- Email: [email protected]
A deficit survival analysis to assess the natural history of uveal melanoma
0021.9681/84
J Chron Dis Vol. 37, No. 6, pp. 481-487. 1984 Printed in Great Bntan. All nghts reserved
Copyright (’ 1984 Pergamon
A DEFICIT SURVIVAL NATURAL HISTORY PHILIP
T.
LAWN,’
DANIEL
M.
$3.00 + 0.00 Press Ltd
ANALYSIS TO ASSESS THE OF UVEAL MELANOMA ALBERT’
and JOHANNA M. SEDDON~
‘Harvard School of Public Health and Dana-Farber Cancer Institute, Division of Biostatistics and Epidemiology, 44 Binney Street. Boston, MA 02115 and ‘Department of Ophthalmology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, 243 Charles Street, Boston, MA 02114, U.S.A. (Recriwd
in rwised
fbrm 14 Ocroher
1983)
Abstract-A deficit survival analysis was conducted using a retrospective series of 230 patients with uveal melanoma. All underwent enucleation as primary treatment for their disease. Considering the three leading prognostic factors (epithelioid cells per high power field, tumor size, and tumor location). median age was plotted against median survival to determine pathways of disease spread, local and metastatic disease phases, consequences of delayed diagnosis, and the timing of enucleation. The subgroup with all three factors favorable could be subdivided into younger and older subsets on the basis of a bimodal age distribution. Only the older subset could be identified as a local disease state where survival following enucleation corresponded to normal population survival. All remaining subgroups were metastatic because of survival deficits compared to the normal population. Disease tended to progress according to three well defined pathways with 68 years required to progress from one subgroup to the next along these paths. These pathways were characterized by constant deficit survivals. suggesting that delays in diagnosis did not translate into loss in survival. Moreover, the value of enucleation as primary therapy can be questioned in the context that it does not appear to alter the natural history of the disease, except for tumors greater than 10 mm in largest dimension which are located anterior to the equator with fewer than 2 epithelioid cells per high power field. Findings suggest that uveal melanoma can be treated by means other than enucleation to allow a chance for prolonged survival with vision preservation.
BACKGROUND
SINCE the beginning of the nineteenth century, the treatment of uveal melanoma has been controversial. In 1810, Travers [l] advised that enucleation gave a patient a chance for prolonged survival by removal of the tumor. Fifty-eight years later, Knapp [2] asserted, on the basis of discussions with mathematicians, that the chances of cure must increase if the tumor is removed at the earliest stages. In a study of 100 cases, Von Graefe [3] recommended early surgeries half-heartedly. Sixteen years later, in 1882, Fuchs [4] advised that enucleation should be performed as early as possible. His recommendation stood as the treatment of choice until recent times. Recent times have seen challenges to this strategy of prompt enucleation. Zimmerman [5,6] postulated in a series of articles that enucleation may actually hasten the spread of disease, as a few cases of metastatic disease were observed among patients initially enucleated for uveal melanoma. Siegel [7] re-examined Zimmerman’s data and concluded that the risk of mortality post-enucleation for uveal melanoma is similar to the risk of mortality post-resection for other solid tumors. Subsequently, Manschot and Van Pepperzeel [8] brought the controversy full circle by recommending prompt enucleation of all eyes containing melanomas. The recommendations of Fuchs to promptly enucleate were followed closely by ophthalmologists treating this disease. There are no definitive data available that pertain This investigation was supported and Grants 5 ROl EY01917
by Grants CA-06516, and 5 SO7 PRO5526 from the National and 5 F32 EY05507-02 from the National Eye Institute.
Cancer
Institute,
482
PHILW T. LAWN LI ul
to the untreated natural history of patients. A randomized clinical trial of early vs delayed enucleation would be ethically difficult to conduct and would require considerable years of follow-up. Consequently, there is no direct evidence of the effect of surgery as a primary therapeutic intervention for uveal melanoma. In the absence of direct evidence, indirect evidence has been sought using deficit survival methodology, devised by Zelen [9], for inferring the natural history of chronic diseases. This technique allows one to address the following questions: (a) (b) (c) (d)
What are the local and metastatic disease phases? What are the pathways of disease progression? What is the consequence of delayed diagnosis? Does the timing of enucleation matter?
This paper
investigates
these questions
SUBJECTS
using
AND
the deficit survival
method.
