Monosomy 3 in Uveal Melanoma: Correlation with Clinical and Histologic Predictors of Survival

Abstract

purpose. To correlate monosomy 3 in uveal melanoma with clinical and histologic prognostic variables and death caused by metastatic disease.

methods. Loss of heterozygosity (LOH) on chromosome 3 was investigated by PCR-based microsatellite analysis in 105 tumors and related to large basal tumor diameter (LBD), ciliary body (CB) involvement, tumor cell type, periodic acid-Schiff (PAS)-positive loops, and death related to metastatic disease. A model relating monosomy 3 to these was created with forward-stepwise logistic regression and used to derive a prognostic index.

results. Monosomy 3 occurred in 54 (51%) tumors and regional chromosome 3 LOH in another six (6%) tumors. Monosomy 3 was associated with epithelioid cells (χ² test, P < 0.001), PAS-positive loops (χ², P = 0.001), LBD (Mann-Whitney test, P = 0.002), CB involvement (χ² test, P = 0.008), and metastasis-related death (log rank analysis, P = 0.0003). The regression coefficients indicated that epithelioid histology was 15 times as influential with each millimeter of increase in LBD. A prognostic score was derived: one point for each LBD category (<7.4, 7.5–12.4, 12.5–17.4, and >17.4 mm) and three points for epithelioid histology. The prevalence of monosomy 3 increased with score, from 0% in 18 tumors scoring less than 4 to 95% in 21 tumors scoring 7.

conclusions. Monosomy 3 correlates with survival but can be predicted only in patients with large epithelioid tumors. The absence of monosomy 3 is predictable only in patients who have small, spindle-cell tumors. In most patients, prediction of monosomy 3 according to tumor size and histology is unreliable.

Uveal melanoma is the most common primary intraocular malignancy in adults. Despite advances in diagnosis and local tumor control, the overall mortality rate remains high, at approximately 50%, because of the development of metastatic disease. ¹ Accurate identification of patients with a high probability of development of metastatic disease is important for clinical management, because metastases are not usually detectable at the time of treatment of the primary tumor and have often reached an advanced stage by the time they cause symptoms. 

Several clinical and histologic features have been correlated with poor prognosis in patients with uveal melanoma. These include large tumor diameter (i.e., >15 mm), ¹ ciliary body involvement, ² the presence of epithelioid cells, ³ and the detection of PAS-positive loop-like structures, which are so-called closed vascular or microvascular loops and channels. ^{4 5}  

Recently, cytogenetic analyses of uveal melanoma have identified common loss of an entire chromosome 3 homologue (monosomy 3) and increased chromosome 8 long arm copy number, often coexisting. ^{6 7 8} These aberrations have shown a greater association with poor prognosis than certain previously recognized clinical and histologic parameters, with monosomy 3 associated with a reduction in the 5-year survival from almost 100% to only 30%. ⁶ Monosomy 3 is reported to be associated both with large basal tumor diameter (LBD) and ciliary body (CB) involvement, ⁶ whereas only a weak relationship has been noted with the presence of epithelioid cells. ⁷ We have found a relationship between monosomy 3 and PAS-positive loops, which has not been previously reported. ⁹  

Our purposes in the present study were to confirm the correlation between monosomy 3 and death caused by metastatic disease and to determine to what extent it is possible to predict monosomy 3 according to the presence of certain clinical and histologic variables. 

Materials and Methods

Patients and Samples

Patients referred to the Liverpool Ocular Oncology Center between January 1998 and February 2000 underwent full preoperative ophthalmic and systemic examination, which included slit lamp and ophthalmoscopic examination and echographic measurement of largest and smallest basal tumor dimensions and tumor thickness. Patients were included in the study if they were to be treated by enucleation or transscleral local resection, which we reserve for tumors we consider unsuitable for radiotherapy or phototherapy. 

At the end of local resection or enucleation, a fresh tumor sample was collected for experimental purposes. Part of each tumor was snap frozen in liquid nitrogen, and the remainder was formalin fixed and embedded in paraffin. Histologic features were recorded from hematoxylin and eosin and PAS-stained sections of paraffin-embedded tumors. 

The research adhered to the tenets of the Declaration of Helsinki. Informed consent was obtained from the subjects after explanation of the nature and possible consequences of the study. The study was approved by the Royal Liverpool University Hospital Ethical Committee. 

