Comparison of clinical and histopathological measurements of longest basal diameter in a scatterplot (A) and a Bland-Altman plot (B).
Comparison of echographic and histopathological measurements of apical tumor height in a scatterplot (A) and a Bland-Altman plot (B).
Comparison of Clinical, Echographic, and Histopathological Measurements From Eyes With Medium-Sized Choroidal Melanoma in the Collaborative Ocular Melanoma StudyCOMS Report No. 21. Arch Ophthalmol. 2003;121(8):1163-1171. doi:10.1001/archopht.121.8.1163
To compare pre-enucleation clinical and echographic measurements with postenucleation histopathological measurements of choroidal melanoma of a size and in a location suitable for iodine 125 brachytherapy.
Cross-sectional study of patients with melanoma classified as medium-sized who were assigned to the enucleation arm of the randomized trial of 125I brachytherapy vs enucleation conducted by the Collaborative Ocular Melanoma Study (COMS) Group.
Measurements of melanoma dimensions by clinical, echographic, and histopathological examinations were compared for 644 eyes.
Tumor longest basal diameter and apical height.
Tumor dimensions were available for 644 (98%) of 660 patients who had unilateral medium-sized choroidal melanoma and were randomly assigned to enucleation. The clinical and histopathological measurements of the longest basal diameter agreed within ±2 mm for 371 eyes (58%). The clinical measurement was less than the histopathological measurement by more than 2 mm in 32 eyes (5%), which occurred more frequently when the tumor was within 2 mm of the optic disc. The echographic and histopathological measurements of apical height agreed within ±2 mm in 579 eyes (90%); the echographic measurement was less than the histopathological measurement in only 2 cases. Features associated with echographic estimates larger than histopathological measurements were greater tumor height and anterior tumor apex location.
Our results suggest that tumor measurements made according to COMS protocol were highly reliable in planning radioactive plaque therapy and monitoring changes in tumor size after such treatment.
TUMOR SIZE is an important prognostic factor in patients with choroidal melanoma.1- 3 Accurate assessment of tumor size is critical in both treatment selection as well as planning and delivery of treatments such as brachytherapy.4 For example, in the Collaborative Ocular Melanoma Study (COMS), the prescription point for brachytherapy for tumors 2.5 to 5.0 mm in apical height was 5 mm from the inner sclera. For tumors 5.1 to 10 mm in apical height, the prescription point was the tumor apex. Plaque size was selected to provide a 2- to 3-mm margin around the tumor base.
Various tumor dimensions have been used to classify tumor size, including apical height, basal dimensions, largest basal dimension, basal area, and tumor volume. Linear tumor dimensions can be estimated using multiple methods of measurement: ophthalmoscopy, echography, computed tomography, magnetic resonance imaging, transillumination, and histopathological examination. However, tumor characteristics may influence the accuracy of tumor measurements.
Previous comparisons of tumor size using different methods have been derived from small series of cases and have revealed various discrepancies in measurements. Char et al5 noted good correlation between clinical, ultrasonographic, and pathological measurements in 51 enucleated eyes but found that the histopathological measurements tended to be smaller than those from the other 2 sources. This phenomenon also was observed by Nicholson et al6 in a comparative study of echographic height with histologic height measurements in 53 patients who also underwent enucleation. On average, the echographic height was 2.2 mm greater than the histopathological height, a difference that the authors attributed to tumor shrinkage induced by preparation and fixation of the eye in the pathology laboratory. Other comparisons have revealed larger estimates of tumor dimensions from transillumination than from histopathological measurements.7- 9 Fixatives, hydration, and embedding in paraffin have been believed to sometimes shrink the tissue by as much as 33%.10 Other factors proposed to account for these discrepancies have included experience of the clinician, confusion between blood and tumor during clinical examination, shadowing caused by highly elevated tumors during gross evaluation and especially during transillumination, 7,8 and variations in histopathological processing, differences in the method of sectioning, or nonrepresentative histopathological sections. A relatively flat extension of tumor cells into the adjacent choroid may be difficult to identify clinically and echographically but may be identified readily during histopathological evaluation. Larger discrepancies tended to be observed with irregularly shaped tumors.8,9 In contrast, Augsburger et al11 described pathological measurements of tumor dimensions that were consistently greater than clinical estimates; they suggested that poor clinical estimation and inadequate orientation for gross sectioning could account for most differences in their cases.
