Clinical Trial of Docosahexaenoic Acid in Patients With Retinitis PigmentosaReceiving Vitamin A Treatment

Objective  To determine whether a therapeutic dose of docosahexaenoic acid (DHA),an ω-3 fatty acid, will slow the course of retinal degeneration in adultpatients with retinitis pigmentosa who are also receiving vitamin A.

Design  Randomized, controlled, double-masked trial of 221 patients, aged 18to 55 years, evaluated over a 4-year interval. Patients were given either1200 mg/d of docosahexaenoic acid or control capsules. All were given 15 000IU/d of vitamin A (given as retinyl palmitate). Randomization considered genetictype and baseline dietary ω-3 fatty acid intake.

Main Outcome Measures  The primary outcome measure was the total point score for the 30-2 programof the Humphrey field analyzer; secondary outcome measures were the totalpoint score for the 30-2 and 30/60-1 programs combined, 30-Hz electroretinogramamplitude, and Early Treatment Diabetic Rentinopathy Study visual acuity.

Results  No significant differences in decline in ocular function were foundbetween the docosahexaenoic acid plus vitamin A (DHA + A) group and controlplus vitamin A (control + A) group over a 4-year interval among all 221 randomizedpatients or among the 208 patients who completed all 4 follow-up visits. Themean annual rate of loss of sensitivity for the Humphrey Field Analyzer 30-2program was 37 dB for the DHA + A group and 38 dB for the control + A group(P = .88). For the Humphrey Field Analyzer 30-2 and30/60-1 programs combined, the mean annual rates of loss of field sensitivitywere 57 dB for the DHA + A group and 60 dB (P = .73)for control + A group. No toxic adverse effects were observed. No significantdifferences by treatment group assignment were observed within genetic typesor within the category of baseline ω-3 fatty acid intake.

Conclusion  In patients assigned to receive 15 000 IU/d of vitamin A, thisrandomized trial showed that 1200 mg/d of docosahexaenoic acid supplementationover a 4-year interval did not, on average, slow the course of disease inpatients with retinitis pigmentosa.

Retinitis pigmentosa has a worldwide prevalence of about 1 in 4000 persons.^1-10 Affectedpatients typically show elevated dark-adaptation thresholds,^11,12 constrictedvisual fields, retinal arteriolar attenuation, intraretinal pigment aroundthe midperiphery, and reduced and delayed electroretinograms (ERGs).^13,14 Histological studies have shown thatvisual loss occurs owing to degeneration of rod and cone photoreceptors.^15,16 A prior randomized, controlled trialshowed that patients treated with vitamin A palmitate, 15 000 IU/d, had,on average, a slower rate of decline of retinal function as monitored by theERG compared with those not receiving this dose.¹⁷

We and others have found that some patients with retinitis pigmentosahave decreased mean plasma and red blood cell (RBC) docosahexaenoic acid (DHA)concentrations.^18-24 Rodouter segment membranes contain exceptionally large amounts of polyunsaturatedfatty acids, especially DHA. In fact, they compose almost half of the esterifiedfatty acids in the outer segment phospholipids.^25-27 Docosahexaenoicacid is a long chain ω-3 fatty acid found particularly in some fish(eg, salmon and tuna). The RBC DHA levels have been strongly correlated withDHA levels in the retina (r = 0.88).²⁸ Theproportion of DHA in outer segment phosphatidylethanolamine (PE) relativeto other fatty acids is typically 10- to 15-fold higher than in nonneuraltissue such as RBC membranes.^25,27 Thespecial lipid composition of rod outer segments is thought to be necessaryto maintain cellular membrane disc fluidity for the normal functioning ofrhodopsin as it changes conformation in the initial stages of phototransduction.^26,29-33 Furthermore,it has been proposed that a concentration gradient of DHA normally existsin the subretinal space between the rod outer segments (higher concentration)and the retinal pigment epithelium (lower concentration) and that the releaseof 11-cis retinal from interphotoreceptor retinoid-bindingprotein is facilitated when interphotoreceptor retinoid-binding protein isexposed to a sufficient DHA concentration in the subretinal space.^34-37

