using
http://dx.doi.org/10.1016/j.clae.2015.01.015 1367-0484/© 2015 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved.
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
a plane mirror ophthalmoscope lit by a candle. In 1875 this phenomenon was described as shadow test. Early retinoscopes with electric lighting used rotating slit and helical filament lamps. A linear filament lamp, which produces a sharp, bright line of light was developed by Jacob Copeland [5,6]. Following investigations lead to better observing systems, control of vergence and handling. Retinoscopy enables measurement of the refractive state. In keratoconus or astigmatic eyes an inhomogeneous light-shadowmovement is visible. At the corneal center the refraction may be slightly myopic in comparison of the periphery [5]. Therefore, in keratoconus retinoscopy typically shows a “against” motion in the center and a “with” motion in the periphery. This opening and closing of the light reflex is called scissors reflex. Literature on retinoscopy especially in keratoconus is limited. In this study we compared retinoscopy with three other diagnostic methods, which are used in clinical routine. Amsler criteria, ORA and Pentacam refer to clinical, biomechanical and tomographic signs of keratoconus and were used as reference. 2. Patients and methods 2.1. Setting This prospective study was conducted at the Department of Ophthalmology, Saarland University Medical Center, Homburg/Saar, Germany. 2.2. Patients Patients were included non-selectively at their first visit to the Homburg Keratoconus Center with an ophthalmological referral for
235
suspected keratoconus or an endocrine referral of thyroid gland disease. Therefore, diagnosis of keratoconus had not been established before and none of the patients had any eye surgery in the past. In addition to the consent to the presented study, the patients gave informed consent to be included into the HKC database. The Helsinki declaration was followed. All patients underwent a full ophthalmological examination including refractometry, slit lamp biomicroscopy, tomography, examination of corneal biomechanics and retinoscopy. Retinoscopy was performed by one experienced pediatric ophthalmologist (BKK) to ensure identical retinoscopic classification of keratoconus. Retinoscopy was performed in a blinded fashion without knowledge about the results of the other diagnostic methods. Measurements were performed without pharmacological stimulation. The data collection included the following: - Diagnosis of keratoconus was established by the Amsler– Krumeich classification [7], based on slit lamp biomicroscopy, astigmatism, corneal power and corneal thickness data from Pentacam (Oculus Optikgeräte GmbH, Heidelberg, Germany). - Retinoscopy: the scissor reflex in retinoscopy was classified into normal (grade 0), low (grade 1), mild (grade 2) or moderate (grade 3). In case of an atypical reflex or no visible movement grade 4 was assigned (Fig. 1). - Rotating slit imaging (Pentacam, Oculus Optikgeräte GmbH, Heidelberg Germany) was used for tomography. It is a clinically established modality to diagnose and follow-up keratoconus patients. A rotating camera with a monochromatic slit light source obtains the tomography of the cornea and anterior eye segment. The Pentacam software provides a series of keratoconus
Fig. 1. Light shadow movement in retinoscopy. In myopia an “against” reflex is visible. The direction of the light movement observed from the retina is different to the light beam. In astigmatism a non-parallel light-shadow-movement is visible. Schematic demonstration of retinoscopy in keratoconus. No homogeneous light-shadow-movement is visible. This effect is called scissors reflex or fish mouth reflex.
236
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
Fig. 2. Distribution of keratoconus classification with 4 classification techniques. Although we rounded down the intermediate stages of TKC (e.g. 1–2 into 1), TKC overestimates in normal and moderate keratoconus stages compared to the other devices.
specific indices such as keratoconus index (KI) and topographic keratoconus classification (TKC). TKC classifies keratoconus into grade 0 (no keratoconus) up to grade 4 (severe keratoconus). In rare cases intermediate values such as 1–2 or 2–3 are displayed. We decided to accept the lower values (e.g. 1–2 to 1). - Keratoconus eyes typically show difference in biomechanical properties. The respective parameters corneal hysteresis (CH) and corneal resistance factor (CRF) are assessed with the ORA (Ocular Response Analyzer, Reichert Ophthalmic Instruments, Depew, NY, USA). The ORA impresses the cornea with an air puff similar to a classical pneumo-tonometer and records waveform of light of a LED reflected off the cornea [8]. Using software version 3.01 a normative database was developed at our institution and these waveform parameters were applied for keratoconus screening. The mathematical characterizations of the waveform signal such as height, slope, width, etc. were compared with the database. Two new keratoconus parameters were developed: KMI: keratoconus match index and KMP: keratoconus match probability. The keratoconus match index (KMI) is a result of a neuronal network calculation of seven parameters. It compares the waveform of the patient’s eye against the waveform classified samples in a
look-up table. The KMP quantifies keratoconus into normal, suspect, mild, moderate or severe [9]. The disposition with the highest percentage of probability was considered as keratoconus stage (from 0 = normal to 4 = severe). 2.3. Statistical analysis Statistical analysis was performed using SPSS software (SPSS version 19.0, IBM, USA). Descriptive evaluation of all data was performed using mean, standard deviation, median and minimum/maximum values. We performed Pearson’s chi-squared test to test the quality of fit, whether or not an observed frequency distribution differs from the theoretical distribution. Cross tables were used to test the matching between categorical variables of different instruments. We applied a power analysis to estimate the required number of patients to include in this study (N ≥ 100). 3. Results One hundred twenty-six eyes of 126 patients were enrolled in this study. Mean age was 37 years (11–75 years). Thirty-three percent of the patients were female. The distribution keratoconus stage
Table 1 Cross tables for keratoconus classification based on retinoscopy reflex, Amsler classification, TKC and KMP for grade 0 (normal), 1 (suspect), 2 (mild), 3 (moderate) and 4 (severe).
