- Email: [email protected]
Analysis of Visual Dysfunctions in HIV-positive Patients Without Retinitis
Analysis of Visual Dysfunctions in HIV-positive Patients Without Retinitis A. J. MUELLER, MD, D. J. PLUMMER, PHD, R. DUA, BS, I. TASKINTUNA, MD, P. A. SAMPLE, PHD, I. GRANT, MD, AND W. R. FREEMAN, MD
• PURPOSE: To investigate visual dysfunctions in ophthalmoscopically normal human immunodefi ciency virus (HlV)-positive patients and to corre late the results to the stage of HIV disease and neuropsychological status. • METHODS: Fifty-one randomly selected eyes (26 right, 25 left) of 51 HIV-positive patients with visual acuity measurements of 20/20 or better and no ophthalmoscopically detectable disorders were prospectively examined using achromatic and short-wavelength automated perimetry, color vi sion testing, and contrast sensitivity testing. CD4 + T-lymphocyte count, presence of systemic infec tion, hemoglobin, hematocrit, serum |$2-microglobulin levels, and results of neuropsychological testing were also analyzed. • RESULTS: On achromatic automated perimetry, 21.6% (11/51) of patients performed abnormally according to the mean defect and 27.5% (14/51) according to the Glaucoma Hemifield Test; 29.4% (15/51) performed abnormally on short-wave length automated perimetry according to the mean
Accepted for publication April 14, 1997. From the Department of Ophthalmology, Shiley Eye Center (Drs Mueller, Plummer, Taskintuna, Sample, and Freeman, and Ms Dua), and the Department of Psychiatry (Dr Grant), University of California San Diego, La Jolla, California. Supported by Deutsche Forschungsgemeinschaft grant Mu 1204/1-2 (Dr Mueller); University of California Universitywide AIDS Research Programs R95-SD-082 (Dr Plummer) and R92-SD-135 (Dr Sample); National Eye Institute grant EYO 7366 (Dr Freeman); and a departmental grant from Research to Prevent Blindness, Inc, New York, New York. Reprint requests to W. R. Freeman, MD, Department of Ophthalmolo gy, Shiley Eye Center, University of California, San Diego, 9415 Campus Point Dr 0946, La Jolla, CA 92093-0946; fax: (619) 534-7985.
158
defect and 23.5% (12/51) according to the Glauco ma Hemifield Test. On contrast sensitivity, 5.9% (3/51) of patients performed abnormally in the 1.5-cycles per degree (cpd) line, 2.0% (2/51) in the 3-cpd line, 23.5% (12/51) in the 6-cpd line, 25.5% (13/51) in the 12-cpd line, and 33.3% (17/51) in the 18-cpd line. On the Farnsworth-Munsell 100hue test, 29.4% (15/51) of patients performed abnormally. After correction for multiple correla tions, two statistically significant correlations were found: sum of log contrast sensitivity with achro matic automated perimetry and sum of log contrast sensitivity with the Farnsworth-Munsell 100-hue test. • CONCLUSIONS: A significant percentage of HIV-positive patients with visual acuity of 20/20 or better and no ophthalmologic evidence of retini tis performed abnormally on visual psychophysical tests. The severity of visual dysfunction was not correlated with the stage of HIV infection or the degree of neuropsychological dysfunction.
P
ATIENTS WITH HUMAN IMMUNODEFICIENCY Vi rus (HIV) are susceptible to several different types of infectious retinopathies1 that may all affect different aspects of visual function. However, various changes in visual function have also been reported in the absence of infectious and noninfectious retinopathies in HIV-positive patients.2'6 Our group has previously reported7 significant losses in visual function in both standard achromatic automated perimetry and short-wavelength automat ed perimetry. In addition to changes measured by
AMERICAN JOURNAL OF OPHTHALMOLOGY 1997;124:158-167
AUGUST 1997
perimetry, we have also demonstrated8 loss in visual contrast sensitivity and changes in central color vision. However, the pattern of visual function loss among a variety of different psychophysical test meth ods has not been explored in a single patient sample. In addition, it is unclear whether decreasing visual function is related to worsening HIV-related illness or neuropsychological status. This prospective study was See also pp. 141-157, 168-180, 181-189, 190-198, 199-205, 227-233, and 234-239.
designed to investigate the pattern of HIV-related visual dysfunctions using four validated visual psychophysical tests and to correlate the results to the stage of HIV infection. We also correlated these findings with a battery of tests performed to measure the severity of HIV-related disease and neuropsychologi cal performance. Each of these tests was performed in each patient within a defined time interval.
METHODS THE PARTICIPANTS FOR THIS STUDY WERE ALL PROSPEC-
tively drawn from a longitudinally followed cohort at the University of California, San Diego, HIV Neurobehavioral Research Center and the AIDS Ocular Research Unit. The inclusion and exclusion criteria for the HIV Neurobehavioral Research Center proto col have been described extensively elsewhere.9 This study was approved by the Human Subjects Commit tee of the University of California, San Diego. All subjects gave informed consent. All patients underwent complete ophthalmologic examination including visual acuity, intraocular pres sure, slit-lamp, and dilated fundus examination by an ophthalmologist (A.J.M., I.T., W.R.F.) who was familiar with the ocular manifestations of AIDS. Inclusion criteria were 20/20 or better normal or corrected-to-normal Snellen visual acuity in each eye using the EDTRS chart,10 normal intraocular pressure (11 to 21 mm Hg by applanation), and normal appearance of the anterior and posterior segment of the eye. With these inclusion criteria, it is not likely that any of the observed visual dysfunctions are related to cataract, glaucoma, or other apparent eye VOL.124, No. 2
diseases. Thus we assumed that detectable losses of visual function are solely related to HIV. Although ruling out persons with visual acuity of worse than 20/20 may bias the population slightly, fewer than 10% of people agreeing to participate in this study failed to meet this criterion. In addition, our popula tion is relatively young (age range, 22 to 52 years), which can explain the high percentage of people with visual acuity of 20/20 or better. Exclusion criteria were diabetes mellitus, arterial hypertension, or a family history of glaucoma or other degenerative retinal disease as well as of any other manifest ocular disease, especially cataract, glaucoma, cytomegalovirus retinitis, vitreal or retinal hemor rhages, or cotton-wool spots in either eye on the date of examination. We also excluded all patients who, within 3 months of testing, had received courses of any medication that might affect visual function. In particular, we checked for any pharmacologically active topical ocular medication (that is, pilocarpine or derivative), chloramphenicol, chloroquine or its derivatives, digoxin or derivatives, ethambutol, gentamicin or other aminoglycosides, suramin sodium, iron in any preparation, oral contraceptives, quinine, corticosteroids, streptomycin, tetracycline or its deriva tives, thioridazine, and vitamins A or D. One eye was selected randomly for achromatic automated perimetry (Humphrey program 24/2), short-wavelength automated perimetry (program 24/2), color vision testing (Farnsworth-Munsell 100hue test), and contrast sensitivity testing (Vistech Consultants, Dayton, Ohio). To rule out fatiguing effects, the tests were performed in random order on the selected eye. We did not test both eyes because total duration of testing averaged approximately 1 hour per eye, and testing a second eye might invali date our results because of excessive fatigue. Achromatic automated perimetry was performed with the standard settings provided by the internal settings of the machine using program 24-2 and a size III (Goldmann) target. A chromatic version of the short-wavelength automated perimetry test has been developed on a modified Humphrey Field Analyzer and is currently used to study primary open-angle glaucoma and diabetic retinopathy.11'14 The test pro cedures were validated by our group and have been previously described.11'12 The test results obtained for both achromatic and short-wavelength automated
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
159
perimetry were corrected for age according to our internal normal database for these tests.11 The other tests were performed according to the instructions of the manufacturers. Age correction for contrast sensitivity was performed according to the manual provided with the test. Age correction for color vision testing was performed according to published data.15 Both tests (contrast sensitivity and Farnsworth-Munsell 100-hue test) have been previ ously validated in our research using HIV-negative control patients.8 All patients also underwent neuropsychological testing, which was performed using a battery of tests based on the extended Halstead-Reiton Battery that has been previously described in detail.16,17 Each group of tests was scored in a masked fashion, and a rating was obtained in each of the following areas: verbal skill, abstraction, psychomotor function, atten tion, learning, memory, and sensory skill. This bat tery assesses verbal language skills, attention and speed of information processing, abstracting ability, cognitive flexibility, complex perceptual motor skills, memory functions (learning, recall), and simple mo tor and simple sensory functions. The neuropsycho logical data from the approximately 50 tests can be summarized in several ways.9 For the purposes of the present study, we focused on the composite or global rating of impairment, which is based on masked ratings by a trained judge (I.G.) of all test data from eight cognitive domains. The global judgment is rendered on a 9-point scale, with 1 being excellent neuropsychological function and 9, very impaired neuropsychological function. A score of 5 indicates mild impairment. For this study, all patients scoring 5 or more were classified as neuropsychologically im paired. Previous research has indicated that such ratings are reliable, sensitive, and specific indicators for brain disease.16 All medical and neuropsychologi cal tests were performed within ±3 months of the time of the visual function testing. Statistical analysis was performed in two steps. In the first step, we calculated the percentage of eyes that performed abnormally separately for each test. For perimetry, we used the clinical standards for cutoffs on the various visual field indices, which are usually 95% for mean defect and 99.5% for the Glaucoma Hemifield Test. For the contrast sensitivity and