Because RBC membrane phospholipid levels are similar in compositionto, and change in parallel with, retinal phospholipid levels with dietarymanipulation,^38-41 weevaluated the DHA content of RBC PE in patients with retinitis pigmentosa.We found that RBC PE DHA levels in our patients were significantly lower,on average, than in subjects with normal vision.²⁴ Inan analysis of longitudinal data from participants in our previous trial ofvitamin A and/or vitamin E for whom we had RBCs available for analysis atyear 3 or 4 of follow-up (n = 61), we found that the decline in 30-Hz ERGamplitude in these patients over a 4-year interval was inversely related tothe RBC PE DHA concentration (P = .03) for all genetictypes combined. These results persisted after controlling for age, sex, genetictype, baseline ERG amplitude, and treatment group assignment in a multipleregression model (P = .05).

The significant positive relationship between RBC PE DHA concentrationand the preservation of ERG amplitude, the known role of DHA in maintainingnormal photoreceptor function, and the evidence that vitamin A slows the averagedecline in ERG amplitude in patients with retinitis pigmentosa prompted usto conduct this randomized, controlled clinical trial to determine whetherorally administered docosahexaenoic acid can halt or slow the course of thetypical forms of retinitis pigmentosa among patients receiving vitamin A (givenas retinyl palmitate) therapy.

We first conducted a pilot study on 35 patients with the typical formsof retinitis pigmentosa who received 600 mg/d (n = 9) of docosahexaenoic acid,1200 mg/d (n = 9) of docosahexaenoic acid, or a placebo control (n = 17) over3 months to determine to what extent these dosages raised the concentrationof RBC PE DHA in the short-term. All participants were also treated with 15 000IU/d of vitamin A. Both dosages of docosahexaenoic acid at least doubled theRBC PE DHA concentration and were safely tolerated in this population. Wechose the higher dosage of docosahexaenoic acid for this trial assuming thatany therapeutic effect would be more readily detected with the higher dosage.

We screened patients with retinitis pigmentosa for eligibility accordingto ocular, dietary, and medical criteria (Table 1). We performed a baseline examination on eligible patientswithin 8 weeks of the screening examination. At baseline, patients were randomlyassigned to 1 of 2 groups—those who received a supplement of 6 capsulesper day each containing 500 mg of fatty acids of which 200 mg was docosahexaenoicacid for a total of 1200 mg/d of docosahexaenoic acid or those who received6 placebo capsules per day containing 500 mg of fatty acids with no docosahexaenoicacid. The docosahexaenoic acid–enriched capsules, provided by MartekBiosciences Corporation, Columbia, Md, contained a vegetable oil from microalgaewith ascorbyl palmitate (<0.001 mg) and tocopherols (<0.001 mg) as antioxidants.The fatty acid composition was as follows: capric, 0 to 10 mg; lauric, 0 to30 mg; myristic, 50 to 100 mg; palmitic, 50 to 100 mg; palmitoleic, 0 to 10mg; stearic, 0 to 10 mg; oleic, 50 to 150 mg; linoleic, 0 to 25 mg; docosahexaenoic,190 to 210 mg; nervonic, 0 to 10 mg; and others, 0 to 15 mg. The control capsules,also provided by Martek Biosciences Corporation, contained a mixture of 50%corn and 50% soybean oil with ascorbyl palmitate (<0.001 mg) and tocopherols(<0.001 mg) as antioxidants. Their fatty acid composition was as follows:palmitic, 45 to 55 mg; stearic, 10 to 20 mg; oleic, 110 to 125 mg; and α-linolenic,15 to 25 mg. Patients were instructed to take 3 capsules in the morning and3 capsules in the evening. Both the docosahexaenoic acid plus vitamin A (DHA+ A)–supplemented group and the control plus vitamin A (control + A)group received 3 g of fatty acids from these capsules; on a 2000-calorie dietthis constituted 1.4% of calories as fat or an increase of 27 cal/d. All patientswere given vitamin A as 15 000 IU of retinyl palmitate in tablets (AkornOphthalmics, Buffalo Grove, Ill) and instructed to take 1 tablet daily withbreakfast. Patients completed a food frequency questionnaire^42,43 andmedical questionnaire at each visit with the aid of a clinical coordinator;they were followed up annually over 4 years (Table 2). Plasma and RBC PE DHA levels were monitored as measuresof compliance,⁴⁴ and serum retinol and retinylester levels as well as serum liver function test results were evaluated tohelp exclude any toxic effects of vitamin A therapy.⁴⁵