Amsler classification
Retinoscopy
KMP
TKC
0 1 2 3 4 0 1 2 3 4 0 1 2 3 4
TKC
KMP
0
1
2
3
4
0
1
2
3
4
0
1
2
3
4
28 10 1 0 0 24 10 1 1 0 33 2 1 1 0
6 14 11 3 2 7 14 13 2 0 5 13 17 1 0
0 7 17 9 2 1 8 15 8 3 0 2 13 19 1
0 1 5 4 2 0 1 2 4 4 0 0 3 7 2
0 1 0 1 2 0 1 1 0 2 0 0 0 3 1
24 11 3 0 0 19 6 6 1 0
2 6 5 2 2 8 12 7 3 4
6 11 9 6 2 5 10 12 4 1
0 5 16 6 4 0 4 6 4 1
0 0 1 3 0 0 0 3 5 2
19 6 6 1 0
8 12 7 3 4
5 10 12 4 1
0 4 6 4 1
0 0 3 5 2
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
classification for Amsler classification, retinoscopy, KMP and TKC is shown in Fig. 2. Table 1 shows cross tables for retinoscopy, Amsler classification, TKC and KMP in the five different stages. Summarizing all stages the classification of retinoscopy and Amsler is congruent in 65 of 126 eyes (51.6%), retinoscopy and TKC in 45 of 124 (36.3%) and retinoscopy and KMP in 19 of 123 eyes (39.8%). Amsler and TKC are congruent in 67 of 124 (54.0%), Amsler and KMP in 59 of 122 (48.4%) and TKC and KMP are congruent in 64 of 120 eyes (53.3%). Retinoscopy underestimates keratoconus stages compared to TKC (35:44) and KMP (33:41), but overestimates compared to Amsler (40:21). Amsler underestimate the keratoconus stage compared to retinoscopy (21:40), TKC (13:44) and KMP (21:42). TKC overestimates compared to KMP (31:25). Fig. 3 demonstrates the results of the cross tables. Four diagrams illustrate the congruence of the four classifications. The respective reference is named on the top of the diagram. Each single diagram shows the coincidence of classification for normal, suspect, mild, moderate and severe keratoconus in percentage. In all four diagrams a high congruence was found in normal eyes – up to 87.0%. The higher the keratoconus stage, the lower is the congruence, e.g. 0% regarding the comparison of retinoscopy and TKC for severe stage. All classification techniques either based on clinical findings (Amsler, retinoscopy) or objective measures (Pentacam, ORA) showed a poor performance in classifying keratoconus. 2 -Test yields the following results: for retinoscopy and Amsler/TKC/KMP 2 was 95.56/69.8/51.28. For Amsler classification and TKC/KMP 2 was 145.7/91.02. For TKC and KMP 2 it was 112.7. All of these respective p-values were <0.001. If classification data were condensed to binary decisions (normal vs. any stage of keratoconus) the respective results are demonstrated in Table 2. The selectivity between retinoscopy and Amsler criteria it was 98.6%, for retinoscopy and TKC it was 88.8%, for retinoscopy and KMP it was 82.4%. For sensitivity and specificity the following values were observed: for retinoscopy and Amsler
237
Table 2 Cross table for binary decision into normal and keratoconus: retinoscopy, Amsler, TKC and KMP for normals (n) and keratoconus (kc). Amsler classification
TKC
KMP
n
kc
n
kc
n
kc
19 7 75.5%
5 18
Retinoscopy n kc Test accuracy
28 1 98.1%
0 23
24 3 81.2%
6 15
KMP n kc Test accuracy
24 2 93.8%
1 21
19 7 75.5%
5 18
TKC n kc Test accuracy
33 2 96.4%
0 21
(98.8%/100%), retinoscopy and TKC (94.0%/80.0%) and retinoscopy and KMP (84.4%/79.1%). 4. Discussion In modern clinical routine classical examination techniques such as retinoscopy are more and more replaced by highly sophisticated topographical and tomographical examination techniques. In special situations, e.g. examination of non-compliant patients, children, disabled patients or lack of high-tech instruments, retinoscopy is still a powerful technique for the diagnosis of keratoconus and should not be considered anachronistic. Tomography systems, such as Pentacam, are used for many years and explored by numerous researchers, either alone or in comparison with other devices [10–14]. The ORA and the Corvis (Oculus Optikgeräte GmbH, Heidelberg, Germany) are relatively new instruments to
Fig. 3. Matching plots for keratoconus classifications based on retinoscopy reflex, Amsler classification, TKC and KMP for grade 0 (normal), 1 (suspect), 2 (mild), 3 (moderate) and 4 (severe). The matching in % is shown for each measurement device in comparison with the other devices. Only in normal eyes a good matching is seen. In the other stages the matching is mostly less than 50%.