Microsatellite Analysis

DNA was extracted from blood and 10-μm frozen tumor sections by standard procedures. ⁹ A 5-μm section from each tumor was stained with hematoxylin and eosin to confirm the histology and to ensure that the specimen was composed of at least 90% tumor cells. Microsatellite analysis, to detect loss of heterozygosity (LOH) on chromosome 3, was undertaken on all tumors, by using ten 3p and ten 3q markers, as previously described. ⁹  

Statistical Analysis

Clinical, histologic, and chromosomal data were recorded in a computerized database. The patients were flagged in the National Cancer Registry, which automatically informed us of any deaths, together with the date and cause of death. The time until death was calculated from the date of local resection or enucleation. 

The results were analyzed on computer (SPSS, ver.10.0; SPSS Science, Chicago, IL). The relationship between monosomy 3 and the standard prognostic variables was determined with the Mann-Whitney test for LBD and the χ² test (with continuity correction) for the other three prognostic variables (cell type, CB involvement, and PAS-positive loops). 

Kaplan-Meier estimates were used to draw survival curves for time to metastasis-related death. Associations between metastasis-related death and risk factors were assessed with log rank analysis for categorical variables (i.e., monosomy 3, cell type, PAS-positive loops, and CB involvement) and with Cox univariate analysis for tumor diameter, a continuous variable. 

Forward stepwise logistic regression was used to derive a model relating monosomy 3 to LBD, CB involvement, epithelioid cell type and PAS-positive loops. ¹⁰ This model was then used to derive a prognostic index. 

Results

Chromosome 3 LOH was detected in 60 (57%) of the 105 tumors studied. The majority of these 60 tumors (54, 90%) showed LOH at all informative loci, indicative of monosomy 3. Regional chromosome 3 LOH, restricted to parts of the p or q arm only, was detected in six patients, one of whom died of metastatic disease. These six tumors were excluded from further analysis in case they represented a distinct subgroup of tumors with a different biological behavior from those with monosomy 3. The remaining sample population consisted of 56 males and 43 females with a mean age of 60.3 ± 15.27 years (SD; range, 25–90). The primary form of treatment was enucleation in 74 patients and local resection in 25 patients. The tumor was located in the right eye in 58 patients and in the left eye in 41 patients. The anterior tumor margin was located anterior to the ora serrata in 44 patients and more posteriorly in 55 patients. The LBD averaged 14.80 ± 3.29 mm (range, 5.1–20.9). In one patient an echographic measurement of LBD was not possible because the tumor showed annular spread around the entire ciliary body. The histologic cell type was spindle B in 21 tumors and mixed or epithelioid in 78 tumors. Microvascular loops were present in 64 tumors. 

By the close of the study, 16 patients had died of metastasis-related disease, and six had died of other causes (bronchial carcinoma, renal failure, heart failure, brain hemorrhage, myocardial infarction, and old age). Figure 1 shows the Kaplan-Meier survival curves for death related to metastatic disease according to monosomy 3, cell type, CB involvement, and PAS-positive loops. Monosomy 3 was associated with an increase in the actuarial rate of metastasis-related death from 0% to 51% (95% confidence interval [CI]: 31%–71%) at 2 years (log rank analysis, P = 0.0003). This association was stronger than that with cell type (log rank analysis, P = 0.14), CB involvement (log rank analysis, P = 0.04), PAS-positive loops (log rank analysis, P = 0.11), or LBD (Cox univariate analysis, P = 0.003). 

Table 1 shows the interrelationships between the five prognostic variables. Monosomy 3 was significantly associated with epithelioid cells, PAS-positive loops, LBD, and CB involvement. The presence of epithelioid cells was associated with PAS-positive loops, but not with tumor size or CB involvement. Tumor size was significantly correlated with CB involvement, but only weakly associated with the presence of PAS-positive loops. There was no association between CB involvement and PAS-positive loops. The prevalence of monosomy 3 according to the other prognostic variables is shown in Table 2 . 