We used the measurement of static perimetric sensitivities (ie, totalpoint score) with the 30-2 program size V target in the Humphrey Field Analyzer(HFA) (Carl Zeiss Ophthalmic Systems, Inc, Dublin, Calif) as the primary outcomemeasure. The size V target was used to minimize the number of locations withfloor effects (sensitivity ≤0 dB). The FASTPAC test strategy was used totest both central (30-2) and peripheral (30/60-1) visual fields in as shorta time as possible.^46-48 Becausethe area of the visual field is statistically related to ERG amplitude⁴⁹ and because our preliminary data derived from the30-Hz ERG amplitude, we used the full-field 30-Hz ERG amplitude as a secondaryoutcome measure. Visual acuity (Early Treatment Diabetic Retinopathy Study[ETDRS])⁵⁰ and the total point score to a sizeV target with the HFA 30-2 and 30/60-1 programs combined were also followedup as secondary outcome measures.

We estimated that 220 patients were needed to provide sufficient power(ie, α = .05, β = .10) to observe a statistically significant difference(29 dB) between mean change in the DHA + A and control + A groups with respectto HFA 30-2 total point score over a 4-year interval and allowing for 5% attrition.The project was approved by the institutional review boards of the MassachusettsEye and Ear Infirmary and Harvard Medical School, Boston, and the study conformedto the Declaration of Helsinki. All patients signed an informed consent formprior to the screening examination and, if eligible, prior to the baselineexamination as well. The trial was funded by the National Eye Institute asa single-center study. A Data and Safety Monitoring Committee selected bythe National Eye Institute met with the investigators prior to the onset ofthe study to approve the protocol and annually thereafter in closed sessionto review the results for both patient safety and efficacy. The study wasplanned to allow 4 years of follow-up for each patient. The Lan-DeMets α-spendingapproach with an O'Brien-Fleming boundary with 5 looks was prespecified⁵¹ as the stopping rule. However, this stopping rulewas just a guideline as the Data and Safety Monitoring Committee needed toconsider all information from several measures of efficacy and safety in makinginformed decisions about stopping the study.

The procedure for randomization considered the estimated dietary intakeand genetic type at the screening examination. A method of block randomizationwas used within 2 groups of ω-3 fatty acid intake (above and below themedian of 0.16 g/d determined from a food frequency questionnaire administeredto patients in our previous therapeutic trial of vitamin A) and 6 genetictypes (dominant without known mutation, dominant with known rhodopsin or rhodopsin/peripherinmutation, recessive, X-linked, isolate, or undetermined [eg, adopted]) or12 strata. Within each stratum, patients were randomized in equal numbersto the DHA + A group and control + A group in blocks of size 4. A separateset of randomization assignments was maintained for each stratum based ona computer-generated set of random numbers that was available only to a programmerwho provided assignment information to the data manager (C.W.D.) on a case-by-casebasis. Group assignment was implemented by the data manager.

All members of the staff in contact with the patients, including theprincipal investigator (E.L.B.), were masked with regard to each patient'streatment group assignment. Each ocular examination was performed withoutreview of previous records. All serum samples were analyzed without knowledgeof treatment group assignment. Patients did not know the contents of the supplementunder study or their treatment group assignment and also agreed not to knowthe course of their retinal degeneration until the end of the study. Treatmentgroup assignments and plasma DHA and RBC PE DHA levels were placed in recordsseparate from that used for ocular examinations as part of masking those incontact with the patients.

