Sunteți pe pagina 1din 5

Vitrectomy for the Diagnosis and

Management of Uveitis of Unknown Cause

Ron Margolis, MD, Oswaldo F. M. Brasil, MD, Careen Y. Lowder, MD, PhD, Rishi P. Singh, MD,
Peter K. Kaiser, MD, Scott D. Smith, MD, MPH, Victor L. Perez, MD, Christine Sonnie, RN,
Jonathan E. Sears, MD
Purpose: To determine the diagnostic yield of tests commonly used for vitreous fluid analysis in eyes with
suspected intraocular infection or malignancy.
Design: Noncomparative interventional case series.
Participants: Forty-four consecutive patients (45 eyes) treated from 1998 through 2006 with posterior
segment inflammation who underwent pars plana vitrectomy for diagnostic purposes.
Intervention: Vitreous specimens obtained via pars plana vitrectomy were analyzed by microbiologic
culture, cytologic analysis, and flow cytometry.
Main Outcome Measures: Diagnostic yield and sensitivity of each test performed on vitreous specimens
and visual outcomes of eyes that underwent diagnostic vitrectomy (DVx).
Results: Preoperative diagnoses were infection in 15 eyes and malignancy in 30 eyes. Overall, vitreous
analysis identified a specific cause in 9 (20%) of 45 eyes. The overall sensitivity of DVx was 63.6%. The
sensitivities of individual tests were: culture, 50%; cytologic analysis, 66.7%; and flow cytometry, 83.3%. The
yields of diagnostic tests were: culture, 5.7%; cytologic analysis, 14.3%; and flow cytometry, 20.6%. Final
diagnoses were infection in 6 eyes, malignancy in 9 eyes, and idiopathic in 30 eyes. Mean visual acuity improved
significantly in the first 6 months after DVx. Visual acuity improved in 60% of eyes, with 37.8% of eyes improving
by 3 lines or more.
Conclusions: Analysis of vitreous fluid by widely available tests is useful in identifying intraocular
infection or malignancy. Most patients experienced a substantial improvement in vision. Ophthalmology
2007;114:18931897 2007 by the American Academy of Ophthalmology.

Determining the cause of uveitis has important implications

for treatment, prognosis, and prediction of systemic disease.
Ophthalmologists rely on clinical findings combined with
systemic laboratory testing and response to treatment to
arrive at a diagnosis. Cases of intraocular inflammation that
fail to respond to immunosuppressive treatment raise suspicion for an indolent infection or a malignancy. In such
cases, a vitreous biopsy often is performed to obtain a
histopathologic diagnosis.
The previously reported yield of diagnostic vitrectomy
(DVx) varies from 14.3% to 61.5%.19 This large range is
the result of patient selection, surgical technique, and vitreous specimen analysis. Slow-growing organisms are difficult to culture, and malignant cells may be difficult to
identify because of the small sample size collected and low
Originally received: September 9, 2006.
Final revision: January 30, 2007.
Accepted: January 31, 2007.
Available online: May 16, 2007.
Manuscript no. 2006-1023.
From the Cole Eye Institute, Cleveland Clinic Foundation, Cleveland,
Presented at: American Academy of Ophthalmology Annual Meeting,
November 2006.
The authors have no conflicts of interest related to the study.
Correspondence to Jonathan E. Sears, MD, Cole Eye Institute, Cleveland
Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195.
2007 by the American Academy of Ophthalmology
Published by Elsevier Inc.

cell viability. Despite these challenges, vitreous biopsy remains the gold standard procedure used to establish a diagnosis and to direct treatment when infection or malignancy
is suspected.
The purpose of this study was to review the authors
experience with DVx in patients with presumed intraocular
infection or malignancy in which careful clinical and laboratory evaluation failed to identify a diagnosis. In particular,
the authors sought to determine the diagnostic yield of DVx
using tests that are widely available and commonly performed for vitreous fluid analysis. A secondary aim was to
study the visual outcomes of patients who underwent DVx.