METHODS
Population This paper is based on a series of 230 patients with uveal melanoma who underwent enucleation at Massachusetts Eye and Ear Infirmary as primary treatment for their disease between 1953 and 1973. All underwent enucleation within 6 months of disease diagnosis. Although 478 patients were seen at this hospital during this time, cases with missing data, inadequate pathology materials, or inadequate follow-up were excluded from this analysis. There did not appear to be any misrepresentation of the original series with respect to homogeneity of population characteristics in the comparison of the 230 included cases to the 248 excluded cases. The study population and details relating to case ascertainment, follow-up, standardization, and endpoints are described elsewhere [l I]. Extensive data relating to histopathology, clinical findings, and demographical information were analyzed. Thirty-seven cases with missing age, tumor location, or tumor size were excluded from that database for analysis in this report. The three leading prognostic factors for survival were the number of epithelioid cells per high power field, size of the tumor at enucleation, and the location of the anterior margin of the tumor. The prognostic value of tumor size and location has been confirmed by other authors such as Shammas and Blodi [lo]. In a companion analyses [l 13, these factors were simultaneously prognostic for disease-free interval, survival including only tumor-related deaths, as well as survival including all causes of death. Methodology These three prognostic factors were incorporated into the deficit survival analysis. Dichotomous levels were chosen for each factor as indicated in Fig. 1. The dichotomous classifications for the number of epithelioid cells per high power field and for the largest dimension were chosen to reflect both the prognostic significance for survival and guarantee of sufficient numbers within subgroups to conduct the analysis. Regarding subgroups, each patient was assigned to one of the eight possible subgroups corresponding to all combinations of the three prognostic factor levels. In the subgroup identification,
Prognostic
factor
Favorable
level
Unfavorable
Epithelioid cells per high power field
6 2.0
> 2.0
Tumor size (largest dimension)
6 10 mm
> 10 mm
Location anterior
of margin
Posterior equator
to
FIG. 1. Prognostic factors-favorable and
Anterior equator
unfavorable
levels.
to
level
A Deficit Survival Analysis
483
a favorable or unfavorable categorization exists for each of the three factors. In this convention, the presence of a letter indicates a favorable factor is present within the subgroup, i.e. e corresponds to tumors with no more than 2.0 epithelioid cells per high power field, s indicates tumors no larger than 10 mm, while a refers to tumors with anterior margin posterior to the equator. The eight possible subgroups are: esa, es, ea, sa, e, s, a, and u; su refers to the subgroup with tumor size level and anterior margin level favorable and epithelioid cell level unfavorable. U refers to the subgroup with all three factors unfavorable. For each of these subgroups, median survival (from all causes of death inclusive of uveal melanoma) and median age were directly computed and then plotted simultaneously for each possible subgroup. The expected survival for a normal population consisting of the 1973 general United States population [12] was also plotted in this display. For each subgroup, the expected survival can be determined from the median age of the subgroup. The deficit survival is then computed by taking the difference between the expected survival and the median survival for each subgroup. Assumptions The deficit survival
method
is based
on the following
four assumptions:
(1) disease progresses from more favorable states to less favorable states over time, (2) disease progression between consecutive states is characterized by a change principally reflected in only one factor at a time, (3) deficit survival cannot decrease in passage from more favorable to less favorable states over time, (4) less favorable states are reached by advancing age from more favorable states. The first two assumptions mean that a patient with uveal melanoma in a particular stage who is not diagnosed, either stays at that stage or will change to another stage which is less favorable in one or more of the prognostic categories (size, cell type, location). These assumptions are reasonable with respect to uveal melanoma. Although the natural history of untreated uveal melanomas is not well known, spontaneous regression has not been reported. Many tumors will grow if followed over time [13]. The last two assumptions imply that the occurrence of the disease or progression does not improve survival or prognosis relative to the normal population. These assumptions also are supported from ongoing clinical studies where patients with uveal melanoma are observed for intervals prior to initiation of tumor directed therapy. The main idea of the method is to connect admissible subgroups (states) and to infer the natural history of the disease from the location of admissible paths with reference to the normal population, Those states with survival near the normal population with respect to remaining years of survival are called local disease states because treatment resulted in a survival outcome comparable to the normal population. Those remaining states that lie below the normal population curve have survival deficits and are thus called metastatic disease states because treatment resulted in survival prognosis that was poorer than the normal population. Passage from local to metastatic states and between metastatic states can be determined from the four rules given above. Transition from local to metastatic states indicates the value of early detection. Passage between metastatic disease states also indicates this as well as the value of early treatment intervention. The change in deficit survival and the time to pass between consecutive states gives the value of early detection and early treatment intervention. For example, to pass between consecutive states with a 10 year increase in deficit survival requiring 5 years to reach the less favorable state, each year of early detection would correspond to a 2 year gain in expected survival. However, no change in deficit survival under the same conditions would suggest that early detection and treatment did not influence survival outcome.
The subgroups survival analysis.