Farnsworth-Munsell 100-hue tests, we used the pub lished and validated data on cutoff values.8 Thus, for achromatic automated perimetry, eyes were assessed as abnormal when the mean defect was less than —2.01 dB.7 For short-wavelength automated perime try, eyes were assessed as abnormal when the mean defect was less than —5.5 dB.7,11 These achromatic automated perimetry mean defect and short-wave length automated perimetry mean defect cutoffs are the calculated and age-corrected 95% confidence interval (CI) of our own normal database.7,11'14 This control group consists of 91 HIV-negative normal subjects with an age distribution (22 to 55 years) comparable to that of the HIV-positive patients.7 The normal limits for our control group are even more stringent than those of the standard Humphrey control group, which consists of patients of all age ranges, including those in higher age groups.18 Al though the mean defect is a good indicator for a generalized defect in the visual field, it may not detect small or localized visual field defects. Thus, both achromatic and short-wavelength automated perime try were also analyzed with the Glaucoma Hemifield Test to obtain information about localized field de fects.7,18 Both the achromatic and the short-wave length automated perimetry Glaucoma Hemifield Test were assessed as abnormal when at least one of the analyzed five sector pairs was outside the 99.5% CI of the above-mentioned normal database.7 The 99.5% CI of both the achromatic and the short-wavelength automated perimetry Glaucoma Hemifield Test was chosen rather than the 95% CI because of a correc tion for multiple comparisons within the visual field. This is explained in detail by Asman and Heijl,19 who developed this technique. For contrast sensitivity, the test results were assessed as abnormal when fewer than patches 1 to 5 were observed correctly for each spatial frequency line.8 For the Farnsworth-Munsell 100-hue test, eyes were assessed as abnormal when the square root of the total error score was great er than 11.22 for patients aged 20 to 29 years, greater than 12.67 for patients aged 30 to 39 years, greater than 13.64 for patients aged 40 to 49 years, or greater than 14-71 for patients aged 50 to 59 years.15
160
OPHTHALMOLOGY
AMERICAN JOURNAL
In the second step of the statistical analysis, a multivariate regression analysis between various test
AUGUST 1997
results was performed both with and without correc tion for multiple correlations based on the Bonferroni inequality.20 For both achromatic automated perimetry and short-wavelength automated perimetry, the mean defect was calculated and entered. For contrast sensitivity, the sum of the log contrast sensitivity values was calculated and entered, and for the Farnsworth-Munsell 100-hue test, the square root of the total error score was calculated and entered.21,22 In addition to the results of the visual function tests, the following medical data were included in the multivariate regression analysis: square root of the CD4+ T-lymphocyte count, presence of any systemic oppor tunistic infection (elsewhere than in the eye), hemo globin, hematocrit, serum P2-microglobulin, and the result of the global neuropsychological score. All statistical analysis was performed using the statistical program JMP 3 (SAS Institute Inc, Cary, North Carolina). For multivariate regression analysis, we reversed the sign of the Farnsworth-Munsell 100-hue test and the neuropsychological testing from + to — because the Farnsworth-Munsell test and the neuropsychological testing have a grading system opposite that of the other tests (achromatic automat ed perimetry, short-wavelength automated perimetry, contrast sensitivity). On the Farnsworth-Munsell 100-hue test and neuropsychological testing, higher (more positive) values indicate poorer test perfor mance than lower (more negative) values do. In contrast, on the other tests (achromatic automated perimetry, short-wavelength automated perimetry, contrast sensitivity) higher (more positive) values indicate better performance than lower (more nega tive) values do. Inverting the values on the Farnsworth-Munsell 100-hue test and neuropsycho logical testing from + to — yields "positive" correla tions when poor performance on one test correlates with poor performance on another test. In contrast, all "negative" correlations between tests indicate that poor performance on one test is correlated with good performance on another test.
RESULTS FIFTY-ONE EYES (26 RIGHT, 25 LEFT) OF 51 HIV-POSITIVE
patients (50 [98%] men, one [2%] woman) were
VOL.124, No. 2
1
.a S a
CM
O
<3
pfi
£g
20 18 16 14 12 10 8 6 4 2 0
0-100 101- 201- 301- 401- 501- 601- 701200 300 400 500 600 700 800 i CD4-count I
FIGURE 1. Distribution of CD4+ T-lymphocyte count. CD4-count = CD4+ T-lymphocyte count.
enrolled in our study. Ages ranged from 26 to 54 years, with a mean ±1 SD of 39.08 ± 6.78 years. Median CD4+ T-lymphocyte count was 173 per mm with a 5%/95% confidence interval of 2/794 per mm (Figure 1). For multivariate regression analysis, the square root of the CD4 T-cell count was entered to achieve a normal distribution. The mean of the square root of the CD4+ T-lymphocyte count was 13.65 ± 7.35 cells per mm. The mean of the hemoglobin was 13.77 ± 1.56 mg/100 ml; of the hematocrit, 40.66% ± 4.90%; and of the 02-microglobulin, 3.66 ± 1.29 mg/1 (Table 1). The normal values of these tests and the standard deviation of the normal values are provided in Table 1. Thirteen (26%) of the 51 patients tested had systemic oppor tunistic infections. Eleven patients (22%) were as sessed as neuropsychologically impaired according to the rating described above. No patient had a neuro psychological impairment score of 6 or higher, indi cating that no patient with severe neuropsychological impairment was included in our study. On achromatic automated perimetry, the mean of the mean defect was —0.40 ± 2.08 dB. On shortwavelength automated perimetry, the mean of the mean defect was —3.94 ±4-17 dB. Mean log contrast sensitivity was 1.60 ± 0.21 at 1.5 spatial cycles per degree (cpd), 1.90 ± 0.21 at 3 cpd, 1.88 ± 0.22 at 6 cpd, 1.74 ± 0.33 at 12 cpd, and 1.34 ± 0.26 at 18 cpd. The mean of the total log for contrast sensitivity was 8.46 ± 0.89. The mean square root of the total error score on the Farnsworth-Munsell 100-hue test was 8.31 ± 3.13 (Table 2).
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
161
TABLE 1. Mean and Standard Deviation of Several Serum Parameters in the Examined Patients Square Root of the CD4* Hemoglobin (mg/100ml)
Hematocrit
Pj-Microglobulin
(cells/pl)
(%)
(mg/l)
13.65 ± 7.35 29.66 ± 18.09
13.77 + 1.56 16.00 ± 2
40.66 i 4.9 47.00 ± 5
3.66 = 1.29 3.50 ± 0.47
T-lymphocyte count
Mean = SD Normal value + SD
TABLE 2. Mean and Standard Deviation of the Results of the Psychophysical Tests Sum of
Mean Defect (dB)
Normal mean ± SD Patient mean ± SD
FarnsworthMunsell
Sensitivity
Log of Contrast Sensitivity
Short-
Log of Contrast
Achromatic Automated Perimetry
Perimetry
1.5
3
6
12
0.26 + 1.75 -0.40 ± 2.08
0.22 ± 3.04 -3.94 ± 4.17
1.82 ± 0.20 1.60 + 0.21
2,09 ± 0.19 1.90 ± 0.21
2.04 + 0.21 1.88+ 0.22
1.80 ± 0.25 1.74 ± 0.33
Values 1.5-18
Square Root
18
cpd
Error Score
1.41 ± 0.25 1.34 ± 0.26
9.05 ± 0.78 8.46 ± 0.89
5.16 ± 2.37 8.31 ± 3.13
for Spatial Frequencies (cpd)
cpd = cycles per degree.
On achromatic automated perimetry, 21.6% (11/ 51) of patients performed abnormally based on their mean defect (mean defect of less than —2.01 dB), and 27.5% (14/51) had a Glaucoma Hemifield Test result outside normal limits. On short-wavelength automated perimetry, 29.4% (15/51) of patients per formed abnormally based on the mean defect (mean defect of less than -5.5 dB), as did 23.5% (12/51) based on the Glaucoma Hemifield Test. The percent age of patients who performed abnormally on con trast sensitivity was 5.9% (3/51) at 1.5 cpd, 2.0% (2/51) at 3 cpd, 23.5% (12/51) at 6 cpd, 25.5% (13/51) at 12 cpd, and 33.3% (17/51) at 18 cpd; 29.4% (15/51) of patients performed abnormally compared with the normal age-dependent square root of the total error score (>11 to 14) on the Farnsworth-Munsell 100-hue test (Figure 2). Seven teen (33%) of the patients were considered normal on all visual function tests, whereas no patient was abnormal on all tests. Five statistically significant correlations between visual function testing were found before correction for multiple correlations based on the Bonferroni
162
inequality: achromatic automated perimetry correlat ed with short-wavelength automated perimetry (r2 = .073, P = .050), the Farnsworth-Munsell 100-hue test with achromatic automated perimetry (r2 = .090, P = .03), sum of log contrast sensitivity with achromatic automated perimetry (r2 = .16, P = .004), sum of log contrast sensitivity with short-wavelength automated perimetry (r2 = .078, P = .048), and sum of log contrast sensitivity with Farnsworth-Munsell 100hue test (r2 = .160, P = .004) (Table 3). After correction for multiple correlations based on the Bonferroni inequality, only two correlations remained statistically significant: sum of log contrast sensitivity with achromatic automated perimetry (r2 = .160, P = .004), and sum of log contrast sensitivity with the Farnsworth-Munsell test (r2 = .160, P = .004) (Table 3). No statistically significant correlation was found between any visual function test and the square root of the CD4+ T-lymphocyte count, presence of a systemic opportunistic infection, hemoglobin, hema tocrit, serum 02-microglobulin, or the global neuropsychological score (Figures 3 and 4).