Patients and Methods

After institutional review board approval, the authors performed a
retrospective analysis of all patients who underwent DVx for
suspected intraocular infection or malignancy. Patients were evaluated at the Cole Eye Institute from 1998 through 2006. Patients
were included if they met the following inclusion criteria: intermediate, posterior, or panuveitis; preoperative diagnosis of uveitis
of unknown cause in which infection or malignancy was suspected; complete systemic laboratory evaluation failed to reveal a
definitive diagnosis; and the indication for surgery was to establish
a diagnosis. Exclusion criteria included acute postoperative or
traumatic endophthalmitis.
Patient demographic data and clinical history were recorded at
ISSN 0161-6420/07/$see front matter


Ophthalmology Volume 114, Number 10, October 2007

the time of initial visit. A complete ophthalmic history and examination were performed by 1 of 2 uveitis specialists (CYL and
VLP). Appropriate ancillary and laboratory tests were performed
according to the differential diagnosis. In addition, the following
information was obtained from the clinical chart: uveitis type,
laterality, suspected preoperative diagnosis, final diagnosis,
follow-up period, preoperative and postoperative best-corrected
Snellen visual acuity, and complications related to DVx.
The DVx was performed by 2 vitreoretinal surgeons (JES and
PKK) using either 20-, 23-, or 25-gauge vitrectomy systems.
Three-port pars plana vitrectomy was performed using a standard
approach and direct visualization of the vitrectomy instruments.
The vitreous specimens were collected in the operating room
through undiluted lines using the vitreous cutter. Undiluted vitreous samples were obtained using a 3-ml syringe connected directly
to the handpiece until the eye was noted to soften visibly. At least
2 ml undiluted vitreous was collected in each case. At that time,
the infusion was started, and diluted vitreous specimen was aspirated into a second syringe. Samples were maintained in the
collection syringe and were delivered to the clinical laboratory
immediately after surgery. Although some eyes underwent multiple diagnostic surgeries, data were collected from the first DVx
Analysis of vitreous specimens by microbiologic culture and
sensitivity, cytologic analysis, and flow cytometry was performed
at the Cleveland Clinic facilities. Tests were ordered according to
the preoperative suspected diagnosis, so that not all tests were
performed on all specimens. Microbiologic analysis was performed in cases with suspected infectious cause by histologic
stains and appropriate culture inoculation. Specific instructions
were given to the microbiology laboratory when fungal or slowgrowing organisms were suspected.
Cytopathologic analysis was used to identify tumor cells in
cases of suspected malignancy. Slides were prepared from undiluted vitreous, stained with Papanicolaous stain, and reviewed by
an experienced cytopathologist. Flow cytometry also was performed to rule out malignancy. Cells were counted and stained
with antibodies to detect cellular surface markers that identify
leukocytes; T and B lymphocytes, including and light-chain
markers; monocytes and macrophages; and lymphocyte activation
according to the usual protocol for a leukemia and lymphoma
panel. Results of flow cytometry were interpreted by an experienced pathologist. Findings of polyclonal B-cell population were
considered to be a nonspecific reactive or inflammatory process.
Final diagnoses were assigned to each case based on all available
information that included clinical course and response to treatment. Cases were categorized as idiopathic uveitis only if no cause
was found and the inflammation subsided with immunosuppression.
The primary outcome measure was diagnostic yield of DVx
overall, and of each of the specific tests used to analyze specimens.
The sensitivity, specificity, and positive and negative predictive

values were calculated for culture, cytologic analysis, and flow

cytometry. Subgroup analysis was performed to identify patient
characteristics in which DVx had a high diagnostic yield. Characteristics tested included patient age (65 years or 65 years and
older), uveitis types (intermediate, posterior, or panuveitis), laterality, preoperative and postoperative diagnoses (infection, malignancy, or idiopathic), and interval between presentation and DVx
(20 days, 20 100 days, or 100 days). Univariate statistical
analysis of continuous and categoric variables was performed
using the t test and the Fisher exact test, respectively. A P value
0.05 was considered statistically significant.
A secondary outcome measure was change in visual acuity
after DVx. No minimal length of follow-up was required. For
statistical analysis, the Snellen visual acuities were converted into
logarithm of the minimum angle of resolution values. Analysis
comparing mean baseline logarithm of the minimum angle of
resolution visual acuity with mean postoperative logarithm of the
minimum angle of resolution visual acuity at 3 months, 6 months,
1 year, and 2 years after surgery was performed using the paired
t test. Subgroup analysis was performed to compare visual outcomes according to final diagnosis using a 1-way analysis of
variance (Bonferroni) and Bartletts test for equal variances.

