- Email: [email protected]
Automated Perimetry and Malingerers
Automated Perimetry and Malingerers Can the Humphrey Be Outwitted? James F. G. Stewart, MB, ChB Background: Through detailed strategies and sophisticated analysis, the Humphrey automated visual field analyzer attempts to indicate if visual field loss is artefactual. Can these measures be outwitted by malingerers? Methods: The author investigated the ease with which motivated individuals (such as are malingerers) could simulate visual field defects consistent with organic neurologic disease on the Humphrey visual field analyzer. Visual field test results were analyzed for characteristic features and compared with visual field tests from patients with documented pituitary tumors. Results: Volunteers, given only broad suggestions as to the visual field they were to simulate, produced conSistent, convincing, neurologic-type field defects, according to textbook descriptions of such fields. These plotted fields were only distinguishable from genuine pituitary tumor Humphrey field tests, in that they more convincingly fitted the classic descriptions ~f visual fields seen with chiasmal compression. Conclusions: The author concludes that single routine Humphrey visual field tests do not show malingerers. An incidental finding of this study was the extent to which Humphrey visual fields from patients with genuine neurologic disease contain field defects with characteristics different from those of the (kinetic) visual field test appearances described in the textbooks. Ophthalmology 1995;102:27-32
This investigation arose from experiences with three young female patients seen over a short period in whom diagnosis of functional visual loss was delayed due to the presence of apparently significant visual field defects on automated perimetry, resulting in unnecessary consultations, investigations, and referrals. Aware of the complex testing and retesting strategies of these visual field analyzers, clinicians had placed excessive value on their results. Seemingly reliable abnormal automated visual field tests proved to be artefactual. I decided to try to find out how easy it is to
Originally received: February 28, 1994. Revision accepted: August 23, 1994. From the Wellington Public Hospital, Wellington, New Zealand. Presented at the Annual Meeting of the Ophthalmological Society of New Zealand, Blenheim, October 1992. Reprint requests to James F. G. Stewart, MB, ChB, Department ofOphthalmology, Western Infirmary, Dumbarton Rd, Glasgow Gil 6NT Scotland.
produce fraudulent, convincing visual field defects on the Humphrey field analyzer.
Method Thirty-one volunteers underwent Humphrey automated perimetry of the visual fields in their right eye using 302 or 24-2 full-threshold strategies. Ten of these participants were medical students and 19 were adults with no association with the health services or physiologic sciences. Subjects were healthy and with one exception had not had visual field testing performed previously. Subjects were tested in the usual manner, with correction of refractive errors as indicated. The patient instructions displayed on the operator's screen were read carefully to each candidate, and any queries as to how to perform the test were answered. Then a brief prepared written statement was read to each patient:
27
Ophthalmology
Volume 102, Number 1, January 1995
"We would like you to pretend that you have a partial right upper visual field defect. This means that you can see bright spots but not dim ones up and to the right of the central spot that you are looking at when you perform the test."
No further information or suggestions were made as to how the test should be undertaken. This defect was chosen to induce our subjects to produce abnormal visual fields similar to, and that might be compared with, those associated with central chiasmal compression from pituitary adenomas. (It is not my hypothesis that hysteric bitemporal hemianopia is a significant form offunctional visual field loss.) Fields obtained were analyzed with fields from 22 eyes (right and left eyes of 11 patients) with documented pituitary adenomas. These fields were chosen somewhat arbitrarily. Some of the patients had undergone previous Goldmann and/or Humphrey perimetry. This group, therefore, is biased toward achieving better reliability indices and visual field tests than the experimental subjects. Some endocrine fields represented recurrence of previously treated disease. Some patients with an endocrine disorder had undergone trans-sphenoidal hypophysectomy during their field testing, but none of these showed significant deterioration in their fields after surgery. We asked two general questions in our analysis: 1. Did each field appear to indicate a genuine neurologic defect? 2. Were the feigned visual field defects significantly different from the genuine test results, and if so, in what way? Fields were first graded subjectively based on the grayscale appearance, with secondary reference to the raw data as required. (We do not recommend this approach, but observe that it is widely used, particularly by nonophthalmologists.) We then recorded the reliability indices (fixation losses and false-positive and false-negative results) as percentages, and these were averaged for each test. Short-term fluctuation, a measure of variability of response, was recorded, as was the probability that this calculation was normal (as provided by the Humphrey analyzer). We recorded whether the gray scale and whether the statistical boxes in the total deviation graphs respected the midline. Finally, an attempt was made to convert the total deviation statistical array to a "convincing defect score," a higher score reflecting a more convincing "neurologic" superior temporal field loss, based on principles of visual field defects caused by pituitary adenomas gleaned from standard textbooks. 1.2 These principles were: 1. The field loss caused by a pituitary adenoma begins superiorly and temporally. 2. The defect respects the midline. 3. The defect then proceeds, in the right eye, clockwise. The "convincing defect score" was calculated from the Humphrey total deviation statistical array as follows. Three values were calculated from the arrays. Value A
28
was the number of probability boxes in the superotemporal quadrant. Value B was the number of times probability boxes situated in the superotemporal quadrant were at the same level of probability or greater (i.e., more densely shaded) than the neighboring loci inferiorly, and inferonasally (a possible score of2 points). A normal locus adjacent to a normal locus did not score a point-O decibels adjacent to always scored a point despite the probability boxes. Values A and B were added together, and from them value C (the number of probability boxes in the nasal hemisphere) was subtracted. (Note: all fields reflected relatively early field losses.) For 30-2 tests, only the central boxes equivalent to the 24-2 test were used to allow comparison between 24-2 and 30-2 tests. Data were compared using the Student's t test.