238
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
measure biomechanical properties [15,16]. ORA in the newest software version 3.0 provided an additional keratoconus diagnostic module, which is under test in many current publications [9,17–20]. No previous study compared retinoscopy with modern keratoconus diagnostic devices. The overall distributions of keratoconus stages of the four different screening indices were similar. Only slight differences in the distribution were found. All four devices showed a comparable number of normal eyes. The distributions of retinoscopy and KMP were similar in all stages. In contrast, Amsler criteria classified mostly into normal, suspects and mild. Only a few cases were defined as moderate or severe. TKC found the lowest rate of severe cases. TKC classified most eyes into normal, only few cases were classified as suspect. All four classification techniques, the clinical based retinoscopy and Amsler as well as the instrument based techniques of tomography and biomechanic measurement, yield keratoconus stages from normal up to severe keratoconus. But up to now there is no gold standard for classification, and therefore each measurement system of grading follows its own philosophy and the results are not highly consistent. Although the distribution of the five stages is similar throughout the different classification systems, the coincidence of the systems differs significantly. Comparing retinoscopy to other techniques the coincidence is best for normals but insufficient for keratoconus stages from suspect to severe. Whereas in normal eyes a coincidence is present up to 89.2%, the coincidence declines rigorously with keratoconus down to 0%. Overall retinoscopy fits mostly with Amsler criteria. Surprisingly, all the classification techniques considered in this paper, the clinically based such as the retinoscopy or Amsler grading as well as the instrument based techniques (Pentacam TKC and ORA KMP) yield unexpectedly inconsistent results. Consequentially they cannot be used interchangeably! As a clinical consequence, for follow-up studies in a longitudinal fashion trend of keratoconus severity can only be quoted properly if the same classification technique is used throughout the monitoring. If, however, we condense our assessment to a binary decision, i.e. whether a situation should be quoted as “normal” or “keratoconus” regardless of the keratoconus stage, all techniques presented in this paper show a good performance and the selectivity of the techniques if comparing one to the other is very high, which means that for this simplified classification (“yes” or “no”) the techniques can be used interchangeably. Labiris et al. compared keratoconus patients and healthy patients with ORA and found no false-positive results in the healthy group, but 22% were classified as suspect. In our study we found 27% of suspect cases with KMP. They encountered 7% false-negatives in the keratoconus group and almost 24% were classified as suspects. They suggested that regarding the high rate of KMP suspects in the normal group, the software shows poor performance, which limits KMP’s clinical value in discriminating KC from normal corneas [9]. The most crucial problem of any study testing diagnostic tools for keratoconus is the unavailability of a real ‘gold standard’ for classifying keratoconus. As long as such a standard is missed a “true” grading still lacks. Furthermore the four classifications are based on different keratoconus attributes, such as biomechanical or topographical measures, retinoscopy and clinical aspects. Therefore, these measurements do not claim comparability. Especially in our HKC we compare the results of all the examinations in case of a preclinical keratoconus not seen yet on the slit lamp. Retinoscopy is a very subjective method to measure. We reduced the interobserver variability by restricting to one experienced examiner performing retinoscopy. Theoretically, scissor reflex occurs only due to differences in optical refraction. They should be low in early keratoconus stages and thus are difficult to detect. However, clinical significant keratoconus can be reliably detected. In this study we
did not address changes in the crystalline lens. In elderly patients the examiner should be aware, that nuclear sclerosis may alter the retinoscopy reflex and mimic the “scissors” reflex. In summary, retinoscopy shows a weak congruence with the other devices in classifying keratoconus stages. It fits mostly with Amsler criteria. Surprisingly other clinically based techniques such as Amsler classification or instrument based techniques such as TKC and KMP do not show a better performance. In contrast, in discriminating normals and keratoconus (stage mild to severe) retinoscopy performs very well compared to all other classification techniques. Therefore, we conclude, that retinoscopy – performed by an experienced examiner – is a reliable option in discriminating keratoconus from normals. It is a low-cost diagnostic technique, which can be used especially for children or non-compliant patients such as individuals with Down’s syndrome. It shows a poor performance in resampling the keratoconus classification of sophisticated instruments such as Pentacam or ORA, but even these instruments cannot be used interchangeably due to inconsistent classification. Therefore, follow-up examinations or monitoring of keratoconus progression should be always based on the same keratoconus classification techniques throughout the study to ensure comparability. In our opinion it is mandatory to form an international expert committee to decide which classification should be defined “the gold standard” for classification of keratoconus. Conflict of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. References [1] Rabinowitz YS. Keratoconus Surv Ophthalmol 1998;42:297–319. [2] Nielsen K, Hjortdal J, Pihlmann M, Corydon TJ. Update on the keratoconus genetics. Acta Ophthalmol 2013;91:106–13. [3] Bühren J, Bischoff G, Kohnen T. Keratoconus: clinical aspects, diagnosis, therapeutic possibilities. Klin Monbl Augenheilkd 2011;228:923–40. [4] Goebels S, Eppig T, Seitz B, Langenbucher A. Detection of early forms of keratoconus – current screening methods. Klin Monbl Augenheilkd 2013;230:998–1004. [5] Jonathan D, Wirtschafter J, Schwartz G. Retinoscopy. In: William T, editor. Duane’s foundation of clinical ophthalmology. JB Lippincott: Philadelphia; 1982 [chapter 37]. [6] Corboy JM. The retinoscopy book. 5th ed. New Jersey: Slack Incorporated; 2003. p. 4–6. [7] Kaufman HE, Baron BA, McDonald MB. The Cornea. 