Forty-five eyes of 44 patients were included in this study. Table 1
summarizes the baseline demographic data according to the surgical indication. Suspected preoperative diagnosis was infection in
15 eyes and malignancy in 30 eyes. Twenty patients were men
(45.4%), and the average age (standard deviation [SD]) was
6513 years (range, 38 87 years). Mean time (SD) from presentation to DVx was 4991098 days (range, 15445 days), and
mean follow-up (SD) was 340369 days (range, 8 1636 days).
Eyes with suspected chronic postoperative endophthalmitis all
underwent cataract extraction a mean (SD) of 13751829 days
(range, 425236 days) before DVx. Intraocular inflammation was
unilateral in 33 cases (73%). Bilateral DVx was performed in 1
patient. The mean number of tests performed on each vitreous
sample was 2.6. Twenty-eight (62.2%) vitreous specimens were
analyzed by culture, cytologic analysis, and flow cytometry. Sixteen (35.6%) specimens were analyzed by 2 of the 3 tests, and 1
specimen was analyzed by culture only (2.2%).

Vitreous Biopsy Analysis

A total of 111 diagnostic tests were performed on the 45 vitreous
specimens, of which 15 demonstrated positive results (13.5%;
Table 2). Of the 45 vitreous specimens analyzed, 9 (20%) specimens demonstrated positive results for infection or malignancy
(Table 3), and 36 (80%) eyes demonstrated negative results for
vitreous analysis (Table 4). Infection was the final clinical diag-

Table 1. Characteristics, Final Diagnosis, and Diagnostic Outcomes of Patients Who Underwent Diagnostic Vitrectomy for
Suspected Intraocular Infection or Malignancy

Indications for


No. of
(% of

No. of

Mean Age

Mean Time to
(Range), Days



No. of


No. Diagnosed
by Diagnostic

15 (33)

68 (4885)

275 (11710)

349 (81636)



30 (67)


64 (3887)

594 (105445)

332 (81178)



45 (100)


65 (3887)

499 (15445)

340 (81636)




Margolis et al Vitrectomy for Uveitis of Unknown Cause

Table 2. Results of Individual Tests used in Vitreous Analysis
Obtained by Diagnostic Vitrectomy

Table 4. Final Diagnosis in Cases with Negative Vitreous

Analysis Results from Diagnostic Vitrectomy


No. Performed

No. Positive Test Results (%)


No. (% of All Cases)

Cytologic analysis
Flow cytometry


2 (5.7)
6 (14.3)
7 (20.6)
15 (13.5)

Idiopathic panuveitis
Propionibacterium acnes endophthalmitis
Fungal endophthalmitis
Acute retinal necrosis
Metastatic testicular lymphoma
Metastatic lung carcinoma

30 (66.7)
1 (2.2)
1 (2.2)
1 (2.2)
1 (2.2)
1 (2.2)
1 (2.2)
36 (80)

nosis in 6 eyes, of which 2 (40%) demonstrated positive culture

results. Infection in the other 4 eyes was later diagnosed in the
following ways: Propionibacterium acnes endophthalmitis was
diagnosed in 1 eye by a second vitrectomy with total capsulectomy, and fungal endophthalmitis, toxoplasmosis, and acute retinal
necrosis later were diagnosed in 1 eye each by the clinical appearance and response to treatment after DVx. Because toxoplasmosis
titers and herpes simplex virus analysis were not performed on
these vitreous biopsies, vitreous testing from the 2 eyes with these
diagnoses was considered truly negative.
Nine eyes had a final clinical diagnosis of malignancy, and 7 of
these (77.8%) were confirmed by cytologic analysis, flow cytometry, or both. One patient with negative vitreous analysis results
later was diagnosed with metastatic lung carcinoma by biopsy of
an iris nodule. A second patient with negative vitreous biopsy
results had a history of testicular lymphoma with brain metastases,
vitreitis, multifocal yellow subretinal lesions, and optic disc
edema. Despite negative vitreous biopsy results, this eye was
categorized as having ocular involvement of systemic lymphoma
based on the characteristic ocular findings.
From the known final diagnoses, statistical analysis was performed of the tests used to evaluate vitreous specimens. The
overall sensitivity of DVx was 63.6%, and its negative predictive
value was 89.5%. Sensitivities and negative predictive values of
individual tests are shown in Table 5. There were no false-positive
results for all tests used, so that all tests were 100% specific, and
the positive predictive value for each test was 100%. These values
did not reach statistical significance. Subgroup analysis could not
identify any patient characteristics that were associated with an
increased diagnostic yield for any test. DVx was complicated by
entry wound-related retinal detachment in 1 patient (2.2%). No
other complications occurred.

