Cone Photopigment BleachingAbnormalities in Diabetes

Ann E. Eisner, Stephen A. Burns, Louis A. Lobes, Jr., and Bernard H. Doft

We have used a color-matching technique to obtain estimates of the optical density of cone photopigmentsas a function of retinal illuminance in patients with insulin-dependent diabetes mellitus (IDDM). Wefound that the half-bleach illuminance of some patients is abnormally high. That is, it takes more lightto bleach an equivalent amount of photopigment in these patients. Since low illuminance color matchesfor these patients are normal, this implies that these patients have normal amounts of photopigment,but the photopigment is not bleaching normally. This result clearly points to abnormalities in the outerretina of these diabetic patients. The most likely causes of this abnormality are either decreases in theability of the cones to absorb light, or an increased rate of regeneration of the cone photopigments.Invest Ophthalmol Vis Sci 28:718-724, 1987

A noninvasive technique for measuring the bleach-ing of the cone photopigments has been developed.1'2

This article describes results obtained when this tech-nique is applied to patients with insulin-dependent di-abetes mellitus. Previous reports have measured ab-normalities in both light and dark adaptation in dia-betic patients.3"7 However, it is difficult to determinewhether the previously-measured abnormalities are dueto dysfunction in the photoreceptors, or to abnormal-ities of neural adaptation mechanisms.8'9 Since thereis ample evidence for diabetes-related abnormalities atmultiple loci, a technique capable of measuring ad-aptational abnormalities at a single locus in the visualsystem is desirable.

A technique based on color-matching, has been de-veloped that is sensitive to the bleaching of the conephotopigments (see previous article1). This techniqueis based on the fact that the absorption spectrum of apigment depends on the concentration of the pig-ment.10 Thus when the concentration of photopigmentin the retina is decreased by bleaching, the photore-ceptors undergo a change in their spectral sensitivity.This change in spectral sensitivity causes a predictablechange in the color-match. We have shown that color-

From the Department of Ophthalmology, University of Pittsburgh,Pittsburgh, Pennsylvania.

Supported in part by NIH-NEI R01-EY04395 and S07 RR05889and by a departmental grant from Research to Prevent Blindness,New York, New York.

Presented in part at the annual meeting of the Association forResearch in Vision and Ophthalmology, 1984, Sarasota, Florida.

Submitted for publication: December 26, 1985.Reprint requests: Ann E. Eisner, Department of Ophthalmology,

Eye and Ear Institute of Pittsburgh, 203 Lothrop Street, Pittsburgh,PA 15213.

matching is a reliable technique for measuring pho-topigment bleaching in normal observers.2

This article presents data showing that more light isrequired to bleach the photopigment of some diabeticpatients, yet these patients have normal cone photo-pigment optical densities. These results limit the pos-sible explanations of the source of the abnormalitymeasured in these patients.

Materials and Methods

Twenty-two volunteers with insulin-dependent di-abetes mellitus (IDDM) were recruited from the pa-tients at the Department of Ophthalmology, Universityof Pittsburgh. All patients had visual acuity 20/40 orbetter, and no visually significant media opacificationsin the eye tested. We did not perform fluorescein an-giography on all patients, so this study does not addressthe possible functional changes associated with para-foveal capillary infarction. Patients with perifovealepiretinal membranes were excluded from this study.Patients were 14-52 years of age. On the basis of retinalexams, the patients were placed into three categories:1) no retinopathy, 2) retinopathy (including both pre-proliferative and proliferative), and 3) patients havingundergone pan-retinal photocoagulation for prolifer-ative retinopathy according to the DRS criteria." Sincethere were wide-spread differences in our test resultsamong patients with similar retinopathy classifications,additional grades were deemed unnecessary. Tables 1and 2 show the distribution of patients by sex, age, andduration, as well as results that are discussed below.The mean age of both the diabetic patients and the 27normal observers (previous paper) was 30. Testing ofthe normal and diabetic population occurred during

