A Prognostic Factor Study of Disease-Free Interval and
Survival Following Enucleation for Uveal Melanoma
Johanna M. Seddon, MD; Daniel M. Albert, MD; Philip
indicators of uveal mela-
\s=b\ Prognostic
noma disease-free interval and survival
following enucleation were evaluated for
267 patients. The median follow-up time
was 17 years. Analysis involved Kaplan\x=req-\
Meier survival curves based on time to
tumor-related deaths and multivariate
proportional hazards analysis, which pro-
vides an estimate of the "force of mortal-
ity." Prognostic indicators considered
included demographic, clinical, and histo-
pathological factors. Cell type was classi-
fied according to the number of epitheli-
oid cells present per high-power field
(HPF) on light microscopy. Various classi-
fications of size of the tumor were com-
pared for prognostic value. The five lead-
ing predictors of survival in order of
importance were as follows: (1) number of
epithelioid cells per HPF, (2) largest
dimension of the tumor, (3) location of the
anterior margin of the tumor, (4) invasion
to the line of transection, and (5) degree
of pigmentation. Risk categories for sur-
vival were constructed based on the three
leading prognostic factors.
(Arch Ophthalmol 1983;101:1894-
1899)
TTThile many studies have explored
' '
factors that predict which uveal
melanomas will metastasize or lead to
death,14 few have considered simulta¬
neous evaluations of these factors.15
Because many of the variables found
to be related to the prognosis of
patients with uveal melanoma in uni-
variate analysis are interrelated, mul-
tivariate analysis is needed to estab¬
lish which variables are independent¬
ly related to prognosis. This study
uses multivariate techniques to iden¬
tify which clinical and histopathologi-
cal variables are the leading indepen¬
dent predictors of the outcome of
uveal melanoma patients treated by
enucleation.
Accepted for publication Feb 7, 1983.
From the Retina Service and the David G.
Cogan Eye Pathology Laboratory, Massachusetts
Eye and Ear Infirmary, and the Department of
Ophthalmology, Harvard Medical School, Boston
(Drs Seddon and Albert and Ms Robinson), and
the Division of Biostatistics and Epidemiology,
Sidney Farber Cancer Institute, Boston (Dr Lav-
in).
Presented to the Association for Research in
Vision and Ophthalmology, Sarasota, Fla, May 4,
1982.
Reprint requests to Massachusetts Eye and
Ear Infirmary, 243 Charles St, Boston, MA 02114
(Dr Seddon).
Downloaded from www.archophthalmol.com at ALL INDIA INST OF MEDICAL SCIENCES, on April 18, 2011
T. Lavin, PhD; Nancy Robinson
Our study differs from other
studies using multivariate tech¬
niques,15 in two ways. First, we evalu¬
ated outcome in terms of the length of
time between enucleation and the
occurrence of métastases and tumor-
related deaths. Most previous studies
examined the proportion of individu¬
als who were alive or métastases free
at one or two points in time after
treatment. Second, cell type was clas¬
sified according to the number of epi¬
thelioid cells present per high-power
field (HPF).
SUBJECTS AND METHODS
Patients who underwent enucleation
with a diagnosis of ciliary body or choroi¬
dal melanoma from 1953 to 1973 were
identified from the records of the Eye
Pathology Laboratory at the Massachu¬
setts Eye and Ear Infirmary (MEEI), Bos¬
ton. Four hundred seventy-eight patients
were identified, and we obtained adequate
follow-up information on 267 of these
patients who are the subjects of this study.
Of the remaining 211, five were reexam-
ined and found not to have melanomas, 12
from outside the United States could not
be located, 30 had absent medical records,
and in 164, the survival status could not be
adequately determined.
Clinical Follow-up Data
Clinical data were obtained from medi¬
cal records and/or contact with physicians
and included the following: demography:
age, sex, and race; clinical evaluation: date
of initial evaluation, involved eye, signs
and symptoms including decreased visual
acuity and pain and/or inflammation; ocu¬
lar findings at clinical diagnosis: visual