Visual Outcomes
Visual acuity outcomes according to final diagnosis are summarized in Table 6. The overall mean preoperative vision was 20/138
for all eyes (range, 20/25light perception). There was no significant difference between the mean preoperative visual acuity of
patients with final diagnosis of infection, malignancy, or idiopathic
uveitis. At 3 months (P0.001) and 6 months (P 0.02) after
Table 3. Final Diagnoses Confirmed by Diagnostic Vitrectomy
Propionibacterium acnes
Aspergillus species
Primary intraocular lymphoma
Systemic NHL

No. (% of
All Cases)

% of Positive Diagnostic
Vitrectomy Results

1 (2.2)


1 (2.2)
6 (13.3)
1 (2.2)


NHL non-Hodgkins lymphoma.

surgery, a statistically significant improvement in visual acuity was

observed. Although mean visual acuities at 1 year and 2 years after
DVx were better than mean preoperative vision, this difference
was not significant.
Subgroup analysis suggested that eyes in the idiopathic group
experienced better visual outcomes throughout the follow-up period compared with eyes with infection or malignancy. However,
this difference was not significant. No significant differences in
visual outcomes were found based on patient age, laterality of
uveitis, uveitis type, and interval between the onset of uveitis and
Overall, 27 (60%) eyes experienced an improvement in vision
after DVx, and 17 eyes (37.8%) improved by 3 Snellen lines or
more at the time of last follow-up. Of eyes that improved by at
least 3 lines of vision, 14 (82.4%) had a final diagnosis of idiopathic uveitis. Vision was unchanged by DVx in 7 (15.6%) eyes,
whereas 11 (24.4%) eyes experienced a worsening in vision.
Causes for loss of vision included cataract in 1 eye, macular edema
in 5 eyes, occlusive vasculitis in 1 eye with idiopathic uveitis,
macular chorioretinal scar in 1 eye with fungal endophthalmitis,
anterior ischemic optic neuropathy in 1 eye, phthisis in an eye with
metastatic testicular lymphoma, and cancer-associated retinopathy in
the setting of systemic non-Hodgkins lymphoma.

In this study, DVx with vitreous analysis established a
diagnosis in 20% of eyes that were suspected of having
posterior segment inflammation resulting from either infectious or malignant causes. In previous studies, DVx was
successful in establishing a final diagnosis in 14% to 61.5%
of cases.19 Diagnostic yield of vitrectomy depends on the
initial patient selection, the number and types of tests used,
and vitrectomy technique. Because DVx may be performed
after treatment with antibiotics or antiinflammatory medication has been attempted, microbial load and malignant
cell counts may be suppressed, yielding negative results.
Additional factors that can affect the diagnostic outcome are
a lag between collecting and processing the specimen and
the experience of the microbiologist and cytopathologist in
analyzing the vitreous. Direct comparison of various studies
therefore is difficult.
Although our diagnostic yield of microbiologic culture
and sensitivity was 5.7%, there were only 2 known falsenegative results, and the sensitivity was 50%. However, it
is possible that with longer follow-up, some cases considered to be idiopathic inflammation might have been


Ophthalmology Volume 114, Number 10, October 2007

Table 5. Statistical Analysis of Laboratory Tests Performed on Vitreous Specimens Obtained by
Diagnostic Vitrectomy

Sensitivity (%)

Specificity (%)

Positive Predictive
Value (%)

Negative Predictive
Value (%)

