Br J Ophthalmol: first published as 10.1136/bjo.71.7.489 on 1 July 1987. Downloaded from http://bjo.bmj.com/ on December 26, 2020 by guest. Protected by copyright.

British Journal of Ophthalmology, 1987, 71, 489-493

A comparison of Goldmann and Humphrey

automated perimetry in patients with glaucoma
G E TROPE AND R BRITTON
From the Department of Ophthalmology, University of Toronto, Toronto General Hospital, Toronto, Ontario,
Canada

SUMMARY Humphrey automated threshold perimetry (Program 30-2) was performed on 42 eyes
of 25 patients with glaucoma to determine both the sensitivity and specificity of automated
perimetry in detecting glaucomatous visual field defects. Automated perimetry sensitivity was
90-38%, while automated perimetry specificity was 91%. Fifty-two patients and a technician took
part in a survey to determine their preference for either test. Patients generally preferred having
Goldmann perimetry. The technician favoured Humphrey automated perimetry. Program 30-2 on
the automated perimeter took 25% longer to perform than Goldmann perimetry.

There are a number of automated perimeters on the

market. These computerised perimeters all perform
visual field tests using preprogrammed and/or pro-
grammable test logics. Some of these perimeters can
determine retinal threshold sensitivity, and their
performance in patients with glaucoma has been
favourably compared to that of standard, manual
perimetry (Fig. 1).'
The Humphrey automated perimeter is a new and
relatively inexpensive automated perimeter that can
perform both relatively rapid suprathreshold static
perimetry for glaucoma screening and threshold
static perimetry. Threshold perimetry allows for
quantification and therefore follow-up of patients
with glaucomatous visual field defects. Recent reports
have indicated that the Humphrey perimeter can be
favourably compared with conventional manual peri-
meters.2' We are not, however, aware of studies
comparing the sensitivity and specificity of this
automated perimeter with standard Goldmann peri-
metry in patients with glaucoma.
The purpose of this study was two-fold. First we
wished to determine patient and technician pre-
ference for either machine. Secondly, we wished to
determine the sensitivity and specificity of Humphrey
automated perimetry using the Goldmann perimeter
as standard in patients with glaucoma.

Correspondence to G E Trope, FRCS cd, Toronto General Hospital,

Department of Ophthalmology, Glaucoma Division, 200 Elizabeth
Street, EN 5-308, Toronto, Ontario, Canada M5G 2C4. Fig. 1 Humphrey automated perimetry.
489
 _, Br J Ophthalmol: first published as 10.1136/bjo.71.7.489 on 1 July 1987. Downloaded from http://bjo.bmj.com/ on December 26, 2020 by guest. Protected by copyright.
490 GE Trope and R Britton

Methods and materials machine was easier to use? (3) Which perimeter
would you prefer to use in future on this particular
This study was divided into two parts. patient?
Part 1. This consisted of a patient and technician The mean test time for both eyes on each perimeter
survey. Fifty-two glaucomatous patients referred to was recorded by the technician.
our visual field service from a number oflocal ophthal- Part 2. One or both eyes of 25 glaucomatous
mologists were randomly allocated to Goldmann patients attending the Toronto General Hospital
perimetry and Humphrey automated perimetry on Glaucoma Service were tested on both perimeters.
the same day. Program 30-2 was used on the Forty-two eyes in total were tested. Patients were
automated perimeter; Goldmann perimetry was randomly allocated to either test. Some of these
performed by standard methods.4 Both eyes were patients took part in part 1 of the study, but were
tested on both machines. included in part 2 of the study (visual field com-
A number of questions were then asked of each parison study) only if the diagnosis of glaucoma had
patient: (1) Which test did you prefer? (2) With been absolutely established in at least one eye in all
which machine was it easier to keep your eyes cases by one of us (GT).
straight? (3) Have you ever had a visual field test?
The technician was then asked to fill out a question- INSTRUMENTATION
naire answering the following questions: (1) With Goldmann perimetry was performed by standard
which perimeter was fixation superior? (2) Which techniques.4 Humphrey automated perimetry was
C E fy PI
s^G4 L tf L:,
tjI-
;I i T
0fI M L'.1-^ I ,I ER
I Te 8 Le E
-IN ';$P_1T '~~.
ElE: :I E. 'I s i CZ E: , PIs"I-HE ¢4 s :.! 4
Fa~PTAP;LE:l C ENTP4
RL 9. LEt .: 7L
.L CE . 8A
TPR T E5. FULL THREe-:-:HC L '.,'sk
PI- L; ES ID * "- %f
::.