Results Table 1 shows the results of sUbjective grading of the grayscale appearance. Sixty-five percent of our volunteers produced field defects that appeared strongly probably to be neurologic. Interestingly, only 23% of the real cases produced fields with this gray-scale appearance. Table 2 compares reliability indices. There was little difference between the two groups; for example, "average reliability indices" ranged from 0% to 23% (standard deviation, 6%) independently in each group, despite the greater visual field testing experience in the endocrine group. Any differences were far from significant. Table 3 compares short-term fluctuation. There was no significant difference between these two similar means. Slightly more of the patients with tumors had abnormally high short-term fluctuation by the Humphrey criteria. Visual fields in approximately half of the malingerer group respected the midline whether judged by gray-scale appearance or statistical boxes versus only those in approximately one quarter of the tumor group, as shown in Table 4. Fifty-five percent ofthe fields in the tumor group definitely did not respect the midline, versus only 13% to 16% of the volunteers' fields. The malingerers produced significantly more "neurologic-type" field defects, as shown in Table 5 (from analysis of the total deviation probability graphs by the "convincing defect scores"). Comparison of results from the medical student volunteers and the nonmedical background "intelligent amateurs" showed no significant difference in any result. Table 1. Gray Scale Appearance Appearance
Malingerers
Endocrine Patients
Normal Abnormal, not neurologic Possible neurologic defect Strongly probably neurologi~
13% 10% 13% 65%
14% 36% 27% 23%
Stewart . Automated Perimetry and Malingerers
Table 2. Reliability Indices*
Table 4. Respect for the Midline*
Indices
Malingerers
Endocrine Patients
Fixation losses False-positives False-negatives Average of reliability indices
8.4% 1.3% 4.5% 5.2%
7.0% 1.5% 4.0% 4.8%
" Converted to percentages and then averaged.
Figure 1 shows a number of visual fields from this project to illustrate how strikingly "neurologic" the faked fields could appear, and how similar they could be to those with genuine pathologic findings.
Discussion Abnormality of the visual fields is one of the two common functional defects seen by ophthalmologists, second in frequency only to decreased visual acuity. Patients may have an abnormality relating to the visual fields, or it may be one of several findings in someone with generalized visual defects. In one study,3 abnormal visual fields were found in 51 % patients with functional visual loss. Our examination techniques (such as holding fingers to be counted to either side of a patient) encourage hemispheric, quadrantic, or peripheral field loss. Currently, many patients with these types of defects will have formal field testing performed by automated perimetry after only crude confrontational field delineation. I wanted to know whether functional disease can be diagnosed in the presence of a convincing abnormal Humphrey visual field test. How easily and how completely can the Humphrey be outwitted? Patients with functional disease are traditionally divided into malingerers and hysterics. Thompson 4 argues that, rather, there is a continuum from what he describes as "suggestible innocents" through to deliberate malingerers. My subjects were "deliberate malingerers," but I believe that my conclusions probably apply to all functional disease. This, however, has not been shown. There are many reasons to anticipate that the Humphrey field analyzer would be difficult to cheat. Threshold testing is compared with normal "hills of vision" in a
Table 3. Short-term Fluctuation Fluctuation
Malingerers
Endocrine Patients
Mean Abnormally high"
1.90 20%
1.97 23%
" Humphrey probabilities: P < 0.05 short-term fluctuation is normal.