Correlations of measurements of tumor thickness by computed tomography, echography, and gross pathological examination have been carried out.12,13 Assessments of interobserver variability5,6,14 have shown good reproducibility of echographic measurements. Based on a study of 10 cases, Seddon et al15 reported that ultrasonographic measurements of height were within ±1 mm of histopathological height. However, transillumination estimates of basal diameter were on average 2 mm greater than the histopathological measurements when the tumor height was greater than 5 mm.
The COMS was designed as a set of 2 randomized trials to evaluate the efficacy of radiotherapy for choroidal melanoma.16 The COMS provides the largest published series of enucleated eyes of patients with choroidal melanoma that were prospectively enrolled, clinically evaluated, and histopathologically processed in a standard fashion. Brachytherapy using iodine 125 was compared with enucleation among 1317 patients who had enrolled by July 31, 1998, when accrual to the trial ended.17 The current investigation used data from that trial to correlate the clinical and histopathological measurements of longest basal tumor diameter and the clinical, echographic and histopathological measurements of tumor apical height. Our aim was to quantify the agreement among these measures, gain an understanding of the difficulties and sources of error associated with these measurements, identify reproducible and reliable measurement techniques, and relate these to brachytherapy parameters.
Prior to initiation of patient evaluation and enrollment, the institutional review board of each participating center reviewed and approved the COMS protocol and informed consent forms. The COMS design and methods have been published previously.16 The COMS Manual of Procedures18 and the COMS Forms Book19 are available. Descriptions of patient characteristics20 and initial mortality findings also have been published for the trial of 125I brachytherapy.17 In the COMS, the linear measurements of apical height and longest basal diameter were used to classify a tumor as small (1-2.5 mm in apical height and >5 mm in longest basal diameter), medium (2.5-10 mm in apical height and ≤16 mm in longest basal diameter), and large (>10 mm in apical height, or ≥2 mm in apical height and >16 mm in longest basal diameter).16 In the COMS trial of 125I brachytherapy for choroidal melanoma classified as medium, all 1317 eyes, regardless of treatment assignment, were evaluated by ophthalmoscopy, photography, and echography before randomization and treatment. All eligible patients gave signed consent prior to enrollment and randomization in the trial. Of 660 eyes with medium-sized tumors randomized to enucleation, prerandomization echograms for all but 16 eyes (2%) were reviewed by the personnel at the COMS Echography Center, and the eyes were processed and subsequently evaluated at the COMS Pathology Center. Findings for these 644 patients with medium-sized choroidal melanoma are analyzed in this report.
Prior to randomization, clinical estimates of tumor size (ie, longest basal diameter and apical height) were made on the basis of the ophthalmologist's evaluation, which included echography. Due to protocol implementation of standardized echography (see the "Echographic Measurements" subsection), the clinical estimate of apical tumor height typically was based on ultrasonography and could be referred to as the "clinical/echographic" estimate. The clinical estimate of longest basal diameter was achieved by a variety of methods. In 1997, a survey was mailed to all COMS ophthalmologists. They were asked to rank, on a scale from 1 to 4, the usefulness of photography, ophthalmoscopy, nomograms(bilateral fundus drawings with corresponding standardized arc and chord lengths, assuming an eye with an inner diameter of 22 mm) that were developed by Thomas A. Rice, MD, 18 and B-scan ultrasonography for the measurement of longest basal diameter. To estimate the frequency of their use of each method, multiple responses were possible. Results of the survey indicated that ophthalmoscopy was the most common method (62%, primary method used; 97%, method ever used). B-scan ultrasonography was ranked as second(18%, primary method used; 90%, method ever used). Similarly, the nomogram also was frequently used (12%, primary method used; 92%, method ever used). Other less commonly used methods included transillumination, fundus photography, fundus drawing, slitlamp biomicroscopy, and A-scan echography.