718

No. 4 BLEACHING ABNORMALITIES IN OABETES / Eisner er ol. 719

Table 1. Clinical data and half-bleach illuminance

Subject/sex

1/F1/F2/M3/M4/F5/M6/F6/F7/F8/F8/F8/F9/F

10/F11/M12/M12/M12/M13/F14/F14/F14/F14/F15/F16/M17/F18/M7/F

10/F11/M12/M19/M14/F20/F14/F14/F21/F21/F20/F22/M

Group

22222222222222222223333333333333

Age(yr)

14151516252728282020212123252628282834353535353646505221252728293435353536363840

Duration(yr)

44

1008799

1417181813102315151524141415151728372414102415212423242418182529

VA

20/1520/1520/2020/1520/1520/1520/2020/2020/2020/2020/2520/2020/2020/1520/2520/2520/2520/1520/2520/1520/2520/1520/2520/2020/2520/4020/2020/2020/3020/2520/2520/3020/3020/2020/3020/2520/4020/4020/1520/20

Macularedema

YYY

YY

Y

Y

YYY

Y

Io(log td)

4.364.444.364.334.62*4.446.09**4.65**4.555.00**4.404.64**4.434.57*4.474.98**4.474.73**4.58*4.554.354.474.424.384.424.69**5.24**4.484.214.264.74**5.27**4.544.404.414.474.86**5.34**4.335.23**

Test date

3/03/841/12/856/09/846/17/868/01/833/15/83

11/04/8212/03/825/16/839/24/83

11/17/8411/17/846/19/853/16/838/04/849/19/839/19/83

12/07/839/17/833/06/863/06/866/04/866/04/864/02/856/13/839/17/844/02/849/13/837/08/838/24/85

12/07/833/04/839/01/833/21/84

12/09/8311/14/8311/14/8311/14/837/10/862/26/83

Eyetested

ODODODODODODODODOSOSODOSODODOSODOSOSOSOSODOSODODODOSOSOSODOSODOSODODODOSODOSODOD

Significantly greater than the normal population at / ) <0 .05, one-tail t- Significantly greater than the normal population at P < 0.01, one-tail t-

the same time period. Informed consent was obtainedfrom each subject after the nature of the test and thestudy were explained fully.

The color matches were performed as previously de-scribed.12 The visual stimulus was a 4 deg bipartitefield. The left side of the field was a 589.6 nm standard,

Table 2. Clinical data and half-bleach for patients with more than two sessions for a given eye

Subject/sex

1/F6/F6/F

12/M14/F14/F7/F

10/F11/M12/M

Group

1112223333

Age(yr)

16283029353522262829

Duration(yr)

69

1015151515102515

VA

20/1520/2020/2020/2520/2520/1520/2020/3020/4020/25

Macularedema

Y

Y

h(logtd)

4.254.384.64**4.62*4.78**4.384.544.58*4.404.65**

Test date

7/10/861/09/836/29/847/12/847/09/867/09/86

10/09/8411/14/833/07/867/12/84

Eyetested

ODODODOSODOSOSODOSOD

* Significantly greater than the normal population at P < 0.05, one-tail t-test.

Significantly greater than the normal population at P< 0.01, one-tail t-test.

720 INVESTIGATIVE OPHTHALMOLOGY 6 VISUAL SCIENCE / April 1987 Vol. 28

5.25

5 0 0

4.75

4.50

4.25

t•

t

sN Dl D2

PATIENT CLASSIFICATION

D3

Fig. 1. The retinal illuminance required to bleach half the pho-topigment (Io) in log td, by patient and observer classification asfollows: N—normal control observers, Dl—diabetic patients withno retinopathy, D2—diabetic patients with retinopathy, D3—diabeticpatients post panretinal laser photocoagulation. The results of patientswho have been retested are shown only if there was a change incategory.

while the right half of the field was a mixture of a 546and a 650 nm light. By turning one knob the observercould alter the color of the mixture field, without al-tering its luminance, while by turning a second knobthe observer could alter the luminance without chang-ing the color. Each patient made ten matches at eachof eight retinal illuminances 260-130,000 td. Duringthis time, the position of the patient's pupil relative tothe entry position of the test light was monitored. Fol-lowing the session, the settings were calibrated,2 andthe log of the average ratio of red to green primaries,the (R-G) ratio, was calculated. From these values theoptical density, the illuminance that causes half of the