AMERICAN JOURNAL OF OPHTHALMOLOGY
AUGUST 1997
60 j
1 i
SO 40 30 20 10 0
5 %
n ? n
<
CO
1-
IO
0-
(5
,-O A C3 C
1
m to O
CD
CO O
CO
o
CO
o
I ■ Abnormal n Normal I
FIGURE 2. Distribution of normal and abnormal per formances on neuropsychological and psychophysical tests. NP = global neuropsychological score; AAP MD = achromatic automated perimetry, mean defect; AAP GHT = achromatic automated perimetry, Glaucoma Hemifield Test; SWAP MD = short-wavelength auto mated perimetry, mean defect; SWAP GHT = shortwavelength automated perimetry, Glaucoma Hemifield Test; CS 1.5 through 18 = contrast sensitivity for spatial frequencies 1.5 to 18 cycles per degree; FM = Farnsworth-Munsell 100-hue test.
DISCUSSION VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS IN THE
absence of infectious retinopathy have been reported in several different areas of visual function (including peripheral visual sensitivity, visual contrast sensitivi ty, and color vision) as well as in electrophysiology 2,8,23 Although some of these studies tried to correlate their results to the C D + T-cell count, none attempted a correlation with a larger battery of medical and neuropsychological tests to characterize exactly the stage of the disease in a single patient. Our study was designed to investigate prospectively the pattern of HIV-related visual dysfunction in a careful ly selected large number of patients using a battery of visual psychophysical tests and to test the hypothesis that the visual dysfunctions seen in HIV-positive patients without infectious retinopathy correlated with declining medical status and progressive HIV disease as measured by a comprehensive set of medi cal and neuropsychological tests. In addition, we VOL.124,
No. 2
carefully screened each enrolled patient for the intake of a comprehensive list of drugs that might affect the tested visual functions. We found that a significant percentage of HIVpositive patients with visual acuity of 20/20 or better and no ophthalmologic evidence of retinitis per formed abnormally on several psychophysical tests compared with age-matched normal control subjects, including achromatic automated perimetry, shortwavelength automated perimetry, contrast sensitivity, and central color vision as measured by the Farnsworth-Munsell 100-hue test. As mentioned in the Methods section, we tested only one eye to reduce potential fatigue effects on these very time-con suming tests. We found statistically significant pos itive correlations between achromatic automated perimetry and short-wavelength automated perimetry, the Farnsworth-Munsell 100-hue test and achromatic automated perimetry, sum of log contrast sensitivity and achromatic automated perimetry, sum of log contrast sensitivity and short-wavelength automated perimetry, and sum of log contrast sensitivity and the Farnsworth-Munsell 100-hue test before correction for multiple correlations. After correction for multiple correlations based on the Bonferroni inequality, only two correlations remained statistically significant: sum of log contrast sensitivity and achromatic automated perimetry and sum of log contrast sensitivity and the Farnsworth-Munsell 100-hue test. This indicates that contrast sensitivity, a relatively easy-to-perform and quick-to-perform test, may have value in screening HIV-positive patients for visual dysfunctions. Four teen (70%) of the 20 patients who performed abnor mally on at least one spatial frequency on contrast sensitivity also performed abnormally on at least one of the other psychophysical tests (achromatic auto mated perimetry mean defect, achromatic automated perimetry Glaucoma Hemifield Test, short-wave length automated perimetry mean defect, short-wave length automated perimetry Glaucoma Hemifield Test, or Farnsworth-Munsell 100-hue test), and 11 patients (55%) performed abnormally on at least two of the other psychophysical tests. W h e n the criteria for performing abnormally on contrast sensitivity testing were more stringent (that is, abnormal perfor mance on at least two spatial frequencies on contrast sensitivity), 13 (93%) of the remaining 14 patients who met this criterion performed abnormally on at
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
163
TABLE 3. Pairwise Correlation of Psychophysical Test Results Sum of the
Achromatic automated perimetry Short-wavelength automated perimetry Sum of the log of contrast sensitivity values 1.5-18 cpd
Achromatic
Short-wavelength
Automated Perimetry
Automated Perimetry
Log of Contrast Sensitivity Values 1.5-18 cpd
—
—
r2 = (P r8 = (P = r* = (P =
.073 .050) .160 .004) .090 .030)
r2 (P = r2 = (P =
.078 .048) .044 .15)
— ^ = .160 (P = .004)
cpd = cycles per degree; r2 = square correlation coefficient (values marked in bold were statistically significant after correction for multiple correlations based on the Bonferroni inequality correction).
least one of the other psychophysical tests, and 10 patients (71%) performed abnormally on at least two of the other psychophysical tests. We found that the results of the psychophysical tests were not related to stage of disease, as deter mined by several validated markers (CD4+ T-lymphocyte count, hemoglobin, hematocrit, Centers for Dis ease Control and Prevention [CDC] classification, serum P2-microglobulin levels) and the global neuropsychological score. In addition, we did not identify a cause for the observed vision changes. This is consis tent with several other studies.2,6'8 Our results further show that at least some of the observed damage is retinal, and we made an effort to rule out clinically measurable, more central damage. However, it is possible that there is damage not measurable by the neuropsychological tests that is contributing to the reductions in visual function. Our findings, which are consistent with those of other reports,2'6,8 are certainly surprising in view of the fact that the incidence of most HIV-related diseases (for example, cytomegalovirus retinitis, noninfectious retinopathy, HIV-related dementia) are directly related to the decline in health as generally measured by CD4+ T-lymphocyte count or CDC classification.1 Plummer and associates7 re cently reported a significant difference between the achromatic and the short-wavelength automated pe rimetry Glaucoma Hemifield tests in patients with low CD4+ T-cell count (<500 cells per (il) compared to those with high CD4+ T-cell count (2:500 cells per |xl). In that study, only a small number of patients fell into the group with high CD + T-lymphocyte
164
count. The authors therefore speculated very carefully on a possible correlation of the reported visual field loss to stage of disease. In the present study, which is more comprehensive in the medical and neuropsy chological characterization of the patients, we were not able to reproduce this difference. The patients examined in the study by Plummer and associates7 were not identical with the patients reported in the present study. However, in the present study, only seven patients had a CD4+ T4-lymphocyte count above 500 cells per |JL1, which limits the power of the statistical analysis. Thus, the different outcomes of the two studies are most likely because of the small numbers of patients with high CD4+ T-lymphocyte count, and therefore no final conclusions can be drawn from either study regarding this point. This study was performed on patients who, under extensive neuropsychological testing,9,16,17 were found to be within the normal global neuropsychological functioning range. Visual abnormalities were based on testing, and we found no correlation between global neuropsychological function and visual dys functions. This result suggests that visual dysfunction may occur partially independently of global neuropsy chological dysfunction, and therefore it is possible that there may be a retinal basis for our results. Indeed, a cohort of patients, similar to our group, that underwent electroretinogram studies24 substantiated that there are clear indications of the presence of retinal dysfunction in HIV-positive individuals with out retinitis. There need to be further studies that investigate the possible locations of changes in the
AMERICAN JOURNAL OF OPHTHALMOLOGY
AUGUST 1997
FniTin uses
mi
FHUCmEIKM HI FUSE KE EltOIS WESTKW ma
HI «u
FOIBU 3t n TEST THE l!ii!
FIGURE 3. Example of a 26-year-old man who had a CD4 + T-lymphocyte count of 377 cells per |JL1 at the time of testing. His hemoglobin was 14.9 mg/100 ml, hematocrit was 43.9%, and P2-microglobulin was 1.5 mg/L No systemic opportunistic infection was detectable. With a neuropsychological test score of 3, he was considered neuropsychologically normal. The following test results were obtained in this patient: the mean defect on achromatic automated perimetry was —4.61 dB (abnormal, top left), and the Glaucoma Hemifield Test on achromatic automated perimetry was abnormal. The mean defect on short wavelength automated perimetry was —10 dB (abnormal), and the Glaucoma Hemifield Test on short-wavelength automated perimetry was abnormal. The patient also showed abnormal performance on contrast sensitivity testing (top right). The total error score on the Farnsworth-Munsell 100-hue test was 76 (normal, bottom left). The visual field defect pattern in the achromatic automated perimetry is nerve fiber bundle defects, which are covered by a generalized reduction in light sensitivity. However, this patient did not show any clinical sign of glaucoma; for example, his intraocular pressure was repeatedly documented as in the normal range, and no optic disk cupping was detected.
visual system. Furthermore, although these patients suffer from no global deficits, there is no question that many HIV-positive patients have subtle neuropsycho logical abnormalities that may not be captured in a global neuropsychological rating; therefore, we can not determine at this time whether the visual changes discovered in this study are causally related or are associated with subtle changes in the central nervous system. Also, our control group was well matched for age but did include more women than the study group did; however, there is no evidence for gender bias in the tests in which our subjects participated. Interestingly, relatively severe visual function im pairments may occur in patients who are still in the
VOL.124, No. 2
early stages of disease (Figure 3). These patients are considered completely asymptomatic. In contrast, patients who have relatively low CD4 + T-lymphocyte counts may demonstrate normal visual functions in all psychophysical tests, even in the presence of mild neuropsychological impairment (Figure 4). Histopathologic studies of patients with AIDS have shown a loss of up to 50% of fibers in the optic nerve with no evidence of retinitis.25 In addition, others have specu lated that cotton-wool spots produce permanent damage in the nerve-fiber layer of the retina.26 In our study, we therefore eliminated patients who had cotton-wool spots at the time of testing. However, it is possible that some of the patients we tested had had
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
165
wwiww
* .ft FDHTIM LOSSES
wtE m mm FUSE *c emu
A A *" A A * ,34
-lB)-g
»H
rat tin urn
32
31 »
31
33
IIS)'
it
21
■
,
31 31
!! 27
»
l!