Yes No Do not know Inot applicable
Malingerers
Endocrine Patients
52%-55% 13%-16% 32%
23%-27% 55% 18%-23%
" Gray scale and total deviation statistical array assessed independently, results combined.
most sophisticated way. Most points are tested by a number of flashes spread over several minutes, interspersed with stimuli at other points. The test is difficult, and requires considerable concentration, allowing little energy left for scheming. Fixation losses and false-negative and false-positive responses are all specifically sought. In particular, the short-term fluctuation test measures the consistency of responses. Ten preselected points are tested twice, and the results are compared. Values have been established for the healthy population, and so an individual's result can be given a probability as to whether it is normal. Younge,S in a wide-ranging thesis concerning automated perimetry (with the Octopus perimeter) in neuroophthalmology, stresses the above features of automated perimetry and writes: "It is much more difficult with the rapid fire, short duration stimuli to be sufficiently consistent in producing an abnormal field deliberately. Falsepositive and false-negative responses, as well as gross inconsistencies in fluctuations alert the observer to the nature of the patient's performance." (p 940) In my experience, this conception that automated perimetry is too devious to be easily out-maneuvered by the malingerer is widespread. I have described the methods by which the Humphrey carefully assesses the consistency of a subject's responses, but on the other hand the analyzer must allow for human variability and the fact that most patients are elderly. In addition, in disease states such as glaucoma it has been clearly shown that short-term fluctuation rises. A perimeter that is too sensitive loses its clinical use. I found that the most obvious example ofthis "blunting of sensitivity" of the Humphrey analyzer was the fact that a group of
Table 5. Convincing Defect Score· Malingererst Mean Significant
Endocrine Patients
18.3%
11.8% P < 0.05
" See text for details.
t The malingerer group produced visual fields that more convincingly fitted standard descriptions of the field defects associated with pituitary adenomas.
29
Ophthalmology
Volume 102, Number 1, January 1995 10TII. IIVlRTTOO
TOT~. [iEVIMT I I~
-\
!.
"..
I
Q
"I ::
I
I
" II
• ••• •••
lit. ,~
Il
" •
••
••
E~
"I !
t
"r "
I
*
"
I" {;
"
"
!, ,i
",:"""
"I::
Q.
"I "
tI
"J. "
I
T·.:: •••
"I"
•
"
!II
"
* ",
OEVICTICtl
. j:
I
"t "
i" I
r -
'IVlRT!(I~
l •• • ,I." ,!
"
,
TOIIl
.'::::::::: ::: ::::
I
,I
""
"'
Q.
.. :::::::::::::::
'ON,
"
Q.
TOTll
IIVIRTTOO
1
fr
!II
1'<', " .
I
10111. O£'IIIlTI(J.
....
.[1 : : •
t · ••• • •
I
'\'
, ""
•
~J'+-i'- -+-"-"-"-I----+--+--=-+---t ......... ...
..........
...... - .. - ..... ..... . "
-:::::::::::::::::: :
, '. :::
... .. -- ....
,
.
"t "
,
....
••
,
!
Figure 1. Which is the genuine pituitary tumor visual field? Gray-scale and total deviation statistical plots from standard Humphrey 24-2 or 30-2 visual field test printouts, Figures with asterisks indicate visual field tests from patients with endocrine disease, All the other figures are feigned fields, (Fig 1 continues,)
30
Stewart . Automated Perimetry and Malingerers
IOIf!.
.. I···· I
OEVIHTlCli
&'
r>-r •• I
I •••
" " ' t"1,··
. . . ,! ;:?, " : . . •
,.
* I
•
I
:
•••
I! • •
''i. •
.
::
~:~
....
·1 &'
&'
••• " ••••• •• • • •••• ,r . "
*
Figure 1 (continued).