RESULTS were examined with respect to age distributions prior to the deficit The most favorable subgroup (esa) had the largest number of patients
PHILII’
484
T.
LAVlV
6’1
9
Subgroups
esa(younger,older);es,ea,so,e,s,o,u Epltheliold cells / HPF Largest tumor dlmenslon Anterior margln
e s
a
c 0
\
12 c es
\
-1,
& E
8
50
x
‘0 E
\
l
l\
‘\
‘\\ \
4
\ \
\\
45
Age
at
diagnosis
( yr )
Medlon FIG. 2. Display
of bimodal age distribution PILI subgroup.
I
I
50
55
age
at
e
5 l\
‘\
‘\.
0
;; w
0
Posterior
‘.
‘\
‘..
5 2.0 5 10mm
‘1 ,;;,
‘\
60
‘.
u
65
dlagnosls
70
( yr )
far the FIG. 3. Deficit
survival
analysis
(MEEI
Series),
(78). An examination of the age distribution revealed a bimodal distribution as displayed in Fig. 2. There were relatively fewer patients between age 50 and 55 than in other age groups. Consequently, in the analysis that follows, the esu subgroup was divided into younger (37 patients) and older (41 patients) subgroups according to the median age of the entire population. Thus, in the analysis that follows, nine subgroups were evaluated. Table 1 displays the number of patients, median survival after enucleation. median age at enucleation, expected survival, and deficit survival for the nine subgroups. Figure 3 presents a deficit survival display of age at enucleation against survival following enucleation for the nine subgroups. The interpretation of this figure provides evidence for: (a) (b) (c) (d)
local and pathways diagnosis timing of
metastatic disease phases. of disease progression, delay consequences, and enucleation.
Local and metastatic
disease states
The nine subgroups cluster into four categories on the basis of the magnitude of the deficit survival. Only the older esa subgroup had remaining survival comparable to the normal population, suggesting that this most favorable subgroup was the only local disease state. The younger esa. ea, and e subgroups each had 9 year survival deficits. This worsened to 14 years for the es, s, and U subgroups and to 18 years for the sa and a subgroups. Thus, all but the older esa subgroup corresponded to metastatic disease states.
A Deficit
Pathways
Survival
Analysis
485
of disease progression
The inference of disease pathways derives from the connection to the deficit survival rules. The following pathways are suggested by these postulates: (1) younger (2) es+sjU (3) sa+a
esa+ea+e+
of subgroups
according
U
The first pathway represents a deterioration in the number of favorable prognostic factors, characterized first by increase in tumor size, then by possible advancement from a location posterior to the equator to an anterior location, and finally marked by an increase in the number of epithelioid cells per high power field to reach the most unfavorable state U. Six to eight years were required to pass between successive states. The passage is indicative of relatively slow tumor growth prior to change in the number of epithelioid cells per high power field. In contrast, the second pathway suggests that small tumors with unfavorable locations can undergo transition in the number of epithelioid cells per high power field prior to demonstration of active tumor growth. The third pathway represents a change in size for tumors located anterior the equator. All three pathways illustrate constant deficit survival, which are essentially parallel to the normal population survival curve. These constant deficit survival paths are indicated by dashed lines in Fig. 3. Diagnosis
delay consequences
The passage times between states illustrate constant deficit survival along three pathways. Thus, earlier detection of disease along a particular pathway did not result in decreased deficit survival, except for the passage from e+ U. There, a 5 year delay in diagnosis (median age 62-67) resulted in a 4 year increase in deficit survival (a loss of 9 years to 13 years compared to the normal population) for patients in the e subgroup. This is in contrast to the remaining pathways where earlier diagnosis did not alter deficit survival along pathways. For these pathways, a l-year delay in diagnosis does not change the deficit survival. It should be noted, however, if the tumor develops additional unfavorable factors with the passage of time, then the remaining years of survival decrease, as one might expect to occur. Timing of enucleation The three constant deficit survival curves suggest that for these pathways, early enucleation did not alter the deficit survival. For all subgroups except the e and older esa subgroups, a year of delay in diagnosis did not increase the deficit survival. For the e subgroup, early enucleation might be regarded as a favorable intervention as a delay leads to increased deficit survival. Thus, earlier enucleation neither increased nor decreased the deficit survival except for the e subgroup. These data suggest that enucleation did not have major impact on the natural history of uveal melanoma. The older esa subgroup had comparable survival to the normal population.
