Further Evaluation of Docosahexaenoic Acid in Patients with Retinitis Pigmentosa Receiving Vitamin a Treatment Subgroup Analyses

CLINICAL SCIENCES Further Evaluation of Docosahexaenoic Acid in Patients With Retinitis Pigmentosa Receiving Vitamin A Treatment Subgroup Analyses

Eliot L. Berson, MD; Bernard Rosner, PhD; Michael A. Sandberg, PhD; Carol Weigel-DiFranco, MA; Ann Moser, BA; Robert J. Brockhurst, MD; K. C. Hayes, DVM, PhD; Chris A. Johnson, PhD; Ellen J. Anderson, RD; Alexander R. Gaudio, MD; Walter C. Willett, MD; Ernst J. Schaefer, MD

Objective: To determine whether docosahexaenoic acid in field sensitivity and electroretinogram amplitude than will slow the course of retinal degeneration in sub- those in the control+A group over the first 2 years (P=.01 groups of patients with retinitis pigmentosa who are re- and P=.03, respectively); these differences were not ob- ceiving vitamin A. served in years 3 and 4 of follow-up or among patients taking vitamin A prior to entry. In the entire cohort, red Design: A cohort of 208 patients with retinitis pigmen- blood cell phosphatidylethanolamine docosahexaenoic tosa, aged 18 to 55 years, were randomly assigned to acid level was inversely related to rate of decline in total 1200 mg of docosahexaenoic acid plus 15000 IU/d of field sensitivity over 4 years (test for trend, P=.05). This vitamin A given as retinyl palmitate (DHA+A group) or was particularly evident over the first 2 years among those control fatty acid plus 15000 IU/d of vitamin A not on vitamin A prior to entry (test for trend, P=.003). (control+A group) and followed up over 4 years. Sev- In the entire control+A group, dietary ␻-3 fatty acid in- enty percent of the patients in each group were taking take was inversely related to loss of total field sensitivity vitamin A, 15000 IU/d, prior to entry. We compared over 4 years (intake, Ͻ0.20 vs Ն0.20 g/d; P=.02). The rates of decline in ocular function in the DHA+A vs duration of vitamin A supplementation prior to entry was control+A groups among the subgroups defined by use inversely related to rate of decline in electroretinogram or nonuse of vitamin A prior to entry. We also deter- amplitude (P=.008). mined whether decline in ocular function was related to red blood cell phosphatidylethanolamine docosahexae- Conclusions: For patients with retinitis pigmentosa be- noic acid level, dietary ␻-3 fatty acid intake, or duration ginning vitamin A therapy, addition of docosahexae- of vitamin A use. Main outcome measures were Hum- noic acid, 1200 mg/d, slowed the course of disease for phrey Field Analyzer visual field sensitivity, 30-Hz elec- 2 years. Among patients on vitamin A for at least 2 years, troretinogram amplitude, and visual acuity. a diet rich in ␻-3 fatty acids (Ն0.20 g/d) slowed the decline in visual field sensitivity. Results: Among patients not taking vitamin A prior to entry, those in the DHA+A group had a slower decline Arch Ophthalmol. 2004;122:1306-1314

E HAVE REPORTED reported taking vitamin A, 15000 IU/d, elsewhere in this prior to entry, whereas 30% did not; issue that oral approximately equal numbers of patients supplementation on and not on vitamin A prior to entry with docosahexa- were in the DHA+A and control+A enoicW acid in a dosage of 1200 mg/d did groups. After 1 year of follow-up, we not, on average, slow the rate of decline unexpectedly noted a highly significant in ocular function over a 4-year interval statistical interaction between the effect of docosahexaenoic acid supplementa- See also page 1297 tion and the status of vitamin A intake prior to entry on change in visual field among 208 patients with retinitis pig- sensitivity (PϽ.01), suggesting the need mentosa who concurrently received vita- for subgroup analyses. Therefore, we fol- min A, 15000 IU/d.1 Among these lowed change in ocular function in these patients randomly assigned to docosa- 4 subgroups over the 3 remaining years hexaenoic acid plus vitamin A (DHA+A of this trial. The present article compares Author affiliations are listed at group) or control fatty acid capsules plus rates of decline in ocular function over 4 the end of this article. vitamin A (control+A group), about 70% years among the subgroups within the