2 .0 5 . 0

Fig. 2. Color matching data for Patient 6. The squares and solidline show the patient's data, the dashes represent plus or minus twostandard deviations of the mean of our normal observers. The estimateof the half-bleach illuminance is shown in the upper left corner.

photopigment to be bleached (the half-bleach illumi-nance or Io), and the low illuminance (baseline) colormatch were calculated by computer as described pre-viously.1'2

Results

Most diabetic patients showed a dependence of theircolor-match on retinal illuminance (the color-matchilluminance effect); that is, as retinal illuminance in-creased, the color-matches required more of the redprimary than matches made at lower illuminances.However, many of the patients required higher retinalilluminances for the color match change than did nor-mal observers (Fig. 1). The half-bleach illuminancevalue (Io) is a quantitative measure of the amount oflight required to bleach one-half of the photopigments.Many of the Io estimates for the diabetic differed sig-nificantly from the normal population (Table 1).Higher half-bleach illuminance values were found evenif the patient was young, had no retinopathy (Patient6 of Table 1, Figs. 2, 3) or visual complaints, and hadnormal Farnsworth Munsell 100-Hue Test results. Infive extreme cases (Patients 10, 12, 18, 19, 22) no color-match illuminance effect was obtained (Figs. 4-6). Thatis, even the highest illuminance did not cause appre-ciable bleaching. Two of these patients were tested justprior to pan-retinal photocoagulation treatment forproliferative diabetic retinopathy (Patients 10, 12). Thelow-illuminance color matches of our diabetic popu-lation (Fig. 7, bottom), were not significantly differentfrom those of the normal population (Fig. 7, top).

Effect of Laser Photocoagulation

Of the five patients tested both prior to and post-photocoagulation with an argon laser, all had an initialdecrease in the half-bleach illuminance following pan-retinal argon laser photocoagulation (Patients 7, 10-13; see Fig. 1). One of the patients with no color-matchilluminance effect prior to photocoagulation had morenormal data post-photocoagulation (Patient 12, Fig.6). For this patient there was no overlap in the highilluminance settings pre- and post-photocoagulation.Two patients (Patients 11, 12) have maintained thelower values on follow up testing. It should be notedthat two patients with normal half-bleach-illuminancesprior to photocoagulation, were still within normallimits after photocoagulation (Patients 7, 19). Althoughthese results may indicate that photocoagulation im-proves photoreceptor function, this benefit may not belong term. Three of the five patients who had no color-match-illuminance effect on at least one test sessionhad undergone argon laser panretinal photocoagulation4 months to 4 yr prior to testing (Patients 10, 19, 22).

No. 4 DLEACHING ABNORMALITIES IN DIADETES / Eisner er ol. 721

Fig. 3. Fundus photograph (left) and fluorescein angiogram (right) for Patient 6.

(The other two patients were tested immediately priorto photocoagulation, Patients 12, 18.)

Retinopathy Classification and Io

The relation between the half-bleach illuminance,Io, and retinopathy is not simple (Fig. 1). Thus far, allpatients showing no evidence of bleaching at high il-luminances have retinopathy severe enough to warrantpanretinal laser photocoagulation by the DRS criteria1 J

and were tested immediately prior to scheduled treat-ment or afterwards. However, abnormally high valuesfor the half-bleach illuminance have been obtainedfrom patients with no retinopathy in either eye, whilenormal values have been obtained from patients justprior to panretinal photocoagulation.

Clinical Follow-up of Selected Patients

Change-is to be expected in a population withIDDM. That is, patients with background retinopathygo on to develop more severe retinopathy and possiblyproliferative retinopathy. However of the five patientswith no color-match-illuminance effect on at least onesession, two went on to develop severe vision loss (Pa-tients 10, 22), one underwent 9 months of sporadichemorrhage (Patient 19), and the other remaining twopatients required laser photocoagulation for prolifer-ative diabetic retinopathy (Patients 12, 18). Patientswho have had more normal Io have not had severevision loss during the testing period. However, 2 yrafter the last test, Patient 13 did have severe hemorrhageand required vitrectomy. This patient is now doingwell, and has 20/30 visual acuity.