All patients were evaluated with standardized echography using prescribed examination techniques and a standard protocol.21- 24 This method required the combined use of standardized A-scan and contact B-scan instrumentation with separate B-scan and A-scan probes.
Eleven representative B-scan and standardized A-scan echograms were obtained from each tumor with the sound beam directed through the apex, perpendicular to both the tumor surface and the inner sclera. Five echograms were obtained with B-scan, 2 in cross-section (transverse view), 2 in radial section (longitudinal view), and 1 in axial orientation. An appropriate reduced-gain setting was used to demonstrate clearly the tumor surface and sclera. Six A-scan echograms were obtained, 3 at the high tissue sensitivity gain setting and 3 at reduced gain. The apical tumor height (ie, clinical/echographic estimate) was determined by the ophthalmologist or echographer at each clinical center.
The 11 echograms of each tumor were sent to the COMS Echography Center for independent analysis.25 Height measurements were made from all but the axial echogram. When the retina was attached, the height was measured as the perpendicular distance from the retina at the tumor apex to the inner sclera. When the retina was detached, the height was measured from the apical tumor surface to the inner sclera. The final height was determined by considering both the quality of the individual echograms and the degree of correlation among them. Using these parameters, a confidence in the final apical height measurement was assigned using a scale of 1 to 4, with 4 representing the highest confidence in the measurement recorded. Typically, echograms were reviewed at the Echography Center after patient enrollment and treatment assignment and often after primary treatment of the choroidal melanoma had been initiated.
Details of the histopathological processing and review are described elsewhere.18,26 Enucleated eyes were processed either in a local laboratory at the COMS clinical center or centrally at the COMS Pathology Center. Transillumination defects were measured at the time of gross examination of eyes prior to processing. The eyes were transilluminated by placing a fiberoptic light probe on the globe and directing the light through the pupil. The tumor cast a shadow on the subadjacent sclera. The outer limits of the defect were traced with a marking pen, and the location of the defect as well as its longest diameter and the one perpendicular to it were recorded. The eyes were then cut along the plane of transection, to include what was judged to be the largest possible tumor diameter.
The eyes were fixed, processed, and sectioned according to a standard protocol.18 Representative slides from tumors processed locally were sent to the COMS Pathology Center for central review. Equivalent sets of slides consisting of hematoxylin-eosin–stained 5- to 8-µm–thick sections were examined by all members of the COMS Pathology Review Committee (initially 5 members, then reduced to 3 members27). The margins of the tumor were located under low(×4) magnification and marked on the slide. The longest basal diameter and tumor height, excluding the sclera, were measured with a caliper independently by each reviewer. The mean of the 3 measurements for each dimension was used in the present analysis.
Bland-Altman plots28 were used to visually inspect the agreement between clinical and histopathological measurements of longest basal diameter and between echographic and histopathological measurements of tumor apical height. These graphical methods also were used to identify differences exceeding the approximate 95% limits of agreement, as defined by the mean difference ±2 SDs of the difference. Correlation between measurements also was assessed by Pearson correlation coefficients. The χ2 statistic was used to assess the bivariate relationships between tumor characteristics and agreement status. Tumor characteristics that were potentially associated with agreement status (P<.20) in the bivariate analyses were included in the multivariable regression analyses. Multiple logistic regression was performed to assess the relationship of various covariates with a dichotomous outcome, indicating agreement or disagreement of measurements within ±2 mm. SAS statistical software (SAS, Inc, Cary, NC) was used for all statistical analyses. Findings reported herein are based on data received at the COMS Coordinating Center by December 31, 2001. All baseline assessments of tumor dimensions from different sources had been completed by that date.
Table 1 shows the distribution of selected clinical, echographic, and histopathological features of 644 patients whose baseline tumor echograms were reviewed by the Echography Center, who were treated by enucleation as assigned, and whose tumor slides were subsequently reviewed by the Pathology Review Committee.