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1306

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 DHA+A and control+A groups defined by use or non- ids (Ͻ5.0%, 5.0%-9.9%, 10.0%-14.9%, and 15.0%-19.9%) in use of vitamin A prior to entry. the entire study population and in the 2 subgroups defined by Data from our previous clinical trial of vitamin A vitamin A status prior to entry. for retinitis pigmentosa showed that the decline in 30-Hz Third, in the entire control+A group (ie, those receiving electroretinogram (ERG) amplitude over a 4-year inter- control fatty acids plus 15000 IU/d of vitamin A regardless of vitamin A status prior to entry), analyses were performed re- val was inversely related to red blood cell phosphati- lating rates of decline in ocular function to level of dietary ␻-3 dylethanolamine docosahexaenoic acid (RBC PE DHA) fatty acid intake (at or above and below the median intake of concentration (P=.03) among 61 patients for whom we 0.20 g/d observed over the course of this study) over the 4-year had RBC levels available for analysis at year 3 or 4.1 Be- period of this study and for the periods years 0 to 2 and 2 to 4. cause RBC DHA levels have been correlated with retinal Fourth, in the entire control+A group, we subdivided pa- DHA levels,2 and DHA concentration is high in the PE tients on the basis of duration of vitamin A intake (15000 IU/d) fraction of retinal phospholipids,3 we evaluated the re- prior to study entry (ie, not on vitamin A prior to entry, on vi- lationship of rate of loss of ocular function to RBC PE tamin A prior to entry for 1-23 months, or on vitamin A prior Ն DHA levels in the present study population. Because RBC to entry for 24 months) and compared rates of decline in ocu- phospholipids vary with manipulation of dietary ␻-3 fatty lar function among these subgroups. 4,5 In addition, in the entire control+A group, we evaluated acids, we also evaluated in the control+A group alone whether a relationship existed between dietary ␻-3 fatty acid whether rate of decline in ocular function could be re- ␻ intake and RBC PE DHA levels and whether vitamin A intake lated to dietary -3 fatty acid intake as a measure of doco- was associated with a change in RBC PE DHA levels over the sahexaenoic acid intake. duration of the study. In our previous trial of vitamin A supplementation for retinitis pigmentosa, we found a significantly slower RESULTS rate of decline in retinal function in those patients randomized to 15000 IU/d of vitamin A than in those not PARTICIPANTS AND PATIENT FLOW on this dosage, but this difference became obvious only after 4 years of follow-up.6 In the present study, we there- Table 1 gives demographic and ocular characteristics fore also evaluated whether the rate of decline in ocular at baseline for the 208 patients seen annually for all 4 function in the control+A group was related to dura- years in the DHA+A vs control+A groups within the sub- tion of vitamin A intake prior to entry. groups defined by status of vitamin A intake prior to entry. In accord with the criteria for randomization, the sub- METHODS groups defined by vitamin A status prior to entry had no significant differences in percentages of patients by ge- Patients with retinitis pigmentosa, aged 18 to 55 years, under- netic type of retinitis pigmentosa and no significant dif- went screening for eligibility according to ocular, dietary, and ferences in dietary intake of ␻-3 fatty acids at baseline. medical criteria and were randomly assigned to a DHA+A or a The study population was about 50% male and 50% fe- control+A group. All eligible patients had typical forms of reti- male, with a comparable distribution in the subgroups nitis pigmentosa with elevated final dark-adaptation thresh- on or not on vitamin A prior to entry. Eleven percent re- olds, retinal arteriolar narrowing, and reduced and delayed ERGs; ported partial hearing loss, with no significant differ- most had intraretinal bone spicule pigment around the midpe- ripheral fundus. Atypical forms such as paravenous, unilateral, ence among the subgroups. or sector retinitis pigmentosa were excluded. Patients with Bardet- Baseline ocular function (ie, mean of screening and Biedl syndrome, Refsum disease, retinitis punctata albescens, or baseline examinations prior to treatment) showed slightly retinitis pigmentosa associated with profound congenital deaf- less visual field sensitivity in the DHA+A vs control+A ness were also excluded. Patients were subdivided according to groups within the subgroups, although the differences whether or not they reported taking 15000 IU/d of vitamin A were not statistically significant in either subgroup. Vi- prior to entry. Study design, eligibility criteria, informed con- sual acuity according to the Early Treatment Diabetic Reti- sent, techniques for evaluation, main outcome measures, method nopathy Study (ETDRS acuity) and ERG amplitudes at of randomization, procedures for masking, and methods of data 1 baseline were also not significantly different between the analyses are described elsewhere. Within 6 to 8 weeks after a DHA+A and control+A groups in either subgroup (Table screening examination, eligible patients were randomly assigned at a baseline examination to 600 mg of oral docosahexae- 1). Within the subgroups defined by vitamin A use prior noic acid twice daily (DHA+A group) or control fatty acid cap- to entry, the DHA+A and control+A groups showed com- sules (control+A group). All were given 1 tablet per day containing parable levels of RBC PE DHA, plasma DHA, serum reti- 15000 IU of vitamin A as retinyl palmitate. The study was ap- nol, and serum retinyl esters prior to treatment. The RBC proved by the institutional review boards of the Massachusetts PE DHA levels (expressed as the mean±SE percentage Eye and Ear Infirmary and Harvard Medical School, Boston, Mass. of total RBC PE fatty acids) were significantly higher The study conformed to the Declaration of Helsinki. among those on vitamin A (4.66%±0.15%) vs those not Subgroup analyses are listed as follows. First, rates of de- on vitamin A prior to entry (4.22%±0.16%) (P=.05). cline in ocular function were compared between the DHA+A For those on vitamin A prior to entry, RBC PE DHA and control+A groups among the subgroups taking vitamin A levels at follow-up were 13.12%±0.26% (n=73) in the prior to entry and between the DHA+A and control+A groups among the subgroups not taking vitamin A prior to entry over DHA+A group and 5.02%±0.21% (n=66) in the control+A the 4-year study. group. For those not on vitamin A prior to entry, RBC PE Second, analyses were performed relating rate of decline DHA levels at follow-up were 11.98%±0.68% (n=29) in in ocular function to category of RBC PE DHA level based on the DHA+A group and 4.31%±0.25% (n=34) in the an equal-interval scale of percentage of total RBC PE fatty ac- control+A group. These RBC levels were consistent with

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1307

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 Table 1. Demographic and Ocular Characteristics at Entry of Patients With Retinitis Pigmentosa*

Stratified by Initial Vitamin A Status (n = 208)

On Vitamin A Prior to Entry Not on Vitamin A Prior to Entry

DHA+A Control + A P DHA+A Control + A P (n = 75) (n = 68) Value† (n = 30) (n = 35) Value† Genetic type, No. (%) Dominant 11 (15) 11 (16) 9 (30) 8 (23) Dominant with mutation 5 (7) 2 (3) 1 (3) 2 (6) Recessive 13 (17) 9 (13) 3 (10) 4 (11) .83 .55 X-linked 5 (7) 5 (7) 0 3 (9) Isolate 37 (49) 35 (51) 13 (43) 15 (43) Undetermined 4 (5) 6 (9) 4 (13) 3 (9) Dietary intake ␻-3 Fatty acids, mean ± SE, g/d 0.20 ± 0.01 0.19 ± 0.01 .74 0.16 ± 0.01 0.16 ± 0.02 .95 Multivitamin use, No. (%) 13 (17) 14 (21) .78 10 (33) 8 (23) .51 Blood values, mean ± SE‡ RBC PE DHA, % 4.91 ± 0.24 4.39 ± 0.17 .08 4.44 ± 0.19 4.04 ± 0.24 .21 Plasma DHA, % 1.67 ± 0.08 1.51 ± 0.07 .14 1.60 ± 0.11 1.53 ± 0.10 .63 Serum retinol, µg/dL 56.6 ± 1.5 56.7 ± 1.7 .97 52.3 ± 2.6 48.4 ± 1.8 .21 Serum retinyl esters, nmol/L 147.1 ± 7.1 145.1 ± 7.2 .85 122.3 ± 8.5 114.6 ± 7.5 .49 Male, No. (%) 38 (51) 38 (56) .65 12 (40) 18 (51) .50 Age, mean ± SE, y 38.1 ± 1.0 36.8 ± 1.3 .43 36.9 ± 1.6 34.5 ± 1.6 .28 Weight, kg, mean ± SE Men 81.7 ± 1.9 84.1 ± 2.0 .37 84.4 ± 4.9 79.2 ± 2.8 .34 Women 64.2 ± 2.1 66.2 ± 2.4 .54 63.9 ± 2.6 64.4 ± 3.9 .92 Baseline ocular function§ HFA total point scores, mean ± SE, dB 30-2 Field 797.0 ± 53.9 869.3 ± 55.2 .35 830.5 ± 84.3 998.5 ± 89.0 .18 Total field 1176 ± 79 1245 ± 86 .55 1276 ± 141 1436 ± 131 .41 ERG amplitudes, µV࿣ 30 Hz 1.29 ± 0.13 (3.63) 1.28 ± 0.14 (3.60) .92 1.48 ± 0.23 (4.39) 1.23 ± 0.17 (3.42) .39 0.5 Hz 2.96 ± 0.13 (19.30) 3.13 ± 0.16 (22.87) .42 3.00 ± 0.24 (20.08) 3.04 ± 0.21 (20.90) .91 ETDRS visual acuity, letters 50.4 ± 1.2 51.2 ± 1.2 .63 53.8 ± 1.7 51.6 ± 1.4 .34