0.4

-0.1 -

2.0 3.0 4.0

Log TdFig. 4. Color matching data for a 29-year-old diabetic patient (Pa-

tient 19) more than 1 yr following panretinal photocoagulation.Symbols as in Figure 2.

2.0

Fig. 5. Color matching data for a 52-year-old diabetic patient (Pa-tient 18) just prior to panretinal photocoagulation. Symbols as inFigure 2.

722 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / April 1987 Vol. 28

2.0 5.0

Fig. 6. Color matching data for a 28-year-old diabetic patient (Pa-tient 12) just prior to (squares), and 73 days post-panretinal laserphotocoagulation (double triangles). Follow-up testing 9 months post-panretinal photocoagulation indicated no further change. Dashes asin Figure 2. The half-bleach illuminance value for the second testwas 4.74.

Variability of the Data

Some, but not all, diabetic observers had more vari-able data than normal control observers, particularlyat the highest illuminance. The variability of the highestilluminance settings decreased for one observer follow-ing photocoagulation. We have not yet tested enough

-o.i o.o

BASELINE MATCH

Fig. 7. The distribution of the baseline (or moderate illuminance)color matches for normal control observers (top) and patients (bot-tom); log (R/G) as in Figure 2.

patients in follow up to estimate the test-retest repeat-ability of our technique over time. Patient 6, who hadno retinopathy (Fig. 2), was tested four times over aperiod of 20 months. The baseline color matches forthis patient varied from 0.025 to 0.045. This patient'sIo varied from 4.38 to 6.09 log td, with three of themeasurements significantly different from normal (witha one-tailed test, P < 0.01). A younger patient, Patient1, with no retinopathy was tested three times in 3 yr,with Io values ranging from 4.25 to 4.44. Four patientswho had undergone panretinal photocoagulation weretested twice and had highly repeatable results. Threeof these (Patients 7, 11, 20), maintained good visionduring the testing period; the other (Patient 13) hadsmall, sporadic hemorrhages. In contrast, an observer(Patient 10) whose vision was reduced to counting fin-gers in the eye tested (and hand motion in the othereye) showed results that changed significantly over timeprior to the hemorrhaging that reduced vision in theeye tested. Both the baseline color match and the half-bleach illuminance changed, with almost no color-match-illuminance effect measured in the test sessionprior to the hemorrhage.

Discussion

Optical Density of the Cone Photopigments

The optical density of the long and middle wave-length-sensitive cone photopigments in patients withIDDM appears to be normal. This is supported by thenormal estimates of the difference between the low andhigh illuminance color matches for those patients witha measurable color-match-illuminance effect. However,since the color matches of many diabetics did not reacha high illuminance asymptote, no optical density es-timate could be obtained for many diabetic observers.However, if the optical densities of diabetic observerswere abnormal, the low illuminance color matchesshould have been abnormal. Although there was scatterof the color matches, the color matches were scatteredsymmetrically around the midpoint of the range of set-tings by normal observers (Fig. 7). This normality ofcolor matches at moderate retinal illuminances is inagreement with previous results.12 The normal baselinecolor matches, along with the normal high-illuminanceassymptote measured for most patients implies thatthe cones have a normal concentration of photopig-ment at moderate illuminances. However, it should benoted that Bresnick et al13 have measured normal colormatches in a patient with an abnormal Stiles-Crawfordeffect due to macular drag. This would not seem toaffect our study group, since none had evidence ofmacular drag, or preretinal membranes. In addition,the measurement of only low-illuminance color-matches in the Bresnick et al study cannot be used to

No. 4 BLEACHING ABNORMALITIES IN DIABETES / Eisner er ol. 720

positively identify patients with normal vs abnormaloptical densities due to the intraobserver variability ofcolor-matches.

