Slitlamp-adapted optical coherence tomogram (gray scale) of the anterior chamber angle (ACA) region. AOD indicates angle-opening distance; C, cornea; S, sclera; ST, scleral thickness; IT, iris thickness; and CB, ciliary body. Dotted lines indicate ACA; solid thin white arrow, AOD; black arrow, ST; dotted arrow, IT; and thick white arrow, scleral spur.
Distribution of the anterior chamber angle (ACA) derived from goniometry with optical coherence tomography (OCT) compared with gonioscopy. Box plots indicate the mean (central point), median (line in the box), 50% confidence interval (box), 90% confidence interval (error bar), and minimum and maximum values (upper and lower points).
Distribution of the angle-opening distance (AOD) derived from goniometry with optical coherence tomography (OCT) compared with gonioscopy. For explanation of symbols, see legend to Figure 2.
Parameters of goniometry with optical coherence tomography (OCT) for the anterior chamber angle (ACA) and angle-opening distance (AOD). The regression line, line of equality (broken line), linear regression equation (y), correlation coefficient (r), P value, and number of cases (n) are displayed.
Wirbelauer C, Karandish A, Häberle H, Pham DT. Noncontact Goniometry With Optical Coherence Tomography. Arch Ophthalmol. 2005;123(2):179-185. doi:10.1001/archopht.123.2.179
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
To assess the value of noncontact goniometry with optical coherence tomography (OCT) compared with current clinical parameters in the evaluation of the anterior chamber angle (ACA).
Prospective observational study of 138 eyes of 109 patients.
The ACA parameters and angle-opening distance (AOD) were measured with slitlamp-adapted OCT goniometry. The iris and scleral thickness and the iris convexity were assessed with OCT. Both ACA and AOD were compared with the clinical parameters of gonioscopy grade, limbal anterior chamber depth (ACD), ultrasonographic central ACD, and lens-axial length (LAL) ratio.
Noncontact goniometry with OCT revealed mean ± SD values of 28° ± 16° for the ACA and 381 ± 234 μm for the AOD. The mean ± SD iris thickness was 369 ± 84 μm, and the scleral thickness at the scleral spur was 943 ± 148 μm. There was a significant correlation (P<.001) with the clinical parameters of gonioscopic grading, limbal ACD, ultrasonographic central ACD, and LAL ratio. The sensitivity and specificity of OCT goniometry to detect an occludable angle were 86% and 95% for ACA and 85% and 90% for AOD, respectively.
Noncontact goniometry with OCT was helpful in evaluating the anterior chamber structures and as a screening modality. Goniometry with OCT could improve the noninvasive clinical assessment and treatment of patients with glaucoma.
Glaucoma is one of the most frequent causes of preventable blindness. Approximately 66.8 million people worldwide are estimated to have primary open-angle glaucoma (POAG) or primary angle-closure glaucoma (PACG), with approximately 6.7 million (10%) being blind from glaucoma.1,2 The use of gonioscopy lenses has helped physicians to understand the normal angle, diagnose angle abnormalities, and treat ocular diseases. Gonioscopy is crucial for the classification of patients with glaucoma,3 and its purpose is to permit visualization of the iridocorneal angle and assess the aqueous outflow through the trabecular meshwork. Patients at risk are those with occludable or narrow angles with a shallow anterior chamber.4- 7 Recent reports8,9 revealed that PACG is a significant cause of ocular morbidity among people of East Asian origin. China and India, as the world’s most populous countries, reveal a higher incidence of occludable angles and risk of developing PACG.10 Therefore, the prospect exists for developing new diagnostic strategies to prevent blindness caused by PACG.1,10,11
The evaluation of gonioscopic findings is subjective, semiquantitative, and dependent on the examiner’s experience.3,12- 14 Although gonioscopic grading systems3,12,13,15 and biometric gonioscopy6 allow semiquantitative measurements of the anterior chamber angle (ACA) width, a precise and objective assessment is not possible with gonioscopy. Ultrasound biomicroscopy (UBM) has been shown to represent the ACA width objectively and quantitatively, but it is time-consuming and relatively invasive when an immersion bath is used.4,5,16 Optical coherence tomography (OCT), a new noninvasive and high-resolution imaging technique, has been shown in experimental and preclinical studies to objectively monitor changes of the anterior segment structures and obtain images of the ACA structures in a noncontact mode.17- 22 In previous clinical evaluations, we demonstrated that slitlamp-adapted OCT enabled reproducible cross-sectional and high-resolution imaging of the cornea23- 26 and ACA.27,28 In this prospective clinical study, noncontact goniometry with OCT was compared with currently used ACA parameters.