my "fakers" produced visual field depressions just visible on the gray scale but too subtle to be declared abnormal by the Humphrey's statistical analysis. My findings in essence were that feigned visual field tests can have appearances strongly suggestive of significant organic neurologic disease, and that they have no characteristics that distinguish them from genuinely abnormal test results. In addition, I found that patients with pituitary adenomas produced very varied field results. Humphrey static threshold perimetry cannot be interpreted in the same way as the Goldmann or tangent screen kinetic perimetry with which we are familiar. For instance, although careful manual perimetry in pituitary disease shows that field defects respect the midline, the Humphrey strategies 24-2 and 30-2, with a limited number of preset test points, do not show this phenomenon. This is illustrated by examples of visual fields used in this study obtained from patients with documented pituitary tumors shown in Figure 2. The characteristics of neurologic field defects as recorded by automated perimetry are yet to be fully described. Mills6 describes many of the problems in interpreting automated perimetry in patients with neurologic disease, warns against using automated perimetry if there is a suspicion of functional disease, but still advocates using a test similar to the Humphrey 30-2 on most patients. Conversations suggest that most neuro-ophthalmologists are only too aware of the problems with automated perimetry, even though very little on the subject has appeared in the literature. Equally, I frequently have witnessed general ophthalmologists struggling to interpret neurologic automated fields, or even to decide whether they areabnormal. The poor respect for the midline shown by our Humphrey pituitary tumor fields makes a mockery of the so-called "neurologic" strategies on these machines that test only either side of the vertical midline. The diagnosis of functional visual defects is not addressed by the Humphrey field analyzer, and I have shown that routine test strategies can produce fields strongly suggestive of organic neuro-ophthalmic disease. Therefore, if functional field loss is in the differential diagnosis, Humphrey automated perimetry is not the appropriate
way to assess the visual fields. Careful manual perimetry produces several typical types offunctional field loss such as "tubular" or "spiral" fields, or reversal and crossing of isopters. Tangent screen testing at more than one distance (carefully adjusting target size to maintain equivalence), testing with a Juler's-type projection target and red/green goggles, or mapping the visual field with both eyes open can produce inconsistent or nonphysiologic fields. Use of several isopters (and a high degree of suspicion) on the Goldmann perimeter can produce documentary evidence of nonphysiologic fields. The Goldmann perimetrist needs to be aware of what is being sought, because inconsistencies easily can be overlooked or "averaged." We tested our subjects only once, and it may be much more difficult to reproduce consistent artefactual field results. It is always good to repeat a Humphrey field test that is proving difficult to interpret. Visual field assessments are SUbjective tests and always must be interpreted in the wider clinical context of the patient's complaints and other findings on examination. Finally, Humphrey field analyzers are becoming increasingly prevalent, and there is strong inducement (if not only to allay their cost) to use them as a routine screening tool. They are appearing in optometrists' offices. We are going to be faced with a growing number of patients whose only abnormality is an abnormal automated visual field, but this test result is an impressive document, printed clearly, referring to "reliability indices," and packed with statistics-a difficult piece of evidence to ignore. I show that such tests may be entirely factitious. Routine use of this machine as a screening device may create a new epidemic of functional disease. In conclusion, malingerers without previous experience of visual field testing easily can produce Humphrey visual fields that appear to reliably reflect organic patterns of visual field loss, and other methods offield analysis should be used if functional visual loss is suspected. In addition, genuine neurologic field defects on the Humphrey analyzer have different characteristics from those produced by the kinetic field tests with which we are familiar, and these features are yet to be fully described.
31
Ophthalmology
Volume 102, Number 1, January 1995
TOHIL
Mill
DEV!HTICtl
IIYIATlotl
....
~
••• ••••• • -r
It
::
I
. ..... 1..i··· i..... ... ..•. : :
i •••• ,
III III
.. ...
• ".111 .. :
~
TOTI(
., .1
I
r ~I_ r:'
\111
.d . . . . .1 :: ,-"
.
-r- . •
• • • •1 • • • •
L
::
{~
T. III 11
:.-;:
•••• ~
I
..
•• ... •••••
......... .. ..,.... ~~
i
•
::1
I
", !~
. ! III
~~
I
t·
~Iq
fit
~
It
•
Figure 2. Automated perimetry of patients with documented pituitary tumors. Gray-scale and total deviation statistical plots from standard Humphrey 24-2 or 30-2 visual field test printouts. Top two figures on left, this field follows the classic description of the field defect associated with a pituitary adenoma. Top two figures on right, many fields, however, are far less clearcut to interpret, such as this one. (See also Fig 1, page 30.) In the remaining figures, the Humphrey fields do not, in particular, "respect the midline."
References I. Glaser JS. Topical diagnosis: the optic chiasm. In: Tasman W, Jaeger EA, eds. Duane's Clinical Ophthalmology, rev ed. Philadelphia: JB Lippincott, 1993; vol. 2, chap. 6, 2-6. 2. Miller NR. Walsh and Hoyt's Clinical Neuro-Ophthalmology, 4th ed. Vol 1. Baltimore: Williams & Wilkins, 1982;120-2.