acuity, presence of iris or choroidal nevi,
rubeosis, intraocular pressure, and history
of glaucoma or other ocular diseases; clini¬
cal description of the tumor; medical histo¬
ry; metastatic workup prior to enucleation;
and presence of primary malignant tumors
other than ocular melanomas.
Survival status, date of death, cause of
death, and metastatic disease status were
determined up to June 1981 by review of
death certificates or locating patients
through telephone listings, voter registra¬
tion books, drivers licenses, and town list¬
ings. Efforts to confirm the outcomes of
the subjects also included the following:
contact with patients or their families by
telephone or letter; review of autopsy and
biopsy reports and/or slides; and contact of
physicians and hospital personnel after
appropriate informed consent was ob¬
tained to explore clinical and laboratory
data.
Outcomes were coded as follows: alive
without known métastases; alive with uve-
al melanoma métastases; dead due to uveal
melanoma métastases; dead due to other
malignant tumor; dead due to causes other
than malignancy.
There were 114 patients with melanoma
métastases among the 267 patients in this
study. The dates of diagnosis of métastases
were determined when possible. Sixty-five
métastases were confirmed by autopsy or
biopsy. Of these 65 specimens, 38 were
available for review, and the diagnosis was
confirmed by two of us (J.M.S. and
D.M. .). Also, 23 patients had a second
primary malignant tumor confirmed by
histopathological examination.
Histopathological Data
Following completion of follow-up, sys¬
tematic histopathological reviews of slides
of enucleation specimens were performed
(J.M.S. and D.M. .). When quantitative
evaluations were made, the results of the
two observers were averaged. Slides were
reevaluated without knowledge of the orig¬
inal cytologie classification or the outcome
of the case. In all but a few instances,
several slides per case were reviewed. The
following information was obtained.
1. Tumor Measurements.—Largest diam¬
eter in millimeters in contact with the
sclera and height of the tumor measured
on the largest section available.
2. Location of the Anterior and Posterior
Margins of the Tumor.—Coded as posterior
to the equator, anterior to the equator
without involvement of the ciliary body,
and anterior to the equator involving the
ciliary body.
3. Cell Type.—Coded numerically as the
number of epithelioid cells present per
HPF averaged for 40 HPFs or coded as
necrotic when accurate classification was
not possible because of the degree of necro¬
sis. Cells were classified as epithelioid only
when all four of the following criteria were
present: large oval nuclei, prominent
nucleoli, abundant cytoplasm, and distinct
cell borders. A HPF is the field obtained
under the microscope at 400X magnifica¬
tion with a field of view of 0.45-mm diam¬
eter.
4. Mitotic Figures.—Number of mitotic
figures per HPF averaged for 40 HPFs.
5. Status of Bruch's Membrane.
6. Extent of Invasion.—Subretinal, reti¬
nal, vitreous, vortex vein, scierai emissary
canal, optic nerve (anterior to or beyond
the lamina cribrosa), and scierai extension
(within sclera, to surface of globe or
beyond, or invasion to the line of transec-
tion with presumed residual tumor in the
orbit).
7. Presence of Retinal Detachment.
8. Presence of Rubcosis.
9. Presence of Choroidal or Iris Nevi.
10. Degree of Inflammation.
11. Degree of Pigmentation.
 The latter two categories were coded and
analyzed as follows: (1) none or minimal- Table 1.—Univariate Analysis With Variables Related to Outcome
no or few foci of inflammation or pigmen¬
Favorable Unfavorable
tation present, (2) moderate—multiple Variable Prognosis Prognosis Value-
foci, and (3) heavy—involvement of almost Age at time of manifestation > 60
the entire tumor. Definitions of grades of inflammation No Yes < .05
Clinicalmanifestation, pain or
pigmentation and inflammation and the No Yes < .05
presence of other variables as previously History of glaucoma
listed were agreed on by two of the authors Invasion of tumor to line of transection No < .0001