A total of 138 eyes of 109 patients were prospectively included in this study. The mean ± SD age of the patients was 66 ± 15 years (range, 23-90 years). The female-male ratio was 66:43. An experienced examiner performed a complete ophthalmologic examination on each patient, including best-corrected visual acuity measured in decimal fraction, manifest refractions, objective refraction (automatic refractor and keratometer, model 559; Carl Zeiss Meditec, Dublin, Calif), slitlamp biomicroscopy, intraocular pressure with applanation tonometry, and indirect ophthalmoscopy. To determine the mean visual acuity, we calculated the logarithm of each of the sample values and then the antilogarithm of this mean. For a distance visual acuity of less than 0.05, a value of 0.01 was considered. Informed consent was obtained from each patient, and all patients were examined in accordance with the tenets of the Declaration of Helsinki of the World Medical Association regarding scientific research on human subjects. No institutional review board approval was requested for this clinical study.
Gonioscopy was performed using a slitlamp (magnification ×20) with a Goldmann 3-mirror contact lens (Haag-Streit, Bern, Switzerland), which provided an indirect reversed image of the angle with topical anesthesia (0.4% oxybuprocaine hydrochloride; Ciba Vision, Grossostheim, Germany). Before placement, a drop of 1% methylcellulose was placed on the corneal curve of the lens for optical continuity. All patients had both eyes examined at 360°, and the nasal and temporal measurements were tabulated. Gonioscopy was performed after all other testing so as not to adversely influence any measures, and the angle was assessed without compression. Gonioscopic grading of the angle in the nasal and temporal quadrants was performed in a darkened room with a standardized, narrow slit beam and the light angled at 45°. The grading included a thorough examination of the ACA structures and the iris configuration. To improve the ACA determination, we used a combination of all 3 current grading systems according to Scheie,12 Shaffer and Schwartz,3,15 and Spaeth.13 In this study, the angle width, which was determined at a point slightly central to a line dropped directly posteriorly from the Schwalbe line, was measured without indentation in 10° increments. A grade of 4 (35°-45°) referred to a wide open angle in which all structures were visible up to the iris root and its attachment to the anterior ciliary body. Grade 3 (20°-35°) referred to a wide open angle up to the scleral spur. In grades 3 and 4, no risk of angle closure existed.3,15 In grade 2 (20°), the angle was narrow with visible trabecular meshwork. In this angle width, a possible risk of closure existed. In grade 1 (<10°), the angle was extremely narrow up to the anterior trabecular meshwork and the Schwalbe line, with a high risk of probable closure. In grade 0 (0°), the angle was closed with iridocorneal contact and no visibility of the ACA structures. The peripheral iris convexity was graded in the nasal and temporal quadrants as regular without significant arching of the iris, convex curved with plateau appearance as steep, or with concave curvature.13
The limbal anterior chamber depth (ACD), as an estimate of the angle width, was determined at the temporal limbus with a slitlamp angle of 60°, according to the Van Herick grading system.29 This semiquantitative grading method allowed us to assess the ACD by comparing it as a fraction of the limbal corneal thickness. A value of 4 described a depth equal to or greater than the corneal thickness; 3, a depth of 50% of corneal thickness; 2, a depth of 25% of corneal thickness; and 1, a depth smaller than 25% of corneal thickness. A closed angle had an absent peripheral anterior chamber.