Outcome data for a given patient for each visit represented the averageof test results from both eyes or for a single eye if data for the other eyewere unavailable. Visual field data (total point scores) were analyzed separatelyfor the central field (HFA 30-2 program) and for total field (HFA 30-2 and30/60-1 programs combined) when both were available. If an eye became pseudophakicafter the baseline visit, visual field data for that eye were analyzed onlyfor those visits prior to cataract surgery. If the total point score for avisual field became zero, the visit at which the zero score was first obtainedwas included in the analyses but subsequent visits were censored. Slopes andchanges from baseline were computed for each patient, and average slopes andmean changes were evaluated for each parameter of ocular function by treatmentgroup. Comparisons by assigned treatment group were also performed withingenetic type and within category (ie, above and below the intake of 0.16 g/d)at baseline of dietary ω-3 fatty acid intake. Longitudinal regressionanalyses were then performed based on PROC MIXED of SAS version 6.12.⁵²

The regression models that we used in this clinical trial generallyhave the following form:

(1) Y_it = a + b₁G_i + b₂t + b₃tG_i + e_it,

where Y_it indicates outcome variable (eg, HFA 30-2 total pointscore) for the ith subject at time t; G_i, indicator variable for treatment group, equals 1 ifpatient i is in group 1, equals 0 if patient i is in group 2; t, time in yearsfrom the screening visit, t equals 0, 1, 2, . . . , 4;e_it, error term that represents within-personvariation for the ith patient at time t; and Corr (e_it1, e_it2) = r. We used PROC MIXED to fit model 1 where both a and b = (b₁, b₂, b₃) are assumed to be random.

Although the baseline value is not explicitly specified as a covariatein equation 1, it is implicitly adjusted for because of the correlation betweenthe error terms for the baseline visit and each follow-up visit. In particular,the marginal model in equation 1 can be rewritten in conditional form as follows:

(2) Y_it = [a(1− R)] + RY_i0 + [b₁(1 − R)G_i] + b₂t + b₃tG_i + u_it,

where Corr(u_it1,u_it2) equals [R/(1 + R)].

Equation 2 is a conditional model. Thus, b₂ can be interpreted as the slope in the control + A group (ie, group2), b₂ + b₃ can be interpreted as the slope in the DHA + A group (ie, group 1),and b₃ is the difference in slope betweenthe DHA + A group and the control + A group after controlling for baselinevalues.⁵³

Two hundred eight patients were seen for all 4 years of follow-up andthe results will focus on these patients. Sensitivity analyses were performedon all 221 patients randomized (1) for all available data and (2) after usingmultiple imputation methods to account for missing data.⁵⁴

From May 13, 1996, to September 26, 1997, we examined 456 patients fromacross the United States to identify 221 patients (1 per family) with retinitispigmentosa who met the preset list of eligibility criteria. Two hundred eightof these patients completed all 4 annual follow-up visits. Baseline characteristicsof these 208 patients for factors used in the stratification of the randomization(ie, genetic type and dietary ω-3 fatty acid intake) are given in Table 3 by assigned treatment group. Alsogiven is baseline information concerning the distribution of demographic factors,ocular function, and laboratory measures of DHA and retinol concentrations.Ninety-seven percent of the eligible patients had intraretinal bone spiculepigmentation around the midperipheral fundus. Eleven percent of patients reportedpartial hearing loss. Six percent of the study population were minorities.With the exception of RBC PE DHA levels, no significant differences in anyof the above patient characteristics were noted between the DHA + A and control+ A groups. Fifty percent of patients in the DHA + A group and 64% of patientsin the control + A group had cataracts in at least 1 eye at baseline (P = .05).

Capsule counts indicated that 92% of the docosahexaenoic acid capsules,92% of the control capsules, and 94% of the vitamin A tablets were consumedover all 4 years. Similar results were seen with returned monthly calendars.One patient in the control + A group diagnosed as having breast cancer justbefore her year 4 visit died 4 months after the year 4 visit. No other patientreported the development of cancer to us during the study. No patient experienceda complete loss of vision in an eye over the course of the study. Furthermore,there was no evidence of systemic illness or a toxic effect that was attributableto the study capsules or to vitamin A during the course of this study basedon complete blood cell counts, liver function assessments, patient responsesto a symptom questionnaire, and on the serum retinol and serum retinyl esterlevels. Two patients were examined by the consulting internist, 1 for gastrointestinaldistress thought subsequently to be due to anxiety and the other for discolorationof the skin that proved subsequently to be related to Cushing syndrome.