DISCUSSION
The debate over enucleation as primary treatment for uveal melanoma has lasted for over 150 years. The opponents of early enucleation represent points of view that surgery does not benefit the patient in terms of prolonged survival and may actually hasten the spread of the disease, reducing the survival. The data presented here suggest that early enucleation sustains normal population survival only for patients with tumors with largest dimension above 10 mm and an anterior margin in front of the equator which have few epithelioid cells. For other more favorable subgroups, enucleation may be delayed to allow for other therapeutic interventions as primary therapy. This finding is similar to that
486
PHILIP
T. LAVIN et al.
reported by Zelen for radical mastectomy and breast cancer where prompt removal of the tumor did not appear to alter survival. The most definitive testing of the benefit of delayed enucleation will require a randomized clinical trial of prompt vs delayed enucleation. However, this study will be very difficult to complete because of ethical considerations. Even if ethical issues could be addressed, completion would require many years due to the low incidence rate (about 1800 new cases per year in the U.S.) and the long survivals associated with uveal melanoma. Consequently, indirect methods of evaluation, provided by the deficit survival method, may provide the only evidence of the value of enucleation as primary treatment for uveal melanoma. For uveal melanoma, this method has certain limitations in making clinical decisions. First, classification into states corresponding to this analysis is not possible preoperatively because the number of epithelioid cells per high power field cannot be determined without performing a pathology assessment. Second, according to these findings, older patients with favorable size and location, who happen to have few epithelioid cells per high power field have been shown to have no deficit survival. This may be due to disease of a more benign natural history, perhaps suggesting that other therapeutic interventions may result in similar outcome. However, the younger patient with either favorable size or location is a candidate for alternative therapeutic interventions as evidenced by the deficit survivals in Table 1, which indicate shortened survivals following enucleation compared to the normal population. It may well be that ophthalmologists can consider alternative forms of therapy in designated subgroups where prompt enucleation might not yield favorable results and the possibility of more favorable outcomes might exist. The data demonstrating the benefit of enucleation is far from definitive, and further studies are required to see if better therapies do indeed exist. In conclusion, these findings explain many of the reasons for the controversies associated with uveal melanoma treatment. All patients with this disease do not uniformly benefit from either prompt diagnosis or enucleation. This differs from many other malignancies where early detection and treatment are associated with prolonged survival. These results should be confirmed by similar analyses of other uveal melanoma populations. Other studies involving long term survival may benefit from the deficit survival method. Here, the deficit survival method was applied in this retrospective series to gain insights not offered by other forms of statistical analysis such as the proportional hazards model or competing risk model. These alternative procedures do not compare survival to the hazards analysis was normal population. In this retrospective series, the proportional performed first to identify the prognostic factors used in this analysis. Thus, the deficit survival method should be regarded as an adjunct procedure to be utilized in the analysis of chronic diseases.
REFERENCES I.
2. 3. 4. 5. 6. 7. 8. 9.
Travers B: On the local diseases termed malignant. Cited by H. Knapp in A Treatise on Intraocular Tumors from Original Clinical Observations and Anatomical Investigations. (Translated by Cole _I.) New York: William Wood. 1839 Knano H: A Treatise on Intraocular Tumors from Original Clinical Observations and Anatomical Inve&igations. (Translated by Cole J.) New York: William Wood, 1869 Von Graefe A: Zusatze ueber intraoculare Tumoren. Arch Oubtbalmol 14: 103, 1868 Fuchs E: Das Sarcom des Uvealtractus. Wien: Wilhelm Braumuller, 1882 Zimmerman LE, McLean IW: An Evaluation of enucleation in the management of weal melanoma. Am J Oubtbalmol 87: 741. 1979 Zimmerman LE, McLean IW, Foster WD: Does enucleation of eye containing a malignant melanoma prevent or accelerate the dissemination of tumor cells? Br J Ouhthalmol 62: 420, 1978 ‘Seigel D. Myers M, Ferris F. III. Sternhorn SC: Survival rates after enucleation of eyes with malignant melanoma. Am J Ophthalmol 87: 761, 1979 Manshot WA, van Pepperreel HA: Choroidal melanoma: enucleation or observation? A new approach. Arch Ophthalmol 89: 7 1, 1980 Zelen M: A hypothesis for the natural time history of breast cancer. Cancer Res 28: 207, 1968
A Deficit IO.
1I. 12. 13.
Survival
Analysis
487
Shammas HF, Blodi FC: Prognostic factors in choroidal and ciliary body melanomas. Arch Ophthalmol 95: 63, 1977 Seddon JM, Albert DM, Lavin PT, Robinson N: A prognostic factor study of disease-free interval and survival following enucleation for uveal melanoma. Arch Ophthalmol 101: 1894-l 899, 1983 Vital Statistics of United States, Mortality Statistics Vol. II. Part A, 1975 McLean IW, Foster. WD, Zimmerman LE: Uveal melanoma: location, size, cell type. and enucleation as risk factors in metastasis. Hum Pathol 13: 123-132, 1982
J Chron Dis Vol. 37, No. 6, pp. 481-487. 1984 Printed in Great Bntan. All nghts reserved
Copyright (’ 1984 Pergamon
A DEFICIT SURVIVAL NATURAL HISTORY PHILIP
T.