Abbreviations: DHA, docosahexaenoic acid; ERG, electroretinogram; ETDRS, Early Treatment Diabetic Retinopathy Study; HFA, Humphrey Field Analyzer; RBC PE, red blood cell phosphatidylethanolamine. SI conversion factor: To convert retinol to micromoles per liter, multiply by 0.0349. *Patients received either 1200 mg/d of docosahexaenoic acid plus 15 000 IU/d of vitamin A (DHA + A) or control capsules plus 15 000 IU/d of vitamin A (control + A). †For continuous variables, we calculated the P value using a t test; for categorical variables, the ␹2 test. ‡The lower normal value for RBC PE DHA level was 3.16%; for plasma DHA level, 0.86%. The normal range for serum retinol level was 38-100 µg/dL; for serum retinyl esters level, 34-380 nmol/L. Sample sizes for DHA and control groups on vitamin A prior to entry were 74 and 66, respectively, for RBC PE DHA level; 74 and 65, respectively, for plasma DHA level; 75 and 68, respectively, for serum retinol level; and 75 and 68, respectively, for retinyl esters level. Sample sizes for DHA and control groups not on vitamin A prior to entry were 30 and 34, respectively, for RBC PE DHA level; 29 and 35, respectively, for plasma DHA level; 30 and 35, respectively, for serum retinol level; and 30 and 35, respectively, for retinyl esters level. §The lower normal value for HFA 30-2 field (size V target) was 2500 dB; for HFA total field (size V target), 4200 dB. The lower normal value for the 30-Hz ERG was 50 µV; for the 0.5-Hz ERG, 350 µV. The lower normal value for ETDRS visual acuity was 60 letters. Fifty-one letters indicates Snellen acuity of 20/30. Sample sizes for the DHA and control groups on vitamin A prior to entry were 74 and 68, respectively, for HFA 30-2; 74 and 68, respectively, for HFA total; 75 and 67, respectively, for 30-Hz ERG; and 44 and 35, respectively, for 0.5-Hz ERG. Sample sizes for the DHA and control groups not on vitamin A prior to entry were 29 and 34, respectively, for HFA 30-2; 28 and 34, respectively, for HFA total; 26 and 33, respectively, for 30-Hz ERG; and 17 and 18, respectively, for 0.5-Hz ERG.

࿣Designated as mean loge ± SE. Geometric means for ERG amplitudes are given in parentheses.

Table 2. Annual Rate of Decline for Measures of Ocular Function by Treatment Group and Vitamin A Status Prior to Entry Over a 4-Year Interval*

On Vitamin A Prior to Entry Not on Vitamin A Prior to Entry P Value DHA + A Control + A P Value† DHA + A Control + A P Value† Interaction† HFA 30-2 field, dB/y 39.41 ± 3.76 (74) 30.26 ± 3.92 (68) .09 30.7 ± 6.48 (29) 52.5 ± 5.99 (34) .01 .002 HFA total field, dB/y‡ 61.01 ± 5.17 (74) 48.13 ± 5.39 (68) .08 47.16 ± 10.56 (28) 82.49 ± 9.58 (34) .01 .001

30-Hz ERG, loge 0.11 ± 0.01 (75) 0.10 ± 0.01 (67) .34 0.08 ± 0.02 (26) 0.14 ± 0.02 (33) .03 .02 % Decline per year§ 10.57 9.23 8.05 12.99 ETDRS visual acuity, 0.67 ± 0.13 (75) 0.68 ± 0.14 (68) .96 0.82 ± 0.26 (29) 0.69 ± 0.24 (34) .70 .62 letters per year

Abbreviations: See Table 1. *Unless otherwise indicated, data are expressed as mean ± SE (number of patients sampled). Patients received either 1200 mg/d of docosahexaenoic acid plus 15 000 IU/d of vitamin A (DHA + A) or control capsules plus 15 000 IU/d of vitamin A (control + A). †Calculated from PROC MIXED (SAS Institute Inc, Cary, NC) analysis comparing rates of decline in both treatment groups. ‡Total field sensitivity consists of 30-2 and 30/60-1 total point scores combined when both are available. §Derived from 100 ϫ [1−exp(mean log change)].