-2 -1 1 1-2
2-1 3
I
-1 -9
I -2
30-
10'
-3 1-3
-2
-4
-2
-4
•J ■?
I
■I -3 ■1
I
1
-I -1
1 - 3 - 4
-1
-2
I
-3
-1
■ 1 1 - 1 1
-2 -I -1
-2
I
-3
-4 -I -1
1
1-2-2
1 -3
2 -3
i
®
L
a
©
20/100 p I-
«
©
3'
-1
20/15
20/20
<$ 2o/ra
®
2
■01
7&. j ^ z», .„ * ®
21
1 •
(T k®
W
-4
1 1 - 2 - 3
I -2-2
wo23 23
-S -1 -4 4
-1
^4.
A " ■ !1 A
31
r
1
(B> lit)
M i»u
®
n
V!t
HSTIMS»9ED 344 RNHI
■ 3>
■m
300-
s &
sons 2W30
O
-! -3
rm ■Evnnim
nnniLm itnas "P<
31
e r < 2i
■r<
y: yellow
I.SI
FIGURE 4. Example of a 43-year-old man who had a CD4 + T-lymphocyte count of 11 cells per |xl at the time of testing. His hemoglobin was 13.3 mg/100 ml, hematocrit was 39.9%, and P2-microglobulin was 2.2 mg/L No systemic opportunistic infection was detectable. With a neuropsychological testing score of 5, he was considered mildly neuropsychologically impaired. This patient performed normally on all visual function tests (mean defect, top left) and Glaucoma Hemifield Test on achromatic automated perimetry, mean defect and Glaucoma Hemifield Test on short-wavelength automated perimetry, contrast sensitivity testing (top right), and the Farnsworth-Munsell 100-hue test (bottom left).
cotton-wool spots that were resolved at the time of testing. We also did not eliminate persons with documented history of cotton-wool spots that were resolved at the time of testing because the implicit assumption that the remaining HIV-positive group has no retinal damage may be incorrect. One way of eliminating this is to recruit a large number of patients relatively soon after seroconversion and track them longitudinally at least every 2 to 3 months for 10 to 15 years, which is impractical. Thus, it could be hypoth esized that our findings are solely because of past presence of cotton-wool spots. However, this is very unlikely because we also found visual changes in patients with high CD4 + T-lymphocyte count, for whom the occurrence of cotton-wool spots is rare. Because visual function loss is not correlated to stage of disease, it could be speculated that it is caused directly as a result of HIV infection in the central nervous system or in the visual pathway, including
166
the retina, or both. This is supported by the fact that several pathology studies have shown cortical changes in the areas associated with visual function (for example, the occipital, parietal, and temporal lobes). Another possible reason for vision loss could be potential side effects of antiviral and other drugs that these patients are required to take, even in the early stage of HIV-disease. It is well known that a variety of drugs may affect visual perception.27 Although we screened carefully and did not include in our study patients who took doses of drugs that are known to affect visual perception, there is insufficient data on the side effects of several new antiretroviral drugs or protease inhibitors, such as dideoxyinosine, 2',3'dideoxycytidine, or stavudine, to rule out completely the possibility that the visual function loss we have found is the result of some of these newer drugs. At this point, we conclude that the cause for our findings can neither be attributed simply to declining
AMERICAN JOURNAL OF OPHTHALMOLOGY
AUGUST 1997
health status, clinically evident cortical or cognitive impairment, or known drug side effects, but is more likely attributable to subtle retinal dysfunctions. To elucidate these findings further, we are currently following up this well-characterized group of patients in a longitudinal study at 3-month intervals to investigate the pattern of change in visual function over time using this same battery of tests.
REFERENCES 1. Jabs DA, Green WR, Fox R, Polk BF, Bartlett JG. Ocular manifestations of acquired immune deficiency syndrome. Ophthalmology 1989;96:1092-1099. 2. Iragui VJ, Kalmijn J, Thai LJ, Grant I. Neurological dysfunc tion in asymptomatic HIV-1-infected men: evidence from evoked potentials. HNRC Group. Electroencephalogr Clin Neurophysiol 1994;92:1-10. 3. Malessa R, Agelink MW, Diener HC. Dysfunction of visual pathways in HIV-1 infection. J Neural Sci 1995;130:82-87. 4. Mueller AJ, Berninger T, Matuschke A, Klauss V, Goebel F-D. Elektroophthalmologische Untersuchungen bei HIVPatienten. Fortschr Ophthalmol 1991;88:712-715. 5. Mueller AJ, Berninger T, Geier S, Lachenmayer B, Klauss V. Elektroophthalmologische und psychophysische Untersuch ungen bei fundoskopisch unauffalligen HIV-Positiven und AIDS-Patienten. Folia Ophthalmol 1992;17:5-9. 6. Mueller AJ, Lachenmayr BJ, Eckstein A, Holzl M. Multiplestimulus presentation and voice control in automated perimetry. Ger J Ophthalmol 1992;1:91-95. 7. Plummer DJ, Sample PA, Arevalo JF, et al. Visual field loss in HIV-positive patients without infectious retinopathy. Am J Ophthalmol 1996;122:542-549. 8. Quiceno JI, Capparelli E, Sadun AA, et al. Visual dysfunc tion without retinitis in patients with acquired immunodefi ciency syndrome. Am J Ophthalmol 1992;113:8-13. 9. Heaton R, Velin R, McCutchan J, et al. Neuropsychological impairment in HIV infection: implications for employment. Psychosom Med 1994;56:8-17. 10. Ferris F, Kassoff A, Bresnick GH, Bailey I. New visual acuity charts for clinical research. Am J Ophthalmol 1982;94: 91-96. 11. Sample PA, Martinez GA, Weinreb RN. Short-wavelength automated perimetry without lens density testing. Am J Ophthalmol 1994;118:632-641.
12 Sample P, Weinreb R. Color perimetry for assessment of primary open angle glaucoma. Invest Ophthalmol Vis Sci 1990;31:1869-1875. 13 Sample P, Weinreb R. Progressive color visual field loss in glaucoma. Invest Ophthalmol Vis Sci 1992;33:240-243. 14. Sample P, Taylor J, Martinez G, Lusky M, Weinreb R. Short-wavelength color visual fields in glaucoma suspects at risk. Am J Ophthalmol 1993;115:225-233. 15 Verriest G, Van Laethem J, Uvijls A. A new assessment of the normal ranges of the Farnsworth-Munsell 100-hue test scores. Am J Ophthalmol 1982;93:635-642. 16. Heaton R, Pendleton M. Use of neuropsychological tests to predict adult patients everyday functioning. J Consult Clin Psychol 1981;49:807-821. 17. Heaton R, Grant I, Butters N, et al. The HNRC 500: neuropsychology of HIV infection at different disease stages. Journal of the International Neuropsychological Society 1995;1:231-251. 18. Heijl A, Lindgren G, Olsson J. Normal variability of static perimetric threshold values across the central visual field. Arch Ophthalmol 1987;105:1544-1549. 19 Asman P, Heijl A. Glaucoma Hemifield Test Automated Visual Field Evaluation. Arch Ophthalmol 1992;110: 812-819. 20. Glantz S. Primer of biostatistics, 3rd ed. New York: McGrawHill, 1992. 21 Knoblauch K. On quantifying the bipolarity and axis of the Farnsworth-Munsell 100-hue test. Invest Ophthalmol Vis Sci 1987;28:707-710. 22 Vingrys A, King-Smith P. A quantitative scoring technique for panel test of color vision. Invest Ophthalmol Vis Sci 1988;29:50-63. 23 Geier S, Nohmeier C, Lachenmayr B, Klauss V, Goebel FD. Deficits in perimetric performance in patients with sympto matic human immunodeficiency virus infection or acquired immunodeficiency syndrome. Am J Ophthalmol 1995; 119: 335-344. 24. Iragui VJ, Kalmijn J, Plummer DJ, Sample PA, Trick GL, Freeman WR. Pattern electroretinograms and visual evoked potentials in HIV infection: evidence of asymptomatic retinal and postretinal impairment in the absence of infectious retinopathy. Neurology 1996;47:1452-1456. 25. Tenhula WN, Shizao X, Madigan MC, Heller K, Freeman WR, Sadun AA. Morphometric comparisons of optic nerve axon loss in AIDS: morphometric comparisons. Am J Ophthalmol 1992;113:14-20. 26 Sadun A, Pepose J, Madigan M, Laycock K, Tenhula W, Freeman W. AIDS-related optic neuropathy: a histological, virological and ultrastructural study. Graefes Arch Clin Exp Ophthalmol 1995;233:387-398. 27, Fraunfelder FT. Drug-induced ocular side effects and drug interactions, 3rd ed. Philadelphia: Febiger, 1989.