LAWN,’
DANIEL
M.
$3.00 + 0.00 Press Ltd
ANALYSIS TO ASSESS THE OF UVEAL MELANOMA ALBERT’
and JOHANNA M. SEDDON~
‘Harvard School of Public Health and Dana-Farber Cancer Institute, Division of Biostatistics and Epidemiology, 44 Binney Street. Boston, MA 02115 and ‘Department of Ophthalmology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, 243 Charles Street, Boston, MA 02114, U.S.A. (Recriwd
in rwised
fbrm 14 Ocroher
1983)
Abstract-A deficit survival analysis was conducted using a retrospective series of 230 patients with uveal melanoma. All underwent enucleation as primary treatment for their disease. Considering the three leading prognostic factors (epithelioid cells per high power field, tumor size, and tumor location). median age was plotted against median survival to determine pathways of disease spread, local and metastatic disease phases, consequences of delayed diagnosis, and the timing of enucleation. The subgroup with all three factors favorable could be subdivided into younger and older subsets on the basis of a bimodal age distribution. Only the older subset could be identified as a local disease state where survival following enucleation corresponded to normal population survival. All remaining subgroups were metastatic because of survival deficits compared to the normal population. Disease tended to progress according to three well defined pathways with 68 years required to progress from one subgroup to the next along these paths. These pathways were characterized by constant deficit survivals. suggesting that delays in diagnosis did not translate into loss in survival. Moreover, the value of enucleation as primary therapy can be questioned in the context that it does not appear to alter the natural history of the disease, except for tumors greater than 10 mm in largest dimension which are located anterior to the equator with fewer than 2 epithelioid cells per high power field. Findings suggest that uveal melanoma can be treated by means other than enucleation to allow a chance for prolonged survival with vision preservation.
BACKGROUND
SINCE the beginning of the nineteenth century, the treatment of uveal melanoma has been controversial. In 1810, Travers [l] advised that enucleation gave a patient a chance for prolonged survival by removal of the tumor. Fifty-eight years later, Knapp [2] asserted, on the basis of discussions with mathematicians, that the chances of cure must increase if the tumor is removed at the earliest stages. In a study of 100 cases, Von Graefe [3] recommended early surgeries half-heartedly. Sixteen years later, in 1882, Fuchs [4] advised that enucleation should be performed as early as possible. His recommendation stood as the treatment of choice until recent times. Recent times have seen challenges to this strategy of prompt enucleation. Zimmerman [5,6] postulated in a series of articles that enucleation may actually hasten the spread of disease, as a few cases of metastatic disease were observed among patients initially enucleated for uveal melanoma. Siegel [7] re-examined Zimmerman’s data and concluded that the risk of mortality post-enucleation for uveal melanoma is similar to the risk of mortality post-resection for other solid tumors. Subsequently, Manschot and Van Pepperzeel [8] brought the controversy full circle by recommending prompt enucleation of all eyes containing melanomas. The recommendations of Fuchs to promptly enucleate were followed closely by ophthalmologists treating this disease. There are no definitive data available that pertain This investigation was supported and Grants 5 ROl EY01917
by Grants CA-06516, and 5 SO7 PRO5526 from the National and 5 F32 EY05507-02 from the National Eye Institute.
Cancer
Institute,
482
PHILW T. LAWN LI ul
to the untreated natural history of patients. A randomized clinical trial of early vs delayed enucleation would be ethically difficult to conduct and would require considerable years of follow-up. Consequently, there is no direct evidence of the effect of surgery as a primary therapeutic intervention for uveal melanoma. In the absence of direct evidence, indirect evidence has been sought using deficit survival methodology, devised by Zelen [9], for inferring the natural history of chronic diseases. This technique allows one to address the following questions: (a) (b) (c) (d)
What are the local and metastatic disease phases? What are the pathways of disease progression? What is the consequence of delayed diagnosis? Does the timing of enucleation matter?
This paper
investigates
these questions
SUBJECTS
using
AND
the deficit survival
method.