The mean (SD) clinical estimate of longest basal diameter was 11.4 (2.7) mm. Upon transillumination of the whole globe, the mean longest tumor diameter was 11.2 (3.8) mm. In contrast, the mean diameter of the tumor along the plane of the section of the transilluminated globe was 9.5 (3.7) mm as compared with the mean histopathological slide measurement of 9.8 (2.8) mm. The mean difference between histopathological and transillumination measurements for 563 eyes was 0.5 (3.5) mm; this excludes 81 eyes that did not allow transmission of light.
Table 2 and Figure 1 depict the agreement between clinical and histopathological slide measurements of longest basal diameter. The mean (SD) difference in measurements was 1.6 (2.4) mm. Inspection of the Bland-Altman plot in Figure 1B indicates that only 27 (4%) of the differences between clinical and histopathological measurements of longest basal diameter exceeded the mean difference by ±2 SDs and that the magnitude of the discrepancy increases somewhat with increasing longest basal diameter.
The clinical and histopathological measurements of longest basal diameters agreed within 2 mm for 371 eyes (58%). The clinical estimate exceeded the histopathological measurement in 241 eyes (37%) and was more than 2 mm less than the histopathological measurement in 32 eyes (5%). The bivariate relationships between these 3 categories of agreement and tumor characteristics were explored for all features listed in Table 1. Table 3 presents the distribution of clinical, echographic, and histopathological features by discrepancies between clinical and histopathological slide measurements of longest basal diameter for those features that appeared to be associated (P<.20) with measurement disagreement. The clinical characteristics associated with clinical estimates larger than histopathological measurements (>2 mm) were larger tumor apical height, increased longest basal diameter, retinal detachment, and posterior location. Clinical estimates smaller than histopathological measurements occurred when the tumor was within 2 mm of the optic disc. The echographic feature of mushroom shape was significantly associated with clinical estimates that were more than 2 mm larger than the histopathological measurement. Clinical estimates larger than histopathological measurements also occurred more frequently following local processing. Histopathological features listed that appeared associated with a greater clinical estimate were presence of moderate or marked macrophages, absence of invasion of the emissary canal, and absence of extrascleral extension. Bivariate analysis of features by agreement category also was performed for subgroups stratified by apical height category(≤5 mm vs >5 mm); similar relationships were observed (data not shown).
Multivariable modeling was performed to identify clinical factors associated with the odds of disagreement (>2 mm) between clinical and histopathological measurements (Table 4). It was observed that the measurement of basal diameter confounds the association between apical height and agreement; thus, only the clinical measurement of longest basal diameter was retained in the model. In a statistical model assessing only the potential clinical predictors of disagreement, longest basal diameter again was the only factor that was statistically significantly associated with increased odds of disagreement. The unadjusted odds of disagreement between clinical and histopathological measurements was 5 times greater for tumors with large (>13 mm) diameters (odds ratio [OR], 5.07; 95% confidence interval[CI], 2.94-8.73) and 3 times greater for tumors with diameters of 9 to 13 mm (OR, 3.06; 95% CI, 1.81-5.18) than for tumors 9 mm or less in diameter. When the echographic features of shape and location of the tumor's posterior border were added to the model, size of longest basal diameter remained significant. Disagreement also was observed more often for mushroom-shaped tumors than for dome-shaped tumors (OR, 1.68; 95% CI, 1.07-2.64).
For the 644 tumors, the measurements of the longest basal diameter and the greatest dimension perpendicular to it differed by more than 2 mm for 218 eyes (34%), indicating more asymmetric tumors. Differences between the clinical and histopathological measurements of longest basal diameter were associated with more asymmetric tumors (OR, 1.11; 95% CI, 1.00-1.22). However, such asymmetry was associated with greater longest basal diameter; only longest basal diameter was retained in the final multivariable model.