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1308

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 reported compliance, ie, 90% of patients reported taking A Control + A Group the study capsules 90% of the time. DHA + A Group 1600 ANALYSES OF OUTCOME MEASURES 1500 1400 Effect of Docosahexaenoic Acid Supplementation 1300 as a Function of Vitamin A Status Prior to Entry 1200 1100 Table 2 lists mean annual rates of decline of central and 1000 total visual field sensitivity, 30-Hz ERG amplitude, and 900

ETDRS acuity over a 4-year interval among the 208 pa- dB Not on Vitamin A Prior to Entry, 800

Mean Total Point Score Among Patients Mean Total 01234 tients in the DHA+A vs control+A groups for those on Year and not on vitamin A prior to entry. We found significant statistical interactions of treatment effects accord- B

ing to vitamin A supplement use prior to entry for vi- 1600 sual field sensitivity and 30-Hz ERG amplitude, suggesting 1500 that treatment group effects were different for those on 1400 vs not on vitamin A prior to entry. For those on vitamin 1300 A prior to entry, the mean annual rates of decline of cen- 1200 tral and total field sensitivity and 30-Hz ERG amplitude 1100 were not significantly different between the DHA+A and 1000 control+A groups. For those not on vitamin A prior to 900 on Vitamin A Prior to Entry, dB on Vitamin A Prior to Entry, entry, mean rates of decline of central and total visual 800

Mean Total Point Score Among Patients Mean Total 01234 field sensitivity and 30-Hz ERG amplitude were signifi- Year cantly lower in the DHA+A vs control+A groups. No significant statistical interaction effect was noted for ETDRS C acuity, and no significant differences in rates of ETDRS 120 acuity decline were noted in subgroup comparisons. 100 Figure 1A shows values (mean±SE) of total visual field sensitivity (total point score for Humphrey Field Ana- 80 lyzer [HFA] 30-2 and 30/60-1 programs combined) by year 60 among those not on vitamin A prior to entry in the DHA+A vs control+A groups. In these subgroups, the difference 40 between the 2 curves was larger during years 1 and 2 and 20 smaller during years 3 and 4. In contrast, as seen in Fig- Not on Vitamin A Prior to Entry, dB/y Not on Vitamin A Prior to Entry, 0 ure 1B, among those on vitamin A prior to entry, the dif- Point Score Decline Among Patients Total ferences between the curves for the DHA+A vs control+A Years 0 - 2 Years 2 - 4 groups were not significantly different for either time pe- Figure 1. Visual field sensitivity by year and treatment group riod, although a slight divergence of the curves was noted (docosahexaenoic acid plus vitamin A [DHA+A group] vs control plus particularly in years 3 and 4. Figure 1C shows the annual vitamin A [control+A group]) for patients not on vitamin A prior to entry (A) rates (mean±SE) among those not on vitamin A prior to and for patients on vitamin A prior to entry (B). Sample sizes for those not on vitamin A prior to entry were 28 for the DHA+A group and 34 for the entry for years 0 to 2 and 2 to 4 for rate of total field sen- control+A group; and for those on vitamin A prior to entry, 75 for the sitivity decline when comparing the DHA+A vs control+A DHA+A group and 68 for the control+A group. Slopes of decline in total field groups. The rate of decline was significantly slower in the sensitivity are shown for the DHA+A vs control+A groups for years 0 to 2 and 2 to 4 among those not on vitamin A prior to entry (C). Limit lines depict DHA+A vs control+A groups for years 0 to 2 (P=.006), standard error values. but was not significantly different for years 2 to 4 (P=.57).

Relationship of Change in Ocular Function test for trend, P=.003). These significant differences were to RBC PE DHA Level not seen from years 2 to 4, although the trends were in the same direction as from years 0 to 2. Among those on Table 3 summarizes that, for the entire study cohort, a vitamin A prior to entry, no significant differences in rate trend toward a significantly faster decline in visual field of loss of central or total HFA sensitivity by RBC PE DHA sensitivity was seen for those with lower compared with level were seen for years 0 to 2 or years 2 to 4 (Table 3). higher RBC PE DHA levels over 4 years (central field test for trend, P=.09; total field test for trend, P=.05). Pa- Effect of Dietary ␻-3 Fatty Acid Intake tients with an RBC PE DHA level of less than 5% of total on Change in Ocular Function RBC PE fatty acids showed a significantly faster rate of loss of total field sensitivity vs those with a level of at least Table 4 shows that among patients in the control+A group 5% (P=.02). Among those not on vitamin A prior to en- on vitamin A prior to entry, the rate of decline in visual try, a significantly faster decline was seen for those with field sensitivity over a 4-year interval for the central (30-2) lower compared with higher RBC PE DHA levels from condition was significantly related to the amount of di- years 0 to 2 (central field test for trend, P=.01; total field etary ␻-3 fatty acid intake; those with an intake of at least

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1309

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 Table 3. Annual Decline in Total Visual Field Sensitivity as a Function of RBC PE DHA Level*

From Year 0 to Year 2 From Year 2 to Year 4 From Year 0 to Year 4 Entire On Vitamin A Not on Vitamin A On Vitamin A Not on Vitamin A RBC PE DHA Level† Study Population Prior to Entry Prior to Entry Prior to Entry Prior to Entry 30-2 Condition, dB Ͻ5.0 44.7 ± 4.1 (68) 30.5 ± 8.0 (34) 39.9 ± 9.5 (25) 44.1 ± 7.9 (38) 60.8 ± 9.6 (33) 5.0-9.9 32.4 ± 4.8 (49) 24.5 ± 7.4 (40) 33.9 ± 12.0 (15) 25.7 ± 8.1 (36) 53.9 ± 17.4 (10) 10.0-14.9 34.4 ± 4.0 (70) 31.2 ± 6.2 (58) 8.8 ± 11.4 (16) 50.6 ± 7.2 (46) 47.6 ± 15.9 (12) 15.0-19.9 34.0 ± 7.9 (18) 25.9 ± 16.0 (8) 1.0 ± 20.3 (5) 30.3 ± 10.6 (21) 34.2 ± 19.4 (8) P value (test for trend)‡ .09 .94 .01 .93 .20 Total condition, dB§ Ͻ5.0 69.6 ± 5.9 (68) 40.1 ± 12.4 (34) 71.1 ± 16.0 (25) 76.0 ± 11.3 (38) 93.9 ± 13.8 (33) 5.0-9.9 53.5 ± 7.1 (48) 42.8 ± 11.8 (40) 57.7 ± 20.0 (15) 42.2 ± 11.6 (36) 81.8 ± 26.5 (10) 10.0-14.9 52.3 ± 5.9 (70) 43.5 ± 10.0 (58) 14.9 ± 18.9 (16) 78.8 ± 10.2 (16) 77.8 ± 23.0 (12) 15.0-19.9 52.8 ± 11.6 (18) 50.8 ± 25.0 (8) 3.7 ± 33.4 (5) 52.9 ± 15.2 (21) 40.1 ± 28.1 (8) P value (test for trend)‡ .05 .73 .003 .75 .11

Abbreviations: See Table 1. *Unless otherwise indicated, data are expressed as mean ± SE (number of patients sampled). †RBC PE DHA levels are expressed as percentages of total RBC PE fatty acids. Values are based on an average of years 1 and 4 for analysis of the entire sample from years 0-4 and on year 1 or year 4 for analyses from years 0-2 and 2-4, respectively. ‡Based on PROC MIXED (SAS Institute Inc, Cary, NC) analysis. §Total condition consists of 30-2 and 30/60-1 conditions when both are available.