Table 3 shows the results of the logistic regression. The regression coefficients (b) indicated that epithelioid histology was approximately 15 times as influential as each millimeter increase in LBD. The prognostic score was derived with the regression formula  

\[\mathrm{Logit}(\mathrm{p})\ {=}\ {-}7.58\ {+}\ 0.29\ (\mathrm{LBD})\ {+}\ 4.25\ (\mathrm{EPI})\]

where LBD is largest basal tumor diameter in millimeters and EPI is the presence of epithelioid cells (0, no; 1, yes). The coefficients were rounded to arrive at the the final scoring system with one point for each LBD category (<7.4, 7.5–12.4, 12.5–17.4, and >17.4 mm) and three points for epithelioid histology. Table 4 shows the prevalence of monosomy 3 according to this prognostic score. The prevalence of monosomy 3 increased with score, from 0% in the 18 tumors with a score of less than 4 to 95% in the 21 tumors with a score of 7. In 61% of cases the score ranged from 4 to 6, so that prediction of monosomy 3 status would not have been accurate. 

Discussion

Monosomy 3 has been shown recently by other investigators to be associated with a poor prognosis for survival after treatment of uveal melanoma. ⁶ Our own data confirm these findings, with monosomy 3 being associated with an increase in the 2-year actuarial rate of metastatic disease from 0% to 51%. Monosomy 3 is better than certain other clinical and histologic factors for predicting metastatic disease and death. ⁶ In our patients, monosomy 3 showed a greater association with metastatic disease than epithelioid histology, PAS-positive loops, CB involvement, and LBD. Because testing for monosomy 3 is not widely performed, it is still often necessary to rely only on clinical and histologic findings when estimating prognosis for survival. We thought it would be of value to measure the associations between the various prognostic factors for survival and to determine to what extent clinical and histologic variables could be used as surrogate predictors of monosomy 3 status. 

The size of our sample allowed a more detailed correlation between recognized clinicopathologic factors and chromosome 3 alterations than has hitherto been reported by Prescher et al. ⁶ and Sisley et al., ⁷ who studied 54 and 49 patients, respectively. Unlike these previous studies, we report the correlation between monosomy 3 and PAS-positive loops. We also describe the prevalence of monosomy 3 according to tumor size, location, and histology, thereby giving some insights into the probability of a patient having monosomy 3. A limitation of our investigation (as with other similar studies) is that the sample is biased by the exclusion of patients treated by phototherapy or radiotherapy, because in such patients tumor samples were not obtained for histologic and molecular genetic analyses. It would not have been ethical to subject such patients to the hazards of tumor biopsy without firm evidence that such a procedure would have improved their care. Our study, however, should include a wider range of tumors than other studies, because we prefer local resection to radiotherapy for the conservative treatment of moderately large tumors (i.e., tumor thickness >5.5 mm and LBD <17 mm). 

Partial deletions of chromosome 3 have been reported in uveal melanoma. ^{9 11} In our present study, six patients with partial deletions of this chromosome were excluded from our statistical analysis because we were unsure of the prognostic significance of these abnormalities. The fact that five of these six patients were still alive at the end of the study seems to support this decision, although follow-up is ongoing. 

Our results regarding the correlation of monosomy 3 with LBD are in keeping with the findings of Prescher et al. ⁶ but not with those of Sisley et al., ⁷ who reported no significant association. The highly statistical correlation between large tumor diameter and monosomy 3 may give the impression that most large melanomas would show this chromosomal abnormality. Our results show, however, that this is not the case, except for tumors greater than 17 mm in diameter; thus we selected this diameter for categorizing large tumors instead of the usual diameter of 15 mm. 

The strong correlation we report between monosomy 3 and CB involvement (P = 0.008) is in keeping with the studies of Prescher et al. (P < 0.001) ⁶ and Sisley et al. (P < 0.0001). ⁷ However, CB involvement just failed to reach significance in the logistic regression model (P = 0.061), so that it was excluded from the prognostic score. Furthermore, we did not find CB tumors to be associated with epithelioid histology or PAS-positive loops. These results differ from those of Rummelt et al. ¹² Discrepancies between results in studies have probably arisen because of differences in the samples analyzed. Choroidal tumors in our sample had a larger median diameter (13.8 mm vs. 11 mm), a higher prevalence of mixed-epithelioid cellularity (75% vs. 48%) and a higher prevalence of closed loops (66% vs. 49%). A possible explanation for the sample differences is that we were more likely than other researchers to treat large choroidal tumors by local resection or enucleation instead of radiotherapy. 