32
3. Keltner JL, May WN, Johnson CA, Post RB. The California syndrome. Functional visual complaints with potential economic impact. Ophthalmology 1985;92:427-35. 4. Thompson HS. Functional visual loss. Am J Ophthalmol 1985; 100:209-13. 5. Younge BR. Computer assisted perimetry in visual pathway disease: neuro-ophthalmic applications. Tr Am Ophth Soc 1984;82:899-942. 6. Mills RP. Automated perimetry in neuro-ophthalmology. Int Ophthalmol Clin 1991;31(4):51-70.
This investigation arose from experiences with three young female patients seen over a short period in whom diagnosis of functional visual loss was delayed due to the presence of apparently significant visual field defects on automated perimetry, resulting in unnecessary consultations, investigations, and referrals. Aware of the complex testing and retesting strategies of these visual field analyzers, clinicians had placed excessive value on their results. Seemingly reliable abnormal automated visual field tests proved to be artefactual. I decided to try to find out how easy it is to
Originally received: February 28, 1994. Revision accepted: August 23, 1994. From the Wellington Public Hospital, Wellington, New Zealand. Presented at the Annual Meeting of the Ophthalmological Society of New Zealand, Blenheim, October 1992. Reprint requests to James F. G. Stewart, MB, ChB, Department ofOphthalmology, Western Infirmary, Dumbarton Rd, Glasgow Gil 6NT Scotland.
produce fraudulent, convincing visual field defects on the Humphrey field analyzer.
Method Thirty-one volunteers underwent Humphrey automated perimetry of the visual fields in their right eye using 302 or 24-2 full-threshold strategies. Ten of these participants were medical students and 19 were adults with no association with the health services or physiologic sciences. Subjects were healthy and with one exception had not had visual field testing performed previously. Subjects were tested in the usual manner, with correction of refractive errors as indicated. The patient instructions displayed on the operator's screen were read carefully to each candidate, and any queries as to how to perform the test were answered. Then a brief prepared written statement was read to each patient:
27
Ophthalmology
Volume 102, Number 1, January 1995
"We would like you to pretend that you have a partial right upper visual field defect. This means that you can see bright spots but not dim ones up and to the right of the central spot that you are looking at when you perform the test."
No further information or suggestions were made as to how the test should be undertaken. This defect was chosen to induce our subjects to produce abnormal visual fields similar to, and that might be compared with, those associated with central chiasmal compression from pituitary adenomas. (It is not my hypothesis that hysteric bitemporal hemianopia is a significant form offunctional visual field loss.) Fields obtained were analyzed with fields from 22 eyes (right and left eyes of 11 patients) with documented pituitary adenomas. These fields were chosen somewhat arbitrarily. Some of the patients had undergone previous Goldmann and/or Humphrey perimetry. This group, therefore, is biased toward achieving better reliability indices and visual field tests than the experimental subjects. Some endocrine fields represented recurrence of previously treated disease. Some patients with an endocrine disorder had undergone trans-sphenoidal hypophysectomy during their field testing, but none of these showed significant deterioration in their fields after surgery. We asked two general questions in our analysis: 1. Did each field appear to indicate a genuine neurologic defect? 2. Were the feigned visual field defects significantly different from the genuine test results, and if so, in what way? Fields were first graded subjectively based on the grayscale appearance, with secondary reference to the raw data as required. (We do not recommend this approach, but observe that it is widely used, particularly by nonophthalmologists.) We then recorded the reliability indices (fixation losses and false-positive and false-negative results) as percentages, and these were averaged for each test. Short-term fluctuation, a measure of variability of response, was recorded, as was the probability that this calculation was normal (as provided by the Humphrey analyzer). We recorded whether the gray scale and whether the statistical boxes in the total deviation graphs respected the midline. Finally, an attempt was made to convert the total deviation statistical array to a "convincing defect score," a higher score reflecting a more convincing "neurologic" superior temporal field loss, based on principles of visual field defects caused by pituitary adenomas gleaned from standard textbooks. 1.2 These principles were: 1. The field loss caused by a pituitary adenoma begins superiorly and temporally. 2. The defect respects the midline. 3. The defect then proceeds, in the right eye, clockwise. The "convincing defect score" was calculated from the Humphrey total deviation statistical array as follows. Three values were calculated from the arrays. Value A
28
was the number of probability boxes in the superotemporal quadrant. Value B was the number of times probability boxes situated in the superotemporal quadrant were at the same level of probability or greater (i.e., more densely shaded) than the neighboring loci inferiorly, and inferonasally (a possible score of2 points). A normal locus adjacent to a normal locus did not score a point-O decibels adjacent to always scored a point despite the probability boxes. Values A and B were added together, and from them value C (the number of probability boxes in the nasal hemisphere) was subtracted. (Note: all fields reflected relatively early field losses.) For 30-2 tests, only the central boxes equivalent to the 24-2 test were used to allow comparison between 24-2 and 30-2 tests. Data were compared using the Student's t test.