METHODS
Population This paper is based on a series of 230 patients with uveal melanoma who underwent enucleation at Massachusetts Eye and Ear Infirmary as primary treatment for their disease between 1953 and 1973. All underwent enucleation within 6 months of disease diagnosis. Although 478 patients were seen at this hospital during this time, cases with missing data, inadequate pathology materials, or inadequate follow-up were excluded from this analysis. There did not appear to be any misrepresentation of the original series with respect to homogeneity of population characteristics in the comparison of the 230 included cases to the 248 excluded cases. The study population and details relating to case ascertainment, follow-up, standardization, and endpoints are described elsewhere [l I]. Extensive data relating to histopathology, clinical findings, and demographical information were analyzed. Thirty-seven cases with missing age, tumor location, or tumor size were excluded from that database for analysis in this report. The three leading prognostic factors for survival were the number of epithelioid cells per high power field, size of the tumor at enucleation, and the location of the anterior margin of the tumor. The prognostic value of tumor size and location has been confirmed by other authors such as Shammas and Blodi [lo]. In a companion analyses [l 13, these factors were simultaneously prognostic for disease-free interval, survival including only tumor-related deaths, as well as survival including all causes of death. Methodology These three prognostic factors were incorporated into the deficit survival analysis. Dichotomous levels were chosen for each factor as indicated in Fig. 1. The dichotomous classifications for the number of epithelioid cells per high power field and for the largest dimension were chosen to reflect both the prognostic significance for survival and guarantee of sufficient numbers within subgroups to conduct the analysis. Regarding subgroups, each patient was assigned to one of the eight possible subgroups corresponding to all combinations of the three prognostic factor levels. In the subgroup identification,
Prognostic
factor
Favorable
level
Unfavorable
Epithelioid cells per high power field
6 2.0
> 2.0
Tumor size (largest dimension)
6 10 mm
> 10 mm
Location anterior
of margin
Posterior equator
to
FIG. 1. Prognostic factors-favorable and
Anterior equator
unfavorable
levels.
to
level
A Deficit Survival Analysis
483
a favorable or unfavorable categorization exists for each of the three factors. In this convention, the presence of a letter indicates a favorable factor is present within the subgroup, i.e. e corresponds to tumors with no more than 2.0 epithelioid cells per high power field, s indicates tumors no larger than 10 mm, while a refers to tumors with anterior margin posterior to the equator. The eight possible subgroups are: esa, es, ea, sa, e, s, a, and u; su refers to the subgroup with tumor size level and anterior margin level favorable and epithelioid cell level unfavorable. U refers to the subgroup with all three factors unfavorable. For each of these subgroups, median survival (from all causes of death inclusive of uveal melanoma) and median age were directly computed and then plotted simultaneously for each possible subgroup. The expected survival for a normal population consisting of the 1973 general United States population [12] was also plotted in this display. For each subgroup, the expected survival can be determined from the median age of the subgroup. The deficit survival is then computed by taking the difference between the expected survival and the median survival for each subgroup. Assumptions The deficit survival
method
is based
on the following
four assumptions:
(1) disease progresses from more favorable states to less favorable states over time, (2) disease progression between consecutive states is characterized by a change principally reflected in only one factor at a time, (3) deficit survival cannot decrease in passage from more favorable to less favorable states over time, (4) less favorable states are reached by advancing age from more favorable states. The first two assumptions mean that a patient with uveal melanoma in a particular stage who is not diagnosed, either stays at that stage or will change to another stage which is less favorable in one or more of the prognostic categories (size, cell type, location). These assumptions are reasonable with respect to uveal melanoma. Although the natural history of untreated uveal melanomas is not well known, spontaneous regression has not been reported. Many tumors will grow if followed over time [13]. The last two assumptions imply that the occurrence of the disease or progression does not improve survival or prognosis relative to the normal population. These assumptions also are supported from ongoing clinical studies where patients with uveal melanoma are observed for intervals prior to initiation of tumor directed therapy. The main idea of the method is to connect admissible subgroups (states) and to infer the natural history of the disease from the location of admissible paths with reference to the normal population, Those states with survival near the normal population with respect to remaining years of survival are called local disease states because treatment resulted in a survival outcome comparable to the normal population. Those remaining states that lie below the normal population curve have survival deficits and are thus called metastatic disease states because treatment resulted in survival prognosis that was poorer than the normal population. Passage from local to metastatic states and between metastatic states can be determined from the four rules given above. Transition from local to metastatic states indicates the value of early detection. Passage between metastatic disease states also indicates this as well as the value of early treatment intervention. The change in deficit survival and the time to pass between consecutive states gives the value of early detection and early treatment intervention. For example, to pass between consecutive states with a 10 year increase in deficit survival requiring 5 years to reach the less favorable state, each year of early detection would correspond to a 2 year gain in expected survival. However, no change in deficit survival under the same conditions would suggest that early detection and treatment did not influence survival outcome.
The subgroups survival analysis.
RESULTS were examined with respect to age distributions prior to the deficit The most favorable subgroup (esa) had the largest number of patients
PHILII’
484
T.