(J.M.S. and D.M.A.) prior to the study and Scierai extension Superficial Deeper
adhered to on review of individual cases. Degree of inflammation Minimal Heavy
Number of epithelioid cells per high-power field Many < .0001
Analysis Mitotic figures Few Many
Location of anterior margin Posterior to Anterior to
The dependent or outcome variables, equator equator
time to the development of uveal melano¬ Largest tumor dimension Small Large < .0001
ma métastases and time to the occurrence "Size" of tumor Small Large
of deaths due to melanoma, were calcu¬ Pigmentation of tumor Minimal Heavy
lated starting from the date of enucleation. *
values associated with logrank test. Multiple levels of individual variables as described in "Methods"
Calculations starting from the date of ini¬ section were analyzed as well as the dichotomous divisions illustrated herein.
tial evaluation were similar since all but 19
patients had enucleation performed within
two months after the initial evaluation. Table 2.—Relationship Between Table 3. Relationship Between
The independent variables tested as prog¬ Mitotic Figures and Epithelioid Cells* Location and Largest Tumor
—

nostic indicators for the previously men¬

Dimension-
tioned outcomes are listed in the "Clinical No. of
Follow-up Data" and "Histopathological Epithelioid No. of Mitotic Location of Anterior
Data" sections. Cells per Figures (%) Largest Margin, No. (%)
Measurements of the tumor were coded High-Power ,-·-. Tumor ,-*-
< 0.1 >0.1
and analyzed in the following three ways: Field_0 Dimension, Anterior to Posterior to
(1) largest tumor dimension (LTD), includ¬ <0.5 102 10 5 _mm_Equatorf_Equator
ing diameter and elevation, was classified (71) (17.5) (10) < 11 24 (25) 90 (60)
0.5-4.9 36 37 29 (75) 59 (40)
as small, less than 11 mm; medium, 11 to 15 > 11 71
(25) (65) (61) Total 95 (100) 149 (100)
mm; and large, greater than 15 mm, (2) >: 5.0 6 10 14
"size" categories similar to the classifica¬ (4) (17.5) (29) *2 =
28.0 and P< .0001.
tion of Warren12 were defined as small, 10 Total 144 57 48 tWith or without ciliary body involvement.
mm or less in largest diameter in contact (100) (100) (100)
with the sclera or 2 mm or less in height;
"V = 83.5 and P< .0001.
medium, 11 to 15 mm in largest diameter
or 3 to 5 mm in height; and large, greater
than 15 mm in diameter or greater than 5
Table 4.—Cox Multivariate Analysis: Leading Prognostic Factors
mm in height, and (3) elevation (rounded to
the nearest whole number) was classified 95% Confidence
as low, 2 mm or less; medium, 3 to 5 mm; Variable RR· Interval PValuet
and high, greater than 5 mm. No. of epithelioid cells per
The number of epithelioid cells present high-power field 1.5-4.1
per HPF, LTD, and the number of mitotic Largest tumor dimension 2.0
figures per HPF were analyzed both as Location of anterior margin 1.7 < .03
continuous variables and as discrete vari¬
Invasion to line of transection 1.04-4.6
ables. Discrete levels of these variables
Pigmentation of tumor 1.8 1.04-3.1 < .05
were determined by evaluating outcome
data and grouping cases with similar sur¬ *RR indicates instantaneous relative risk.
vival rates. The following divisions were tP value associated with the standardized regression coefficient. Most influential variables in model have
used in the univariate analysis: (1) epithe¬ smallest associated values.
lioid cells per HPF: none, 0.1 to 0.49, 0.5 to
1.9, 2.0 to 4.9, 5.0 or more; (2) LTD: less variables was conducted for logrank test¬ variables. For the Cox model analysis, the
than 11 mm, 11 to 15 mm, and greater than ing in this univariate analysis. Confidence following divisions were made. The num¬
15 mm, and further subdivisions including limits for the life-table estimates were ber of epithelioid cells per HPF was di¬
7 or 13,14 to 15,16 to 18,
less, 8 to 10,11 to constructed using the method of Green¬ vided into two groups (<2.0 and >2.0), and
and 18 mm or more; and (3) mitotic figures wood" to compute SEs. the LTD was divided into two groups (<12
per HPF: none, 0.1 or less, greater than 0.1. Since variables that predict survival mm and >12 mm). These divisions were
Divisions for the other variables were the may be interrelated, those variables signif¬ made by maximizing the discrimination of
same as those listed previously. icantly associated with survival on univar¬ the log likelihood function. Location of the
Kaplan-Meier survival curves were con¬ iate analysis (Table 1) as well as others anterior margin was classified as anterior
structed for both time to métastases and thought to have clinical significance by to the equator (with or without involve¬
time to melanoma-related deaths'* for all these and other investigators (rupture of ment of the ciliary body) and posterior to
the independent variables. For these two Bruch's membrane or optic nerve invasion) the equator. Pigmentation was classified
end points, patients dying of causes other were then analyzed simultaneously using as minimal or moderate to heavy. Mitotic
than uveal melanoma were considered to proportional hazards analysis of Cox.18 figures were grouped as 0.1 or less or
be censored observations in the analysis. This analysis estimates the "force of mor¬ greater than 0.1.
The logrank test16 was performed to deter¬ tality" and determines the effect of each Interactions were modeled for the fol¬
mine whether the differences in survival factor as a contributor to the instanta¬ lowing pairs of variables: LTD and loca¬
among various levels of these variables neous relative risk of uveal melanoma tion, LTD and "cell type," LTD and age,
when considered individually were statisti¬ métastases and death while permitting location and cell type, location and age, and
cally significant. Categorical grouping of adjustment for the effects of all other cell type and age. Step-up and step-down