The mean (SD) difference between clinical measurement of tumor apical height (using echography) and echographic assessment made by the Echography Center was 0.1 (0.5) mm. Table 5 depicts the agreement between clinical, echographic, and histopathological measurements of tumor apical height. The mean difference between echographic and histopathological height measurements was 0.8 (1.0) mm; Figure 2 shows agreement between these measurements. Inspection of the Bland-Altman plot in Figure 2B indicates that only 28 (4%) of the differences between echographic and histopathological measurements of apical height exceeded ±2 SDs of the mean difference. These 28 cases were reviewed by the COMS Echography Center Director (S.F.B.); the larger discrepancies were associated with earlier years of the COMS (between 1987 and 1991) when the apical height was measured to the outer scleral surface. Other possible reasons for discrepancies included confusion of a shallow retinal detachment with the tumor surface, echograms obtained obliquely rather than through the apex of the tumor, and poor echogram quality.
The bivariate relationships between the agreement categories and tumor characteristics were explored for all features listed in Table 1. Table 6 lists the factors potentially associated (P<.20) with discrepancies between echographic and histopathological measurement of apical height. The echographic and histopathological measurements of tumor apical height agreed within 2 mm for 579 eyes (89.9%). The echographic estimates exceeded the histopathological measurement in 63 eyes (9.8%) and were less than the histopathological measurement by more than 2 mm in only 2 eyes (0.3%). Echographic estimates larger than histopathological measurements occurred more frequently following local processing. Clinical features associated with greater measurements were greater tumor apical height, anterior location of the anterior border of the tumor, mushroom shape, and anterior location of the tumor apex. Histopathological features that were associated with a greater echographic estimate were anterior location and diffuse infiltration. The echographic estimate was less likely to be greater than the histopathological measurement when the Bruch membrane was ruptured.
Multivariable modeling was performed to identify factors associated with the odds of disagreement (>2 mm) between echographic and histopathological measurements (Table 7). Due to the small numbers, the 2 cases for which echography underestimated the microscopic slide measurement were combined with the agreement category and compared with cases in which echographic measurement exceeded the histopathological measurement. It was observed that the characteristic most strongly associated (P<.001) with greater echographic measurements (>2 mm) was greater tumor apical height. After adjusting for tumor apical height, neither tumor apex location nor shape was statistically significantly associated with greater echographic measurement. The odds of disagreement between echographic and histopathological measurements was almost 3 times more likely for tumors with larger apical height (>5 mm) than for tumors with an apical height of 5 mm or less (OR, 2.81; 95% CI, 1.38-5.73).
Upon review of echograms at the Echography Center, 2 (0.3%) and 28 (4.3%) of 644 eyes were judged to have definite or suspected extrascleral extension, respectively. The extension was not measurable in the 2 eyes with definite extension. Initially, the extent of extension was not recorded on the data collection form for eyes with suspected extension. For 6 later cases, the extension measured 0.5 mm, 1 mm, and 2.5 mm, respectively, in 3 cases, and was not measurable in 3 cases. Upon histopathological examination, extrascleral extension was detected in only 1 of the 2 eyes determined at the Echography Center to have definite extension.
Only 5 cases were recorded as having extrascleral extension at the time of transillumination measurement of the enucleated globe. Three of these cases also were judged to have extrascleral extension upon histopathological review. Histopathological examination of the microscopic slides revealed 11 cases with extrascleral extension with presumed residual tumor in the orbit and 7 cases with extension without evidence of transection. Only 1 of these eyes was judged to have extrascleral extension on echographic examination.
Overall, agreement was high between clinical and histopathological slide measurement of tumor longest basal dimension. In our series, measurements more often disagreed for tumors that were more than 9 mm in largest basal diameter, with a posterior border located behind the equator, or with mushroom-shaped growth.
Our findings differ from those of previous reports. Polivogianis et al8 reported that longest tumor diameter by transillumination was significantly larger than that by histopathological diameter on average. This difference was attributed to possible subretinal fluid, shadowing caused by highly elevated tumors, or inadequate sectioning technique. In our series, the mean (SD) difference between diameter measurements by histopathological examination and by transillumination of the enucleated globe was 0.5 (3.5) mm. Augsburger et al11 described consistent underestimation of clinical tumor size in relation to histopathological dimensions. We observed a mean difference of 1.6 (2.4) mm between clinical and microslide measurements of diameter; in 241 eyes (37%), the clinical measurement was greater than the histopathological dimension by more than 2 mm.