Contact ultrasonographic biometry was performed in 123 eyes with a 10-MHz A-mode hard-tipped probe (A+ Auto-Scan; Sonomed Inc, Lake Success, NY) mounted on a tonometer (Haag-Streit), which was set to the intraocular pressure. The mean of 5 measurements was further processed. The lens-axial length (LAL) ratio (lens/axial length × 10) was also calculated from the axial eye length and lens thickness.30
All patients were examined with a slitlamp-adapted OCT system (AS-OCT; 4Optics AG, Lübeck, Germany). This clinical medical device was in conformity with the essential requirements based on certification by DIN EN ISO 9001, DIN EN ISO 46001, and DIN EN ISO 13485. To enhance the OCT image for the anterior eye segment, we used a light source with a superluminescent diode (SLD-561; Superlum Diodes Ltd, Moscow, Russia) that operated at a wavelength of λ = 1310 nm with a bandwidth of 50 nm and an intensity of the incident light of less than 200 μW.26,27 This corresponded to a longitudinal resolution in air of 15 μm and approximately 11 μm within the ocular tissue. The sample arm of the interferometer and the scanning module were integrated in the projected slit of a standard clinical slitlamp (SL-3C; Topcon Corp, Tokyo, Japan) as described previously.23- 27 This enabled us to reliably adjust the OCT infrared light on the structures to be examined.
For the tomographic representation of the ACA, the reflected light was analyzed, and the intensity was converted to logarithmic gray-scaled images, which were simultaneously recorded during the measurements (Figure 1). The total acquisition time depended on the line scan frequency of 60 Hz and on the variable range of scans from 100 to 400. The focus diameter of the measurement beam was 20 μm, the lateral width of the scan was 6.0 mm, and the depth of each scan in air was 2.0 mm, with a scanning data acquisition time of 2 seconds. The resulting cross-sectional tomographic images had 360 × 200 pixels with a digital-sampling increment of 5.6 μm axially and 30 μm laterally. The lateral resolution was limited by the separation between 2 adjacent scans and the total scan width.
All OCT measurements were performed perpendicularly to the ocular surface with the slitlamp aligned at a 45° angle. This allowed a perpendicular radial projection of the ACA region and reduced possible image distortions. The nasal and temporal angles were studied owing to ease of access in the horizontal meridian. Furthermore, to reduce the influence of the peripheral iris structures, we performed all measurements with the same lighting conditions used for gonioscopy. The cross-sectional OCT image with the best quality was further analyzed.
The determination of the ACA and the angle opening distance (AOD) was adapted from UBM to the OCT method.31- 33 All measurements were performed manually with a specific program that enabled the creation of an angle and measured the distance between the optical signals with the highest reflectivity at the tissue boundaries (OCTeval, version 1.1; 4Optics AG). On the basis of different acoustic or optical tissue signal densities, the most important landmark and reference point for all measurements was the exact location of the scleral spur, which can be consistently defined by its peaked outline, slight projection into the anterior chamber, and appearance at the posterior border of the trabecular meshwork.31- 33 In OCT, imaging of the scleral spur, which is composed of a ring of collagen fibers that run parallel to the limbus, revealed a highly reflective structure with a reflectivity similar to the sclera (Figure 1).20 The AOD was defined as the distance from the posterior corneal surface to the anterior iris perpendicular to a line drawn along the trabecular meshwork at a distance of 500 μm from the scleral spur. The posterior cornea and the opposite peripheral iris were also the measurement points to create the arms of the ACA, with the apex lying in the angle recess (Figure 1). The iris convexity was graded from anterior cross-sectional OCT images with the slitlamp set at a 0° angle, according to the classification proposed by Spaeth et al.13,34 Furthermore, the scleral thickness was measured perpendicularly to the episcleral surface from the optical scleral spur, and the iris thickness was recorded at 500 μm anterior to the scleral spur. The optical delay values obtained were then divided by the group refractive index (n = 1.33) to obtain the geometric distances between the tissue interfaces. Then OCT goniometry was compared with the clinical parameters of semiquantitative gonioscopic grading, limbal ACD, ultrasonic central ACD, and the LAL ratio. All measurement points were evaluated separately.
All results are presented as mean ± SD and range. The statistical comparison to evaluate the association between the individual clinical parameters and OCT goniometry was performed with linear regression analysis and the Pearson correlation coefficient. P < .01 was considered statistically significant.
Sensitivity and specificity were calculated for the detection of occludable angles (ie, a gonioscopy grade of ≤2). The discriminant numbers for OCT goniometry were an ACA lower than 22° or an AOD lower than 290 μm. For the clinical parameters, the cutoff values were grade 2 for the limbal ACD,29 2.7 mm for the central ACD,35 and 2.2 for the LAL ratio.30
In this study, a wide variation of clinical and ACA changes were assessed. The demographics of the study patients are given in Table 1. Four patients had acute angle-closure glaucoma, 3 patients had Posner-Schlossmann syndrome, 1 patient had Rieger syndrome, and 1 patient had aphakia. There was a closed ACA in 11 eyes (8%).