Downloaded from www.archophthalmol.com at ALL INDIA INST OF MEDICAL SCIENCES, on April 18, 2011
 100 100

80
None

<0.5
60 0.5-1.9
<
*
40
2.0-4.9

20 >5.0

20

Years
Fig 1.—Relationship between probability of not dying of melanoma and
number of epithelioid cells present per high-power field. (Survival
curves were based on uveal melanoma metastatic deaths, and those
dying of other causes contributed survival data only until time of
death.)
Years
Fig 2.—Relationship between probability of not dying of melanoma and
largest tumor dimension. (Survival curves were based on uveal mela¬
noma metastatic deaths, and those dying of other causes contributed
survival data only until time of death.)

100

Posterior
to Equator

<

Ciliary Body
20
-

0
12 16 20

Years Years
Fig 3. Relationship between probability of not dying of melanoma and Fig 4.—Relationship between probability of not dying of melanoma and
location of anterior margin of tumor. "Anterior to equator" includes
—

invasion of tumor to line of transection. (Survival curves were based on

anterior tumors not involving ciliary body. "Ciliary body" includes uveal melanoma metastatic deaths, and those dying of other causes
anterior tumors involving ciliary body. (Survival curves were based on contributed survival data only until time of death.)
uveal melanoma metastatic deaths, and those dying of other causes
contributed survival data only until time of death.)

procedures were used to confirm the inclu¬
sion of significant covariates and the
exclusion of nonsignificant factors from
the final model. This process provided a
successive selection and ranking of the
independent variables according to their
relative importance as determined by the
values. No time-dependent proportional
hazard analysis was performed.
Instantaneous relative risk (slope of the
survival curve on a log scale) was then
determined for each variable from the
resulting maximum likelihood estimates.
The instantaneous relative risk for a vari¬
able with two levels is the ratio of the risk
of melanoma death for patients with unfa¬
vorable values of the variable divided by
the risk of death for patients with favor¬
able values of the variable, controlling for
the effects of other prognostic variables.
race was known, there were 243
whites and one black. The right eye
was involved in 142 cases (53%) and
the left eye in 125 cases (47% ).
Twenty-one cases (8%) were
nosed routine eye examination, 190
on
patients (71%) had decreased visual
acuity, and 33 patients (12% ) had pain
and/or inflammation. Forty-one pa¬
tients (15%) had other symptoms,
such as photopsia (in addition to the
previously mentioned symptoms in
some cases). Two hundred thirty
patients had enucleation performed at
the MEEI, while the remaining 37
patients were pathology referrals.
Survival Data General
—
causes. The number of patients alive
four years following enucleation was
193; eight years, 143; 12 years, 92; 16
years, 68; and 20 years, 35.
diag¬ Twenty-five percent of the métasta¬
ses occurred within \xk years of the
enucleation, 50% within 3]/2 years,
and 75% within eight years, with a
range of less than one month to 27
years. Métastases involved the liver in
61% of cases. The median time from
métastases to death was 113 days,
with 75% of deaths occurring within
ten months after detection of metas-
tases, (range, less than one month to
five years). The overall 5-, 10-, and
15-year survival rates based on mela¬
noma related deaths and correspond¬

RESULTS
Demographic, Baseline Data
The median age at enucleation of
the 267 patients with ciliary body or
choroidal melanoma was 58.0 years
for men and 59.5 years for women.
There were 123 men (46%) and 144
women (54%). Of the patients whose
Follow-up time ranged from eight ing standard errors were 74% ± 3%,
to 28 years, with a median of 17 years. 63% ±3%, and 55% ±3%, respec¬
Eighty-nine patients (33%) were alive tively.
at the close of the study and four of
these had melanoma métastases. One Survival Data Related to
hundred seventy-seven patients (66% ) Prognostic Factors
were dead, 110 due to melanoma Univariate Analysis.—Results of the
métastases, 23 due to métastases from analyses very similar for both
are
another tumor, and 44 due to other time to métastases and time to

Downloaded from www.archophthalmol.com at ALL INDIA INST OF MEDICAL SCIENCES, on April 18, 2011
 ~H_
H Minimal

Moderate

Heavy

20 r

16 20

Years
Fig 5.—Top left, Relationship between probability of not dying of
melanoma and degree of pigmentation of tumor. (Survival curves were
based on uveal melanoma metastatic deaths and those dying of other
causes contributed survival data only until time of death.)