2nd ed. Boston: Butterworth-Heinemann Ltd; 1998. p. 560–4. [8] Goebels SC, Seitz B, Langenbucher A. Precision of ocular response analyzer. Curr Eye Res 2012;37:689–93. [9] Labiris G, Gatzioufas Z, Sideroudi H, Giarmoukakis A, Kozobolis V, Seitz B. Biomechanical diagnosis of keratoconus: evaluation of the keratoconus match index and the keratoconus match probability. Acta Ophthalmol (Copenh) 2013;91:258–62. ˜ [10] Pinero DP, Alió JL, Alesón A, Escaf Vergara M, Miranda M. Corneal volume, pachymetry, and correlation of anterior and posterior corneal shape in subclinical and different stages of clinical keratoconus. J Cataract Refract Surg 2010;36:814–25. [11] Fontes BM, Ambrósio Jr R, Salomão M, Velarde GC, Nosé W. Biomechanical and tomographic analysis of unilateral keratoconus. J Refract Surg 2010;26:677–81. [12] Nilforoushan M, Speaker M, Marmor M, Abramson J, Tullo W, Morschauser D, Latkany R. Comparative evaluation of refractive surgery candidates with Placido topography, Orbscan II, Pentacam, and wavefront analysis. J Cataract Refract Surg 2008;34:623–31. [13] Chen D, Lam AKC. Intrasession and intersession repeatability of the Pentacam system on posterior corneal assessment in the normal human eye. J Cataract Refract Surg 2007;33:448–54. [14] Kovács I, Miháltz K, Németh J, Nagy ZZ. Anterior chamber characteristics of keratoconus assessed by rotating Scheimpflug imaging. J Cataract Refract Surg 2010;36:1101–6. [15] Nemeth G, Hassan Z, Csutak A, Szalai E, Berta A, Modis Jr L. Repeatability of ocular biomechanical data measurements with a Scheimpflug-based noncontact device on normal corneas. J Refract Surg 2013;29:558–63. [16] Valbon BF, Ambrósio-Jr R, Fontes BM, Alves MR. Effects of age on corneal deformation by non-contact tonometry integrated with an ultra-high-speed (UHS) Scheimpflug camera. Arq Bras Oftalmol 2013;76:229–32.
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239 [17] Luce D. Determining in-vivo biomechanical properties of the cornea with an ocular response analyzer. J Cataract Refract Surg 2005;31:156–62. [18] Shah S, Laiquzzaman M, Bhojwani R, Mantry S, Cunliffe I. Assessment of biomechanical properties of the cornea with the ocular response analyzer in normal and keratoconic eyes. Invest Ophthalmol Vis Sci 2007;48: 3026–31.
239
[19] Schweitzer C, Roberts CJ, Mahmoud AM, Colin J, Maurice-Tison S, Kerautret J. Screening of forme fruste keratoconus with the ocular response analyzer. Invest Ophthalmol Vis Sci 2010;51:2403–10. [20] Kirwan C, Omalley D, O’ Keefe M. Corneal hysteresis and corneal resistance factor in keratectasia: finding using the Reichert ocular response analyser. Opthalmologica 2008;222:334–7.
Contents lists available at ScienceDirect
Contact Lens & Anterior Eye journal homepage: www.elsevier.com/locate/clae
Can retinoscopy keep up in keratoconus diagnosis? Susanne Goebels a,∗ , Barbara Käsmann-Kellner a , Timo Eppig b , Berthold Seitz a , Achim Langenbucher b a b
Department of Ophthalmology, Saarland University Medical Center, Kirrberger Strasse 100, Bldg. 22, 66421 Homburg, Germany Experimental Ophthalmology, Saarland University, Kirrberger Strasse 100, Bldg. 22, 66421 Homburg, Germany
a r t i c l e
i n f o
Article history: Received 23 October 2014 Received in revised form 27 January 2015 Accepted 29 January 2015 Keywords: Retinoscopy Tomography ORA Keratoconus Scissors reflex
a b s t r a c t Purpose: To evaluate the diagnostic potential of retinoscopy in comparison with Amsler-grading, Pentacam and Ocular Response Analyzer (ORA) in classifying keratoconus stages. Methods: Clinical examination, retinoscopy, Pentacam and ORA were performed in 126 patients. Data of Amsler, retinoscopy, topographic keratoconus classification (TKC) of Pentacam and keratoconus match probability (KMP) of ORA were analyzed. Each of these four classification techniques quotes keratoconus into stage 0 (normal) to 4 (severe). Descriptive analysis and cross tables were used to compare the different devices. Results: For retinoscopy the distribution in the five keratoconus grades normal/suspect/ mild/moderate/severe (in numbers) was 34/33/34/17/8. For Amsler it was 37/36/35/12/4, for TKC 38/17/34/31/4, for KMP 32/34/32/15/9. The cross tables show large classification differences of all devices. Overall, classification of retinoscopy and Amsler/TKC/KMP is congruent in 51.6%/36.3%/39.8% of the cases. Of all eyes, Amsler was congruent with TKC/KMP in 54.0%/48.4%, and TKC and KMP were congruent in 53.3%. In a binary decision (normal vs. any stage of mild/moderate/severe) matching between retinoscopy and Amsler/TKC/KMP was 98.6%/88.8%/82.4%. Sensitivity/specificity for retinoscopy and Amsler, TKC, KMP was 98.8%/94.0%, 84.4%/100% and 80.0%/79.1%. Conclusions: The congruence of keratoconus classification was very poor of all the techniques tested in our study. This applies to objective measures such as TKC, KMP as well as clinical classification techniques such as Amsler and retinoscopy. Compared to TKC and KMP, retinoscopy underestimates keratoconus stages. In contrast, the performance of binary decisions (normal vs. keratoconus) shows a high sensitivity and specificity. Retinoscopy, however, showed a clear clinical use in confirming the diagnosis of keratoconus. © 2015 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved.