FI X' T I LGEE E 0Ori*

C .
OLIL1£fTr f tt'J'- 044 E1D 42
i.. C1 F I7
FALSE PuC, ERIF4'F 0- I
,c-ALSE NEG EPPC QG'4 0-.
I_;Z_:] I_':JPT!^ ~tj C IFfr

"
t.4
'.':
K K

LLt1:
.211
Fig. 2 A normal Humphrey visual
field. sJWXITHI 4 OL, OF EJ.2ECTE
P4CD. - DEFECTsCEPTH IN 0: j
.
,,
s: 'si?4
12 CIE `2ENTPRL REF L.VEV''

..... ............... .....T

... t: 8: ." z ..... >

..
4 44
* -;- t- t -t °---- .Sm~+sj
wolw-
HTPuIFllIH R
NgXt-
;N'TPUMENI`
s--
's, ^~~~~~~~~~ ..is...^'^
....

.ss..2...5^.s
 Br J Ophthalmol: first published as 10.1136/bjo.71.7.489 on 1 July 1987. Downloaded from http://bjo.bmj.com/ on December 26, 2020 by guest. Protected by copyright.
A comparison of Goldmann and Humphrey automatedperimetry in patients with glaucoma 491

performed with Program 30-2. This program tests the by HeijI and Drance.' 5 The criteria for abnormality
central 300 field at 76 points. Points tested are 6° with the Humphrey automated perimeter were: (1)
apart. Initially, this automated perimeter determines All peripheral test points were excluded to avoid lens
retinal threshold at four primary points-i.e., one in rim artefact. (2) Test points immediately above and
each quadrant 10° from fixation. It does so by below the blind spot were excluded. (3) A 5 db
presenting a stimulus which is very bright (supra- difference between two neighbouring points in con-
threshold). The intensity is then reduced in 4 db junction with a 10 db difference in mirror image
decrements until it is not seen. The stimulus then points between the upper and lower hemifield were
increases in 2 db increments until it is seen again. This considered positive glaucoma visual field defects.
value is recorded as the patient's retinal threshold. (4) One or more points 15 db below the point of
The four initial thresholds are then used as starting highest retinal sensitivity was considered an abnormal
points to determine the thresholds of neighbouring visual field.
points by means of the same bracketing technique. Sensitivity and specificity results were analysed
These points are used in a similar manner to deter- according to standard methods.6 The sensitivity of
mine retinal thresholds of neighbouring points until Humphrey automated perimetry was defined as the
the central 30° area has been fully tested. If a value number of abnormal Humphrey visual fields divided
varies from expected by more than 4 db, it is by the number of Goldmann abnormal visual fields
remeasured automatically. expressed as a percentage. Humphrey automated
The Humphrey automated perimeter Program 30-2 perimetry specificity was determined by dividing the
uses a background illumination of 31-5 apostilbs with number of normal Humphrey automated perimetry
a 4 mm2 stimulus size (equivalent to Goldmann fields by the number of normal Goldmann visual
perimeter size III target); the appropriate near fields expressed as a percentage.
correction for age is used.
The nasal step was not tested in this study as the Results
nasal step program had not yet been installed in the
perimeter. Fixation is monitored both by the tech- Part 1. Table 1 reviews the results of the patients
nician and indirectly by the machine. surveyed. Sixty per cent of the patients preferred
The blind spot is located at the beginning of the Goldmann perimetry. Patients generally thought
test, and during the test approximately 10% of the fixation was easier to maintain with the Goldmann
stimuli are presented in the blind spot. If fixation is perimeter. Fifty four per cent of patients, however,
accurate, the patient will not see any of the stimuli had previously had a Goldmann visual field test.
presented in the region of the blind spot. If the None of the patients had previously undergone
patient responds that he does see a significant automated visual field testing.
number of stimuli presented to the blind spot, a Table 2 reviews the results of the technician
beeping sound is emitted by the machine to alert the survey. Fixation appeared to be similar to the
technician to encourage patient fixation. The number technician during both of the tests. The technician
of times fixation is lost is typed out on the result sheet. found automated perimetry overall a little easier to
The test continues to run even if fixation is very poor. perform than Goldmann perimetry. Technician pre-
The machine also assesses patient reliability by ference strongly favoured Humphrey perimetry.
occasionally producing just the sound associated with The mean test time for both eyes was 26-2 minutes
the light stimulus without actually presenting the
light. If the patient responds to this sound, by
pressing the response button, this is considered a Table 1 Fifty-two patient responses
false positive result. Periodically a suprathreshold
stimulus is presented to the patient that he/she should Goldmann Humphrey Same
easily see. If the patient fails to respond, this indicates Patient preference 60% 17% 23%
lack of attention, a false negative result. Fixation loss Fixation ease 44% 21% 35%
and false positives and negatives are all printed out Previous field test 54% 0% -