Fig 6.—Top right, Relationship between probability of not dying of

melanoma and risk class category (see "Risk Categories" section and
Table 6). (Survival curves were based on uveal melanoma metastatic
deaths, and those dying of other causes contributed survival data only
until time of death.)

Fig 7.—Bottom, Relationship between probability of not dying of

melanoma and size of tumor. Small was defined as 10 mm or less in
largest diameter in contact with sclera or 2 mm or less in height;
medium, as 11 to 15 mm in diameter in contact with sclera or 3 to 5 mm
in height; and large, as greater than 15 mm in diameter in contact with
sclera or greater than 5 mm in height. (Survival curves were based on
uveal melanoma metastatic deaths, and those dying of other causes
contributed survival data only until time of death.)

tumor-related deaths. This is to be
expected since the median time from
métastases to death was less than
four months. Results will, therefore,
be presented for tumor related deaths
only.
The 12 variables listed in Table 1
were related to survival when ana¬
lyzed univariately (P < .05). In addi¬
tion, tumor invasion of a scierai emis¬
sary canal was an important prognos¬
tic indicator for the occurrence of
métastases. Location of the anterior
tumor margin was of prognostic sig¬
nificance, but location of the anterior
margin was not.
Interrelationships Among Independent
Variables.—Some variables related to
survival on univariate analysis were
interrelated. In this study, many of
these relationships were statistically
significant, with values less than
.05. For example, tumors with a high¬
er number of epithelioid cells were
also larger, more anteriorly located,
and had more mitotic figures (Table
2). Larger tumors were more likely to
invade to the line of transection, be
anteriorly located (Table 3), and occur
in patients with a history of glauco¬
ma. Anteriorly located tumors were
also more likely to invade to the line
of transection. Tumors involving the
ciliary body had more mitotic figures
per HPF. Older patients were more
likely to have tumors with more
thelioid cells and ciliary body involve¬
ment. Tumors with largest dimension
(>15 mm) were more likely to be
found in older patients; however,
there was no consistent relationship
between age and size in the other size
categories.
Multivariate Analysis.—Multivariate
analysis was performed to determine
which of these interrelated variables
were independently related to out¬
come. Cox proportional hazard model
showed the following five variables in
combination best predicted outcome
in terms of both time to métastases
and time to tumor-related deaths (Ta¬
ble 4). These same factors were also
prognostic for time to death from any
cause: (1) number of epithelioid cells
per HPF, (2) LTD, (3) location of the
anterior margin of the tumor, (4)
invasion to the line of transection, and
(5) pigmentation of the tumor.
None of the six interactions
referred to in the "Methods" section
was found to be statistically signifi¬
cant.
Survival Curves.—Kaplan-Meier sur¬
vival curves for various levels of the
five leading factors are illustrated
(Figs 1 through 5), and the corre¬
sponding survival data are shown in
Table 5.
Figure 1 shows the probability of
not dying of melanoma métastases
epi- dependent on the number of epitheli-
oid cells present. The higher the num¬
ber of these cells, the worse the prog¬
nosis. The outcome was particularly
poor above two epithelioid cells
present per HPF. Only two tumors in
the category "none" were predomi¬
nantly spindle A.
The larger the size of the tumor
based on LTD, the worse the progno¬
sis (Fig 2). The five-year survival rate
for tumors less than 11 mm was 89 ±
3% in contrast to a five-year survival
rate of 35 ± 9% for tumors greater
than 15 mm. Tumors involving the
ciliary body had the worst prognosis
up to about 12 years, after which those
choroidal tumors not involving the
ciliary body but anterior to the equa¬
tor had the poorest outcome (Fig 3).
(Most tumors extending to the ciliary
body in this study also involved the
choroid; only six did not include cho¬
roid.)
Prognosis was worse for tumors
that histologically invaded to the line
of transection with presumed residual
tumor in the orbit (Fig 4). Also, more
heavily pigmented tumors had a
worse prognosis (Fig 5).
Risk Categories
Risk profiles were constructed to
estimate the prognosis of patients
based on the set of tumor characteris¬
tics present at the time of enucleation.
The three most important determi-