Bleaching of the Cone Photopigments

Some diabetic patients require a higher than normalretinal illuminance to bleach their cone photopigments.We have considered several hypotheses to explain thisresult. The first, and simplest, is that a filter could bereducing the amount of light reaching the cones. Aspectrally neutral filter should cause a shape invarianttranslation of the entire color match vs illuminancecurve to the right (see Fig. 1, previous paper). This issimilar to what we measure in the diabetic patients,since we find no shift in the color-matches at moderateretinal illuminances. Thus, we tested the implicationsof this hypothesis. If the high Io was due to an increaseddensity of preretinal filters, then the view of the retinashould be affected by the filter. If we measure a 0.5 logunit increase in Io, then the view should be equivalentto viewing through a 0.5 log unit filter, and the amountof light exiting the eye would be decreased by 1 logunit. One of us (a retina specialist) viewed the retinaof test subjects through a series of filters and found thatfilters of 0.6-1.0 log units would be easily detected.Similarly, such dense preretinal filters would effectfundus photography. Since the dynamic range of thefilm used in color photography is about 1 log unit, thepresence of such preretinal filters would necessitatelarge changes in the flash intensities of the fundus cam-eras (Fig. 3). No such changes were required. Spectrally-selective filters would cause differential filtering of thered and green primaries and, therefore, are not con-sistent with the normal baseline color matches. Weconclude that a preretinal filter is an unlikely expla-nation for the high Io.

A second possibility is that eye pathology could resultin a receptoral abnormality that mimics the filter hy-pothesis. For instance, if there are changes in the conesresulting in less efficient capture of light, then higherilluminance could be required to cause a change in thecolor match. Similarly, there could be a screening pig-ment in the inner segments of the cones. Or, the lightcapturing ability of the cones could be diminished bylocal environmental changes, such as changes in therefractive index of the retina. Our test of the fundu-scopic view of the retina cannot address this particularhypothesis, since the changes are assumed to be local.However, any such changes must be reversible, as ourpre- and post-photocoagulation data show, and be ableto decrease the light capture ability of the photorecep-tors by at least 1 log unit. Although this is a large effectfor local phenomena, we cannot reject this hypothesiswith our data.

A third possibility is that an abnormality of photo-pigment kinetics could cause the photopigments in di-abetic patients to be regenerated faster than in normalobservers. A small increase in the rate of regenerationcould cause an increased half-bleach illuminance, whilea large increase in rate could cause the lack of color-match illuminance effect we see in some patients. Thishypothesis has the advantage that it could be controlledby metabolic factors, which are likely to be abnormalin diabetic observers. A possibly related finding is thatyoung, diabetic patients can have abnormally highrenal clearances.14 Finally, the hypothesis of abnormalkinetics has the advantage that an imbalance is re-versible, and would be consistent with the observedchanges pre- and post-photocoagulation. We plan totest these hypotheses by measuring the rate of changeof the color match with time following changes in ret-inal illuminance.

Caveats

At this time we can make no strong statement aboutour test results and patient prognosis. The relation be-tween half-bleach illuminance and retinopathy may bestronger than our data indicate because we cannot testsome patients with severe retinopathy, eg patients withsevere hemorrhage. By recruiting additional diabeticpatients who do not have retinopathy, and by followingthe natural history of those who do, we may betterunderstand the relation of photopigment abnormalitiesand long-term prognosis of diabetic patients.

Implications

Our results also affect interpretation of clinical testingof light and dark adaptation. For instance, diabetic ob-servers have been found to have dark adaptation ab-normalities, such as elevated final thresholds or a longertime required to reach threshold.3"5 Many tests of darkadaptation are performed following exposure to ableaching or adaptation light. Since we found that di-abetic observers can require more light than normalobservers to bleach a comparable proportion of pho-topigment, then these dark adaptation tests may notbe appropriate for comparing recovery times in thetwo populations. More important, the mechanisms re-sponsible for causing changes in dark adaptation indiabetics remain uncertain. Our data imply a receptoralabnormality in at least some diabetic observers. Thisabnormality could be due to either an increased rateof photopigment regeneration, or a decreased abilityof the cones to absorb light. The decreased light capturehypothesis is consistent with the increased absolutethresholds of diabetic observers.3'4 However, a loss ofsensitivity alone cannot explain the delayed recoveryof visual acuity following exposure to bright lights,6'715