Results Table 1 shows the results of sUbjective grading of the grayscale appearance. Sixty-five percent of our volunteers produced field defects that appeared strongly probably to be neurologic. Interestingly, only 23% of the real cases produced fields with this gray-scale appearance. Table 2 compares reliability indices. There was little difference between the two groups; for example, "average reliability indices" ranged from 0% to 23% (standard deviation, 6%) independently in each group, despite the greater visual field testing experience in the endocrine group. Any differences were far from significant. Table 3 compares short-term fluctuation. There was no significant difference between these two similar means. Slightly more of the patients with tumors had abnormally high short-term fluctuation by the Humphrey criteria. Visual fields in approximately half of the malingerer group respected the midline whether judged by gray-scale appearance or statistical boxes versus only those in approximately one quarter of the tumor group, as shown in Table 4. Fifty-five percent ofthe fields in the tumor group definitely did not respect the midline, versus only 13% to 16% of the volunteers' fields. The malingerers produced significantly more "neurologic-type" field defects, as shown in Table 5 (from analysis of the total deviation probability graphs by the "convincing defect scores"). Comparison of results from the medical student volunteers and the nonmedical background "intelligent amateurs" showed no significant difference in any result. Table 1. Gray Scale Appearance Appearance
Malingerers
Endocrine Patients
Normal Abnormal, not neurologic Possible neurologic defect Strongly probably neurologi~
13% 10% 13% 65%
14% 36% 27% 23%
Stewart . Automated Perimetry and Malingerers
Table 2. Reliability Indices*
Table 4. Respect for the Midline*
Indices
Malingerers
Endocrine Patients
Fixation losses False-positives False-negatives Average of reliability indices
8.4% 1.3% 4.5% 5.2%
7.0% 1.5% 4.0% 4.8%
" Converted to percentages and then averaged.
Figure 1 shows a number of visual fields from this project to illustrate how strikingly "neurologic" the faked fields could appear, and how similar they could be to those with genuine pathologic findings.
Discussion Abnormality of the visual fields is one of the two common functional defects seen by ophthalmologists, second in frequency only to decreased visual acuity. Patients may have an abnormality relating to the visual fields, or it may be one of several findings in someone with generalized visual defects. In one study,3 abnormal visual fields were found in 51 % patients with functional visual loss. Our examination techniques (such as holding fingers to be counted to either side of a patient) encourage hemispheric, quadrantic, or peripheral field loss. Currently, many patients with these types of defects will have formal field testing performed by automated perimetry after only crude confrontational field delineation. I wanted to know whether functional disease can be diagnosed in the presence of a convincing abnormal Humphrey visual field test. How easily and how completely can the Humphrey be outwitted? Patients with functional disease are traditionally divided into malingerers and hysterics. Thompson 4 argues that, rather, there is a continuum from what he describes as "suggestible innocents" through to deliberate malingerers. My subjects were "deliberate malingerers," but I believe that my conclusions probably apply to all functional disease. This, however, has not been shown. There are many reasons to anticipate that the Humphrey field analyzer would be difficult to cheat. Threshold testing is compared with normal "hills of vision" in a
Table 3. Short-term Fluctuation Fluctuation
Malingerers
Endocrine Patients
Mean Abnormally high"
1.90 20%
1.97 23%
" Humphrey probabilities: P < 0.05 short-term fluctuation is normal.