Downloaded from www.archophthalmol.com at ALL INDIA INST OF MEDICAL SCIENCES, on April 18, 2011
 Table 5.—Uveal Melanoma Survival Data for Categories of Leading Prognostic Factors Table 7.—Location and Largest
Tumor Dimension: Independent
% Alive ± SE
No. of
Predictors of Survival
Patients' 5 Year 10 Year 15 Year
Location of
Number of Epithelioid Cells per High-Power Field Anterior Margin,
None 70 85 ± 4 80 ± % Survival'
86 ± 5 81 ± 6
0.5-1.9 65 77 66 ± 6 60 ± 7 Largest Anterior
Tumor to Posterior
37 55 + 8 25t
Dimension, Equa- to
301 20t mm tort Equator Value}
Necrotic 63 ± 17 17t 17t < 11 78 92 .05
Largest Dimension, mm >11 51 77 .001
40 92 ± 4 89 value§ < .01 < .001
8-10 88 78 70 ± 6
Only five-year survival rates are listed to illus¬
58 80 57 ± 7 50 + 8 trate differences.
tWith or without ciliary body involvement.
16-18 23 31t 24t %P values refer to differences in overall survival
curves based on melanoma metastatic deaths
19+ 22 ± 15
dependent on location (logrank test), within each
Location of Anterior Margin size category.
Posterior to equator 151 86+3 76 §P values refer to differences in overall survival
Anterior to equator curves based on melanoma metastatic deaths
Without ciliary body dependent on largest tumor dimension (logrank test),
involvement 45 69 45 ± 9 21t within each location category.
With ciliary body
involvement 39 ± 7 34t
Invasion to Line of Transection
should be considered when interpret¬
217 81 ± 3 69 + 3 ing these results. The following five
Yes 25 ± 8 21t
characteristics were determined for
Pigmentation eligible patients who were not in¬
68 85 81 ± 5 cluded in the study: age, sex, size of
Moderate 60 the tumor based on the LTD, extra-
Heavy 89 68 50 ± 6 46 scleral extension, and location of the
*
Survival data based on melanoma-related deaths. Patients dying of other causes were considered to be tumor. There were no significant dif¬
censored observations and contributed survival data only until their time of death. Total number of patients for ferences for the first four variables
each variable might not add to 267, since some characteristics were not able to be determined for all between this group and the 267
patients. patients who were included in the
tFive patients survived beyond the specific time period.
study. Thus, the study population is
likely to be representative of the
Table 6.—Survival Data for Three Risk Classes
entire group of patients with uveal
melanoma meeting the inclusion cri¬
% Alive ± SEt teria with respect to these character¬
istics. However, patients included in
Risk Class' No. of Patients 5 Year 10 Year 15 Year Value* the study had more tumors with ante¬
Good 92 85 83
rior margins located posterior to the
Fair 119 75 65 + 5 50 ± 6 < .0001
Poor 46 31 10§
equator. This raises the possibility
that tumors with a worse prognosis
"Good means patients with favorable values for all three variables; fair, patients with either favorable
values for "cell type" or largest tumor dimension (LTD) and one unfavorable variable; poor, patients with
(ie, more anteriorly located tumors)
were lost to follow-up.
unfavorable values for both "cell type" and LTD.
tSurvival data based on melanoma related deaths. Patients dying of other causes were considered to be Most medical records that were
censored observations and contributed survival data only until their time of death. available were not complete with
tP value associated with logrank test. respect to all of the independent vari¬
§Number of patients surviving beyond the specific time period indicated is less than five. ables analyzed. Therefore, complete
ascertainment of variables such as eye
nants of outcome were divided into type or LTD was favorable, and (3) involved, manifesting signs and symp¬
favorable and unfavorable categories. poor, those with unfavorable values toms, visual acuity, and history of
Less than two epithelioid cells was for both cell type and LTD. Survival glaucoma was not possible.
favorble, LTD 12 mm or less was curves were constructed for these Determination of cause of death is
favorable, and the location of anterior three groups based on melanoma- always subject to error, since death
margin being posterior to the equator related deaths (Fig 6, Table 6). For certificates and hospital discharge
was favorable. The location was unfa¬ example, if a tumor is in a poor-risk diagnoses are not always accurate and
vorable if it was anterior to the equa¬ category, has many epithelioid cells, is metastatic workups may be incom¬
tor. Two of these factors—size and medium or large, and extends anterior plete. We made every attempt to
location—can be estimated clinically. to the equator, the probability of sur¬ obtain biopsy or autopsy specimens or
Patients in the study were then viving (without melanoma metastatic reports and confirmed 65 of 115
grouped into risk classes using the death) five years after enucleation is métastases.
following criteria: (1) good, those with about 31% ± 7%. Comparison among studies is diffi¬
favorable values for all three vari¬ cult because of the lack of an objec¬
COMMENT
ables, (2) fair, those with at least one tive, reproducible classification of cell
unfavorable variable, but either cell Some features of our methods type. The percentage of the various