The manifest spherical equivalent refraction was 0.13 ± 3.46 diopters (D) (range, –14.13 to 9.38 D). The best-corrected visual acuity was 0.39 ± 0.31 (decimal fraction) (range, 0.01-1.25). The intraocular pressure was 21 ± 11 mm Hg (range, 8-75 mm Hg). The values for gonioscopic grading, the limbal ACD, the ultrasonographic axial eye length, central ACD, and the calculated LAL ratio, are given in Table 2.
The OCT value was 28° ± 16° (range, 0°-68°) for ACA and 381 ± 234 μm (range, 0-1154 μm) for AOD (Table 2). The iris thickness was 369 ± 84 μm (range, 173-766 μm), and the scleral thickness at the optical scleral spur was 943 ± 148 μm (range, 641-1528 μm).
The comparison of the clinical parameters with OCT goniometry revealed a significant correlation (Table 3). The correlation was highest for the ACA and AOD compared with gonioscopy, with values of 0.85 (P<.001) and 0.80 (P<.001), respectively (Figure 2 and Figure 3). There was a significant negative correlation compared with the LAL ratio (Table 3).
In Table 4, the OCT and clinical parameters are given according to the gonioscopy grades. There was a linear relationship between gonioscopy and OCT values, limbal ACD, and central ACD, with a decrease in lower angle widths. The mean LAL ratio increased progressively from 1.95 in patients with wide open angles to 2.36 in patients with closed angles (Table 4). The comparison of the ACA and AOD revealed a good correlation between angular and linear parameters (Figure 4), with an increasing correlation in occludable and narrow angles (Table 5). Because of changes in the peripheral iris shape, 3 cases (1.08%) were manifested as an ACA of 0°, with an AOD ranging from 76 to 141 μm, which revealed partial drainage of the trabecular meshwork.
The distribution of the shape of the peripheral iris is given in Table 6. There was a positive correlation of 0.66 (P<.001) between OCT goniometry and the Spaeth specification.13,34 However, the correlation for the qualitative iris configuration was lower than for the quantitative OCT goniometry values of the ACA and depth.
For the ACA, the detection of occludable angles revealed a sensitivity of 86% (105/122) and a specificity of 95% (147/154) (Table 7). The AOD had a sensitivity of 85% (104/122) and a specificity of 90% (139/154). The results suggest that the determination of the ACA or AOD with OCT goniometry was more effective in detecting an occludable angle compared with the clinical parameters.
The main objective of gonioscopy is the visualization of the ACA, and the most common goal of gonioscopy is to determine whether the angle is open, closed, or occludable. In this study, we assessed a wide variation of ACA changes and confirmed the clinical validity of noncontact OCT goniometry. The anterior segment geometry provided ACA and iris convexity as angular parameters and AOD, iris thickness, and scleral thickness as linear parameters. Thus, it was possible to follow angular and linear biometric changes in the angle region, and the values correlated closely with the clinical parameters of ACA estimation such as gonioscopy, limbal ACD, and central ultrasonographic ACD.
The consistency of gonioscopic angle grading systems has been demonstrated,36,37 but the subjective nature of grading of ACAs and the lack of clearly defined cutoffs between normal and abnormal angles make comparisons difficult. Furthermore, gonioscopy grades provide only rough estimates, and the various angle grades can merge into one another, with a great variability in the normal appearance of the angle recess.13 One of the other disadvantages in the use of a goniolens is the direct contact with the cornea, which can lead to changes in corneal curvature and shifts in aqueous humor in the anterior chamber. This can cause a certain amount of distortion, which might influence the final grading of the angle and reduce the accuracy of the gonioscopic evaluation. To improve the gonioscopic grading, we used a modification of the methods described by Scheie,12 Shaffer and Schwartz,3,15 and Spaeth.13 The descriptive angular approach based on anatomic landmarks viewed on gonioscopy and the numerical estimation of the angle width was more objective.12 These values were then compared with ACA and AOD assessed with OCT goniometry. Additionally, the convexity of the iris was assessed as described by Spaeth et al.13,34 Previous ultrasound studies of healthy patients revealed an ACA of 30° ± 11°, an AOD at 500 μm from the scleral spur of 347 ± 181 μm, a peripheral iris thickness of 372 ± 58 μm, and a scleral thickness of 938 ± 58 μm.33 The biometric values of the ACA structures determined with OCT goniometry in our study compared favorably with the reported UBM values.33,38 Thus, OCT goniometry was revealed to be a useful tool for the noninvasive evaluation of the ACA, eliminating the risk of observer bias associated with clinical grading schemes.