on the visual field chart together with a grey scale,

decibel scale, and a scale that indicates deviation Table 2 Technician survey
of retinal threshold from expected decibel values
(Fig. 2). Goldmann Humphrey Same
ANALYSIS OF VISUAL FIELDS Fixation 15% 15% 70%
For part 2 of this study (sensitivity and specificity) we Technical ease 10% 40% 50%
Preference 13% 67% 20%
used decibel criteria modified from papers published
 .'*;¢CSh^-sk!ft,Be:M£ Br J Ophthalmol: first published as 10.1136/bjo.71.7.489 on 1 July 1987. Downloaded from http://bjo.bmj.com/ on December 26, 2020 by guest. Protected by copyright.
492 G E Trope and R Britton
Table 3 Sensitivity and specificity of Humphrey perimetry were 10 normal Humphrey fields compared with 11
using Goldmann perimetry as standard normal Goldmann fields (91%).
Humphrey automated perimetry sensitivity: Discussion
28 lOO0- =90-3%
31 1 The major purpose of this study was to determine the
Humphrey automated perimetry specificity: sensitivity and specificity of Humphrey automated
10 x100=91
= 91% perimetry to detect glaucomatous visual field defects
11 1 using the Goldmann perimeter as the standard. The
results of this section of the study indicate that
Program 30-2 is both highly sensitive and specific for
for Goldmann perimetry and 32*6 minutes for auto- detecting glaucomatous visual field defects. The
mated perimetry. sensitivity and specificity results reported here are
Part2. Table 3 reviews the sensitivity and specificity similar to those reported for the Octopus perimeter
of Humphrey automated perimetry Program 32. using Tubinger perimetry as the standard. '
Sensitivity tests showed that the Humphrey auto- Although the Humphrey automated perimeter is
mated perimeter detected 28 abnormal glauco- both sensitive and specific for glaucomatous visual
matous visual fields out of a total of 31 abnormal field defects, it does have various drawbacks when
fields (90.3%). Specificity tests showed that there compared with Goldmann perimetry. Firstly, auto-
CENTR i-L 30 - 2 THFFeiE: S; H O L E T ESST
STIMULUS II I m WHITE. ECKOND 31.5 R$S NPME
BLIND SPOT CHECK SIZE IrI 10 S8-53-40 BIRTHOATE O2-14-10
FrXATION TRRGT CENTRRL DATE DR-O9-9S5 TIME 112O:52 AM
STRATEGY FULL FRON PRIOR DRTP PUPIL DIAMETER VA
REFERENCE TEST ORTE 09-OR-85 S ?X USED +3.50 DS MX DEG
R I. GHT FIXATI2J4 LOSSES 5.
3
UiESTI-NS ASKED 540
FALSE P05 ERRORS 1 ,17
FPALSE NEC ERPR S G-1l
". FLTTATION 5.9ED0
Q06 TEST TIME OO011724

..... :-

Fig. 3 An abnormal Humphrey NO.