Prior to treatment, the DHA + A and control + A groups showed comparablelevels of plasma DHA (Table 3).At follow-up, mean ± SE plasma DHA percentages (mean of all follow-upmeasurements) of total plasma fatty acids were 5.12% ± 0.12% and 1.61%± 0.07% for the DHA + A group vs the control + A group, respectively;mean change from baseline for the DHA + A group was significantly differentfrom the mean change from baseline for the control + A group (P<.001). By year 1 the mean plasma DHA percentage of total fattyacids was 3-fold higher in the DHA + A group than in the control + A groupand remained at that level for the 4 years of follow-up; no significant increasein plasma DHA percentage was noted in the control + A group.

Prior to treatment, the mean ± SE RBC PE DHA percentage of totalRBC PE fatty acids was significantly higher in the DHA + A group (4.77% ±0.18%) vs the control + A group (4.27% ± 0.14%) (P = .03) (Table 3). Atyear 1 of follow-up the mean ± SE RBC PE DHA percentages of total RBCPE fatty acids for the DHA + A group vs the control + A group were 12.62%± 0.28% and 4.90% ± 0.18%; at year 4 the mean RBC PE DHA percentagesof total fatty acids were 12.83% ± 0.05% and 4.66% ± 0.18% forthe 2 groups, respectively. Averaging years 1 and 4, the mean ± SEchange from baseline for the DHA + A group (7.95 ± 0.30) and mean changefrom baseline for the control + A group (0.51 ± 0.11) were both significantlydifferent from zero (P<.001) and were significantlydifferent from each other (P<.001). The RBC PEDHA values ranged from 2% to 18% of the total RBC PE fatty acids prior totreatment in both groups and at follow-up from 3% to 17% of the total RBCPE fatty acids in the DHA + A group and from 2% to 10% in the control + Agroup (Figure 1).

Table 4 lists serum triglycerideand cholesterol levels at screening and year 4 for the 2 groups. The differencein slopes for the 2 groups was significant for the levels of triglycerides,total cholesterol, and low-density lipoprotein cholesterol.

Figure 2 shows mean ±SE values by year and treatment group for central field (HFA 30-2 program)sensitivity, total field (HFA 30-2 and 30/60-1 programs combined) sensitivity,30-Hz ERG amplitude, and ETDRS visual acuity. Similar declines can be seenin all outcome measures in the DHA + A group vs the control + A group. Meanchange analyses from time zero (mean of screening and baseline values) didnot reveal statistically significant differences between the 2 treatment groupsfor any of these 4 measures.

Table 5 summarizes the meanannual rates of decline of visual field sensitivity to the HFA 30-2 programand the HFA 30-2 and 30/60-1 programs combined, of 30-Hz ERG amplitude, andof ETDRS visual acuity for the DHA + A group vs the control + A group amongthe 208 patients followed up for each of 4 annual visits. No significant differencesin rates of change were observed between these 2 groups. Both groups lost,on average, about 37 to 38 dB per year to the HFA 30-2 program condition and57 to 60 dB to the HFA 30-2 and 30/60-1 programs combined. These total pointscore declines correspond to losses of approximately 0.5 dB and 0.4 dB peryear, respectively, for an average location in the visual field. Over 4 years,analysis of 30-Hz ERGs showed that the mean annual rates of decline of remainingfunction were 9.92% in the DHA + A group and 10.49% in the control + A group,which was not statistically significantly different (P =.64). These analyses were also performed on all available data, includingpatients with partial follow-up, but with missing values left as missing andafter using multiple imputation methods to account for missing data amongpatients with incomplete follow-up. No substantive differences were notedbetween results obtained with these additional approaches compared with thosein Table 5. A similar analysisas in Table 5 was performed for0.5-Hz ERG amplitudes based on the 55% of patients with quantifiable responses;mean rate of decline for the DHA + A group (11.8%) was not significantly differentfrom that for the control + A group (10.4%) (P =.54).