Cytologic analysis
Flow cytometry





infectious. There are several possible factors that could

have contributed to falsely negative culture results. Cultures of slow-growing organisms have an inherently low
yield because of the indolent nature of the species. Also,
the eye with P. acnes endophthalmitis and negative culture results was treated with antibiotics before the initial
vitrectomy, which may have decreased the microbial load
so that an insufficient or nonviable microbial sample was
obtained. Although some studies achieved a diagnostic
yield approaching 100% from culture,7,8 it is difficult to
compare these directly with the present results because of
the small number of patients and different organisms
Cytologic analysis had a yield of 14.3% compared with
the 10% to 70% diagnostic yield of cytologic analysis
reported in previous studies. The sensitivity of cytologic
analysis was 66.7%, which is at the higher end of the range
of previous studies (31% 64%).8,9 All 6 cases of primary
intraocular lymphoma were confirmed by cytologic analysis, and 1 of 3 intraocular metastases was detected. In the
authors experience, the most important variable for an
accurate cytologic evaluation is rapid processing of the
specimen. Although not used at their institution, previous
authors have advocated additional means of increasing cell
viability and diagnostic yield of cytologic analysis, including discontinuing systemic corticosteroids before biopsy,10
obtaining cells from the most dense portion of the infiltrate
and the subretinal pigment epithelium space,9 and using
special culture media rather than saline before specimen
Flow cytometry had a diagnostic yield of 14.7% (5 of
34 eyes) and the highest sensitivity of all tests (83.3%),
which is consistent with findings of Davis et al.8,11 Zaldivar et al9 diagnosed primary intraocular lymphoma in 6
of 10 eyes by flow cytometry of the first vitreous biopsy.
Two eyes required repeat vitrectomy, and 1 patient had

large-cell lymphoma with positive cytologic and negative

flow cytometry results. Although not included in the
present analysis, malignancy was diagnosed by a second
DVx in 1 patient. Multiple vitreous biopsies seem to
increase the diagnostic yield, although no study has been
carried out to quantify the added benefit.
Flow cytometry confirmed 1 case of primary intraocular
lymphoma in which the cytologic results were negative.
Conversely, cytologic analysis confirmed a case of primary
intraocular lymphoma that had negative flow cytometric
results. When both cytologic and flow cytometry examinations were used in vitreous analysis, all cases of malignancy
were detected. The 2 cases in which intraocular metastases
were missed were evaluated by cytologic analysis alone
without flow cytometry. Combined analysis by cytologic
examination and flow cytometry seems to be more sensitive
than using either test alone in diagnosing malignancy. However, both techniques require a sufficient number of cells
and an experienced cytopathologist.
The authors routinely sent vitreous fluid for analysis
by culture, cytologic analysis, and flow cytometry because these tests and the specialists who interpret them
are available at their institution. Recent studies reported
the use of innovative techniques to increase the yield of
DVx. Specifically, polymerase chain reaction has been
used for the analysis of gene rearrangement in the diagnosis of lymphoma, as well as in the detection of bacterial and viral DNA.6 8 Determination of antibodies in
vitreous fluid has been reported for the diagnosis of
infections.3,7,8 Measurement of intraocular cytokines has
been studied in the diagnosis of malignancy.1215 Although some of the current specimens were analyzed by
polymerase chain reaction or interleukin levels, the authors did not include these data because of the small
number of test results available. Evolving methods for
the laboratory diagnosis of infection and malignancy may

Table 6. Visual Outcomes of Patients Who Underwent Diagnostic Vitrectomy

Final Diagnosis

Vision (Snellen),
n 45

Mean 3-Month
Vision (Snellen),
n 45

Mean 6-Month
Vision (Snellen),
n 35

Mean 1-Year
Vision (Snellen),
n 24

Mean 2-Year
Vision (Snellen),
n 19


20/55 (P0.01)*

20/65 (P 0.02)

20/90 (P 0.56)

20/100 (P 0.61)

*P values comparing visual outcomes at postoperative follow-up time points relative to preoperative vision.