Yes No Do not know Inot applicable
Malingerers
Endocrine Patients
52%-55% 13%-16% 32%
23%-27% 55% 18%-23%
" Gray scale and total deviation statistical array assessed independently, results combined.
most sophisticated way. Most points are tested by a number of flashes spread over several minutes, interspersed with stimuli at other points. The test is difficult, and requires considerable concentration, allowing little energy left for scheming. Fixation losses and false-negative and false-positive responses are all specifically sought. In particular, the short-term fluctuation test measures the consistency of responses. Ten preselected points are tested twice, and the results are compared. Values have been established for the healthy population, and so an individual's result can be given a probability as to whether it is normal. Younge,S in a wide-ranging thesis concerning automated perimetry (with the Octopus perimeter) in neuroophthalmology, stresses the above features of automated perimetry and writes: "It is much more difficult with the rapid fire, short duration stimuli to be sufficiently consistent in producing an abnormal field deliberately. Falsepositive and false-negative responses, as well as gross inconsistencies in fluctuations alert the observer to the nature of the patient's performance." (p 940) In my experience, this conception that automated perimetry is too devious to be easily out-maneuvered by the malingerer is widespread. I have described the methods by which the Humphrey carefully assesses the consistency of a subject's responses, but on the other hand the analyzer must allow for human variability and the fact that most patients are elderly. In addition, in disease states such as glaucoma it has been clearly shown that short-term fluctuation rises. A perimeter that is too sensitive loses its clinical use. I found that the most obvious example ofthis "blunting of sensitivity" of the Humphrey analyzer was the fact that a group of
Table 5. Convincing Defect Score· Malingererst Mean Significant
Endocrine Patients
18.3%
11.8% P < 0.05
" See text for details.
t The malingerer group produced visual fields that more convincingly fitted standard descriptions of the field defects associated with pituitary adenomas.
29
Ophthalmology
Volume 102, Number 1, January 1995 10TII. IIVlRTTOO
TOT~. [iEVIMT I I~
-\
!.
"..
I
Q
"I ::
I
I
" II
• ••• •••
lit. ,~
Il
" •
••
••
E~
"I !
t
"r "
I
*
"
I" {;
"
"
!, ,i
",:"""
"I::
Q.
"I "
tI
"J. "
I
T·.:: •••
"I"
•
"
!II
"
* ",
OEVICTICtl
. j:
I
"t "
i" I
r -
'IVlRT!(I~
l •• • ,I." ,!
"
,
TOIIl
.'::::::::: ::: ::::
I
,I
""
"'
Q.
.. :::::::::::::::
'ON,
"
Q.
TOTll
IIVIRTTOO
1
fr
!II
1'<', " .
I
10111. O£'IIIlTI(J.
....
.[1 : : •
t · ••• • •
I
'\'
, ""
•
~J'+-i'- -+-"-"-"-I----+--+--=-+---t ......... ...
..........
...... - .. - ..... ..... . "
-:::::::::::::::::: :
, '. :::
... .. -- ....
,
.
"t "
,
....
••
,
!
Figure 1. Which is the genuine pituitary tumor visual field? Gray-scale and total deviation statistical plots from standard Humphrey 24-2 or 30-2 visual field test printouts, Figures with asterisks indicate visual field tests from patients with endocrine disease, All the other figures are feigned fields, (Fig 1 continues,)
30
Stewart . Automated Perimetry and Malingerers
IOIf!.
.. I···· I
OEVIHTlCli
&'
r>-r •• I
I •••
" " ' t"1,··
. . . ,! ;:?, " : . . •
,.
* I
•
I
:
•••
I! • •
''i. •
.
::
~:~
....
·1 &'
&'
••• " ••••• •• • • •••• ,r . "
*
Figure 1 (continued).