Early studies14 that compared ultrasonography and gonioscopy revealed a high variability and a weak correlation in the assessment of the angle width. The correlation of gonioscopic grading with the width of the ACA on B-scan ultrasonography ranged from 0.641 to 0.791.14 Furthermore, a high variability was found, particularly in patients with narrow angles, which was partly related to plateau iris configurations.14 Pavlin et al31- 33 introduced high-resolution UBM, which enabled the recognition of the scleral spur as a highly reflective structure compared with the tissues of the ciliary body and cornea. This reference point of the human ACA configuration was proposed for more reproducible determinations and revealed a high correlation between gonioscopy and UBM in the determination of the angle width and peripheral iris curvature.16,34 However, a comparison of UBM and Scheimpflug photography for the determination of the angle width revealed only a relatively low correlation of 0.64.39 The variability was 3.1° (14%) for Scheimpflug photography and 4.7° (23%) for UBM in 20 healthy patients.39 Although Scheimpflug methods are noninvasive, the low resolution precluded an exact visualization of the angle structures and the scleral spur with possible optical distortions.39 In another UBM study,40 the mean intraobserver reproducibility to determine the ACA was 6.97%, and the AOD at 500 μm from the scleral spur was 7.27%. However, the interobserver reproducibility varied considerably in UBM and was also affected by subjective interpretation of the visualized anatomic landmarks. Therefore, previous authors16 have concluded that contact UBM may not become the standard method for characterizing the ACA, since it is too costly and time-consuming. In a recent OCT study,28 we confirmed a high reproducibility for the optical measurements of the ACA structures with values of 5.4% for the ACA and 4.5% for the AOD with high interobserver agreement. However, similar to UBM,40 the measured parameters can vary considerably and can be affected by the subjective interpretation of the images of anatomic landmarks. Furthermore, the angular approach of estimating the angle width by gonioscopy has been shown to slightly overestimate the width compared with UBM,16 which could also reflect some differences observed with OCT goniometry.
The AOD pertained to the area of the trabecular meshwork and is an important measure to assess aqueous outflow facility. The slight quantitative differences in width between the ACA and the AOD values were predominantly caused by the peripheral iris convexity, particularly a steep peripheral iris or plateau iris configuration with a narrowing of the angle recess, a forward angulation of the iris root, and concomitant deeper ACD.13,16,41,42 However, noncontact OCT goniometry might better represent the iris convexity than UBM, since in the latter a certain extent of corneal pressure from the use of an immersion bath can distort the peripheral iris configuration.16 Although quantitative measurements of ACA and AOD should be performed perpendicularly to avoid optical distortions,43 in this study the qualitative evaluation of the iris convexity was improved from anteriorly positioned radial OCT images.