1. Introduction Keratoconus is an ectatic non-inflammatory corneal disorder, in which the cornea forms a conic shape due to thinning of the corneal stroma. This corneal thinning normally induces irregular astigmatism, myopia and protrusion, leading to a mild to marked deterioration of visual performance [1]. Keratoconus is usually bilateral, mostly diagnosed in the 2nd or 3rd decade of life with an incidence of 55:100.000 [1–3], and more prominent in males than females. Diagnosis of keratoconus is easy in the presence of clinical signs such as corneal ectasia or thinning, scars, nodes, Vogt’s striae
∗ Corresponding author at: Department of Ophthalmology, Saarland University Medical Center, Kirrberger Strasse 100, Bldg. 22, 66421 Homburg/Saar, Germany. Tel.: +49 6841 1621240; fax: +49 6841 1622440. E-mail address: [email protected] (S. Goebels).
or Fleischer ring. In early stages (forme fruste) or unclear cases additional diagnostic tools are necessary. High-end technology devices are available to assist the examiner in the diagnosis and staging of keratoconus. Topography or tomography systems, aberrometry or assessment of corneal biomechanics as well as dedicated classification systems are available and actually in focus of research [4]. Beside these highly developed devices classical retinoscopy has been successfully established for decades to classify keratoconus. Although retinoscopy requires training, when having command on it, it is a valuable tool in children and/or handicapped or otherwise non-compliant patients or if sophisticated instruments are not available, it may be the case in developing regions. As keratoconus may be associated with some genetic syndromes as Down syndrome, Leber’s amaurosis or CDL-syndrome, keratoconus needs to be excluded or confirmed in patients who cannot fixate properly. The first idea of retinoscopy evolved in 1859 [5]. Sir William Bowman noted a linear shadow seen in an astigmatic eye using
http://dx.doi.org/10.1016/j.clae.2015.01.015 1367-0484/© 2015 British Contact Lens Association. Published by Elsevier Ltd. All rights reserved.
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
a plane mirror ophthalmoscope lit by a candle. In 1875 this phenomenon was described as shadow test. Early retinoscopes with electric lighting used rotating slit and helical filament lamps. A linear filament lamp, which produces a sharp, bright line of light was developed by Jacob Copeland [5,6]. Following investigations lead to better observing systems, control of vergence and handling. Retinoscopy enables measurement of the refractive state. In keratoconus or astigmatic eyes an inhomogeneous light-shadowmovement is visible. At the corneal center the refraction may be slightly myopic in comparison of the periphery [5]. Therefore, in keratoconus retinoscopy typically shows a “against” motion in the center and a “with” motion in the periphery. This opening and closing of the light reflex is called scissors reflex. Literature on retinoscopy especially in keratoconus is limited. In this study we compared retinoscopy with three other diagnostic methods, which are used in clinical routine. Amsler criteria, ORA and Pentacam refer to clinical, biomechanical and tomographic signs of keratoconus and were used as reference. 2. Patients and methods 2.1. Setting This prospective study was conducted at the Department of Ophthalmology, Saarland University Medical Center, Homburg/Saar, Germany. 2.2. Patients Patients were included non-selectively at their first visit to the Homburg Keratoconus Center with an ophthalmological referral for
235
suspected keratoconus or an endocrine referral of thyroid gland disease. Therefore, diagnosis of keratoconus had not been established before and none of the patients had any eye surgery in the past. In addition to the consent to the presented study, the patients gave informed consent to be included into the HKC database. The Helsinki declaration was followed. All patients underwent a full ophthalmological examination including refractometry, slit lamp biomicroscopy, tomography, examination of corneal biomechanics and retinoscopy. Retinoscopy was performed by one experienced pediatric ophthalmologist (BKK) to ensure identical retinoscopic classification of keratoconus. Retinoscopy was performed in a blinded fashion without knowledge about the results of the other diagnostic methods. Measurements were performed without pharmacological stimulation. The data collection included the following: - Diagnosis of keratoconus was established by the Amsler– Krumeich classification [7], based on slit lamp biomicroscopy, astigmatism, corneal power and corneal thickness data from Pentacam (Oculus Optikgeräte GmbH, Heidelberg, Germany). - Retinoscopy: the scissor reflex in retinoscopy was classified into normal (grade 0), low (grade 1), mild (grade 2) or moderate (grade 3). In case of an atypical reflex or no visible movement grade 4 was assigned (Fig. 1). - Rotating slit imaging (Pentacam, Oculus Optikgeräte GmbH, Heidelberg Germany) was used for tomography. It is a clinically established modality to diagnose and follow-up keratoconus patients. A rotating camera with a monochromatic slit light source obtains the tomography of the cornea and anterior eye segment. The Pentacam software provides a series of keratoconus
Fig. 1. Light shadow movement in retinoscopy. In myopia an “against” reflex is visible. The direction of the light movement observed from the retina is different to the light beam. In astigmatism a non-parallel light-shadow-movement is visible. Schematic demonstration of retinoscopy in keratoconus. No homogeneous light-shadow-movement is visible. This effect is called scissors reflex or fish mouth reflex.