We also performed randomized comparisons by genetic type and by dietaryintake of ω-3 fatty acids. No significant differences by treatment groupassignment were observed for either the primary or the secondary outcome measureswithin the dominant, recessive, X-linked, or isolate forms of retinitis pigmentosaor within the category of baseline dietary ω-3 fatty acid intake (aboveand below intake of 0.16 g/d) (data not shown).

The present trial shows that in adult patients with retinitis pigmentosa,assigned to 15 000 IU/d of vitamin A, a daily supplement of 1200 mg ofdocosahexaenoic acid over a 4-year interval does not, on average, slow thecourse of retinal degeneration. Both the DHA + A and control + A groups lost37 to 38 dB of visual field sensitivity to the HFA 30-2 program (size V target),58 to 60 dB of visual field sensitivity to the HFA 30-2 and 30/60-1 programscombined (size V target), 10% of remaining 30-Hz cone ERG amplitude, and 0.7letter of ETDRS visual acuity per year. No significant toxic adverse effectswere noted over the course of this trial both for docosahexaenoic acid orvitamin A supplementation. Similarly, docosahexaenoic acid supplementation(400 mg/d) was associated with no identifiable safety risks over 4 years among23 males with X-linked retinitis pigmentosa.⁵⁵ Nosignificant treatment effect was noted within any of the genetic types. Serumtriglyceride levels increased by about 25% in the control + A group as describedpreviously.⁴⁵ In both groups the high-densitylipoprotein cholesterol level increased and the low-density lipoprotein cholesterollevel decreased or remained stable. On balance, changes in these values wouldnot increase the risk for heart disease.

Good separation as monitored by mean RBC PE DHA levels was achievedbetween the DHA + A group vs the control + A group. The DHA + A group hada mean RBC PE DHA level 2- to 3-fold greater than the control + A group at1 year and 4 years. At follow-up, averaging years 1 and 4, the mean RBC PEDHA level increased by 0.51% from baseline for the control + A group (significantlydifferent from zero, P<.001) and increased 7.95%for the DHA + A group. How this RBC PE DHA change in the control + A groupmay have influenced their results and, thus, the comparison of results tothe DHA + A group is unclear.

The lack of a significant difference in rate of loss of ocular functionbetween the DHA + A and control + A groups precludes any general recommendationof DHA supplementation for patients with retinitis pigmentosa receiving vitaminA. Further investigation of these data from this trial or other trials mayhelp to clarify whether any subgroups of patients with retinitis pigmentosawill benefit from a combination of docosahexaenoic acid plus vitamin A supplementation.

Correspondence: Eliot L. Berson, MD, Berman-Gund Laboratory for theStudy of Retinal Degenerations, 243 Charles St, Boston, MA 02114.

Submitted for publication July 11, 2003; final revision received January8, 2004; accepted April 19, 2004.

This study was supported by grant U10EY11030 from the National Eye Institute,Bethesda, Md, and in part by the Foundation Fighting Blindness, Owings Mills,Md.

We thank the study patients and their families and gratefully acknowledgethe following individuals who contributed to the conduct of this trial: TinaSkop-Chaput, Michele Berry, Melissa Stillberger, Peggy Rodriguez, Kevin McDermott,Linda Berard, Heather Lee, Susana Chung, Shyana Harper, Anna Maria Baglieri,Ciara Rice, Cathy Lonergan, Suzanne Dalton, Marion McPhee, Martin Van Denburgh,Anita Liu, David Jones, and Sherrie Kaplan, PhD.

Members of the Data and Safety Monitoring Committee were Marian Fisher,PhD (chair); George Bresnick, MD; Baruch A. Brody, PhD; Barry Davis, MD, PhD;Natalie Kurinij, PhD (ex officio); Carol M. Mangione, MD; James Olson, PhD(1996-1999); Peter R. Pavan, MD (2000-2003); Sander J. Robins, MD; PamelaSample, PhD; and Barbara A. Underwood, PhD (2001-2003).