Table 4. Annual Decline in Visual Field Sensitivity in the Control Group as a Function of Dietary ␻-3 Fatty Acid Intake Among Patients on Vitamin A Prior to Entry*

Dietary ␻-3 Intake† Years 0 to 4 Years 0 to 2 Years 2 to 4 30-2 Condition, dB Ͻ0.20 g/d 39.2 ± 5.5 (35) 32.9 ± 8.2 (35) 44.5 ± 7.5 (35) Ն0.20 g/d 20.8 ± 5.7 (33) 14.3 ± 8.4 (33) 20.1 ± 7.7 (33) P value‡ .02 .12 .03 Total condition, dB§ Ͻ0.20 g/d 57.8 ± 7.0 (35) 44.7 ± 12.4 (35) 72.9 ± 10.5 (35) Ն0.20 g/d 37.9 ± 7.2 (33) 25.0 ± 12.7 (33) 39.3 ± 10.8 (33) P value‡ .05 .27 .03

*Unless otherwise indicated, data are expressed as mean ± SE (number of patients sampled). Patients received control capsules plus 15 000 IU/d of vitamin A. †Expressed as the mean of grams per day of all visits. ‡Based on PROC MIXED (SAS Institute Inc, Cary, NC) analysis. §Total condition consists of the sum of 30-2 and 30/60-1 conditions when both are available.

0.20 g/d had a 40% to 50% slower rate of decline com- 13.0% for those not on vitamin A prior to entry, 11.6% pared with those with intake of less than 0.20 g/d (P=.02). for those on vitamin A for 1 to 23 months prior to entry, A similar result was seen for the total (HFA 30-2 and 30/ and 7.9% for those on vitamin A for 24 or more months 60-1 combined) condition (P=.05). Similar trends were prior to entry. There was a significant linear trend with seen for years 0 to 2 and 2 to 4, although the differences duration of intake of vitamin A prior to entry (test for trend, were only significant for the latter period (P=.03). The same P=.008). A similar trend was seen for central field and total pattern was seen for the 30-2 and total conditions among field sensitivity decline. With respect to total field sensi- those not on vitamin A prior to entry, but the differences tivity, patients in the control+A group showed an 84-dB were not significant (data not shown). annual decline during the study for those not on vitamin A prior to entry, a 42-dB decline during the study for those Effect of Duration of Vitamin A Intake taking vitamin A for 1 to 23 months prior to entry, and a on Change in Ocular Function 62-dB decline for those taking vitamin A for 2 or more years prior to entry (test for trend, P=.08). Table 5 lists data for patients in the entire control+A group categorized by the number of years on vitamin A, 15000 OTHER ANALYSES IU/d, prior to entry. Although their initial 30-Hz ERG amplitudes did not differ significantly, those on vitamin A for Relationship of ␻-3 Fatty Acid Intake 2 or more years prior to entry had the slowest rate of de- to RBC PE DHA Levels cline during the study, whereas those not on vitamin A prior to entry had the fastest rate of decline. Estimated an- Figure 2 shows that a significant relationship existed nual rates of decline of remaining retinal function were between dietary ␻-3 fatty acid intake and RBC PE DHA

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1310

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 Table 5. Relationship Between Duration of Vitamin A Intake Prior to Entry and 30-Hz ERG Amplitude Decline in the Control Group*

Duration of Intake of Vitamin A Before Entry

None 1-23 mo Ն24 mo Mean of screening and baseline in both eyes 30-Hz ERG† 1.23 ± 0.18 (33) 1.22 ± 0.21 (24) 1.31 ± 0.18 (43) Geometric mean, µV‡ 3.42 3.40 3.69 Year 1 30-Hz ERG† 1.06 ± 0.19 (33) 1.10 ± 0.22 (24) 1.28 ± 0.19 (43) Geometric mean, µV‡ 2.87 3.01 3.58 Year 2 30-Hz ERG† 0.85 ± 0.20 (33) 0.91 ± 0.21 (24) 1.13 ± 0.18 (43) Geometric mean, µV‡ 2.35 2.49 3.09 Year 3 30-Hz ERG† 0.77 ± 0.20 (33) 0.81 ± 0.20 (24) 1.01 ± 0.19 (43) Geometric mean, µV‡ 2.15 2.24 2.74 Year 4 30-Hz ERG† 0.68 ± 0.20 (33) 0.75 ± 0.22 (24) 1.03 ± 0.19 (43) Geometric mean, µV‡ 1.97 2.13 2.80

Estimate of slope, %/y§ 13.0 11.6 7.9

Abbreviations: See Table 1. *Patients received control capsules plus 15 000 IU/d of vitamin A. †Expressed as mean loge ± SE in microvolts (number of patients sampled). ‡Lower normal value for 30-Hz ERG was 50 µV. §P = .008 (based on test for trend of slopes, where not on vitamin A prior to entry = 1; on vitamin A for 1-23 months prior to entry = 2; and on vitamin A for Ն24 months prior to entry = 3).