LAVlV
6’1
9
Subgroups
esa(younger,older);es,ea,so,e,s,o,u Epltheliold cells / HPF Largest tumor dlmenslon Anterior margln
e s
a
c 0
\
12 c es
\
-1,
& E
8
50
x
‘0 E
\
l
l\
‘\
‘\\ \
4
\ \
\\
45
Age
at
diagnosis
( yr )
Medlon FIG. 2. Display
of bimodal age distribution PILI subgroup.
I
I
50
55
age
at
e
5 l\
‘\
‘\.
0
;; w
0
Posterior
‘.
‘\
‘..
5 2.0 5 10mm
‘1 ,;;,
‘\
60
‘.
u
65
dlagnosls
70
( yr )
far the FIG. 3. Deficit
survival
analysis
(MEEI
Series),
(78). An examination of the age distribution revealed a bimodal distribution as displayed in Fig. 2. There were relatively fewer patients between age 50 and 55 than in other age groups. Consequently, in the analysis that follows, the esu subgroup was divided into younger (37 patients) and older (41 patients) subgroups according to the median age of the entire population. Thus, in the analysis that follows, nine subgroups were evaluated. Table 1 displays the number of patients, median survival after enucleation. median age at enucleation, expected survival, and deficit survival for the nine subgroups. Figure 3 presents a deficit survival display of age at enucleation against survival following enucleation for the nine subgroups. The interpretation of this figure provides evidence for: (a) (b) (c) (d)
local and pathways diagnosis timing of
metastatic disease phases. of disease progression, delay consequences, and enucleation.
Local and metastatic
disease states
The nine subgroups cluster into four categories on the basis of the magnitude of the deficit survival. Only the older esa subgroup had remaining survival comparable to the normal population, suggesting that this most favorable subgroup was the only local disease state. The younger esa. ea, and e subgroups each had 9 year survival deficits. This worsened to 14 years for the es, s, and U subgroups and to 18 years for the sa and a subgroups. Thus, all but the older esa subgroup corresponded to metastatic disease states.
A Deficit
Pathways
Survival
Analysis
485
of disease progression
The inference of disease pathways derives from the connection to the deficit survival rules. The following pathways are suggested by these postulates: (1) younger (2) es+sjU (3) sa+a
esa+ea+e+
of subgroups
according
U
The first pathway represents a deterioration in the number of favorable prognostic factors, characterized first by increase in tumor size, then by possible advancement from a location posterior to the equator to an anterior location, and finally marked by an increase in the number of epithelioid cells per high power field to reach the most unfavorable state U. Six to eight years were required to pass between successive states. The passage is indicative of relatively slow tumor growth prior to change in the number of epithelioid cells per high power field. In contrast, the second pathway suggests that small tumors with unfavorable locations can undergo transition in the number of epithelioid cells per high power field prior to demonstration of active tumor growth. The third pathway represents a change in size for tumors located anterior the equator. All three pathways illustrate constant deficit survival, which are essentially parallel to the normal population survival curve. These constant deficit survival paths are indicated by dashed lines in Fig. 3. Diagnosis
delay consequences
The passage times between states illustrate constant deficit survival along three pathways. Thus, earlier detection of disease along a particular pathway did not result in decreased deficit survival, except for the passage from e+ U. There, a 5 year delay in diagnosis (median age 62-67) resulted in a 4 year increase in deficit survival (a loss of 9 years to 13 years compared to the normal population) for patients in the e subgroup. This is in contrast to the remaining pathways where earlier diagnosis did not alter deficit survival along pathways. For these pathways, a l-year delay in diagnosis does not change the deficit survival. It should be noted, however, if the tumor develops additional unfavorable factors with the passage of time, then the remaining years of survival decrease, as one might expect to occur. Timing of enucleation The three constant deficit survival curves suggest that for these pathways, early enucleation did not alter the deficit survival. For all subgroups except the e and older esa subgroups, a year of delay in diagnosis did not increase the deficit survival. For the e subgroup, early enucleation might be regarded as a favorable intervention as a delay leads to increased deficit survival. Thus, earlier enucleation neither increased nor decreased the deficit survival except for the e subgroup. These data suggest that enucleation did not have major impact on the natural history of uveal melanoma. The older esa subgroup had comparable survival to the normal population.