724 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / April 1987 Vol. 28

nor the increased time required to reach absolutethreshold. These previous findings are consistent withabnormalities of neural processing in individuals withIDDM. That is, typical tests of visual function are sen-sitive to abnormalities of visual function at a numberof loci, but cannot provide information concerning thefunction of specific portions of the visual system. Thepresence of a photoreceptoral abnormality makes re-sults from such tests difficult to interpret, since thephotoreceptor signal sent for later neural processingcan be abnormal.

In conclusion, we have measured abnormalities inthe bleaching of cone photopigments of some patientswith insulin dependent diabetes mellitus. These ab-normalities are most likely to arise from either abnor-malities in the collection of light by the photoreceptors,or due to abnormalities of photopigment kinetics. Al-though it is clear that many more patients need to bestudied to understand the relation of these abnormal-ities to diabetic retinopathy and vision loss, the presenceof the outer retinal abnormality is clear.

Key words: photoreceptor, diabetic retinopathy, cone, colorvision, diabetes

Acknowledgments

We would like to thank Dr. Alvin Eisner for commentsand criticisms of previous versions of this manuscript andBeverly Bober for aid in performing the experiments.

References

1. Burns SA, Eisner AE, Lobes LA, and Doft BH: A Psychophysicaltechnique for measuring cone photopigment bleaching. InvestOphthalmol Vis Sci 28:711, 1987.

2. Burns SA and Eisner AE: Color matching at high illuminances:the color-match area effect and photopigment bleaching. J OptSocAm A 2:698, 1985.

3. Aspinall PA: Rod-cone interaction: some indirect evidence. ActaOphthalmol 55:294, 1977.

4. Wepman B, Sokol S, and Price J: The effects of photocoagulationon the electroretinogram and dark adaptation in diabetic reti-nopathy. Fourteenth ISCERG Symposium, 139-147, 1977.

5. Henson DB and North RV: Dark adaptation in diabetes mellitus.Br J Ophthalmol 63:539, 1979.

6. Begg IS, Broome SJ, and Schulzer M: Photostress recovery timein type I diabetics. ARVO Abstracts. Invest Ophthalmol Vis Sci22(Suppl):68, 1982.

7. Russell P, Sekuler R, and Fetkenhour CL: Visual function afterpanretinal photocoagulation (RPR). ARVO Abstracts. InvestOphthalmol Vis Sci 24(Suppl):81, 1983.

8. Hood DC, and Greenstein VC: An approach to testing alternativehypotheses of changes in visual sensitivity due to retinal disease.Invest Ophthalmol Vis Sci 23:96, 1982.

9. Greenstein VC, Hood DC, and Campell CJ: The use of flash-on-flash paradigm to assess sensitivity changes due to retinal dis-ease. Invest Ophthalmol Vis Sci 23:102, 1982.

10. Dartnall HJA: The Visual Pigments, London, Methuen, 1957.

11. Diabetic Retinopathy Study Research Group: Photocoagulationtreatment of proliferative diabetic retinopathy. Ophthalmology88:583, 1981.

12. Lakowski R, Aspinall PA, and Kinnear PR: Association betweencolour vision losses and diabetes mellitus. Ophthalmic Res 4:145, 1972/73.

13. Bresnick GH, Smith VC, and Pokorny J: Visual function ab-normalities in macular heterotopia caused by proliferative dia-betic retinopathy. Am J Ophthalmol 92:85, 1981.

14. Ellis D, Becker DJ, Daneman MB, Lobes L, and Drash AL:Proteinuria in children with insulin-dependent diabetes: Rela-tionship to duration of disease, metabolic control, and retinalchanges. J Pediatr 102:673, 1983.

15. Glaser JS, Savino PJ, Sumers KD, McDonald SA, and KnightonRW: The Photostress test in the clinical assessment of visualfunction. Am J Ophthalmol 83:255, 1977.