my "fakers" produced visual field depressions just visible on the gray scale but too subtle to be declared abnormal by the Humphrey's statistical analysis. My findings in essence were that feigned visual field tests can have appearances strongly suggestive of significant organic neurologic disease, and that they have no characteristics that distinguish them from genuinely abnormal test results. In addition, I found that patients with pituitary adenomas produced very varied field results. Humphrey static threshold perimetry cannot be interpreted in the same way as the Goldmann or tangent screen kinetic perimetry with which we are familiar. For instance, although careful manual perimetry in pituitary disease shows that field defects respect the midline, the Humphrey strategies 24-2 and 30-2, with a limited number of preset test points, do not show this phenomenon. This is illustrated by examples of visual fields used in this study obtained from patients with documented pituitary tumors shown in Figure 2. The characteristics of neurologic field defects as recorded by automated perimetry are yet to be fully described. Mills6 describes many of the problems in interpreting automated perimetry in patients with neurologic disease, warns against using automated perimetry if there is a suspicion of functional disease, but still advocates using a test similar to the Humphrey 30-2 on most patients. Conversations suggest that most neuro-ophthalmologists are only too aware of the problems with automated perimetry, even though very little on the subject has appeared in the literature. Equally, I frequently have witnessed general ophthalmologists struggling to interpret neurologic automated fields, or even to decide whether they areabnormal. The poor respect for the midline shown by our Humphrey pituitary tumor fields makes a mockery of the so-called "neurologic" strategies on these machines that test only either side of the vertical midline. The diagnosis of functional visual defects is not addressed by the Humphrey field analyzer, and I have shown that routine test strategies can produce fields strongly suggestive of organic neuro-ophthalmic disease. Therefore, if functional field loss is in the differential diagnosis, Humphrey automated perimetry is not the appropriate
way to assess the visual fields. Careful manual perimetry produces several typical types offunctional field loss such as "tubular" or "spiral" fields, or reversal and crossing of isopters. Tangent screen testing at more than one distance (carefully adjusting target size to maintain equivalence), testing with a Juler's-type projection target and red/green goggles, or mapping the visual field with both eyes open can produce inconsistent or nonphysiologic fields. Use of several isopters (and a high degree of suspicion) on the Goldmann perimeter can produce documentary evidence of nonphysiologic fields. The Goldmann perimetrist needs to be aware of what is being sought, because inconsistencies easily can be overlooked or "averaged." We tested our subjects only once, and it may be much more difficult to reproduce consistent artefactual field results. It is always good to repeat a Humphrey field test that is proving difficult to interpret. Visual field assessments are SUbjective tests and always must be interpreted in the wider clinical context of the patient's complaints and other findings on examination. Finally, Humphrey field analyzers are becoming increasingly prevalent, and there is strong inducement (if not only to allay their cost) to use them as a routine screening tool. They are appearing in optometrists' offices. We are going to be faced with a growing number of patients whose only abnormality is an abnormal automated visual field, but this test result is an impressive document, printed clearly, referring to "reliability indices," and packed with statistics-a difficult piece of evidence to ignore. I show that such tests may be entirely factitious. Routine use of this machine as a screening device may create a new epidemic of functional disease. In conclusion, malingerers without previous experience of visual field testing easily can produce Humphrey visual fields that appear to reliably reflect organic patterns of visual field loss, and other methods offield analysis should be used if functional visual loss is suspected. In addition, genuine neurologic field defects on the Humphrey analyzer have different characteristics from those produced by the kinetic field tests with which we are familiar, and these features are yet to be fully described.
31
Ophthalmology
Volume 102, Number 1, January 1995
TOHIL
Mill
DEV!HTICtl
IIYIATlotl
....
~
••• ••••• • -r
It
::
I
. ..... 1..i··· i..... ... ..•. : :
i •••• ,
III III
.. ...
• ".111 .. :
~
TOTI(
., .1
I
r ~I_ r:'
\111
.d . . . . .1 :: ,-"
.
-r- . •
• • • •1 • • • •
L
::
{~
T. III 11
:.-;:
•••• ~
I
..
•• ... •••••
......... .. ..,.... ~~
i
•
::1
I
", !~
. ! III
~~
I
t·
~Iq
fit
~
It
•
Figure 2. Automated perimetry of patients with documented pituitary tumors. Gray-scale and total deviation statistical plots from standard Humphrey 24-2 or 30-2 visual field test printouts. Top two figures on left, this field follows the classic description of the field defect associated with a pituitary adenoma. Top two figures on right, many fields, however, are far less clearcut to interpret, such as this one. (See also Fig 1, page 30.) In the remaining figures, the Humphrey fields do not, in particular, "respect the midline."
References I. Glaser JS. Topical diagnosis: the optic chiasm. In: Tasman W, Jaeger EA, eds. Duane's Clinical Ophthalmology, rev ed. Philadelphia: JB Lippincott, 1993; vol. 2, chap. 6, 2-6. 2. Miller NR. Walsh and Hoyt's Clinical Neuro-Ophthalmology, 4th ed. Vol 1. Baltimore: Williams & Wilkins, 1982;120-2.
32
3. Keltner JL, May WN, Johnson CA, Post RB. The California syndrome. Functional visual complaints with potential economic impact. Ophthalmology 1985;92:427-35. 4. Thompson HS. Functional visual loss. Am J Ophthalmol 1985; 100:209-13. 5. Younge BR. Computer assisted perimetry in visual pathway disease: neuro-ophthalmic applications. Tr Am Ophth Soc 1984;82:899-942. 6. Mills RP. Automated perimetry in neuro-ophthalmology. Int Ophthalmol Clin 1991;31(4):51-70.