The strong correlation between monosomy 3 and the presence of epithelioid cells (P < 0.001) in our study differs from the results of Prescher et al. (P = 0.02) ⁶ and Sisley et al. (P = 0.02). ⁷ This disparity may be due to the small numbers of tumors with epithelioid cells in previous studies or to variation in the way tumor cell type was categorized, this being a subjective procedure. Our results indicate that the prediction of monosomy 3 is greatly enhanced by the inclusion of tumor cell type in the prognostic model. 

To our knowledge, we are the first group to report the statistical association between monosomy 3 and PAS-positive loops. ⁹ However, this factor did not reach significance in the logistic regression, probably because the presence of epithelioid cells was more strongly associated with monosomy 3. 

An accurate method of estimating the prognosis for survival in patients with uveal melanoma would be valuable, as with other cancers. First, accurate prognostication would identify high-risk patients who might benefit from screening for metastases or adjuvant systemic immunotherapy or chemotherapy. These measures would be more cost effective if targeted more accurately at high-risk patients. Second, such prognostication would enhance any evaluation of the influence of any treatment or screening on survival. Third, patients with a good prognosis could be given reassurance about their survival, thereby improving their quality of life. Finally, patients with clinical metastatic disease unresponsive to treatment have told us that they appreciated having been informed of their poor prognosis at an early stage, because this knowledge enabled them to make administrative and other preparations while they still felt well. We do not think, however, that we can advocate monosomy 3 testing as a routine procedure until the results of a more formal evaluation are available, showing that this test provides significant benefit to the patient. 

In conclusion, monosomy 3 indicates a high probability of metastasis-related death within the first few years of treatment for uveal melanoma. If this prognostic factor cannot be determined by cytogenetic or molecular genetic studies, its presence or absence can be predicted only in large epithelioid tumors and small spindle-cell tumors, respectively. Only a minority of tumors show such characteristics. Thus, in most patients prediction of monosomy 3 according to tumor size and histology is unreliable. 

 Supported by Grant 98-53b from the Guide Dogs for the Blind Association, United Kingdom, and Grant 2641 from The University of Liverpool Research Development Fund.

 Submitted for publication February 12, 2002; revised July 12, 2002; accepted August 2, 2002.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Kaplan-Meier curves showing actuarial rates of metastasis-related death according to presence of (A) monosomy 3, (B) epithelioid cells, (C) CB involvement, and (D) PAS-positive loops. Absence and presence of the adverse factor are represented by dashed and solid lines, respectively. Ticks: censored patients.

Table 1.

 

Interrelationships between Monosomy 3 and Other Factors Prognostic of Survival

	Monosomy 3	LBD	CB Involvement	Epithelioid
LBD	0.002	—	—	—
CB involvement	0.008	<0.001	—	—
Epithelioid	<0.001	0.972	0.364	—
Loops	0.001	0.062	1.000	<0.001

Table 2.

 

Monosomy 3 According to Prognostic Variables

Variables	n	Monosomy 3
				Absent (n)	Present (n)	% (95% CI)
LBD (mm)
<7.5	1	1	0	—
7.5–12.4	25	13	12	48 (27–69)
12.5–17.4	48	27	21	44 (29–58)
>17.4	24	3	21	88 (73–100)
CB involvement
No	55	32	23	42 (28–55)
Yes	44	13	31	71 (56–85)
Epithelioid
No	21	20	1	5 (0–15)
Yes	78	25	53	68 (57–79)
Loops
No	35	24	11	31 (15–48)
Yes	64	21	43	67 (55–79)

Table 3.

 

Logistic Regression Analysis of Prognostic Variables for Monosomy 3

Variable	b	SE (b)	P
LBD	0.29	0.09	0.001
Epithelioid cells	4.25	1.11	<0.001
Constant	−7.580	1.800	<0.001

Table 4.

 

Monosomy 3 According to Prognostic Score

Score	n	Monosomy 3
				Absent (n)	Present (n)	% (95% CI)
1	0	0	0	—
2	3	3	0	0 (0–71)
3	15	15	0	0 (0–22)
4	4	3	1	25 (1–81)
5	22	10	12	55 (32–76)
6	33	12	21	64 (46–81)
7	21	1	20	95 (76–100)

The authors thank Gary Cheetham for data management and Ian Campbell for statistical support.