236
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
Fig. 2. Distribution of keratoconus classification with 4 classification techniques. Although we rounded down the intermediate stages of TKC (e.g. 1–2 into 1), TKC overestimates in normal and moderate keratoconus stages compared to the other devices.
specific indices such as keratoconus index (KI) and topographic keratoconus classification (TKC). TKC classifies keratoconus into grade 0 (no keratoconus) up to grade 4 (severe keratoconus). In rare cases intermediate values such as 1–2 or 2–3 are displayed. We decided to accept the lower values (e.g. 1–2 to 1). - Keratoconus eyes typically show difference in biomechanical properties. The respective parameters corneal hysteresis (CH) and corneal resistance factor (CRF) are assessed with the ORA (Ocular Response Analyzer, Reichert Ophthalmic Instruments, Depew, NY, USA). The ORA impresses the cornea with an air puff similar to a classical pneumo-tonometer and records waveform of light of a LED reflected off the cornea [8]. Using software version 3.01 a normative database was developed at our institution and these waveform parameters were applied for keratoconus screening. The mathematical characterizations of the waveform signal such as height, slope, width, etc. were compared with the database. Two new keratoconus parameters were developed: KMI: keratoconus match index and KMP: keratoconus match probability. The keratoconus match index (KMI) is a result of a neuronal network calculation of seven parameters. It compares the waveform of the patient’s eye against the waveform classified samples in a
look-up table. The KMP quantifies keratoconus into normal, suspect, mild, moderate or severe [9]. The disposition with the highest percentage of probability was considered as keratoconus stage (from 0 = normal to 4 = severe). 2.3. Statistical analysis Statistical analysis was performed using SPSS software (SPSS version 19.0, IBM, USA). Descriptive evaluation of all data was performed using mean, standard deviation, median and minimum/maximum values. We performed Pearson’s chi-squared test to test the quality of fit, whether or not an observed frequency distribution differs from the theoretical distribution. Cross tables were used to test the matching between categorical variables of different instruments. We applied a power analysis to estimate the required number of patients to include in this study (N ≥ 100). 3. Results One hundred twenty-six eyes of 126 patients were enrolled in this study. Mean age was 37 years (11–75 years). Thirty-three percent of the patients were female. The distribution keratoconus stage
Table 1 Cross tables for keratoconus classification based on retinoscopy reflex, Amsler classification, TKC and KMP for grade 0 (normal), 1 (suspect), 2 (mild), 3 (moderate) and 4 (severe).
Amsler classification
Retinoscopy
KMP
TKC
0 1 2 3 4 0 1 2 3 4 0 1 2 3 4
TKC
KMP
0
1
2
3
4
0
1
2
3
4
0
1
2
3
4
28 10 1 0 0 24 10 1 1 0 33 2 1 1 0
6 14 11 3 2 7 14 13 2 0 5 13 17 1 0
0 7 17 9 2 1 8 15 8 3 0 2 13 19 1
0 1 5 4 2 0 1 2 4 4 0 0 3 7 2
0 1 0 1 2 0 1 1 0 2 0 0 0 3 1
24 11 3 0 0 19 6 6 1 0
2 6 5 2 2 8 12 7 3 4
6 11 9 6 2 5 10 12 4 1
0 5 16 6 4 0 4 6 4 1
0 0 1 3 0 0 0 3 5 2
19 6 6 1 0
8 12 7 3 4
5 10 12 4 1
0 4 6 4 1
0 0 3 5 2
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
classification for Amsler classification, retinoscopy, KMP and TKC is shown in Fig. 2. Table 1 shows cross tables for retinoscopy, Amsler classification, TKC and KMP in the five different stages. Summarizing all stages the classification of retinoscopy and Amsler is congruent in 65 of 126 eyes (51.6%), retinoscopy and TKC in 45 of 124 (36.3%) and retinoscopy and KMP in 19 of 123 eyes (39.8%). Amsler and TKC are congruent in 67 of 124 (54.0%), Amsler and KMP in 59 of 122 (48.4%) and TKC and KMP are congruent in 64 of 120 eyes (53.3%). Retinoscopy underestimates keratoconus stages compared to TKC (35:44) and KMP (33:41), but overestimates compared to Amsler (40:21). Amsler underestimate the keratoconus stage compared to retinoscopy (21:40), TKC (13:44) and KMP (21:42). TKC overestimates compared to KMP (31:25). Fig. 3 demonstrates the results of the cross tables. Four diagrams illustrate the congruence of the four classifications. The respective reference is named on the top of the diagram. Each single diagram shows the coincidence of classification for normal, suspect, mild, moderate and severe keratoconus in percentage. In all four diagrams a high congruence was found in normal eyes – up to 87.0%. The higher the keratoconus stage, the lower is the congruence, e.g. 0% regarding the comparison of retinoscopy and TKC for severe stage. All classification techniques either based on clinical findings (Amsler, retinoscopy) or objective measures (Pentacam, ORA) showed a poor performance in classifying keratoconus. 2 -Test yields the following results: for retinoscopy and Amsler/TKC/KMP 2 was 95.56/69.8/51.28. For Amsler classification and TKC/KMP 2 was 145.7/91.02. For TKC and KMP 2 it was 112.7. All of these respective p-values were <0.001. If classification data were condensed to binary decisions (normal vs. any stage of keratoconus) the respective results are demonstrated in Table 2. The selectivity between retinoscopy and Amsler criteria it was 98.6%, for retinoscopy and TKC it was 88.8%, for retinoscopy and KMP it was 82.4%. For sensitivity and specificity the following values were observed: for retinoscopy and Amsler