Margolis et al Vitrectomy for Uveitis of Unknown Cause

represent significant improvement over conventional testing and may increase diagnostic yield in the future.
Diagnostic vitrectomy resulted in a significant improvement in visual acuity within the first 6 months after surgery.
Although the improvement in vision at the 1-year and
2-year follow-up periods was not significant, this may represent a selection bias because patients with poor vision
were more likely to follow-up than those with good visual
outcomes. Overall, 60% of eyes experienced an improvement in vision after vitrectomy, and more than one third
improved by 3 lines or more at the last follow-up. Vision
remained unchanged in 15.6% of eyes, whereas 24.4%
experienced a worsening in vision. Causes for decreased
vision were not related to DVx, but rather to the underlying
cause of the uveitis. Although difficult to compare, these
results are consistent with those reported in a review by
Becker and Davis16 of 1575 patients who underwent therapeutic vitrectomy. The authors found that visual acuity
improved in 68%, remained unchanged in 20%, and worsened in 12% of patients.16 The current data suggest that
patients with idiopathic uveitis experienced the greatest
visual benefit from vitrectomy as compared with patients
with intraocular infection or malignancy. A likely explanation is that in eyes with idiopathic uveitis, vitrectomy
cleared the inflammatory debris, and the inflammation improved with immunomodulating therapy. Intraocular infection or malignancy, however, resulted in severe complications and visual loss. The ability to predict visual outcome
based on the results of DVx is of potential value to both
surgeon and patient.
This study had a number of limitations. First, patients
were followed up for a variable period. It is possible that
final diagnoses, particularly of cases considered idiopathic,
as well as visual outcomes would have changed given
adequate follow-up. Furthermore, sensitivities, specificities,
and predictive values are based on the assumption that
culture, cytologic analysis, and flow cytometry are gold
standard tests, which may not be correct. In this setting, the
specificity of these tests is by definition 100% because other
ancillary tests were not used. Finally, because of the small
number of positive biopsy results, these results should be
interpreted with caution.
This study demonstrates that DVx with analysis of vitreous fluid by widely available tests is useful in diagnosing
infection and malignancy in cases of uveitis of unknown
cause. In addition, DVx resulted in improved visual acuity
in most patients. As surgical techniques and diagnostic
methods improve, the benefit from DVx in these challenging cases likely will increase.

1. Priem H, Verbraeken H, de Laey JJ. Diagnostic problems in
chronic vitreous inflammation. Graefes Arch Clin Exp Ophthalmol 1993;231:453 6.
2. Palexas GN, Green WR, Goldberg MF, Ding Y. Diagnostic
pars plana vitrectomy: report of a 21-year retrospective study.
Trans Am Ophthalmol Soc 1995;93:281308.
3. Verbraeken H. Diagnostic vitrectomy and chronic uveitis.
Graefes Arch Clin Exp Ophthalmol 1996;234(suppl):
4. Liu K, Klintworth GK, Dodd LG. Cytologic findings in vitreous fluids: analysis of 74 specimens. Acta Cytol 1999;43:
201 6.
5. Akpek EK, Ahmed I, Hochberg FH, et al. Intraocular-central
nervous system lymphoma: clinical features, diagnosis, and
outcomes. Ophthalmology 1999;106:180510.
6. Coupland SE, Bechrakis NE, Anastassiou G, et al. Evaluation
of vitrectomy specimens and chorioretinal biopsies in the
diagnosis of primary intraocular lymphoma in patients with
Masquerade syndrome. Graefes Arch Clin Exp Ophthalmol
2003;241:860 70.
7. Mruthyunjaya P, Jumper JM, McCallum R, et al. Diagnostic yield of vitrectomy in eyes with suspected posterior
segment infection or malignancy. Ophthalmology 2002;
8. Davis JL, Miller DM, Ruiz P. Diagnostic testing of vitrectomy
specimens. Am J Ophthalmol 2005;140:8229.
9. Zaldivar RA, Martin DF, Holden JT, Grossniklaus HE. Primary intraocular lymphoma. Clinical, cytologic, and flow cytometric analysis. Ophthalmology 2004;111:17627.
10. Whitcup SM, Chan CC, Buggage RR, et al. Improving the
diagnostic yield of vitrectomy for intraocular lymphoma
[letter]. Arch Ophthalmol 2000;118:446.
11. Davis JL, Viciana AL, Ruiz P. Diagnosis of intraocular
lymphoma by flow cytometry. Am J Ophthalmol 1997;124:
12. Merle-Beral H, Davis F, Cassoux N, et al. Biological diagnosis of primary intraocular lymphoma. Br J Haematol 2004;
124:469 73.
13. Chan CC. Molecular pathology of primary intraocular lymphoma. Trans Am Ophthalmol Soc 2003;101:27592.
14. Akpek EK, Maca SM, Christen WG, Foster CS. Elevated
vitreous interleukin-10 level is not diagnostic of intraocularcentral nervous system lymphoma. Ophthalmology 1999;106:
15. Whitcup SM, Stark-Vancs V, Wittes RE, et al. Association of
interleukin 10 in the vitreous and cerebrospinal fluid and
primary central nervous system lymphoma. Arch Ophthalmol
1997;115:1157 60.
16. Becker M, Davis J. Vitrectomy in the treatment of uveitis.
Am J Ophthalmol 2005;140:1096 105.