Authors Interactive™ Comments and questions regarding this article may be addressed to authors via THE JOURNAL'S Web Site. Selected comments, questions, and author responses are posted on Authors Interactive™ at http://www.ajo.com/
VOL.124,
No. 2
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
167
• PURPOSE: To investigate visual dysfunctions in ophthalmoscopically normal human immunodefi ciency virus (HlV)-positive patients and to corre late the results to the stage of HIV disease and neuropsychological status. • METHODS: Fifty-one randomly selected eyes (26 right, 25 left) of 51 HIV-positive patients with visual acuity measurements of 20/20 or better and no ophthalmoscopically detectable disorders were prospectively examined using achromatic and short-wavelength automated perimetry, color vi sion testing, and contrast sensitivity testing. CD4 + T-lymphocyte count, presence of systemic infec tion, hemoglobin, hematocrit, serum |$2-microglobulin levels, and results of neuropsychological testing were also analyzed. • RESULTS: On achromatic automated perimetry, 21.6% (11/51) of patients performed abnormally according to the mean defect and 27.5% (14/51) according to the Glaucoma Hemifield Test; 29.4% (15/51) performed abnormally on short-wave length automated perimetry according to the mean
Accepted for publication April 14, 1997. From the Department of Ophthalmology, Shiley Eye Center (Drs Mueller, Plummer, Taskintuna, Sample, and Freeman, and Ms Dua), and the Department of Psychiatry (Dr Grant), University of California San Diego, La Jolla, California. Supported by Deutsche Forschungsgemeinschaft grant Mu 1204/1-2 (Dr Mueller); University of California Universitywide AIDS Research Programs R95-SD-082 (Dr Plummer) and R92-SD-135 (Dr Sample); National Eye Institute grant EYO 7366 (Dr Freeman); and a departmental grant from Research to Prevent Blindness, Inc, New York, New York. Reprint requests to W. R. Freeman, MD, Department of Ophthalmolo gy, Shiley Eye Center, University of California, San Diego, 9415 Campus Point Dr 0946, La Jolla, CA 92093-0946; fax: (619) 534-7985.
158
defect and 23.5% (12/51) according to the Glauco ma Hemifield Test. On contrast sensitivity, 5.9% (3/51) of patients performed abnormally in the 1.5-cycles per degree (cpd) line, 2.0% (2/51) in the 3-cpd line, 23.5% (12/51) in the 6-cpd line, 25.5% (13/51) in the 12-cpd line, and 33.3% (17/51) in the 18-cpd line. On the Farnsworth-Munsell 100hue test, 29.4% (15/51) of patients performed abnormally. After correction for multiple correla tions, two statistically significant correlations were found: sum of log contrast sensitivity with achro matic automated perimetry and sum of log contrast sensitivity with the Farnsworth-Munsell 100-hue test. • CONCLUSIONS: A significant percentage of HIV-positive patients with visual acuity of 20/20 or better and no ophthalmologic evidence of retini tis performed abnormally on visual psychophysical tests. The severity of visual dysfunction was not correlated with the stage of HIV infection or the degree of neuropsychological dysfunction.
P
ATIENTS WITH HUMAN IMMUNODEFICIENCY Vi rus (HIV) are susceptible to several different types of infectious retinopathies1 that may all affect different aspects of visual function. However, various changes in visual function have also been reported in the absence of infectious and noninfectious retinopathies in HIV-positive patients.2'6 Our group has previously reported7 significant losses in visual function in both standard achromatic automated perimetry and short-wavelength automat ed perimetry. In addition to changes measured by
AMERICAN JOURNAL OF OPHTHALMOLOGY 1997;124:158-167
AUGUST 1997
perimetry, we have also demonstrated8 loss in visual contrast sensitivity and changes in central color vision. However, the pattern of visual function loss among a variety of different psychophysical test meth ods has not been explored in a single patient sample. In addition, it is unclear whether decreasing visual function is related to worsening HIV-related illness or neuropsychological status. This prospective study was See also pp. 141-157, 168-180, 181-189, 190-198, 199-205, 227-233, and 234-239.
designed to investigate the pattern of HIV-related visual dysfunctions using four validated visual psychophysical tests and to correlate the results to the stage of HIV infection. We also correlated these findings with a battery of tests performed to measure the severity of HIV-related disease and neuropsychologi cal performance. Each of these tests was performed in each patient within a defined time interval.
METHODS THE PARTICIPANTS FOR THIS STUDY WERE ALL PROSPEC-
tively drawn from a longitudinally followed cohort at the University of California, San Diego, HIV Neurobehavioral Research Center and the AIDS Ocular Research Unit. The inclusion and exclusion criteria for the HIV Neurobehavioral Research Center proto col have been described extensively elsewhere.9 This study was approved by the Human Subjects Commit tee of the University of California, San Diego. All subjects gave informed consent. All patients underwent complete ophthalmologic examination including visual acuity, intraocular pres sure, slit-lamp, and dilated fundus examination by an ophthalmologist (A.J.M., I.T., W.R.F.) who was familiar with the ocular manifestations of AIDS. Inclusion criteria were 20/20 or better normal or corrected-to-normal Snellen visual acuity in each eye using the EDTRS chart,10 normal intraocular pressure (11 to 21 mm Hg by applanation), and normal appearance of the anterior and posterior segment of the eye. With these inclusion criteria, it is not likely that any of the observed visual dysfunctions are related to cataract, glaucoma, or other apparent eye VOL.124, No. 2
diseases. Thus we assumed that detectable losses of visual function are solely related to HIV. Although ruling out persons with visual acuity of worse than 20/20 may bias the population slightly, fewer than 10% of people agreeing to participate in this study failed to meet this criterion. In addition, our popula tion is relatively young (age range, 22 to 52 years), which can explain the high percentage of people with visual acuity of 20/20 or better. Exclusion criteria were diabetes mellitus, arterial hypertension, or a family history of glaucoma or other degenerative retinal disease as well as of any other manifest ocular disease, especially cataract, glaucoma, cytomegalovirus retinitis, vitreal or retinal hemor rhages, or cotton-wool spots in either eye on the date of examination. We also excluded all patients who, within 3 months of testing, had received courses of any medication that might affect visual function. In particular, we checked for any pharmacologically active topical ocular medication (that is, pilocarpine or derivative), chloramphenicol, chloroquine or its derivatives, digoxin or derivatives, ethambutol, gentamicin or other aminoglycosides, suramin sodium, iron in any preparation, oral contraceptives, quinine, corticosteroids, streptomycin, tetracycline or its deriva tives, thioridazine, and vitamins A or D. One eye was selected randomly for achromatic automated perimetry (Humphrey program 24/2), short-wavelength automated perimetry (program 24/2), color vision testing (Farnsworth-Munsell 100hue test), and contrast sensitivity testing (Vistech Consultants, Dayton, Ohio). To rule out fatiguing effects, the tests were performed in random order on the selected eye. We did not test both eyes because total duration of testing averaged approximately 1 hour per eye, and testing a second eye might invali date our results because of excessive fatigue. Achromatic automated perimetry was performed with the standard settings provided by the internal settings of the machine using program 24-2 and a size III (Goldmann) target. A chromatic version of the short-wavelength automated perimetry test has been developed on a modified Humphrey Field Analyzer and is currently used to study primary open-angle glaucoma and diabetic retinopathy.11'14 The test pro cedures were validated by our group and have been previously described.11'12 The test results obtained for both achromatic and short-wavelength automated
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
159
perimetry were corrected for age according to our internal normal database for these tests.11 The other tests were performed according to the instructions of the manufacturers. Age correction for contrast sensitivity was performed according to the manual provided with the test. Age correction for color vision testing was performed according to published data.15 Both tests (contrast sensitivity and Farnsworth-Munsell 100-hue test) have been previ ously validated in our research using HIV-negative control patients.8 All patients also underwent neuropsychological testing, which was performed using a battery of tests based on the extended Halstead-Reiton Battery that has been previously described in detail.16,17 Each group of tests was scored in a masked fashion, and a rating was obtained in each of the following areas: verbal skill, abstraction, psychomotor function, atten tion, learning, memory, and sensory skill. This bat tery assesses verbal language skills, attention and speed of information processing, abstracting ability, cognitive flexibility, complex perceptual motor skills, memory functions (learning, recall), and simple mo tor and simple sensory functions. The neuropsycho logical data from the approximately 50 tests can be summarized in several ways.9 For the purposes of the present study, we focused on the composite or global rating of impairment, which is based on masked ratings by a trained judge (I.G.) of all test data from eight cognitive domains. The global judgment is rendered on a 9-point scale, with 1 being excellent neuropsychological function and 9, very impaired neuropsychological function. A score of 5 indicates mild impairment. For this study, all patients scoring 5 or more were classified as neuropsychologically im paired. Previous research has indicated that such ratings are reliable, sensitive, and specific indicators for brain disease.16 All medical and neuropsychologi cal tests were performed within ±3 months of the time of the visual function testing. Statistical analysis was performed in two steps. In the first step, we calculated the percentage of eyes that performed abnormally separately for each test. For perimetry, we used the clinical standards for cutoffs on the various visual field indices, which are usually 95% for mean defect and 99.5% for the Glaucoma Hemifield Test. For the contrast sensitivity and