Ͻ A level for the patients in the control+A group (P .001). + The analysis shows that a 0.20-g/d dietary intake of ␻-3 fatty acids corresponds to an RBC PE DHA level of ap- 9 8 proximately 5%. 7 6 Association Between Vitamin A Intake 5 and RBC PE DHA Levels 4 3 2 In the control+A group (ie, those receiving control fatty r = 0.53; P < .001 acids plus vitamin A, 15000 IU/d), irrespective of vita- RBC PE Fatty Acids , % of Total 1 0 min A status prior to entry, mean RBC PE DHA level in- Group 0 0.10.20.30.4 0.5 0.6 0.7 ω creased significantly from 4.27% (weighted average of RBC PE DHA Among Patients in the Control Dietary -3 Fatty Acid Intake, g/d 4.39% and 4.04%; Table 1) at baseline to 4.78% at year Ͻ Figure 2. Linear regression of red blood cell phosphatidylethanolamine 4 (mean±SE increase, 0.51%±0.11%; P .001). Al- docosahexaenoic acid (RBC PE DHA) level on dietary ␻-3 fatty acid intake though intake of dietary ␻-3 fatty acids increased from based on 103 patients in the control plus vitamin A–treated (control+A) 0.18 to 0.26 g/d in the control+A group, a significant rise group. Each point represents the mean of 3 measurements of RBC PE DHA from baseline in RBC PE DHA remained after adjusting level and 5 measurements of dietary ␻-3 fatty acid intake over the course of ␻ the study. The correlation (lower right) indicates that 28% (r 2)ofthe for change in intake of dietary -3 fatty acids as mea- variation in RBC PE DHA level can be explained by the variation in dietary sured by the average intake of dietary ␻-3 fatty acids for ␻-3 fatty acid intake. Based on the regression line, a dietary ␻-3 fatty acid all follow-up visits minus the intake of dietary ␻-3 fatty intake of 0.20 g/d corresponds to an RBC PE DHA value of approximately 5% acids at baseline (P=.001). of total RBC PE fatty acids.

COMMENT field sensitivity at baseline (Table 1) and the amount of decline over the first 2 years (Figure 1A), the DHA+A The present study shows that among patients with reti- group lost, on average, 2% (25 dB/1276 dB), whereas the nitis pigmentosa not previously taking vitamin A, those control+A group lost, on average, 10% (ie, 141 dB/1436 who receive a new supplementation of a combination of dB); the saving was, therefore, about 8% over 2 years. This docosahexaenoic acid plus vitamin A have a signifi- beneficial effect did not persist beyond year 2. For those cantly slower rate of decline in visual field sensitivity and already taking 15000 IU/d of vitamin A prior to the on- 30-Hz ERG amplitude than those given vitamin A alone set of the trial, docosahexaenoic acid supplementation for 2 years. For this 2-year period, among those not al- did not provide additional benefit. ready taking vitamin A prior to entry, those given doco- Among the entire study population, those with lower sahexaenoic acid along with vitamin A lost, on average, RBC PE DHA levels (Ͻ5% of total RBC PE fatty acids) 25 dB of sensitivity, whereas those given vitamin A alone had a significantly faster rate of decline in visual field sen- lost, on average, 141 dB. Considering their total visual sitivity than those with higher RBC PE DHA levels over

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1311

ROS

IRBP ROL RAL Site 2 Site 2 Site 2 RAL ROL Site 1 RAL Site 1 Site 1

IPM

IRBP

Site 2 Site 2 Site 2 ROL

RAL [Palmitate] Site 1 Site 1 ROL Site 1

ROL RAL

RPE

DHA Palmitate

Figure 3. Model for interphotoreceptor retinoid-binding protein (IRBP)–mediated targeting of retinoids to their sites of action in the eye. The model is basedon the following 3 observations: (1) docosahexaenoic acid (DHA) is enriched in photoreceptor rather than retinal pigment epithelium (RPE); (2) this fatty acid constitutes a large fraction of fatty acids bound to IRBP endogenously; and (3) the addition of docosahexaenoic acid switches IRBP retinoid-binding site 2 from a state in which it possesses a high affinity for 11-cis retinal (RAL) to a state in which it is incapable of binding this ligand. It is proposed that when IRBP is localized near RPE, its fatty acid–binding site is occupied with fatty acids that are enriched in these cells, such as palmitate. Under these conditions, IRBP adopts a conformation (shaded) that confers a high affinity for RAL on retinoid-binding site 2 (bottom left corner of this scheme). When IRBP moves across the interphotoreceptor matrix (IPM) to the vicinity of the photoreceptors, its fatty acid content is readjusted according to the fatty acids prevalent in these cells to contain a large fraction of DHA. Binding of DHA switches the protein to a conformation (unshaded) in which site 2 has a very low affinity for RAL, resulting in rapid release of this ligand. Site 2 then becomes occupied with all-trans-retinol (ROL), the affinity for which remains high in the presence of DHA. The process is reversed when IRBP moves back to the proximity of RPE. The overall outcomes of the cycle are that RAL moves to photoreceptors, ROL is transported to RPE, and the fatty acids transfer down their concentration gradients. (The affinity of IRBP’s retinoid-binding site 1 for ROL and RAL is only marginally affected by fatty acids. This site may function simply as a buffer that nonspecifically allows for elevated concentrations of retinoids in the IPM.) ROS indicates rod outer segment. Reprinted with permission from Wolf.10 Used with permission from the International Life Sciences Institute, Washington, DC.

the 4 years of the study. Similarly, a significant inverse retinal (ie, the active form of vitamin A) that is essential relationship has been shown by others between rate of for photoreceptor survival.13,14 Because the RBC DHA level loss of cone ERG amplitude over 4 years and RBC DHA is thought to reflect the retinal level of DHA,2 we pro- level among patients with X-linked retinitis pigmen- pose that an RBC PE DHA fatty acid level less than 5% tosa.7 It has been proposed that a concentration gradi- may indicate that the subretinal level of DHA is insuffi- ent of DHA normally exists in the subretinal space be- cient to release 11-cis retinal from IRBP in patients with tween the rod outer segments (higher concentration) and retinitis pigmentosa, thereby contributing to death of pho- the retinal pigment epithelium (lower concentration) and toreceptor cells. It follows that increasing the RBC PE DHA that the release of 11-cis retinal from interphotorecep- level to 5% or higher through docosahexaenoic acid tor retinoid-binding protein (IRBP) is facilitated when supplementation or eating foods rich in ␻-3 fatty acids IRBP is exposed to sufficient DHA in the subretinal space7-9 may be necessary to increase the DHA level in the sub- (Figure 3 ). It is known that the DHA concentration is retinal space enough to allow release of 11-cis retinal from reduced in the outer retina in canine11 and murine12 mod- IRBP. els of hereditary retinal degeneration. Whether degen- Alternatively, increasing the concentration of 11- eration leads to loss of DHA or vice versa is not estab- cis retinal in the retina by supplementing with vitamin lished, but a lack of DHA may impair release of 11-cis A may facilitate incorporation of DHA into the retina. In