W-ITHIN 4 De OF EXPECTED 1
DEFECT DEPTH IN DS
NO. THRESHOLD SI DRP
visualfield. (NO. SNCV'3R0 TIME
.25 D05 .- CENITRAL REF LE'VEL

SJt 6 | r X

tf v z IO s §
9 , & ) § 8 , 9 4
'
e a 5 a .¢ t .< X 4 ,,,X =

9 oif v.7_e__&.

t N 8 -fv .. w y .

} § 5 -tt ! r

a t r

t § ; ¢v rtS

126~ ~ ~
W_<
w~~~
_ s + - **_ 4 _~~lo- wo
.
H UP H
Evg
RE JL a:L.
N INSTRUMEP-MT' I1 _ tC
ta tc
J
 Br J Ophthalmol: first published as 10.1136/bjo.71.7.489 on 1 July 1987. Downloaded from http://bjo.bmj.com/ on December 26, 2020 by guest. Protected by copyright.
A comparison of Goldmann and Humphrey automated perimetry in patients with glaucoma 493

mated perimetry takes about 25% longer to perform fixation. Static profiles can also be tested. Other
than Goldmann manual perimetry. Further, patients programs also exist which allow one to test areas not
clearly prefer Goldmann perimetry to automated tested by Program 30-2. Clearly, however, the use of
perimetry, but this preference may be biased by the Program 30-2 with one or more of these other test
fact that nearly half the patients had previously programs will add significantly to the already fairly
undergone only Goldmann perimetry. The lengthy test period. These combined programs will,
prolonged test time, the rapid appearance and however, allow for very accurate assessment of the
disappearance of the light stimulus, and the fact that visual field in glaucoma.
the test continues to run despite detection of poor We thank Mrs Mena Cali and other technicians at the Toronto
fixation suggest to us that this automated perimeter in General Hospital for performing visual field tests. We also thank
no way overcomes many of the major obstacles to Dr Clive Mortimer for his help and Mrs Gail Taylor-Cole for typing
accurate visual field assessment such as accurate the manuscript.
fixation, maintenance of concentration, and patient References
fatigue. A major advantage of the Humphrey auto- 1 Heijl A, Drance SM. A clinical comparison of three computerized
mated perimeter, however, is that it can be per- automatic perimeters in the detection of glaucoma defects. Arch
formed by less highly trained technicians. It also Ophthalmol 1981; 99: 832-6.
overcomes the tedium of manual perimetry, and 2 Beck RW, Bergstrom TJ, Lichter PR. A clinical comparison
of visual field testing with a new automated perimeter, the
operator error is completely avoided.' Humphrey field analyzer and the Goldmann perimeter. Ophthal-
It is important to note that the use of only Program mology 1985; 92: 77-82.
30-2 on the Humphrey automated perimeter, 3 Mogil LG, Abramovsky-Kaplan I, Rosenthal S, Podos SM.
although highly sensitive and specific for glauco- Comparison of Goldmann, Humphrey and Octopus perimetry in
glaucoma. Invest Ophthalmol Vis Sci 1985; 26: 225.
matous visual field defects, will not detect visual field 4 Rock WJ, Drance SM, Morgan RW. A modification of Armaly
defects close to fixation nor in the nasal step area visual field screening technique for glaucoma. Can J Ophthalmol
beyond 30°. Furthermore, it will not detect defects 1981; 6: 283-90.
between the 76 points tested in the central 300 (Fig. 5 Heijl A. Computerized perimetry. Trans Ophthalmol Soc UK
1984; 104: 76-87.
3). The machine does, however, have a separate 6 Sommer A. Epidemiology and statistics for the ophthalmologist.
program to test the nasal step area and a macular New York: Oxford University Press, 1980: 11.
threshold program to test the visual field close to Acceptedfor publication 27August 1986.