DISCUSSION
The debate over enucleation as primary treatment for uveal melanoma has lasted for over 150 years. The opponents of early enucleation represent points of view that surgery does not benefit the patient in terms of prolonged survival and may actually hasten the spread of the disease, reducing the survival. The data presented here suggest that early enucleation sustains normal population survival only for patients with tumors with largest dimension above 10 mm and an anterior margin in front of the equator which have few epithelioid cells. For other more favorable subgroups, enucleation may be delayed to allow for other therapeutic interventions as primary therapy. This finding is similar to that
486
PHILIP
T. LAVIN et al.
reported by Zelen for radical mastectomy and breast cancer where prompt removal of the tumor did not appear to alter survival. The most definitive testing of the benefit of delayed enucleation will require a randomized clinical trial of prompt vs delayed enucleation. However, this study will be very difficult to complete because of ethical considerations. Even if ethical issues could be addressed, completion would require many years due to the low incidence rate (about 1800 new cases per year in the U.S.) and the long survivals associated with uveal melanoma. Consequently, indirect methods of evaluation, provided by the deficit survival method, may provide the only evidence of the value of enucleation as primary treatment for uveal melanoma. For uveal melanoma, this method has certain limitations in making clinical decisions. First, classification into states corresponding to this analysis is not possible preoperatively because the number of epithelioid cells per high power field cannot be determined without performing a pathology assessment. Second, according to these findings, older patients with favorable size and location, who happen to have few epithelioid cells per high power field have been shown to have no deficit survival. This may be due to disease of a more benign natural history, perhaps suggesting that other therapeutic interventions may result in similar outcome. However, the younger patient with either favorable size or location is a candidate for alternative therapeutic interventions as evidenced by the deficit survivals in Table 1, which indicate shortened survivals following enucleation compared to the normal population. It may well be that ophthalmologists can consider alternative forms of therapy in designated subgroups where prompt enucleation might not yield favorable results and the possibility of more favorable outcomes might exist. The data demonstrating the benefit of enucleation is far from definitive, and further studies are required to see if better therapies do indeed exist. In conclusion, these findings explain many of the reasons for the controversies associated with uveal melanoma treatment. All patients with this disease do not uniformly benefit from either prompt diagnosis or enucleation. This differs from many other malignancies where early detection and treatment are associated with prolonged survival. These results should be confirmed by similar analyses of other uveal melanoma populations. Other studies involving long term survival may benefit from the deficit survival method. Here, the deficit survival method was applied in this retrospective series to gain insights not offered by other forms of statistical analysis such as the proportional hazards model or competing risk model. These alternative procedures do not compare survival to the hazards analysis was normal population. In this retrospective series, the proportional performed first to identify the prognostic factors used in this analysis. Thus, the deficit survival method should be regarded as an adjunct procedure to be utilized in the analysis of chronic diseases.
REFERENCES I.
2. 3. 4. 5. 6. 7. 8. 9.
Travers B: On the local diseases termed malignant. Cited by H. Knapp in A Treatise on Intraocular Tumors from Original Clinical Observations and Anatomical Investigations. (Translated by Cole _I.) New York: William Wood. 1839 Knano H: A Treatise on Intraocular Tumors from Original Clinical Observations and Anatomical Inve&igations. (Translated by Cole J.) New York: William Wood, 1869 Von Graefe A: Zusatze ueber intraoculare Tumoren. Arch Oubtbalmol 14: 103, 1868 Fuchs E: Das Sarcom des Uvealtractus. Wien: Wilhelm Braumuller, 1882 Zimmerman LE, McLean IW: An Evaluation of enucleation in the management of weal melanoma. Am J Oubtbalmol 87: 741. 1979 Zimmerman LE, McLean IW, Foster WD: Does enucleation of eye containing a malignant melanoma prevent or accelerate the dissemination of tumor cells? Br J Ouhthalmol 62: 420, 1978 ‘Seigel D. Myers M, Ferris F. III. Sternhorn SC: Survival rates after enucleation of eyes with malignant melanoma. Am J Ophthalmol 87: 761, 1979 Manshot WA, van Pepperreel HA: Choroidal melanoma: enucleation or observation? A new approach. Arch Ophthalmol 89: 7 1, 1980 Zelen M: A hypothesis for the natural time history of breast cancer. Cancer Res 28: 207, 1968
A Deficit IO.
1I. 12. 13.
Survival
Analysis
487
Shammas HF, Blodi FC: Prognostic factors in choroidal and ciliary body melanomas. Arch Ophthalmol 95: 63, 1977 Seddon JM, Albert DM, Lavin PT, Robinson N: A prognostic factor study of disease-free interval and survival following enucleation for uveal melanoma. Arch Ophthalmol 101: 1894-l 899, 1983 Vital Statistics of United States, Mortality Statistics Vol. II. Part A, 1975 McLean IW, Foster. WD, Zimmerman LE: Uveal melanoma: location, size, cell type. and enucleation as risk factors in metastasis. Hum Pathol 13: 123-132, 1982