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1312

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 this regard, the control+A group (ie, patients receiving entry showed a significantly smaller loss of remaining control fatty acids plus vitamin A) showed a significant 30-Hz ERG amplitude per year (7.9%) compared with rise in their RBC PE DHA level during the course of the those not on vitamin A prior to entry (13%) or on vita- study that was independent of change in dietary ␻-3 fatty min A for less than 2 years (11.6%) (Table 5). A similar acid intake. This may partially explain why the treat- trend was seen for visual field results. These data sug- ment effect of docosahexaenoic acid was significant in gest that, in the absence of docosahexaenoic acid supple- the first 2 years of the study only among those who were mentation, at least 2 years are required before the ben- not taking vitamin A prior to entry. Those taking vita- eficial effect of vitamin A on the course of retinitis min A prior to entry may have already elevated their RBC pigmentosa is fully achieved. PE DHA level sufficiently to modify the course of their Together the results show that docosahexaenoic acid disease. Indeed, the mean level of RBC PE DHA was sig- supplementation shortens the interval for vitamin A to nificantly higher among patients on vs those not on vi- take its full effect and supports a recommendation that tamin A prior to entry. most adult patients with the typical forms of retinitis pig- A significant relationship was also observed be- mentosa who start vitamin A therapy at a dosage of 15000 tween the level of dietary ␻-3 fatty acid intake and the IU/d should also take docosahexaenoic acid, 1200 mg/d rates of decline of central and total field sensitivity among for 2 years (600 mg twice a day). No evidence was found patients in the control+A group on vitamin A prior to of a continued benefit of docosahexaenoic acid supple- entry (Table 4). A level of dietary intake of ␻-3 fatty ac- mentation beyond the first 2 years, leading to the rec- ids of at least 0.20 g/d, which corresponds to an RBC PE ommendation that patients receiving vitamin A should DHA level of about 5% or greater (Figure 2), was asso- stop docosahexaenoic acid supplementation after 2 years. ciated with a slower rate of decline in visual field sensi- A slight tendency toward adversity was noted in years 3 tivity than a level of intake of less than 0.20 g/d. These and 4 among patients on vitamin A, 15000 IU/d, prior findings support the hypothesis that patients with reti- to entry in the DHA+A group (Figure 1B), further sup- nitis pigmentosa taking vitamin A would benefit from porting discontinuation of docosahexaenoic acid therapy maintaining an average dietary intake of ␻-3 fatty acids at a dosage of 1200 mg/d after 2 years. of at least 0.20 g/d. This is equivalent to one to two 84- It must be emphasized that these conclusions are to 112-g (3- to 4-ounce) servings per week of fish rich based on group averages, and, therefore, no assurance in ␻-3 fatty acids, such as salmon, tuna, mackerel, sar- can be given that a specific patient will benefit from this dines, or herring.15 treatment. This study did not include patients younger The clinical significance of an average dietary in- than 18 years or those receving a dosage of less than 1200 take of ␻-3 fatty acids of at least 0.20 g/d among patients mg/d of docosahexaenoic acid, and therefore no formal with retinitis pigmentosa who receive vitamin A supple- recommendation can be made for younger patients or for ments can only be estimated. If the rates of decline of cen- a smaller dosage. The study also did not include pa- tral visual field sensitivity observed during this study per- tients with central visual field sensitivity in the HFA of sist long-term, and if the average central field sensitivity less than 250 dB with a size V test light or patients with of a patient aged 37 years is about 869 dB (Table 1, best-corrected visual acuity of less than 20/100 in both control+A group on vitamin A prior to entry), we pro- eyes, and therefore no formal recommendation can be pose that an average patient on vitamin A who main- made for such patients. Because patients were advised tains an intake of ␻-3 fatty acids of 0.20 g/d after 37 years to stop docosahexaenoic acid and vitamin A therapy dur- of age would be expected to lose about 21 dB per year ing pregnancy and because of the increased risk of birth (Table 4) and, therefore, would lose virtually all central defects among patients receiving high-dose vitamin A field sensitivity by 78 years of age (ie, [869/21]+37). In supplementation,16 patients who are pregnant or plan- contrast, an average patient receiving vitamin A with di- ning to become pregnant should not take this combina- etary intake of ␻-3 fatty acids of less than 0.20 g/d after tion of supplements. 37 years of age would be expected to lose 39 dB per year In addition, these data suggest that adult patients (Table 4) and would lose virtually all central field sen- with retinitis pigmentosa who have taken 15000 IU/d of sitivity by 59 years of age (ie, [869/39]+37). Therefore, vitamin A for at least 2 years should eat 1 to 2 servings maintenance of a dietary intake of ␻-3 fatty acids of at of fish rich in ␻-3 fatty acids weekly (equivalent to an least 0.20 g/d after 37 years of age could result in an es- average food intake of ␻-3 fatty acids of at least 0.20 g/d) timated 19 years of additional vision for the average pa- to maintain an RBC PE DHA level of at least 5% of total tient with retinitis pigmentosa who combines vitamin A, RBC PE fatty acids. A test to determine RBC PE DHA level 15000 IU/d, with a diet rich in ␻-3 fatty acids. is not widely available. As an alternative, physicians should In a previous trial of vitamin A for retinitis pigmen- consider obtaining a fasting RBC DHA level about 3 tosa, we reported that those receiving vitamin A palmi- months (ie, considering RBC turnover) after patients start tate, 15000 IU/d, showed an 8.3% annual rate of decline eating ␻-3-rich fish and periodically thereafter. We have in remaining 30-Hz ERG amplitude over 6 years com- observed a high correlation between RBC PE DHA and pared with a 10.0% annual decline among those in the RBC DHA levels in patients with retinitis pigmentosa not trace control group receiving vitamin A at a dosage of on docosahexaenoic acid supplementation (r=0.96, 75 IU/d.6 The beneficial effect of 15000 IU/d of vitamin n=49); an RBC DHA level of 4% of RBC total lipid fatty A was best seen after patients had received this dosage acids will ensure with 95% confidence that the RBC PE for several years.6 In the present 4-year study, patients DHA level is at least 5% in a given patient (E.L.B., B.R., on 15000 IU/d of vitamin A for 2 or more years prior to M.A.S., C.W.-D., and A.M., unpublished data, 2004).