Farnsworth-Munsell 100-hue tests, we used the pub lished and validated data on cutoff values.8 Thus, for achromatic automated perimetry, eyes were assessed as abnormal when the mean defect was less than —2.01 dB.7 For short-wavelength automated perime try, eyes were assessed as abnormal when the mean defect was less than —5.5 dB.7,11 These achromatic automated perimetry mean defect and short-wave length automated perimetry mean defect cutoffs are the calculated and age-corrected 95% confidence interval (CI) of our own normal database.7,11'14 This control group consists of 91 HIV-negative normal subjects with an age distribution (22 to 55 years) comparable to that of the HIV-positive patients.7 The normal limits for our control group are even more stringent than those of the standard Humphrey control group, which consists of patients of all age ranges, including those in higher age groups.18 Al though the mean defect is a good indicator for a generalized defect in the visual field, it may not detect small or localized visual field defects. Thus, both achromatic and short-wavelength automated perime try were also analyzed with the Glaucoma Hemifield Test to obtain information about localized field de fects.7,18 Both the achromatic and the short-wave length automated perimetry Glaucoma Hemifield Test were assessed as abnormal when at least one of the analyzed five sector pairs was outside the 99.5% CI of the above-mentioned normal database.7 The 99.5% CI of both the achromatic and the short-wavelength automated perimetry Glaucoma Hemifield Test was chosen rather than the 95% CI because of a correc tion for multiple comparisons within the visual field. This is explained in detail by Asman and Heijl,19 who developed this technique. For contrast sensitivity, the test results were assessed as abnormal when fewer than patches 1 to 5 were observed correctly for each spatial frequency line.8 For the Farnsworth-Munsell 100-hue test, eyes were assessed as abnormal when the square root of the total error score was great er than 11.22 for patients aged 20 to 29 years, greater than 12.67 for patients aged 30 to 39 years, greater than 13.64 for patients aged 40 to 49 years, or greater than 14-71 for patients aged 50 to 59 years.15
160
OPHTHALMOLOGY
AMERICAN JOURNAL
In the second step of the statistical analysis, a multivariate regression analysis between various test
AUGUST 1997
results was performed both with and without correc tion for multiple correlations based on the Bonferroni inequality.20 For both achromatic automated perimetry and short-wavelength automated perimetry, the mean defect was calculated and entered. For contrast sensitivity, the sum of the log contrast sensitivity values was calculated and entered, and for the Farnsworth-Munsell 100-hue test, the square root of the total error score was calculated and entered.21,22 In addition to the results of the visual function tests, the following medical data were included in the multivariate regression analysis: square root of the CD4+ T-lymphocyte count, presence of any systemic oppor tunistic infection (elsewhere than in the eye), hemo globin, hematocrit, serum P2-microglobulin, and the result of the global neuropsychological score. All statistical analysis was performed using the statistical program JMP 3 (SAS Institute Inc, Cary, North Carolina). For multivariate regression analysis, we reversed the sign of the Farnsworth-Munsell 100-hue test and the neuropsychological testing from + to — because the Farnsworth-Munsell test and the neuropsychological testing have a grading system opposite that of the other tests (achromatic automat ed perimetry, short-wavelength automated perimetry, contrast sensitivity). On the Farnsworth-Munsell 100-hue test and neuropsychological testing, higher (more positive) values indicate poorer test perfor mance than lower (more negative) values do. In contrast, on the other tests (achromatic automated perimetry, short-wavelength automated perimetry, contrast sensitivity) higher (more positive) values indicate better performance than lower (more nega tive) values do. Inverting the values on the Farnsworth-Munsell 100-hue test and neuropsycho logical testing from + to — yields "positive" correla tions when poor performance on one test correlates with poor performance on another test. In contrast, all "negative" correlations between tests indicate that poor performance on one test is correlated with good performance on another test.
RESULTS FIFTY-ONE EYES (26 RIGHT, 25 LEFT) OF 51 HIV-POSITIVE
patients (50 [98%] men, one [2%] woman) were
VOL.124, No. 2
1
.a S a
CM
O
<3
pfi
£g
20 18 16 14 12 10 8 6 4 2 0
0-100 101- 201- 301- 401- 501- 601- 701200 300 400 500 600 700 800 i CD4-count I
FIGURE 1. Distribution of CD4+ T-lymphocyte count. CD4-count = CD4+ T-lymphocyte count.
enrolled in our study. Ages ranged from 26 to 54 years, with a mean ±1 SD of 39.08 ± 6.78 years. Median CD4+ T-lymphocyte count was 173 per mm with a 5%/95% confidence interval of 2/794 per mm (Figure 1). For multivariate regression analysis, the square root of the CD4 T-cell count was entered to achieve a normal distribution. The mean of the square root of the CD4+ T-lymphocyte count was 13.65 ± 7.35 cells per mm. The mean of the hemoglobin was 13.77 ± 1.56 mg/100 ml; of the hematocrit, 40.66% ± 4.90%; and of the 02-microglobulin, 3.66 ± 1.29 mg/1 (Table 1). The normal values of these tests and the standard deviation of the normal values are provided in Table 1. Thirteen (26%) of the 51 patients tested had systemic oppor tunistic infections. Eleven patients (22%) were as sessed as neuropsychologically impaired according to the rating described above. No patient had a neuro psychological impairment score of 6 or higher, indi cating that no patient with severe neuropsychological impairment was included in our study. On achromatic automated perimetry, the mean of the mean defect was —0.40 ± 2.08 dB. On shortwavelength automated perimetry, the mean of the mean defect was —3.94 ±4-17 dB. Mean log contrast sensitivity was 1.60 ± 0.21 at 1.5 spatial cycles per degree (cpd), 1.90 ± 0.21 at 3 cpd, 1.88 ± 0.22 at 6 cpd, 1.74 ± 0.33 at 12 cpd, and 1.34 ± 0.26 at 18 cpd. The mean of the total log for contrast sensitivity was 8.46 ± 0.89. The mean square root of the total error score on the Farnsworth-Munsell 100-hue test was 8.31 ± 3.13 (Table 2).
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
161
TABLE 1. Mean and Standard Deviation of Several Serum Parameters in the Examined Patients Square Root of the CD4* Hemoglobin (mg/100ml)
Hematocrit
Pj-Microglobulin
(cells/pl)
(%)
(mg/l)
13.65 ± 7.35 29.66 ± 18.09
13.77 + 1.56 16.00 ± 2
40.66 i 4.9 47.00 ± 5
3.66 = 1.29 3.50 ± 0.47
T-lymphocyte count
Mean = SD Normal value + SD
TABLE 2. Mean and Standard Deviation of the Results of the Psychophysical Tests Sum of
Mean Defect (dB)
Normal mean ± SD Patient mean ± SD
FarnsworthMunsell
Sensitivity
Log of Contrast Sensitivity
Short-
Log of Contrast
Achromatic Automated Perimetry
Perimetry
1.5
3
6
12
0.26 + 1.75 -0.40 ± 2.08
0.22 ± 3.04 -3.94 ± 4.17
1.82 ± 0.20 1.60 + 0.21
2,09 ± 0.19 1.90 ± 0.21
2.04 + 0.21 1.88+ 0.22
1.80 ± 0.25 1.74 ± 0.33
Values 1.5-18
Square Root
18
cpd
Error Score
1.41 ± 0.25 1.34 ± 0.26
9.05 ± 0.78 8.46 ± 0.89
5.16 ± 2.37 8.31 ± 3.13
for Spatial Frequencies (cpd)
cpd = cycles per degree.
On achromatic automated perimetry, 21.6% (11/ 51) of patients performed abnormally based on their mean defect (mean defect of less than —2.01 dB), and 27.5% (14/51) had a Glaucoma Hemifield Test result outside normal limits. On short-wavelength automated perimetry, 29.4% (15/51) of patients per formed abnormally based on the mean defect (mean defect of less than -5.5 dB), as did 23.5% (12/51) based on the Glaucoma Hemifield Test. The percent age of patients who performed abnormally on con trast sensitivity was 5.9% (3/51) at 1.5 cpd, 2.0% (2/51) at 3 cpd, 23.5% (12/51) at 6 cpd, 25.5% (13/51) at 12 cpd, and 33.3% (17/51) at 18 cpd; 29.4% (15/51) of patients performed abnormally compared with the normal age-dependent square root of the total error score (>11 to 14) on the Farnsworth-Munsell 100-hue test (Figure 2). Seven teen (33%) of the patients were considered normal on all visual function tests, whereas no patient was abnormal on all tests. Five statistically significant correlations between visual function testing were found before correction for multiple correlations based on the Bonferroni
162
inequality: achromatic automated perimetry correlat ed with short-wavelength automated perimetry (r2 = .073, P = .050), the Farnsworth-Munsell 100-hue test with achromatic automated perimetry (r2 = .090, P = .03), sum of log contrast sensitivity with achromatic automated perimetry (r2 = .16, P = .004), sum of log contrast sensitivity with short-wavelength automated perimetry (r2 = .078, P = .048), and sum of log contrast sensitivity with Farnsworth-Munsell 100hue test (r2 = .160, P = .004) (Table 3). After correction for multiple correlations based on the Bonferroni inequality, only two correlations remained statistically significant: sum of log contrast sensitivity with achromatic automated perimetry (r2 = .160, P = .004), and sum of log contrast sensitivity with the Farnsworth-Munsell test (r2 = .160, P = .004) (Table 3). No statistically significant correlation was found between any visual function test and the square root of the CD4+ T-lymphocyte count, presence of a systemic opportunistic infection, hemoglobin, hema tocrit, serum 02-microglobulin, or the global neuropsychological score (Figures 3 and 4).