Acute angle-closure glaucoma is considered to result from an abnormal biometric configuration of the anterior eye segment.37,44,45 The ocular biometric traits involve a shallow anterior chamber, a thick lens, an anterior lens position, a short axial length, a small corneal diameter, and a small radius of corneal curvature, all resulting in small and crowded anterior eye segments.11,35,44 Current guidelines recommend that in angles with less than 20° an occlusion is possible, whereas in angles with less than 10° occlusion is probable.3,4 Thus, angle closure is an anatomic disorder, and OCT imaging of the anterior eye segment might improve its detection noninvasively. The OCT goniometry allowed a consistent evaluation of the structural changes compared with gonioscopy, with an increasing consistency as the angle narrowed, and was revealed to predict patients with occludable angles at risk of acute angle closure. With finer grading and scaling in OCT goniometry, the sensitivity to detect relevant changes increased compared with clinical diagnostic methods.46 The higher resolution of OCT goniometry seems particularly important in narrow or closed angles with low ACA or AOD values. Nevertheless, particular caution should be warranted when interpreting only quantitative differences, and OCT goniometry changes of more than 3° or 45 μm are not instrument dependent in high-quality images.28
Previous studies36 confirmed a linear relationship between the central ACD and gonioscopic grading, and this relationship was similar for OCT goniometry parameters. The progressive changes of the anterior chamber in the different patient groups studied were also confirmed by increasing values for the ultrasonographic LAL ratio, which defined the relationship between lens, iris, and cornea and thus indirectly the status of the ACA. Previous studies30,45 described values of 1.91 for healthy patients and more than 2.2 in patients with acute or chronic angle closure. In this study, the LAL ratio increased from 1.95 in open angles to 2.36 in closed angles, and there was a significant negative correlation compared with the parameters of OCT goniometry. These changes indicate that the lens with an anterior positioning and thickening was the major anatomic risk factor for the narrowing of the angle.30,35,47
One of the aims of recent population-based studies has been to evaluate possible screening modalities for angle-closure glaucoma.11,35,37,44,47 In white patients, only approximately 6.5% have angular width lower than 20°, with 5.9% having a steep peripheral curvature of the iris.13 The rate of patients at risk for angle closure is lower than 2% in the Western hemisphere15,29 but is considerably higher in East Asian countries.8,9,35,37,44,47 The incidence of PACG is more frequent in East Asian populations, since evidence exists that the iris joins the scleral wall more anteriorly in this population.37 Furthermore, the incidence increases with increasing age and hyperopia.11,37,44 Therefore, angle-closure glaucoma is a major public health problem and a leading cause of blindness in Asia, with large populations at risk.8- 10 In previous studies35 that concerned the performance of screening methods, only ultrasonographic measurement of central ACD with a cutoff depth of 2.70 mm provided an adequate sensitivity of 77% and a specificity of 87% compared with gonioscopy in the evaluation of narrow or critically narrow angles in PACG. More recently, biometric gonioscopy had a sensitivity of 75% and a specificity of 93% compared with the Spaeth angle classification.6 In this study, we set the criteria for OCT goniometry to detect an occludable angle at less than 22° for ACA and less than 290 μm for AOD, which is similar to values reported for UBM in patients at risk for angle-closure glaucoma.45 Although the sensitivity was similar for ACA and AOD, with values of 85% to 86%, the specificity was slightly better for ACA, reaching a value of 95%. The sensitivity and specificity seemed better for OCT goniometry than for hand-light tests, limbal ACD, or central ultrasonographic ACD35,41 and thus suggest an improved detection of occludable angles. A significant positive correlation also existed between angular and linear parameters of the ACA region. Our results revealed an increasing correlation in occludable and narrow or closed angles, which further improved the diagnostic accuracy in these conditions. Therefore, OCT goniometry with its noncontact format and improved handling could be used as an objective screening method on a population basis and could be implemented in epidemiologic studies of the ACA. Further refinements with reduced costs and adaptation in a portable system will make OCT goniometry more affordable for these purposes.
In summary, in this study we assessed a wide variation of changes under clinical circumstances to validate OCT goniometry compared with the classification of the angle landmarks with gonioscopy and other glaucoma-relevant clinical parameters. Noncontact OCT goniometry was helpful in evaluating the anterior chamber structures and could improve the noninvasive clinical assessment and treatment of patients with glaucoma. Although OCT goniometry cannot completely replace microscopic evaluation of the ACA anatomy and pigmentation with a gonioscopic lens, it has the potential to supplement or even replace quantification with current gonioscopic grading systems and as a screening modality for the presence of an occludable drainage angle. Furthermore, this imaging method could improve the knowledge of dynamic changes of the eye involved in angle closure38,47 and could also be helpful in evaluating the effects of glaucoma surgery, such as iridotomy in angle closure7 or development of malignant glaucoma following filtration surgery.27 Further developments will enable a combined rapid measurement of the central corneal thickness24 and angle assessment with improved imaging in the evaluation of patients with glaucoma.
Correspondence: Christopher Wirbelauer, MD, Klinik für Augenheilkunde, Vivantes Klinikum Neukölln, Rudower Str 48, D-12351 Berlin, Germany (firstname.lastname@example.org).
Submitted for Publication: December 29, 2003; final revision received June 10, 2004; accepted June 24, 2004.
Financial Disclosure: None.
Funding/Support: This work was supported by the Herbert Funke-Stiftung, Berlin.