237
Table 2 Cross table for binary decision into normal and keratoconus: retinoscopy, Amsler, TKC and KMP for normals (n) and keratoconus (kc). Amsler classification
TKC
KMP
n
kc
n
kc
n
kc
19 7 75.5%
5 18
Retinoscopy n kc Test accuracy
28 1 98.1%
0 23
24 3 81.2%
6 15
KMP n kc Test accuracy
24 2 93.8%
1 21
19 7 75.5%
5 18
TKC n kc Test accuracy
33 2 96.4%
0 21
(98.8%/100%), retinoscopy and TKC (94.0%/80.0%) and retinoscopy and KMP (84.4%/79.1%). 4. Discussion In modern clinical routine classical examination techniques such as retinoscopy are more and more replaced by highly sophisticated topographical and tomographical examination techniques. In special situations, e.g. examination of non-compliant patients, children, disabled patients or lack of high-tech instruments, retinoscopy is still a powerful technique for the diagnosis of keratoconus and should not be considered anachronistic. Tomography systems, such as Pentacam, are used for many years and explored by numerous researchers, either alone or in comparison with other devices [10–14]. The ORA and the Corvis (Oculus Optikgeräte GmbH, Heidelberg, Germany) are relatively new instruments to
Fig. 3. Matching plots for keratoconus classifications based on retinoscopy reflex, Amsler classification, TKC and KMP for grade 0 (normal), 1 (suspect), 2 (mild), 3 (moderate) and 4 (severe). The matching in % is shown for each measurement device in comparison with the other devices. Only in normal eyes a good matching is seen. In the other stages the matching is mostly less than 50%.
238
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239
measure biomechanical properties [15,16]. ORA in the newest software version 3.0 provided an additional keratoconus diagnostic module, which is under test in many current publications [9,17–20]. No previous study compared retinoscopy with modern keratoconus diagnostic devices. The overall distributions of keratoconus stages of the four different screening indices were similar. Only slight differences in the distribution were found. All four devices showed a comparable number of normal eyes. The distributions of retinoscopy and KMP were similar in all stages. In contrast, Amsler criteria classified mostly into normal, suspects and mild. Only a few cases were defined as moderate or severe. TKC found the lowest rate of severe cases. TKC classified most eyes into normal, only few cases were classified as suspect. All four classification techniques, the clinical based retinoscopy and Amsler as well as the instrument based techniques of tomography and biomechanic measurement, yield keratoconus stages from normal up to severe keratoconus. But up to now there is no gold standard for classification, and therefore each measurement system of grading follows its own philosophy and the results are not highly consistent. Although the distribution of the five stages is similar throughout the different classification systems, the coincidence of the systems differs significantly. Comparing retinoscopy to other techniques the coincidence is best for normals but insufficient for keratoconus stages from suspect to severe. Whereas in normal eyes a coincidence is present up to 89.2%, the coincidence declines rigorously with keratoconus down to 0%. Overall retinoscopy fits mostly with Amsler criteria. Surprisingly, all the classification techniques considered in this paper, the clinically based such as the retinoscopy or Amsler grading as well as the instrument based techniques (Pentacam TKC and ORA KMP) yield unexpectedly inconsistent results. Consequentially they cannot be used interchangeably! As a clinical consequence, for follow-up studies in a longitudinal fashion trend of keratoconus severity can only be quoted properly if the same classification technique is used throughout the monitoring. If, however, we condense our assessment to a binary decision, i.e. whether a situation should be quoted as “normal” or “keratoconus” regardless of the keratoconus stage, all techniques presented in this paper show a good performance and the selectivity of the techniques if comparing one to the other is very high, which means that for this simplified classification (“yes” or “no”) the techniques can be used interchangeably. Labiris et al. compared keratoconus patients and healthy patients with ORA and found no false-positive results in the healthy group, but 22% were classified as suspect. In our study we found 27% of suspect cases with KMP. They encountered 7% false-negatives in the keratoconus group and almost 24% were classified as suspects. They suggested that regarding the high rate of KMP suspects in the normal group, the software shows poor performance, which limits KMP’s clinical value in discriminating KC from normal corneas [9]. The most crucial problem of any study testing diagnostic tools for keratoconus is the unavailability of a real ‘gold standard’ for classifying keratoconus. As long as such a standard is missed a “true” grading still lacks. Furthermore the four classifications are based on different keratoconus attributes, such as biomechanical or topographical measures, retinoscopy and clinical aspects. Therefore, these measurements do not claim comparability. Especially in our HKC we compare the results of all the examinations in case of a preclinical keratoconus not seen yet on the slit lamp. Retinoscopy is a very subjective method to measure. We reduced the interobserver variability by restricting to one experienced examiner performing retinoscopy. Theoretically, scissor reflex occurs only due to differences in optical refraction. They should be low in early keratoconus stages and thus are difficult to detect. However, clinical significant keratoconus can be reliably detected. In this study we