AMERICAN JOURNAL OF OPHTHALMOLOGY
AUGUST 1997
60 j
1 i
SO 40 30 20 10 0
5 %
n ? n
<
CO
1-
IO
0-
(5
,-O A C3 C
1
m to O
CD
CO O
CO
o
CO
o
I ■ Abnormal n Normal I
FIGURE 2. Distribution of normal and abnormal per formances on neuropsychological and psychophysical tests. NP = global neuropsychological score; AAP MD = achromatic automated perimetry, mean defect; AAP GHT = achromatic automated perimetry, Glaucoma Hemifield Test; SWAP MD = short-wavelength auto mated perimetry, mean defect; SWAP GHT = shortwavelength automated perimetry, Glaucoma Hemifield Test; CS 1.5 through 18 = contrast sensitivity for spatial frequencies 1.5 to 18 cycles per degree; FM = Farnsworth-Munsell 100-hue test.
DISCUSSION VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS IN THE
absence of infectious retinopathy have been reported in several different areas of visual function (including peripheral visual sensitivity, visual contrast sensitivi ty, and color vision) as well as in electrophysiology 2,8,23 Although some of these studies tried to correlate their results to the C D + T-cell count, none attempted a correlation with a larger battery of medical and neuropsychological tests to characterize exactly the stage of the disease in a single patient. Our study was designed to investigate prospectively the pattern of HIV-related visual dysfunction in a careful ly selected large number of patients using a battery of visual psychophysical tests and to test the hypothesis that the visual dysfunctions seen in HIV-positive patients without infectious retinopathy correlated with declining medical status and progressive HIV disease as measured by a comprehensive set of medi cal and neuropsychological tests. In addition, we VOL.124,
No. 2
carefully screened each enrolled patient for the intake of a comprehensive list of drugs that might affect the tested visual functions. We found that a significant percentage of HIVpositive patients with visual acuity of 20/20 or better and no ophthalmologic evidence of retinitis per formed abnormally on several psychophysical tests compared with age-matched normal control subjects, including achromatic automated perimetry, shortwavelength automated perimetry, contrast sensitivity, and central color vision as measured by the Farnsworth-Munsell 100-hue test. As mentioned in the Methods section, we tested only one eye to reduce potential fatigue effects on these very time-con suming tests. We found statistically significant pos itive correlations between achromatic automated perimetry and short-wavelength automated perimetry, the Farnsworth-Munsell 100-hue test and achromatic automated perimetry, sum of log contrast sensitivity and achromatic automated perimetry, sum of log contrast sensitivity and short-wavelength automated perimetry, and sum of log contrast sensitivity and the Farnsworth-Munsell 100-hue test before correction for multiple correlations. After correction for multiple correlations based on the Bonferroni inequality, only two correlations remained statistically significant: sum of log contrast sensitivity and achromatic automated perimetry and sum of log contrast sensitivity and the Farnsworth-Munsell 100-hue test. This indicates that contrast sensitivity, a relatively easy-to-perform and quick-to-perform test, may have value in screening HIV-positive patients for visual dysfunctions. Four teen (70%) of the 20 patients who performed abnor mally on at least one spatial frequency on contrast sensitivity also performed abnormally on at least one of the other psychophysical tests (achromatic auto mated perimetry mean defect, achromatic automated perimetry Glaucoma Hemifield Test, short-wave length automated perimetry mean defect, short-wave length automated perimetry Glaucoma Hemifield Test, or Farnsworth-Munsell 100-hue test), and 11 patients (55%) performed abnormally on at least two of the other psychophysical tests. W h e n the criteria for performing abnormally on contrast sensitivity testing were more stringent (that is, abnormal perfor mance on at least two spatial frequencies on contrast sensitivity), 13 (93%) of the remaining 14 patients who met this criterion performed abnormally on at
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
163
TABLE 3. Pairwise Correlation of Psychophysical Test Results Sum of the
Achromatic automated perimetry Short-wavelength automated perimetry Sum of the log of contrast sensitivity values 1.5-18 cpd
Achromatic
Short-wavelength
Automated Perimetry
Automated Perimetry
Log of Contrast Sensitivity Values 1.5-18 cpd
—
—
r2 = (P r8 = (P = r* = (P =
.073 .050) .160 .004) .090 .030)
r2 (P = r2 = (P =
.078 .048) .044 .15)
— ^ = .160 (P = .004)
cpd = cycles per degree; r2 = square correlation coefficient (values marked in bold were statistically significant after correction for multiple correlations based on the Bonferroni inequality correction).
least one of the other psychophysical tests, and 10 patients (71%) performed abnormally on at least two of the other psychophysical tests. We found that the results of the psychophysical tests were not related to stage of disease, as deter mined by several validated markers (CD4+ T-lymphocyte count, hemoglobin, hematocrit, Centers for Dis ease Control and Prevention [CDC] classification, serum P2-microglobulin levels) and the global neuropsychological score. In addition, we did not identify a cause for the observed vision changes. This is consis tent with several other studies.2,6'8 Our results further show that at least some of the observed damage is retinal, and we made an effort to rule out clinically measurable, more central damage. However, it is possible that there is damage not measurable by the neuropsychological tests that is contributing to the reductions in visual function. Our findings, which are consistent with those of other reports,2'6,8 are certainly surprising in view of the fact that the incidence of most HIV-related diseases (for example, cytomegalovirus retinitis, noninfectious retinopathy, HIV-related dementia) are directly related to the decline in health as generally measured by CD4+ T-lymphocyte count or CDC classification.1 Plummer and associates7 re cently reported a significant difference between the achromatic and the short-wavelength automated pe rimetry Glaucoma Hemifield tests in patients with low CD4+ T-cell count (<500 cells per (il) compared to those with high CD4+ T-cell count (2:500 cells per |xl). In that study, only a small number of patients fell into the group with high CD + T-lymphocyte
164
count. The authors therefore speculated very carefully on a possible correlation of the reported visual field loss to stage of disease. In the present study, which is more comprehensive in the medical and neuropsy chological characterization of the patients, we were not able to reproduce this difference. The patients examined in the study by Plummer and associates7 were not identical with the patients reported in the present study. However, in the present study, only seven patients had a CD4+ T4-lymphocyte count above 500 cells per |JL1, which limits the power of the statistical analysis. Thus, the different outcomes of the two studies are most likely because of the small numbers of patients with high CD4+ T-lymphocyte count, and therefore no final conclusions can be drawn from either study regarding this point. This study was performed on patients who, under extensive neuropsychological testing,9,16,17 were found to be within the normal global neuropsychological functioning range. Visual abnormalities were based on testing, and we found no correlation between global neuropsychological function and visual dys functions. This result suggests that visual dysfunction may occur partially independently of global neuropsy chological dysfunction, and therefore it is possible that there may be a retinal basis for our results. Indeed, a cohort of patients, similar to our group, that underwent electroretinogram studies24 substantiated that there are clear indications of the presence of retinal dysfunction in HIV-positive individuals with out retinitis. There need to be further studies that investigate the possible locations of changes in the
AMERICAN JOURNAL OF OPHTHALMOLOGY
AUGUST 1997
FniTin uses
mi
FHUCmEIKM HI FUSE KE EltOIS WESTKW ma
HI «u
FOIBU 3t n TEST THE l!ii!
FIGURE 3. Example of a 26-year-old man who had a CD4 + T-lymphocyte count of 377 cells per |JL1 at the time of testing. His hemoglobin was 14.9 mg/100 ml, hematocrit was 43.9%, and P2-microglobulin was 1.5 mg/L No systemic opportunistic infection was detectable. With a neuropsychological test score of 3, he was considered neuropsychologically normal. The following test results were obtained in this patient: the mean defect on achromatic automated perimetry was —4.61 dB (abnormal, top left), and the Glaucoma Hemifield Test on achromatic automated perimetry was abnormal. The mean defect on short wavelength automated perimetry was —10 dB (abnormal), and the Glaucoma Hemifield Test on short-wavelength automated perimetry was abnormal. The patient also showed abnormal performance on contrast sensitivity testing (top right). The total error score on the Farnsworth-Munsell 100-hue test was 76 (normal, bottom left). The visual field defect pattern in the achromatic automated perimetry is nerve fiber bundle defects, which are covered by a generalized reduction in light sensitivity. However, this patient did not show any clinical sign of glaucoma; for example, his intraocular pressure was repeatedly documented as in the normal range, and no optic disk cupping was detected.
visual system. Furthermore, although these patients suffer from no global deficits, there is no question that many HIV-positive patients have subtle neuropsycho logical abnormalities that may not be captured in a global neuropsychological rating; therefore, we can not determine at this time whether the visual changes discovered in this study are causally related or are associated with subtle changes in the central nervous system. Also, our control group was well matched for age but did include more women than the study group did; however, there is no evidence for gender bias in the tests in which our subjects participated. Interestingly, relatively severe visual function im pairments may occur in patients who are still in the
VOL.124, No. 2
early stages of disease (Figure 3). These patients are considered completely asymptomatic. In contrast, patients who have relatively low CD4 + T-lymphocyte counts may demonstrate normal visual functions in all psychophysical tests, even in the presence of mild neuropsychological impairment (Figure 4). Histopathologic studies of patients with AIDS have shown a loss of up to 50% of fibers in the optic nerve with no evidence of retinitis.25 In addition, others have specu lated that cotton-wool spots produce permanent damage in the nerve-fiber layer of the retina.26 In our study, we therefore eliminated patients who had cotton-wool spots at the time of testing. However, it is possible that some of the patients we tested had had
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
165
wwiww
* .ft FDHTIM LOSSES
wtE m mm FUSE *c emu
A A *" A A * ,34
-lB)-g
»H
rat tin urn
32
31 »
31
33
IIS)'
it
21
■
,
31 31
!! 27
»
l!