Downloaded from www.archophthalmol.com at ALL INDIA INST OF MEDICAL SCIENCES, on April 18, 2011
types of cells required for each sub¬
type in the classification of Callen¬
der19 is unclear. Present criteria of cell
type are subjective and lead to varia¬
tions among different observers of the
same tumor.2021 We have classified cell
type by the number of epithelioid cells
present in a tumor per HPF on light
microscopy. Such classification is
more quantitative than the method of
Callender, but it still depends on the
observer's definition of an epithelioid
cell. Assessments of cell types based
on cytologie features of tumor cells
are also being explored.21 Reliability
checks and a comparison of predictive
value in terms of survival between old
and new techniques are required be¬
fore one classification can be consid¬
ered an improvement over the other.
With these méthodologie points in
mind, our list of the most important
prognostic factors is worthy of com¬
ment.
First, by classifying tumors accord¬
ing to the number of epithelioid cells
present, our results demonstrate that
tumors that might be grouped togeth¬
er as mixed cell in the classification of
Callender have different prognoses
dependent on the number of epitheli¬
oid cells present. Consistent with Zim¬
merman and McLean22 and McLean et
al,1 we are convinced that cell type is
the most important predictor of the
outcome of patients with uveal mela¬
noma.
Second, our study is the first to
compare the prognostic value of three
classifications of size among tumors
of all sizes with and without extra-
scleral extension. The largest
dimension of the tumor was the best
prognostic indicator (Fig 2). "Size"
classified similar to the definition of
Warren" was not nearly as effective
in predicting outcome (Fig 7). Nor was
height of the tumor an independent
predictor of outcome. Our findings are
consistent with previous reports.ls
Third, we found that the location of
the anterior margin of the tumor was
an important independent predictor
of outcome. Others who found that
anteriorly located tumors have a
worse prognosis1·3 believed that this
correlation was not an independent
one but was due to other factors,
especially size. Our data confirm that
larger tumors are more likely to have
their anterior margins located anteri¬
or to the equator (Table 3). However,
in this study each variable, LTD, and
location, has an independent effect on
outcome. Choroidal tumors anterior
to the equator, many of which involve
the ciliary body, have a worse progno¬
sis regardless of size (Table 7). The
Downloaded from www.archophthalmol.com at ALL INDIA INST OF MEDICAL SCIENCES, on April 18, 2011
reason for this is unclear. Our data
show that tumors extending anterior
to the equator are more likely to
invade to the line of transection, have
more epithelioid cells, and, if the cili¬
ary body is involved, have more mitot¬
ic activity. Such findings suggest a
more malignant potential for these
tumors with perhaps more rapid
growth. Raivio9 found that patients
with a melanoma of the ciliary body
have a worse prognosis than patients
with tumors involving only the cho¬
roid. He postulated that the close con¬
nection with ciliary vessels and con¬
stant movement of the ciliary muscle
may facilitate the spread of tumor
cells into the bloodstream. The impor¬
tance of anterior location should be
confirmed by multivariate analysis of
other populations of patients with
uveal melanoma.
Fourth, our study placed less
emphasis on some factors reported as
important variables by others.111 This
can be explained in part by the multi¬
variate analysis that controls for
"confounding" by identifying vari¬
ables correlated with a "true" prog¬
nostic indicator and also related to
outcome. For example, in our data, the
number of mitotic figures and number
of epithelioid cells per HPF were
interrelated (Table 2). Mitotic activity
was related to outcome by univariate
analysis but became statistically non¬
significant in multivariate analysis,
because the variable most related to
survival was the number of epitheli¬
oid cells. Within each category of cell
type, mitotic activity was not signifi¬
cantly related to outcome. The rela¬
tionship between mitotic figures and