did not address changes in the crystalline lens. In elderly patients the examiner should be aware, that nuclear sclerosis may alter the retinoscopy reflex and mimic the “scissors” reflex. In summary, retinoscopy shows a weak congruence with the other devices in classifying keratoconus stages. It fits mostly with Amsler criteria. Surprisingly other clinically based techniques such as Amsler classification or instrument based techniques such as TKC and KMP do not show a better performance. In contrast, in discriminating normals and keratoconus (stage mild to severe) retinoscopy performs very well compared to all other classification techniques. Therefore, we conclude, that retinoscopy – performed by an experienced examiner – is a reliable option in discriminating keratoconus from normals. It is a low-cost diagnostic technique, which can be used especially for children or non-compliant patients such as individuals with Down’s syndrome. It shows a poor performance in resampling the keratoconus classification of sophisticated instruments such as Pentacam or ORA, but even these instruments cannot be used interchangeably due to inconsistent classification. Therefore, follow-up examinations or monitoring of keratoconus progression should be always based on the same keratoconus classification techniques throughout the study to ensure comparability. In our opinion it is mandatory to form an international expert committee to decide which classification should be defined “the gold standard” for classification of keratoconus. Conflict of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. References [1] Rabinowitz YS. Keratoconus Surv Ophthalmol 1998;42:297–319. [2] Nielsen K, Hjortdal J, Pihlmann M, Corydon TJ. Update on the keratoconus genetics. Acta Ophthalmol 2013;91:106–13. [3] Bühren J, Bischoff G, Kohnen T. Keratoconus: clinical aspects, diagnosis, therapeutic possibilities. Klin Monbl Augenheilkd 2011;228:923–40. [4] Goebels S, Eppig T, Seitz B, Langenbucher A. Detection of early forms of keratoconus – current screening methods. Klin Monbl Augenheilkd 2013;230:998–1004. [5] Jonathan D, Wirtschafter J, Schwartz G. Retinoscopy. In: William T, editor. Duane’s foundation of clinical ophthalmology. JB Lippincott: Philadelphia; 1982 [chapter 37]. [6] Corboy JM. The retinoscopy book. 5th ed. New Jersey: Slack Incorporated; 2003. p. 4–6. [7] Kaufman HE, Baron BA, McDonald MB. The Cornea. 2nd ed. Boston: Butterworth-Heinemann Ltd; 1998. p. 560–4. [8] Goebels SC, Seitz B, Langenbucher A. Precision of ocular response analyzer. Curr Eye Res 2012;37:689–93. [9] Labiris G, Gatzioufas Z, Sideroudi H, Giarmoukakis A, Kozobolis V, Seitz B. Biomechanical diagnosis of keratoconus: evaluation of the keratoconus match index and the keratoconus match probability. Acta Ophthalmol (Copenh) 2013;91:258–62. ˜ [10] Pinero DP, Alió JL, Alesón A, Escaf Vergara M, Miranda M. Corneal volume, pachymetry, and correlation of anterior and posterior corneal shape in subclinical and different stages of clinical keratoconus. J Cataract Refract Surg 2010;36:814–25. [11] Fontes BM, Ambrósio Jr R, Salomão M, Velarde GC, Nosé W. Biomechanical and tomographic analysis of unilateral keratoconus. J Refract Surg 2010;26:677–81. [12] Nilforoushan M, Speaker M, Marmor M, Abramson J, Tullo W, Morschauser D, Latkany R. Comparative evaluation of refractive surgery candidates with Placido topography, Orbscan II, Pentacam, and wavefront analysis. J Cataract Refract Surg 2008;34:623–31. [13] Chen D, Lam AKC. Intrasession and intersession repeatability of the Pentacam system on posterior corneal assessment in the normal human eye. J Cataract Refract Surg 2007;33:448–54. [14] Kovács I, Miháltz K, Németh J, Nagy ZZ. Anterior chamber characteristics of keratoconus assessed by rotating Scheimpflug imaging. J Cataract Refract Surg 2010;36:1101–6. [15] Nemeth G, Hassan Z, Csutak A, Szalai E, Berta A, Modis Jr L. Repeatability of ocular biomechanical data measurements with a Scheimpflug-based noncontact device on normal corneas. J Refract Surg 2013;29:558–63. [16] Valbon BF, Ambrósio-Jr R, Fontes BM, Alves MR. Effects of age on corneal deformation by non-contact tonometry integrated with an ultra-high-speed (UHS) Scheimpflug camera. Arq Bras Oftalmol 2013;76:229–32.
S. Goebels et al. / Contact Lens & Anterior Eye 38 (2015) 234–239 [17] Luce D. Determining in-vivo biomechanical properties of the cornea with an ocular response analyzer. J Cataract Refract Surg 2005;31:156–62. [18] Shah S, Laiquzzaman M, Bhojwani R, Mantry S, Cunliffe I. Assessment of biomechanical properties of the cornea with the ocular response analyzer in normal and keratoconic eyes. Invest Ophthalmol Vis Sci 2007;48: 3026–31.
239
[19] Schweitzer C, Roberts CJ, Mahmoud AM, Colin J, Maurice-Tison S, Kerautret J. Screening of forme fruste keratoconus with the ocular response analyzer. Invest Ophthalmol Vis Sci 2010;51:2403–10. [20] Kirwan C, Omalley D, O’ Keefe M. Corneal hysteresis and corneal resistance factor in keratectasia: finding using the Reichert ocular response analyser. Opthalmologica 2008;222:334–7.