-2 -1 1 1-2
2-1 3
I
-1 -9
I -2
30-
10'
-3 1-3
-2
-4
-2
-4
•J ■?
I
■I -3 ■1
I
1
-I -1
1 - 3 - 4
-1
-2
I
-3
-1
■ 1 1 - 1 1
-2 -I -1
-2
I
-3
-4 -I -1
1
1-2-2
1 -3
2 -3
i
®
L
a
©
20/100 p I-
«
©
3'
-1
20/15
20/20
<$ 2o/ra
®
2
■01
7&. j ^ z», .„ * ®
21
1 •
(T k®
W
-4
1 1 - 2 - 3
I -2-2
wo23 23
-S -1 -4 4
-1
^4.
A " ■ !1 A
31
r
1
(B> lit)
M i»u
®
n
V!t
HSTIMS»9ED 344 RNHI
■ 3>
■m
300-
s &
sons 2W30
O
-! -3
rm ■Evnnim
nnniLm itnas "P<
31
e r < 2i
■r<
y: yellow
I.SI
FIGURE 4. Example of a 43-year-old man who had a CD4 + T-lymphocyte count of 11 cells per |xl at the time of testing. His hemoglobin was 13.3 mg/100 ml, hematocrit was 39.9%, and P2-microglobulin was 2.2 mg/L No systemic opportunistic infection was detectable. With a neuropsychological testing score of 5, he was considered mildly neuropsychologically impaired. This patient performed normally on all visual function tests (mean defect, top left) and Glaucoma Hemifield Test on achromatic automated perimetry, mean defect and Glaucoma Hemifield Test on short-wavelength automated perimetry, contrast sensitivity testing (top right), and the Farnsworth-Munsell 100-hue test (bottom left).
cotton-wool spots that were resolved at the time of testing. We also did not eliminate persons with documented history of cotton-wool spots that were resolved at the time of testing because the implicit assumption that the remaining HIV-positive group has no retinal damage may be incorrect. One way of eliminating this is to recruit a large number of patients relatively soon after seroconversion and track them longitudinally at least every 2 to 3 months for 10 to 15 years, which is impractical. Thus, it could be hypoth esized that our findings are solely because of past presence of cotton-wool spots. However, this is very unlikely because we also found visual changes in patients with high CD4 + T-lymphocyte count, for whom the occurrence of cotton-wool spots is rare. Because visual function loss is not correlated to stage of disease, it could be speculated that it is caused directly as a result of HIV infection in the central nervous system or in the visual pathway, including
166
the retina, or both. This is supported by the fact that several pathology studies have shown cortical changes in the areas associated with visual function (for example, the occipital, parietal, and temporal lobes). Another possible reason for vision loss could be potential side effects of antiviral and other drugs that these patients are required to take, even in the early stage of HIV-disease. It is well known that a variety of drugs may affect visual perception.27 Although we screened carefully and did not include in our study patients who took doses of drugs that are known to affect visual perception, there is insufficient data on the side effects of several new antiretroviral drugs or protease inhibitors, such as dideoxyinosine, 2',3'dideoxycytidine, or stavudine, to rule out completely the possibility that the visual function loss we have found is the result of some of these newer drugs. At this point, we conclude that the cause for our findings can neither be attributed simply to declining
AMERICAN JOURNAL OF OPHTHALMOLOGY
AUGUST 1997
health status, clinically evident cortical or cognitive impairment, or known drug side effects, but is more likely attributable to subtle retinal dysfunctions. To elucidate these findings further, we are currently following up this well-characterized group of patients in a longitudinal study at 3-month intervals to investigate the pattern of change in visual function over time using this same battery of tests.
REFERENCES 1. Jabs DA, Green WR, Fox R, Polk BF, Bartlett JG. Ocular manifestations of acquired immune deficiency syndrome. Ophthalmology 1989;96:1092-1099. 2. Iragui VJ, Kalmijn J, Thai LJ, Grant I. Neurological dysfunc tion in asymptomatic HIV-1-infected men: evidence from evoked potentials. HNRC Group. Electroencephalogr Clin Neurophysiol 1994;92:1-10. 3. Malessa R, Agelink MW, Diener HC. Dysfunction of visual pathways in HIV-1 infection. J Neural Sci 1995;130:82-87. 4. Mueller AJ, Berninger T, Matuschke A, Klauss V, Goebel F-D. Elektroophthalmologische Untersuchungen bei HIVPatienten. Fortschr Ophthalmol 1991;88:712-715. 5. Mueller AJ, Berninger T, Geier S, Lachenmayer B, Klauss V. Elektroophthalmologische und psychophysische Untersuch ungen bei fundoskopisch unauffalligen HIV-Positiven und AIDS-Patienten. Folia Ophthalmol 1992;17:5-9. 6. Mueller AJ, Lachenmayr BJ, Eckstein A, Holzl M. Multiplestimulus presentation and voice control in automated perimetry. Ger J Ophthalmol 1992;1:91-95. 7. Plummer DJ, Sample PA, Arevalo JF, et al. Visual field loss in HIV-positive patients without infectious retinopathy. Am J Ophthalmol 1996;122:542-549. 8. Quiceno JI, Capparelli E, Sadun AA, et al. Visual dysfunc tion without retinitis in patients with acquired immunodefi ciency syndrome. Am J Ophthalmol 1992;113:8-13. 9. Heaton R, Velin R, McCutchan J, et al. Neuropsychological impairment in HIV infection: implications for employment. Psychosom Med 1994;56:8-17. 10. Ferris F, Kassoff A, Bresnick GH, Bailey I. New visual acuity charts for clinical research. Am J Ophthalmol 1982;94: 91-96. 11. Sample PA, Martinez GA, Weinreb RN. Short-wavelength automated perimetry without lens density testing. Am J Ophthalmol 1994;118:632-641.
12 Sample P, Weinreb R. Color perimetry for assessment of primary open angle glaucoma. Invest Ophthalmol Vis Sci 1990;31:1869-1875. 13 Sample P, Weinreb R. Progressive color visual field loss in glaucoma. Invest Ophthalmol Vis Sci 1992;33:240-243. 14. Sample P, Taylor J, Martinez G, Lusky M, Weinreb R. Short-wavelength color visual fields in glaucoma suspects at risk. Am J Ophthalmol 1993;115:225-233. 15 Verriest G, Van Laethem J, Uvijls A. A new assessment of the normal ranges of the Farnsworth-Munsell 100-hue test scores. Am J Ophthalmol 1982;93:635-642. 16. Heaton R, Pendleton M. Use of neuropsychological tests to predict adult patients everyday functioning. J Consult Clin Psychol 1981;49:807-821. 17. Heaton R, Grant I, Butters N, et al. The HNRC 500: neuropsychology of HIV infection at different disease stages. Journal of the International Neuropsychological Society 1995;1:231-251. 18. Heijl A, Lindgren G, Olsson J. Normal variability of static perimetric threshold values across the central visual field. Arch Ophthalmol 1987;105:1544-1549. 19 Asman P, Heijl A. Glaucoma Hemifield Test Automated Visual Field Evaluation. Arch Ophthalmol 1992;110: 812-819. 20. Glantz S. Primer of biostatistics, 3rd ed. New York: McGrawHill, 1992. 21 Knoblauch K. On quantifying the bipolarity and axis of the Farnsworth-Munsell 100-hue test. Invest Ophthalmol Vis Sci 1987;28:707-710. 22 Vingrys A, King-Smith P. A quantitative scoring technique for panel test of color vision. Invest Ophthalmol Vis Sci 1988;29:50-63. 23 Geier S, Nohmeier C, Lachenmayr B, Klauss V, Goebel FD. Deficits in perimetric performance in patients with sympto matic human immunodeficiency virus infection or acquired immunodeficiency syndrome. Am J Ophthalmol 1995; 119: 335-344. 24. Iragui VJ, Kalmijn J, Plummer DJ, Sample PA, Trick GL, Freeman WR. Pattern electroretinograms and visual evoked potentials in HIV infection: evidence of asymptomatic retinal and postretinal impairment in the absence of infectious retinopathy. Neurology 1996;47:1452-1456. 25. Tenhula WN, Shizao X, Madigan MC, Heller K, Freeman WR, Sadun AA. Morphometric comparisons of optic nerve axon loss in AIDS: morphometric comparisons. Am J Ophthalmol 1992;113:14-20. 26 Sadun A, Pepose J, Madigan M, Laycock K, Tenhula W, Freeman W. AIDS-related optic neuropathy: a histological, virological and ultrastructural study. Graefes Arch Clin Exp Ophthalmol 1995;233:387-398. 27, Fraunfelder FT. Drug-induced ocular side effects and drug interactions, 3rd ed. Philadelphia: Febiger, 1989.
Authors Interactive™ Comments and questions regarding this article may be addressed to authors via THE JOURNAL'S Web Site. Selected comments, questions, and author responses are posted on Authors Interactive™ at http://www.ajo.com/
VOL.124,
No. 2
VISUAL DYSFUNCTIONS IN HIV-POSITIVE PATIENTS
167