Agreement was very high between clinical and Echography Center estimates of apical tumor height, reflecting the studywide use of echography in making the final clinical estimate. Most clinicians rely on echography for this measurement. Previous investigators6,13,14 have used B-scan for tumor localization and then relied solely upon standardized A-scan, using reduced gain, to measure tumor height. Our protocol differed in that we used transverse and longitudinal B-scan echograms and all A-scan echograms to measure tumor height. This method facilitated measurement at the tumor apex and reliable identification of the inner sclera to provide the final measurement. Reasons for discrepancies between echographic and histopathological measurements included poor quality of echograms or a measurement that was not through the apex of the tumor.
This series provided the opportunity to assess a large number of enucleated eyes that were evaluated using the COMS standard protocols. Our study shows that possible sources of error in measurement exist at the time of clinical estimation, gross evaluation, and histopathological assessment. Our findings suggest that clinicians should take extra care in assessing tumors with larger basal diameters, apices located anterior to the equator, or mushroom shape when considering treatment options. This study also has implications for brachytherapy planning. In the COMS, we chose the size of the 125I plaque to allow for a 2- to 3-mm margin beyond the tumor as determined by clinical measurement, that is, a plaque of diameter 4 to 6 mm larger than the longest basal diameter. An 125I plaque that was 4 mm greater in diameter than the clinical measurement of longest basal diameter, if accurately placed, would have encompassed the choroidal melanoma and histopathological extension in 95% of eyes so treated; an 125I plaque that was 6 mm greater than the longest basal diameter would have encompassed the tumor mass plus extension in all eyes. Echographic measurement of apical height exceeded the histopathological measurement of apical height by more than 1 mm in 229 eyes (35.6%) and was greater than 2 mm in 63 eyes (9.8%). Thus, the target volume would have been fully irradiated. Underestimation of apical height by more than 2 mm may have occurred in 2 eyes (0.3%) based on a greater measurement from histopathological examination than from echography. Only about 1% of tumors might have been judged ineligible for this COMS clinical trial based on tumor size if the histopathological measurements of longest basal diameter and apical height could have been known before enrollment. It should be noted that clinical and echographic measurements were made in increments of 0.1 mm, whereas histopathological measurements were made to 1 mm; this introduces a slight artifact in assessing agreement of these measurements that serves to increase slightly the level of disagreement.
Tumor dimensions were key criteria in assessing eligibility for COMS clinical trials, for planning radiation dosimetry, including selection of the prescription point and plaque size, and for assessing tumor growth after 125I brachytherapy. Because of the randomized design of the COMS trial of 125I vs enucleation, studies of clinical and histopathological measurements of apical height and longest basal diameter in eyes enucleated with medium-sized melanoma provide information applicable to tumors being considered for alternative treatments, such as brachytherapy. Our findings demonstrate that these measurements made according to the COMS protocol were highly reliable in planning radioactive plaque therapy. Thus, undertreatment of tumors assigned to 125I brachytherapy was unlikely.
Corresponding author and reprints: Marie Diener-West, PhD, COMS Coordinating Center, Wilmer Clinical Trials and Biometry, 550 N Broadway, Ninth Floor, Baltimore, MD 21205 (e-mail: firstname.lastname@example.org).
Submitted for publication August 15, 2002; final revision received April 16, 2003; accepted May 8, 2003.
The COMS has received support from the National Eye Institute and the National Cancer Institute through cooperative agreements EY06253, EY06257, EY06258, EY06259, EY06020, EY06264, EY06265, EY06266, EY06268, EY06269, EY06270, EY06274, EY06275, EY06276, EY06279, EY06280, EY06282, EY06283, EY06284, EY06287, EY06288, EY06289, EY06291, EY06839, EY06843, EY06844, EY06848, EY06858, and EY06899 with the National Institutes of Health, Bethesda, Md.