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1313

©2004 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/30/2021 The present study also supports a previous recom- REFERENCES mendation6 that most adults with the typical forms of retinitis pigmentosa should continue to take 15000 IU/d of vitamin A palmitate under medical supervision to slow 1. Berson EL, Rosner B, Sandberg MA, et al. Clinical trial of docosahexaenoic acid in patients with retinitis pigmentosa receiving vitamin A treatment. Arch Oph- the course of their condition. It should be noted that the thalmol. 2004;122:1297-1305. precursor of vitamin A, betacarotene, is not predictably 2. Sarkadi-Nady E, Wijendran V, Diau GY, et al. The influence of prematurity and converted into vitamin A; therefore, betacarotene is not long chain polyunsaturate supplementation in 4-week adjusted age baboon neo- a suitable substitute for vitamin A palmitate in the con- nate brain and related tissues. Pediatr Res. 2003;54:244-252. text of this treatment regimen. 3. Anderson RE, Benolken RM, Dudley PA, Landis DJ, Wheeler TG. Proceedings: polyunsaturated fatty acids of photoreceptor membranes. Exp Eye Res. 1974; 18:205-213. Submitted for publication July 14, 2003; final revision re- 4. Neuringer M, Anderson GJ, Connor WE. The essentiality of ␻-3 fatty acids for ceived January 8, 2004; accepted April 19, 2004. the development and function of the retina and brain. Annu Rev Nutr. 1988;8: From the Berman-Gund Laboratory for the Study of 517-541. Retinal Degenerations, Harvard Medical School, Massa- 5. Birch DG, Birch EE, Hoffman DR, Uauy RD. Retinal development in very-low- birth-weight infants fed diets differing in omega-3 fatty acids. Invest Ophthal- chusetts Eye and Ear Infirmary (Drs Berson, Rosner, Sand- mol Vis Sci. 1992;33:2365-2376. berg, Brockhurst, and Gaudio and Mss Weigel-DiFranco and 6. Berson EL, Rosner B, Sandberg MA, et al. A randomized trial of vitamin A and Anderson), the Department of Nutrition, Harvard School of vitamin E supplementation for retinitis pigmentosa. Arch Ophthalmol. 1993;111: Public Health (Dr Willett), and the Lipid Metabolism Labo- 761-772. 7. Hoffman DR, Locke KG, Wheaton DH, Fish GE, Spencer R, Birch DG. A ran- ratory, Jean Mayer US Department of Agriculture Human domized, placebo-controlled clinical trial of docosahexaenoic acid supple- Nutrition Research, Center on Aging at Tufts University (Dr mentation for X-linked retinitis pigmentosa. Am J Ophthalmol. 2004;137:704- Schaefer), Boston, Mass; the Kennedy Krieger Institute, Per- 718. oxisomal Diseases Laboratory, Baltimore, Md (Ms Moser); 8. Chen Y, Houghton LA, Brenna JT, Noy N. Docosahexaenoic acid modulates the the Foster Biomedical Research Laboratory, Brandeis Uni- interactions of the interphotoreceptor retinoid-binding protein with 11-cis- retinal. J Biol Chem. 1996;271:20507-20515. versity, Waltham, Mass (Dr Hayes); and the Devers Eye In- 9. Chen Y, Saari JC, Noy N. Interactions of all-trans-retinol and long-chain fatty ac- stitute, Portland, Ore (Dr Johnson). The authors have no ids with interphotoreceptor retinoid-binding protein. Biochemistry. 1993;32: relevant financial interest in this article. 11311-11318. This study was supported by grant U10EY11030 from 10. Wolf G. Transport of retinoids by the interphotoreceptor retinoid-binding pro- the National Eye Institute, Bethesda, Md, and in part by the tein. Nutr Rev. 1998;56:156-158. 11. Aguirre GD, Acland GM, Maude MB, Anderson RE. Diets enriched in docosa- Foundation Fighting Blindness, Owings Mills, Md. hexaenoic acid fail to correct progressive rod-cone degeneration (prcd) pheno- We thank the study patients and their families and grate- type. Invest Ophthalmol Vis Sci. 1997;38:2387-2407. fully acknowledge the following individuals who contrib- 12. Anderson RE, Maude MB, Bok D. Low docosahexaenoic acid levels in rod outer uted to the conduct of this trial: Tina Skop-Chaput, Michele segment membranes of mice with rds/peripherin and P2 16L peripherin muta- tions. Invest Ophthalmol Vis Sci. 2001;42:1715-1720. Berry, Melissa Stillberger, Peggy Rodriguez, Kevin McDer- 13. Dowling JE, Wald G. The biological function of vitamin A acid. Proc Natl Acad mott, Linda Berard, Heather Lee, Susana Chung, Shyana Sci U S A. 1960;46:587-608. Harper, Anna Marie Baglieri, Ciara Rice, Cathy Loner- 14. Dowling JE, Gibbons IR. The effect of vitamin A deficiency on the fine structure gan, Suzanne Dalton, Marion McPhee, Martin Van Den- of the retina. In: Smelser G, ed. The Structure of the Eye. Orlando, Fla: Academic burgh, Sherrie Kaplan, PhD, Anita Liu, and David Jones. Press Inc; 1961:85-99. 15. Kris-Etherton PM, Harris WS, Appel LJ. Fish consumption, fish oil, omega-3 fatty Correspondence: Eliot L. Berson, MD, Berman-Gund acids, and cardiovascular disease. Circulation. 2002;106:2747-2757. Laboratory for the Study of Retinal Degenerations, 243 16. Rothman KJ, Moore LL, Singer MR, Nguyen US, Mannino S, Milunsky A. Tera- Charles St, Boston, MA 02114. togenicity of high vitamin A intake. N Engl J Med. 1995;333:1369-1373.

(REPRINTED) ARCH OPHTHALMOL / VOL 122, SEP 2004 WWW.ARCHOPHTHALMOL.COM 1314