outcome, therefore, was confounded
by the number of epithelioid cells
present. Mitotic activity, however,
was found to be important in a study
that also employed multivariate tech¬
niques.1 In that report, only smaller
tumors were analyzed, and this may
account for the differences in our
results.
Similar to the previously described
example, the relationship between the
three variables, the clinical manifes¬
tation of pain and inflammation, the
degree of inflammation on histopath¬
ological examination, and a history of
glaucoma with survival were all
explained by the size of the tumor
(LTD). These variables were not prog¬
nostic indicators when analyzed mul-
tivariately because the true relation¬
ship for all three was between LTD
and outcome.
This study was supported in part by grants 5
F32 EY05507 (Dr Seddon) and 5 ROI EY01917 (Dr
Albert) from the National Eye Institute Bethes-
da, Md, grant CA-06516 from the National Can¬
cer Institute (Dr Lavin), Bethesda,
Md, and grant
5 S07 PR05526, Biomédical Research Support
Grant, from the Sidney Farber Cancer Institute,
Boston (Dr Lavin).
William Fine, MS, provided follow-up proce¬
dures and Ellen Hertzmark, MA, helped with
statistical analysis.
References
1. McLean IW, Foster WD, Zimmerman LE:
Prognostic factors in small malignant melano-
mas of choroid and ciliary body. Arch Ophthal-
mol 1977;95:48-58.
2. Shammas HF, Blodi FC: Prognostic factors
in choroidal and ciliary body melanomas. Arch
Ophthalmol 1977;95:63-69.
3. Packard RBS: Pattern of mortality in cho-
roidal malignant melanoma. Br J Ophthalmol
1980;64:565-575.
4. McLean IW, Shields JA: Prognostic value of
32P uptake in posterior uveal melanomas. Oph-
thalmology 1980;87:543-548.
5. Affeldt JC, Minckler DS, Azen SP, et al:
Prognosis in uveal melanoma with extrascleral
extension. Arch Ophthalmol 1980;98:1975-1979.
6. Greer CH, Buckley C, Buckley J, et al: An
Australian choroidal melanoma survey: Factors
affecting survival following enucleation. Aust J
Ophthalmol 1981;9:255-261.
7. Jensen OA: Malignant melanoma of the
human uvea: Recent follow-up of cases in Den-
mark, 1943-1952. Acta Ophthalmol 1970;48:1113\x=req-\
1128.
8. Jensen OA: Malignant melanoma of the
uvea in Denmark, 1943-1952: A clinical, histo-
pathologic and prognostic study. Acta Ophthal-
mol Suppl 1963;75:1-220.
9. Raivio I: Uveal melanoma in Finland. Acta
Ophthalmol Suppl 1977;133:1-64.
10. Flocks M, Gerende JH, Zimmerman LE:
The size and shape of malignant melanomas of
the choroid and ciliary body in relation to prog-
nosis and histologic characteristics. Trans Am
Acad Ophthalmol Otolaryngol 1955;59:740-758.
11. Barr CC, Sipperley JO, Nicholson DH:
Small melanomas of the choroid. Arch Ophthal-
mol 1978;96:1580-1582.
12. Thomas JV, Green WR, Maumenee AE:
Small choroidal melanomas: A long term follow\x=req-\
up study. Arch Ophthalmol 1979;97:861-864.
13. Davidorf FH, Lang JR: The natural history
of malignant melanoma of the choroid: Small vs
large tumors. Trans Am Acad Ophthalmol Otolar-
yngol 1975;79:310-320.
14. Warren RM: Prognosis of malignant mela-
nomas of the choroid and ciliary body, in Blodi
FC (ed): Current Concepts in Ophthalmology. St
Louis, CV Mosby Co, 1974, vol 4, p 160.
15. Kaplan EL, Meier P: Nonparametric esti-
mation from incomplete observations. J Am Stat
Assoc 1958;53:457-481.
16. Peto R, Peto J: Asymptotically efficient
rank invariant test procedures. J R Stat Soc
1972;135:185-206.
17. Greenwood M: The natural duration of
cancer. Rep Health Soc Subj 1926;33:1-26.
18. Cox DR: Regression models and life-tables.
J R Stat Soc Series B 1972;34:187-220.
19. Callender GR: Malignant melanotic tu-
mors of the eye: A study of histologic types in 111
cases. Trans Am Acad Ophthalmol Otolaryngol
1931;36:131-140.
20. Gass JDM: Problems in the differential
diagnosis of choroidal nevi and malignant mela-
nomas. Trans Am Acad Ophthalmol Otolaryngol
1977;83:19-48.
21. Gamel JW, McLean IW: Quantitative anal-
ysis of the Callender classification of uveal mela-
noma cells. Arch Ophthalmol 1977;95:686-691.
22. Zimmerman LE, McLean IW: Changing
concepts of the prognosis and management of
small malignant melanomas of the choroid.
Trans Ophthalmol Soc UK 1975;95:487-494.